Merge from emacs-23
[bpt/emacs.git] / lisp / progmodes / cc-engine.el
1 ;;; cc-engine.el --- core syntax guessing engine for CC mode
2
3 ;; Copyright (C) 1985, 1987, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
4 ;; 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
5 ;; Free Software Foundation, Inc.
6
7 ;; Authors: 2001- Alan Mackenzie
8 ;; 1998- Martin Stjernholm
9 ;; 1992-1999 Barry A. Warsaw
10 ;; 1987 Dave Detlefs
11 ;; 1987 Stewart Clamen
12 ;; 1985 Richard M. Stallman
13 ;; Maintainer: bug-cc-mode@gnu.org
14 ;; Created: 22-Apr-1997 (split from cc-mode.el)
15 ;; Version: See cc-mode.el
16 ;; Keywords: c languages oop
17
18 ;; This file is part of GNU Emacs.
19
20 ;; GNU Emacs is free software: you can redistribute it and/or modify
21 ;; it under the terms of the GNU General Public License as published by
22 ;; the Free Software Foundation, either version 3 of the License, or
23 ;; (at your option) any later version.
24
25 ;; GNU Emacs is distributed in the hope that it will be useful,
26 ;; but WITHOUT ANY WARRANTY; without even the implied warranty of
27 ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
28 ;; GNU General Public License for more details.
29
30 ;; You should have received a copy of the GNU General Public License
31 ;; along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>.
32
33 ;;; Commentary:
34
35 ;; The functions which have docstring documentation can be considered
36 ;; part of an API which other packages can use in CC Mode buffers.
37 ;; Otoh, undocumented functions and functions with the documentation
38 ;; in comments are considered purely internal and can change semantics
39 ;; or even disappear in the future.
40 ;;
41 ;; (This policy applies to CC Mode as a whole, not just this file. It
42 ;; probably also applies to many other Emacs packages, but here it's
43 ;; clearly spelled out.)
44
45 ;; Hidden buffer changes
46 ;;
47 ;; Various functions in CC Mode use text properties for caching and
48 ;; syntactic markup purposes, and those of them that might modify such
49 ;; properties but still don't modify the buffer in a visible way are
50 ;; said to do "hidden buffer changes". They should be used within
51 ;; `c-save-buffer-state' or a similar function that saves and restores
52 ;; buffer modifiedness, disables buffer change hooks, etc.
53 ;;
54 ;; Interactive functions are assumed to not do hidden buffer changes,
55 ;; except in the specific parts of them that do real changes.
56 ;;
57 ;; Lineup functions are assumed to do hidden buffer changes. They
58 ;; must not do real changes, though.
59 ;;
60 ;; All other functions that do hidden buffer changes have that noted
61 ;; in their doc string or comment.
62 ;;
63 ;; The intention with this system is to avoid wrapping every leaf
64 ;; function that do hidden buffer changes inside
65 ;; `c-save-buffer-state'. It should be used as near the top of the
66 ;; interactive functions as possible.
67 ;;
68 ;; Functions called during font locking are allowed to do hidden
69 ;; buffer changes since the font-lock package run them in a context
70 ;; similar to `c-save-buffer-state' (in fact, that function is heavily
71 ;; inspired by `save-buffer-state' in the font-lock package).
72
73 ;; Use of text properties
74 ;;
75 ;; CC Mode uses several text properties internally to mark up various
76 ;; positions, e.g. to improve speed and to eliminate glitches in
77 ;; interactive refontification.
78 ;;
79 ;; Note: This doc is for internal use only. Other packages should not
80 ;; assume that these text properties are used as described here.
81 ;;
82 ;; 'category
83 ;; Used for "indirection". With its help, some other property can
84 ;; be cheaply and easily switched on or off everywhere it occurs.
85 ;;
86 ;; 'syntax-table
87 ;; Used to modify the syntax of some characters. It is used to
88 ;; mark the "<" and ">" of angle bracket parens with paren syntax, and
89 ;; to "hide" obtrusive characters in preprocessor lines.
90 ;;
91 ;; This property is used on single characters and is therefore
92 ;; always treated as front and rear nonsticky (or start and end open
93 ;; in XEmacs vocabulary). It's therefore installed on
94 ;; `text-property-default-nonsticky' if that variable exists (Emacs
95 ;; >= 21).
96 ;;
97 ;; 'c-is-sws and 'c-in-sws
98 ;; Used by `c-forward-syntactic-ws' and `c-backward-syntactic-ws' to
99 ;; speed them up. See the comment blurb before `c-put-is-sws'
100 ;; below for further details.
101 ;;
102 ;; 'c-type
103 ;; This property is used on single characters to mark positions with
104 ;; special syntactic relevance of various sorts. Its primary use is
105 ;; to avoid glitches when multiline constructs are refontified
106 ;; interactively (on font lock decoration level 3). It's cleared in
107 ;; a region before it's fontified and is then put on relevant chars
108 ;; in that region as they are encountered during the fontification.
109 ;; The value specifies the kind of position:
110 ;;
111 ;; 'c-decl-arg-start
112 ;; Put on the last char of the token preceding each declaration
113 ;; inside a declaration style arglist (typically in a function
114 ;; prototype).
115 ;;
116 ;; 'c-decl-end
117 ;; Put on the last char of the token preceding a declaration.
118 ;; This is used in cases where declaration boundaries can't be
119 ;; recognized simply by looking for a token like ";" or "}".
120 ;; `c-type-decl-end-used' must be set if this is used (see also
121 ;; `c-find-decl-spots').
122 ;;
123 ;; 'c-<>-arg-sep
124 ;; Put on the commas that separate arguments in angle bracket
125 ;; arglists like C++ template arglists.
126 ;;
127 ;; 'c-decl-id-start and 'c-decl-type-start
128 ;; Put on the last char of the token preceding each declarator
129 ;; in the declarator list of a declaration. They are also used
130 ;; between the identifiers cases like enum declarations.
131 ;; 'c-decl-type-start is used when the declarators are types,
132 ;; 'c-decl-id-start otherwise.
133 ;;
134 ;; 'c-awk-NL-prop
135 ;; Used in AWK mode to mark the various kinds of newlines. See
136 ;; cc-awk.el.
137
138 ;;; Code:
139
140 (eval-when-compile
141 (let ((load-path
142 (if (and (boundp 'byte-compile-dest-file)
143 (stringp byte-compile-dest-file))
144 (cons (file-name-directory byte-compile-dest-file) load-path)
145 load-path)))
146 (load "cc-bytecomp" nil t)))
147
148 (cc-require 'cc-defs)
149 (cc-require-when-compile 'cc-langs)
150 (cc-require 'cc-vars)
151
152 ;; Silence the compiler.
153 (cc-bytecomp-defun buffer-syntactic-context) ; XEmacs
154
155 \f
156 ;; Make declarations for all the `c-lang-defvar' variables in cc-langs.
157
158 (defmacro c-declare-lang-variables ()
159 `(progn
160 ,@(apply 'nconc
161 (mapcar (lambda (init)
162 `(,(if (elt init 2)
163 `(defvar ,(car init) nil ,(elt init 2))
164 `(defvar ,(car init) nil))
165 (make-variable-buffer-local ',(car init))))
166 (cdr c-lang-variable-inits)))))
167 (c-declare-lang-variables)
168
169 \f
170 ;;; Internal state variables.
171
172 ;; Internal state of hungry delete key feature
173 (defvar c-hungry-delete-key nil)
174 (make-variable-buffer-local 'c-hungry-delete-key)
175
176 ;; The electric flag (toggled by `c-toggle-electric-state').
177 ;; If t, electric actions (like automatic reindentation, and (if
178 ;; c-auto-newline is also set) auto newlining) will happen when an electric
179 ;; key like `{' is pressed (or an electric keyword like `else').
180 (defvar c-electric-flag t)
181 (make-variable-buffer-local 'c-electric-flag)
182
183 ;; Internal state of auto newline feature.
184 (defvar c-auto-newline nil)
185 (make-variable-buffer-local 'c-auto-newline)
186
187 ;; Included in the mode line to indicate the active submodes.
188 ;; (defvar c-submode-indicators nil)
189 ;; (make-variable-buffer-local 'c-submode-indicators)
190
191 (defun c-calculate-state (arg prevstate)
192 ;; Calculate the new state of PREVSTATE, t or nil, based on arg. If
193 ;; arg is nil or zero, toggle the state. If arg is negative, turn
194 ;; the state off, and if arg is positive, turn the state on
195 (if (or (not arg)
196 (zerop (setq arg (prefix-numeric-value arg))))
197 (not prevstate)
198 (> arg 0)))
199
200 ;; Dynamically bound cache for `c-in-literal'.
201 (defvar c-in-literal-cache t)
202
203 \f
204 ;; Basic handling of preprocessor directives.
205
206 ;; This is a dynamically bound cache used together with
207 ;; `c-query-macro-start' and `c-query-and-set-macro-start'. It only
208 ;; works as long as point doesn't cross a macro boundary.
209 (defvar c-macro-start 'unknown)
210
211 (defsubst c-query-and-set-macro-start ()
212 (if (symbolp c-macro-start)
213 (setq c-macro-start (save-excursion
214 (c-save-buffer-state ()
215 (and (c-beginning-of-macro)
216 (point)))))
217 c-macro-start))
218
219 (defsubst c-query-macro-start ()
220 (if (symbolp c-macro-start)
221 (save-excursion
222 (c-save-buffer-state ()
223 (and (c-beginning-of-macro)
224 (point))))
225 c-macro-start))
226
227 (defun c-beginning-of-macro (&optional lim)
228 "Go to the beginning of a preprocessor directive.
229 Leave point at the beginning of the directive and return t if in one,
230 otherwise return nil and leave point unchanged.
231
232 Note that this function might do hidden buffer changes. See the
233 comment at the start of cc-engine.el for more info."
234 (when c-opt-cpp-prefix
235 (let ((here (point)))
236 (save-restriction
237 (if lim (narrow-to-region lim (point-max)))
238 (beginning-of-line)
239 (while (eq (char-before (1- (point))) ?\\)
240 (forward-line -1))
241 (back-to-indentation)
242 (if (and (<= (point) here)
243 (looking-at c-opt-cpp-start))
244 t
245 (goto-char here)
246 nil)))))
247
248 (defun c-end-of-macro ()
249 "Go to the end of a preprocessor directive.
250 More accurately, move the point to the end of the closest following
251 line that doesn't end with a line continuation backslash - no check is
252 done that the point is inside a cpp directive to begin with.
253
254 Note that this function might do hidden buffer changes. See the
255 comment at the start of cc-engine.el for more info."
256 (while (progn
257 (end-of-line)
258 (when (and (eq (char-before) ?\\)
259 (not (eobp)))
260 (forward-char)
261 t))))
262
263 (defun c-syntactic-end-of-macro ()
264 ;; Go to the end of a CPP directive, or a "safe" pos just before.
265 ;;
266 ;; This is normally the end of the next non-escaped line. A "safe"
267 ;; position is one not within a string or comment. (The EOL on a line
268 ;; comment is NOT "safe").
269 ;;
270 ;; This function must only be called from the beginning of a CPP construct.
271 ;;
272 ;; Note that this function might do hidden buffer changes. See the comment
273 ;; at the start of cc-engine.el for more info.
274 (let* ((here (point))
275 (there (progn (c-end-of-macro) (point)))
276 (s (parse-partial-sexp here there)))
277 (while (and (or (nth 3 s) ; in a string
278 (nth 4 s)) ; in a comment (maybe at end of line comment)
279 (> there here)) ; No infinite loops, please.
280 (setq there (1- (nth 8 s)))
281 (setq s (parse-partial-sexp here there)))
282 (point)))
283
284 (defun c-forward-over-cpp-define-id ()
285 ;; Assuming point is at the "#" that introduces a preprocessor
286 ;; directive, it's moved forward to the end of the identifier which is
287 ;; "#define"d (or whatever c-opt-cpp-macro-define specifies). Non-nil
288 ;; is returned in this case, in all other cases nil is returned and
289 ;; point isn't moved.
290 ;;
291 ;; This function might do hidden buffer changes.
292 (when (and c-opt-cpp-macro-define-id
293 (looking-at c-opt-cpp-macro-define-id))
294 (goto-char (match-end 0))))
295
296 (defun c-forward-to-cpp-define-body ()
297 ;; Assuming point is at the "#" that introduces a preprocessor
298 ;; directive, it's moved forward to the start of the definition body
299 ;; if it's a "#define" (or whatever c-opt-cpp-macro-define
300 ;; specifies). Non-nil is returned in this case, in all other cases
301 ;; nil is returned and point isn't moved.
302 ;;
303 ;; This function might do hidden buffer changes.
304 (when (and c-opt-cpp-macro-define-start
305 (looking-at c-opt-cpp-macro-define-start)
306 (not (= (match-end 0) (c-point 'eol))))
307 (goto-char (match-end 0))))
308
309 \f
310 ;;; Basic utility functions.
311
312 (defun c-syntactic-content (from to paren-level)
313 ;; Return the given region as a string where all syntactic
314 ;; whitespace is removed or, where necessary, replaced with a single
315 ;; space. If PAREN-LEVEL is given then all parens in the region are
316 ;; collapsed to "()", "[]" etc.
317 ;;
318 ;; This function might do hidden buffer changes.
319
320 (save-excursion
321 (save-restriction
322 (narrow-to-region from to)
323 (goto-char from)
324 (let* ((parts (list nil)) (tail parts) pos in-paren)
325
326 (while (re-search-forward c-syntactic-ws-start to t)
327 (goto-char (setq pos (match-beginning 0)))
328 (c-forward-syntactic-ws)
329 (if (= (point) pos)
330 (forward-char)
331
332 (when paren-level
333 (save-excursion
334 (setq in-paren (= (car (parse-partial-sexp from pos 1)) 1)
335 pos (point))))
336
337 (if (and (> pos from)
338 (< (point) to)
339 (looking-at "\\w\\|\\s_")
340 (save-excursion
341 (goto-char (1- pos))
342 (looking-at "\\w\\|\\s_")))
343 (progn
344 (setcdr tail (list (buffer-substring-no-properties from pos)
345 " "))
346 (setq tail (cddr tail)))
347 (setcdr tail (list (buffer-substring-no-properties from pos)))
348 (setq tail (cdr tail)))
349
350 (when in-paren
351 (when (= (car (parse-partial-sexp pos to -1)) -1)
352 (setcdr tail (list (buffer-substring-no-properties
353 (1- (point)) (point))))
354 (setq tail (cdr tail))))
355
356 (setq from (point))))
357
358 (setcdr tail (list (buffer-substring-no-properties from to)))
359 (apply 'concat (cdr parts))))))
360
361 (defun c-shift-line-indentation (shift-amt)
362 ;; Shift the indentation of the current line with the specified
363 ;; amount (positive inwards). The buffer is modified only if
364 ;; SHIFT-AMT isn't equal to zero.
365 (let ((pos (- (point-max) (point)))
366 (c-macro-start c-macro-start)
367 tmp-char-inserted)
368 (if (zerop shift-amt)
369 nil
370 ;; If we're on an empty line inside a macro, we take the point
371 ;; to be at the current indentation and shift it to the
372 ;; appropriate column. This way we don't treat the extra
373 ;; whitespace out to the line continuation as indentation.
374 (when (and (c-query-and-set-macro-start)
375 (looking-at "[ \t]*\\\\$")
376 (save-excursion
377 (skip-chars-backward " \t")
378 (bolp)))
379 (insert ?x)
380 (backward-char)
381 (setq tmp-char-inserted t))
382 (unwind-protect
383 (let ((col (current-indentation)))
384 (delete-region (c-point 'bol) (c-point 'boi))
385 (beginning-of-line)
386 (indent-to (+ col shift-amt)))
387 (when tmp-char-inserted
388 (delete-char 1))))
389 ;; If initial point was within line's indentation and we're not on
390 ;; a line with a line continuation in a macro, position after the
391 ;; indentation. Else stay at same point in text.
392 (if (and (< (point) (c-point 'boi))
393 (not tmp-char-inserted))
394 (back-to-indentation)
395 (if (> (- (point-max) pos) (point))
396 (goto-char (- (point-max) pos))))))
397
398 (defsubst c-keyword-sym (keyword)
399 ;; Return non-nil if the string KEYWORD is a known keyword. More
400 ;; precisely, the value is the symbol for the keyword in
401 ;; `c-keywords-obarray'.
402 (intern-soft keyword c-keywords-obarray))
403
404 (defsubst c-keyword-member (keyword-sym lang-constant)
405 ;; Return non-nil if the symbol KEYWORD-SYM, as returned by
406 ;; `c-keyword-sym', is a member of LANG-CONSTANT, which is the name
407 ;; of a language constant that ends with "-kwds". If KEYWORD-SYM is
408 ;; nil then the result is nil.
409 (get keyword-sym lang-constant))
410
411 ;; String syntax chars, suitable for skip-syntax-(forward|backward).
412 (defconst c-string-syntax (if (memq 'gen-string-delim c-emacs-features)
413 "\"|"
414 "\""))
415
416 ;; Regexp matching string limit syntax.
417 (defconst c-string-limit-regexp (if (memq 'gen-string-delim c-emacs-features)
418 "\\s\"\\|\\s|"
419 "\\s\""))
420
421 ;; Regexp matching WS followed by string limit syntax.
422 (defconst c-ws*-string-limit-regexp
423 (concat "[ \t]*\\(" c-string-limit-regexp "\\)"))
424
425 ;; Holds formatted error strings for the few cases where parse errors
426 ;; are reported.
427 (defvar c-parsing-error nil)
428 (make-variable-buffer-local 'c-parsing-error)
429
430 (defun c-echo-parsing-error (&optional quiet)
431 (when (and c-report-syntactic-errors c-parsing-error (not quiet))
432 (c-benign-error "%s" c-parsing-error))
433 c-parsing-error)
434
435 ;; Faces given to comments and string literals. This is used in some
436 ;; situations to speed up recognition; it isn't mandatory that font
437 ;; locking is in use. This variable is extended with the face in
438 ;; `c-doc-face-name' when fontification is activated in cc-fonts.el.
439 (defvar c-literal-faces
440 (append '(font-lock-comment-face font-lock-string-face)
441 (when (facep 'font-lock-comment-delimiter-face)
442 ;; New in Emacs 22.
443 '(font-lock-comment-delimiter-face))))
444
445 (defsubst c-put-c-type-property (pos value)
446 ;; Put a c-type property with the given value at POS.
447 (c-put-char-property pos 'c-type value))
448
449 (defun c-clear-c-type-property (from to value)
450 ;; Remove all occurrences of the c-type property that has the given
451 ;; value in the region between FROM and TO. VALUE is assumed to not
452 ;; be nil.
453 ;;
454 ;; Note: This assumes that c-type is put on single chars only; it's
455 ;; very inefficient if matching properties cover large regions.
456 (save-excursion
457 (goto-char from)
458 (while (progn
459 (when (eq (get-text-property (point) 'c-type) value)
460 (c-clear-char-property (point) 'c-type))
461 (goto-char (next-single-property-change (point) 'c-type nil to))
462 (< (point) to)))))
463
464 \f
465 ;; Some debug tools to visualize various special positions. This
466 ;; debug code isn't as portable as the rest of CC Mode.
467
468 (cc-bytecomp-defun overlays-in)
469 (cc-bytecomp-defun overlay-get)
470 (cc-bytecomp-defun overlay-start)
471 (cc-bytecomp-defun overlay-end)
472 (cc-bytecomp-defun delete-overlay)
473 (cc-bytecomp-defun overlay-put)
474 (cc-bytecomp-defun make-overlay)
475
476 (defun c-debug-add-face (beg end face)
477 (c-save-buffer-state ((overlays (overlays-in beg end)) overlay)
478 (while overlays
479 (setq overlay (car overlays)
480 overlays (cdr overlays))
481 (when (eq (overlay-get overlay 'face) face)
482 (setq beg (min beg (overlay-start overlay))
483 end (max end (overlay-end overlay)))
484 (delete-overlay overlay)))
485 (overlay-put (make-overlay beg end) 'face face)))
486
487 (defun c-debug-remove-face (beg end face)
488 (c-save-buffer-state ((overlays (overlays-in beg end)) overlay
489 (ol-beg beg) (ol-end end))
490 (while overlays
491 (setq overlay (car overlays)
492 overlays (cdr overlays))
493 (when (eq (overlay-get overlay 'face) face)
494 (setq ol-beg (min ol-beg (overlay-start overlay))
495 ol-end (max ol-end (overlay-end overlay)))
496 (delete-overlay overlay)))
497 (when (< ol-beg beg)
498 (overlay-put (make-overlay ol-beg beg) 'face face))
499 (when (> ol-end end)
500 (overlay-put (make-overlay end ol-end) 'face face))))
501
502 \f
503 ;; `c-beginning-of-statement-1' and accompanying stuff.
504
505 ;; KLUDGE ALERT: c-maybe-labelp is used to pass information between
506 ;; c-crosses-statement-barrier-p and c-beginning-of-statement-1. A
507 ;; better way should be implemented, but this will at least shut up
508 ;; the byte compiler.
509 (defvar c-maybe-labelp)
510
511 ;; New awk-compatible version of c-beginning-of-statement-1, ACM 2002/6/22
512
513 ;; Macros used internally in c-beginning-of-statement-1 for the
514 ;; automaton actions.
515 (defmacro c-bos-push-state ()
516 '(setq stack (cons (cons state saved-pos)
517 stack)))
518 (defmacro c-bos-pop-state (&optional do-if-done)
519 `(if (setq state (car (car stack))
520 saved-pos (cdr (car stack))
521 stack (cdr stack))
522 t
523 ,do-if-done
524 (throw 'loop nil)))
525 (defmacro c-bos-pop-state-and-retry ()
526 '(throw 'loop (setq state (car (car stack))
527 saved-pos (cdr (car stack))
528 ;; Throw nil if stack is empty, else throw non-nil.
529 stack (cdr stack))))
530 (defmacro c-bos-save-pos ()
531 '(setq saved-pos (vector pos tok ptok pptok)))
532 (defmacro c-bos-restore-pos ()
533 '(unless (eq (elt saved-pos 0) start)
534 (setq pos (elt saved-pos 0)
535 tok (elt saved-pos 1)
536 ptok (elt saved-pos 2)
537 pptok (elt saved-pos 3))
538 (goto-char pos)
539 (setq sym nil)))
540 (defmacro c-bos-save-error-info (missing got)
541 `(setq saved-pos (vector pos ,missing ,got)))
542 (defmacro c-bos-report-error ()
543 '(unless noerror
544 (setq c-parsing-error
545 (format "No matching `%s' found for `%s' on line %d"
546 (elt saved-pos 1)
547 (elt saved-pos 2)
548 (1+ (count-lines (point-min)
549 (c-point 'bol (elt saved-pos 0))))))))
550
551 (defun c-beginning-of-statement-1 (&optional lim ignore-labels
552 noerror comma-delim)
553 "Move to the start of the current statement or declaration, or to
554 the previous one if already at the beginning of one. Only
555 statements/declarations on the same level are considered, i.e. don't
556 move into or out of sexps (not even normal expression parentheses).
557
558 If point is already at the earliest statement within braces or parens,
559 this function doesn't move back into any whitespace preceding it; it
560 returns 'same in this case.
561
562 Stop at statement continuation tokens like \"else\", \"catch\",
563 \"finally\" and the \"while\" in \"do ... while\" if the start point
564 is within the continuation. If starting at such a token, move to the
565 corresponding statement start. If at the beginning of a statement,
566 move to the closest containing statement if there is any. This might
567 also stop at a continuation clause.
568
569 Labels are treated as part of the following statements if
570 IGNORE-LABELS is non-nil. (FIXME: Doesn't work if we stop at a known
571 statement start keyword.) Otherwise, each label is treated as a
572 separate statement.
573
574 Macros are ignored \(i.e. skipped over) unless point is within one, in
575 which case the content of the macro is treated as normal code. Aside
576 from any normal statement starts found in it, stop at the first token
577 of the content in the macro, i.e. the expression of an \"#if\" or the
578 start of the definition in a \"#define\". Also stop at start of
579 macros before leaving them.
580
581 Return:
582 'label if stopped at a label or \"case...:\" or \"default:\";
583 'same if stopped at the beginning of the current statement;
584 'up if stepped to a containing statement;
585 'previous if stepped to a preceding statement;
586 'beginning if stepped from a statement continuation clause to
587 its start clause; or
588 'macro if stepped to a macro start.
589 Note that 'same and not 'label is returned if stopped at the same
590 label without crossing the colon character.
591
592 LIM may be given to limit the search. If the search hits the limit,
593 point will be left at the closest following token, or at the start
594 position if that is less ('same is returned in this case).
595
596 NOERROR turns off error logging to `c-parsing-error'.
597
598 Normally only ';' and virtual semicolons are considered to delimit
599 statements, but if COMMA-DELIM is non-nil then ',' is treated
600 as a delimiter too.
601
602 Note that this function might do hidden buffer changes. See the
603 comment at the start of cc-engine.el for more info."
604
605 ;; The bulk of this function is a pushdown automaton that looks at statement
606 ;; boundaries and the tokens (such as "while") in c-opt-block-stmt-key. Its
607 ;; purpose is to keep track of nested statements, ensuring that such
608 ;; statements are skipped over in their entirety (somewhat akin to what C-M-p
609 ;; does with nested braces/brackets/parentheses).
610 ;;
611 ;; Note: The position of a boundary is the following token.
612 ;;
613 ;; Beginning with the current token (the one following point), move back one
614 ;; sexp at a time (where a sexp is, more or less, either a token or the
615 ;; entire contents of a brace/bracket/paren pair). Each time a statement
616 ;; boundary is crossed or a "while"-like token is found, update the state of
617 ;; the PDA. Stop at the beginning of a statement when the stack (holding
618 ;; nested statement info) is empty and the position has been moved.
619 ;;
620 ;; The following variables constitute the PDA:
621 ;;
622 ;; sym: This is either the "while"-like token (e.g. 'for) we've just
623 ;; scanned back over, 'boundary if we've just gone back over a
624 ;; statement boundary, or nil otherwise.
625 ;; state: takes one of the values (nil else else-boundary while
626 ;; while-boundary catch catch-boundary).
627 ;; nil means "no "while"-like token yet scanned".
628 ;; 'else, for example, means "just gone back over an else".
629 ;; 'else-boundary means "just gone back over a statement boundary
630 ;; immediately after having gone back over an else".
631 ;; saved-pos: A vector of either saved positions (tok ptok pptok, etc.) or
632 ;; of error reporting information.
633 ;; stack: The stack onto which the PDA pushes its state. Each entry
634 ;; consists of a saved value of state and saved-pos. An entry is
635 ;; pushed when we move back over a "continuation" token (e.g. else)
636 ;; and popped when we encounter the corresponding opening token
637 ;; (e.g. if).
638 ;;
639 ;;
640 ;; The following diagram briefly outlines the PDA.
641 ;;
642 ;; Common state:
643 ;; "else": Push state, goto state `else'.
644 ;; "while": Push state, goto state `while'.
645 ;; "catch" or "finally": Push state, goto state `catch'.
646 ;; boundary: Pop state.
647 ;; other: Do nothing special.
648 ;;
649 ;; State `else':
650 ;; boundary: Goto state `else-boundary'.
651 ;; other: Error, pop state, retry token.
652 ;;
653 ;; State `else-boundary':
654 ;; "if": Pop state.
655 ;; boundary: Error, pop state.
656 ;; other: See common state.
657 ;;
658 ;; State `while':
659 ;; boundary: Save position, goto state `while-boundary'.
660 ;; other: Pop state, retry token.
661 ;;
662 ;; State `while-boundary':
663 ;; "do": Pop state.
664 ;; boundary: Restore position if it's not at start, pop state. [*see below]
665 ;; other: See common state.
666 ;;
667 ;; State `catch':
668 ;; boundary: Goto state `catch-boundary'.
669 ;; other: Error, pop state, retry token.
670 ;;
671 ;; State `catch-boundary':
672 ;; "try": Pop state.
673 ;; "catch": Goto state `catch'.
674 ;; boundary: Error, pop state.
675 ;; other: See common state.
676 ;;
677 ;; [*] In the `while-boundary' state, we had pushed a 'while state, and were
678 ;; searching for a "do" which would have opened a do-while. If we didn't
679 ;; find it, we discard the analysis done since the "while", go back to this
680 ;; token in the buffer and restart the scanning there, this time WITHOUT
681 ;; pushing the 'while state onto the stack.
682 ;;
683 ;; In addition to the above there is some special handling of labels
684 ;; and macros.
685
686 (let ((case-fold-search nil)
687 (start (point))
688 macro-start
689 (delims (if comma-delim '(?\; ?,) '(?\;)))
690 (c-stmt-delim-chars (if comma-delim
691 c-stmt-delim-chars-with-comma
692 c-stmt-delim-chars))
693 c-in-literal-cache c-maybe-labelp after-case:-pos saved
694 ;; Current position.
695 pos
696 ;; Position of last stmt boundary character (e.g. ;).
697 boundary-pos
698 ;; The position of the last sexp or bound that follows the
699 ;; first found colon, i.e. the start of the nonlabel part of
700 ;; the statement. It's `start' if a colon is found just after
701 ;; the start.
702 after-labels-pos
703 ;; Like `after-labels-pos', but the first such position inside
704 ;; a label, i.e. the start of the last label before the start
705 ;; of the nonlabel part of the statement.
706 last-label-pos
707 ;; The last position where a label is possible provided the
708 ;; statement started there. It's nil as long as no invalid
709 ;; label content has been found (according to
710 ;; `c-nonlabel-token-key'. It's `start' if no valid label
711 ;; content was found in the label. Note that we might still
712 ;; regard it a label if it starts with `c-label-kwds'.
713 label-good-pos
714 ;; Symbol just scanned back over (e.g. 'while or 'boundary).
715 ;; See above.
716 sym
717 ;; Current state in the automaton. See above.
718 state
719 ;; Current saved positions. See above.
720 saved-pos
721 ;; Stack of conses (state . saved-pos).
722 stack
723 ;; Regexp which matches "for", "if", etc.
724 (cond-key (or c-opt-block-stmt-key
725 "\\<\\>")) ; Matches nothing.
726 ;; Return value.
727 (ret 'same)
728 ;; Positions of the last three sexps or bounds we've stopped at.
729 tok ptok pptok)
730
731 (save-restriction
732 (if lim (narrow-to-region lim (point-max)))
733
734 (if (save-excursion
735 (and (c-beginning-of-macro)
736 (/= (point) start)))
737 (setq macro-start (point)))
738
739 ;; Try to skip back over unary operator characters, to register
740 ;; that we've moved.
741 (while (progn
742 (setq pos (point))
743 (c-backward-syntactic-ws)
744 ;; Protect post-++/-- operators just before a virtual semicolon.
745 (and (not (c-at-vsemi-p))
746 (/= (skip-chars-backward "-+!*&~@`#") 0))))
747
748 ;; Skip back over any semicolon here. If it was a bare semicolon, we're
749 ;; done. Later on we ignore the boundaries for statements that don't
750 ;; contain any sexp. The only thing that is affected is that the error
751 ;; checking is a little less strict, and we really don't bother.
752 (if (and (memq (char-before) delims)
753 (progn (forward-char -1)
754 (setq saved (point))
755 (c-backward-syntactic-ws)
756 (or (memq (char-before) delims)
757 (memq (char-before) '(?: nil))
758 (eq (char-syntax (char-before)) ?\()
759 (c-at-vsemi-p))))
760 (setq ret 'previous
761 pos saved)
762
763 ;; Begin at start and not pos to detect macros if we stand
764 ;; directly after the #.
765 (goto-char start)
766 (if (looking-at "\\<\\|\\W")
767 ;; Record this as the first token if not starting inside it.
768 (setq tok start))
769
770 ;; The following while loop goes back one sexp (balanced parens,
771 ;; etc. with contents, or symbol or suchlike) each iteration. This
772 ;; movement is accomplished with a call to scan-sexps approx 130 lines
773 ;; below.
774 (while
775 (catch 'loop ;; Throw nil to break, non-nil to continue.
776 (cond
777 ((save-excursion
778 (and macro-start ; Always NIL for AWK.
779 (progn (skip-chars-backward " \t")
780 (eq (char-before) ?#))
781 (progn (setq saved (1- (point)))
782 (beginning-of-line)
783 (not (eq (char-before (1- (point))) ?\\)))
784 (looking-at c-opt-cpp-start)
785 (progn (skip-chars-forward " \t")
786 (eq (point) saved))))
787 (goto-char saved)
788 (if (and (c-forward-to-cpp-define-body)
789 (progn (c-forward-syntactic-ws start)
790 (< (point) start)))
791 ;; Stop at the first token in the content of the macro.
792 (setq pos (point)
793 ignore-labels t) ; Avoid the label check on exit.
794 (setq pos saved
795 ret 'macro
796 ignore-labels t))
797 (throw 'loop nil))
798
799 ;; Do a round through the automaton if we've just passed a
800 ;; statement boundary or passed a "while"-like token.
801 ((or sym
802 (and (looking-at cond-key)
803 (setq sym (intern (match-string 1)))))
804
805 (when (and (< pos start) (null stack))
806 (throw 'loop nil))
807
808 ;; The PDA state handling.
809 ;;
810 ;; Refer to the description of the PDA in the opening
811 ;; comments. In the following OR form, the first leaf
812 ;; attempts to handles one of the specific actions detailed
813 ;; (e.g., finding token "if" whilst in state `else-boundary').
814 ;; We drop through to the second leaf (which handles common
815 ;; state) if no specific handler is found in the first cond.
816 ;; If a parsing error is detected (e.g. an "else" with no
817 ;; preceding "if"), we throw to the enclosing catch.
818 ;;
819 ;; Note that the (eq state 'else) means
820 ;; "we've just passed an else", NOT "we're looking for an
821 ;; else".
822 (or (cond
823 ((eq state 'else)
824 (if (eq sym 'boundary)
825 (setq state 'else-boundary)
826 (c-bos-report-error)
827 (c-bos-pop-state-and-retry)))
828
829 ((eq state 'else-boundary)
830 (cond ((eq sym 'if)
831 (c-bos-pop-state (setq ret 'beginning)))
832 ((eq sym 'boundary)
833 (c-bos-report-error)
834 (c-bos-pop-state))))
835
836 ((eq state 'while)
837 (if (and (eq sym 'boundary)
838 ;; Since this can cause backtracking we do a
839 ;; little more careful analysis to avoid it:
840 ;; If there's a label in front of the while
841 ;; it can't be part of a do-while.
842 (not after-labels-pos))
843 (progn (c-bos-save-pos)
844 (setq state 'while-boundary))
845 (c-bos-pop-state-and-retry))) ; Can't be a do-while
846
847 ((eq state 'while-boundary)
848 (cond ((eq sym 'do)
849 (c-bos-pop-state (setq ret 'beginning)))
850 ((eq sym 'boundary) ; isn't a do-while
851 (c-bos-restore-pos) ; the position of the while
852 (c-bos-pop-state)))) ; no longer searching for do.
853
854 ((eq state 'catch)
855 (if (eq sym 'boundary)
856 (setq state 'catch-boundary)
857 (c-bos-report-error)
858 (c-bos-pop-state-and-retry)))
859
860 ((eq state 'catch-boundary)
861 (cond
862 ((eq sym 'try)
863 (c-bos-pop-state (setq ret 'beginning)))
864 ((eq sym 'catch)
865 (setq state 'catch))
866 ((eq sym 'boundary)
867 (c-bos-report-error)
868 (c-bos-pop-state)))))
869
870 ;; This is state common. We get here when the previous
871 ;; cond statement found no particular state handler.
872 (cond ((eq sym 'boundary)
873 ;; If we have a boundary at the start
874 ;; position we push a frame to go to the
875 ;; previous statement.
876 (if (>= pos start)
877 (c-bos-push-state)
878 (c-bos-pop-state)))
879 ((eq sym 'else)
880 (c-bos-push-state)
881 (c-bos-save-error-info 'if 'else)
882 (setq state 'else))
883 ((eq sym 'while)
884 ;; Is this a real while, or a do-while?
885 ;; The next `when' triggers unless we are SURE that
886 ;; the `while' is not the tailend of a `do-while'.
887 (when (or (not pptok)
888 (memq (char-after pptok) delims)
889 ;; The following kludge is to prevent
890 ;; infinite recursion when called from
891 ;; c-awk-after-if-for-while-condition-p,
892 ;; or the like.
893 (and (eq (point) start)
894 (c-vsemi-status-unknown-p))
895 (c-at-vsemi-p pptok))
896 ;; Since this can cause backtracking we do a
897 ;; little more careful analysis to avoid it: If
898 ;; the while isn't followed by a (possibly
899 ;; virtual) semicolon it can't be a do-while.
900 (c-bos-push-state)
901 (setq state 'while)))
902 ((memq sym '(catch finally))
903 (c-bos-push-state)
904 (c-bos-save-error-info 'try sym)
905 (setq state 'catch))))
906
907 (when c-maybe-labelp
908 ;; We're either past a statement boundary or at the
909 ;; start of a statement, so throw away any label data
910 ;; for the previous one.
911 (setq after-labels-pos nil
912 last-label-pos nil
913 c-maybe-labelp nil))))
914
915 ;; Step to the previous sexp, but not if we crossed a
916 ;; boundary, since that doesn't consume an sexp.
917 (if (eq sym 'boundary)
918 (setq ret 'previous)
919
920 ;; HERE IS THE SINGLE PLACE INSIDE THE PDA LOOP WHERE WE MOVE
921 ;; BACKWARDS THROUGH THE SOURCE.
922
923 ;; This is typically fast with the caching done by
924 ;; c-(backward|forward)-sws.
925 (c-backward-syntactic-ws)
926
927 (let ((before-sws-pos (point))
928 ;; Set as long as we have to continue jumping by sexps.
929 ;; It's the position to use as end in the next round.
930 sexp-loop-continue-pos
931 ;; The end position of the area to search for statement
932 ;; barriers in this round.
933 (sexp-loop-end-pos pos))
934
935 ;; The following while goes back one sexp per iteration.
936 (while
937 (progn
938 (unless (c-safe (c-backward-sexp) t)
939 ;; Give up if we hit an unbalanced block. Since the
940 ;; stack won't be empty the code below will report a
941 ;; suitable error.
942 (throw 'loop nil))
943
944 ;; Check if the sexp movement crossed a statement or
945 ;; declaration boundary. But first modify the point
946 ;; so that `c-crosses-statement-barrier-p' only looks
947 ;; at the non-sexp chars following the sexp.
948 (save-excursion
949 (when (setq
950 boundary-pos
951 (cond
952 ((if macro-start
953 nil
954 (save-excursion
955 (when (c-beginning-of-macro)
956 ;; Set continuation position in case
957 ;; `c-crosses-statement-barrier-p'
958 ;; doesn't detect anything below.
959 (setq sexp-loop-continue-pos (point)))))
960 ;; If the sexp movement took us into a
961 ;; macro then there were only some non-sexp
962 ;; chars after it. Skip out of the macro
963 ;; to analyze them but not the non-sexp
964 ;; chars that might be inside the macro.
965 (c-end-of-macro)
966 (c-crosses-statement-barrier-p
967 (point) sexp-loop-end-pos))
968
969 ((and
970 (eq (char-after) ?{)
971 (not (c-looking-at-inexpr-block lim nil t)))
972 ;; Passed a block sexp. That's a boundary
973 ;; alright.
974 (point))
975
976 ((looking-at "\\s\(")
977 ;; Passed some other paren. Only analyze
978 ;; the non-sexp chars after it.
979 (goto-char (1+ (c-down-list-backward
980 before-sws-pos)))
981 ;; We're at a valid token start position
982 ;; (outside the `save-excursion') if
983 ;; `c-crosses-statement-barrier-p' failed.
984 (c-crosses-statement-barrier-p
985 (point) sexp-loop-end-pos))
986
987 (t
988 ;; Passed a symbol sexp or line
989 ;; continuation. It doesn't matter that
990 ;; it's included in the analyzed region.
991 (if (c-crosses-statement-barrier-p
992 (point) sexp-loop-end-pos)
993 t
994 ;; If it was a line continuation then we
995 ;; have to continue looping.
996 (if (looking-at "\\\\$")
997 (setq sexp-loop-continue-pos (point)))
998 nil))))
999
1000 (setq pptok ptok
1001 ptok tok
1002 tok boundary-pos
1003 sym 'boundary)
1004 ;; Like a C "continue". Analyze the next sexp.
1005 (throw 'loop t)))
1006
1007 sexp-loop-continue-pos) ; End of "go back a sexp" loop condition.
1008 (goto-char sexp-loop-continue-pos)
1009 (setq sexp-loop-end-pos sexp-loop-continue-pos
1010 sexp-loop-continue-pos nil))))
1011
1012 ;; ObjC method def?
1013 (when (and c-opt-method-key
1014 (setq saved (c-in-method-def-p)))
1015 (setq pos saved
1016 ignore-labels t) ; Avoid the label check on exit.
1017 (throw 'loop nil))
1018
1019 ;; Handle labels.
1020 (unless (eq ignore-labels t)
1021 (when (numberp c-maybe-labelp)
1022 ;; `c-crosses-statement-barrier-p' has found a colon, so we
1023 ;; might be in a label now. Have we got a real label
1024 ;; (including a case label) or something like C++'s "public:"?
1025 ;; A case label might use an expression rather than a token.
1026 (setq after-case:-pos (or tok start))
1027 (if (looking-at c-nonlabel-token-key) ; e.g. "while" or "'a'"
1028 (setq c-maybe-labelp nil)
1029 (if after-labels-pos ; Have we already encountered a label?
1030 (if (not last-label-pos)
1031 (setq last-label-pos (or tok start)))
1032 (setq after-labels-pos (or tok start)))
1033 (setq c-maybe-labelp t
1034 label-good-pos nil))) ; bogus "label"
1035
1036 (when (and (not label-good-pos) ; i.e. no invalid "label"'s yet
1037 ; been found.
1038 (looking-at c-nonlabel-token-key)) ; e.g. "while :"
1039 ;; We're in a potential label and it's the first
1040 ;; time we've found something that isn't allowed in
1041 ;; one.
1042 (setq label-good-pos (or tok start))))
1043
1044 ;; We've moved back by a sexp, so update the token positions.
1045 (setq sym nil
1046 pptok ptok
1047 ptok tok
1048 tok (point)
1049 pos tok))) ; Not nil (for the while loop).
1050
1051 ;; If the stack isn't empty there might be errors to report.
1052 (while stack
1053 (if (and (vectorp saved-pos) (eq (length saved-pos) 3))
1054 (c-bos-report-error))
1055 (setq saved-pos (cdr (car stack))
1056 stack (cdr stack)))
1057
1058 (when (and (eq ret 'same)
1059 (not (memq sym '(boundary ignore nil))))
1060 ;; Need to investigate closer whether we've crossed
1061 ;; between a substatement and its containing statement.
1062 (if (setq saved (if (looking-at c-block-stmt-1-key)
1063 ptok
1064 pptok))
1065 (cond ((> start saved) (setq pos saved))
1066 ((= start saved) (setq ret 'up)))))
1067
1068 (when (and (not ignore-labels)
1069 (eq c-maybe-labelp t)
1070 (not (eq ret 'beginning))
1071 after-labels-pos
1072 (or (not label-good-pos)
1073 (<= label-good-pos pos)
1074 (progn
1075 (goto-char (if (and last-label-pos
1076 (< last-label-pos start))
1077 last-label-pos
1078 pos))
1079 (looking-at c-label-kwds-regexp))))
1080 ;; We're in a label. Maybe we should step to the statement
1081 ;; after it.
1082 (if (< after-labels-pos start)
1083 (setq pos after-labels-pos)
1084 (setq ret 'label)
1085 (if (and last-label-pos (< last-label-pos start))
1086 ;; Might have jumped over several labels. Go to the last one.
1087 (setq pos last-label-pos)))))
1088
1089 ;; Have we got "case <expression>:"?
1090 (goto-char pos)
1091 (when (and after-case:-pos
1092 (not (eq ret 'beginning))
1093 (looking-at c-case-kwds-regexp))
1094 (if (< after-case:-pos start)
1095 (setq pos after-case:-pos))
1096 (if (eq ret 'same)
1097 (setq ret 'label)))
1098
1099 ;; Skip over the unary operators that can start the statement.
1100 (while (progn
1101 (c-backward-syntactic-ws)
1102 ;; protect AWK post-inc/decrement operators, etc.
1103 (and (not (c-at-vsemi-p (point)))
1104 (/= (skip-chars-backward "-+!*&~@`#") 0)))
1105 (setq pos (point)))
1106 (goto-char pos)
1107 ret)))
1108
1109 (defun c-crosses-statement-barrier-p (from to)
1110 "Return non-nil if buffer positions FROM to TO cross one or more
1111 statement or declaration boundaries. The returned value is actually
1112 the position of the earliest boundary char. FROM must not be within
1113 a string or comment.
1114
1115 The variable `c-maybe-labelp' is set to the position of the first `:' that
1116 might start a label (i.e. not part of `::' and not preceded by `?'). If a
1117 single `?' is found, then `c-maybe-labelp' is cleared.
1118
1119 For AWK, a statement which is terminated by an EOL (not a \; or a }) is
1120 regarded as having a \"virtual semicolon\" immediately after the last token on
1121 the line. If this virtual semicolon is _at_ from, the function recognizes it.
1122
1123 Note that this function might do hidden buffer changes. See the
1124 comment at the start of cc-engine.el for more info."
1125 (let ((skip-chars c-stmt-delim-chars)
1126 lit-range)
1127 (save-excursion
1128 (catch 'done
1129 (goto-char from)
1130 (while (progn (skip-chars-forward skip-chars to)
1131 (< (point) to))
1132 (cond
1133 ((setq lit-range (c-literal-limits from)) ; Have we landed in a string/comment?
1134 (goto-char (cdr lit-range)))
1135 ((eq (char-after) ?:)
1136 (forward-char)
1137 (if (and (eq (char-after) ?:)
1138 (< (point) to))
1139 ;; Ignore scope operators.
1140 (forward-char)
1141 (setq c-maybe-labelp (1- (point)))))
1142 ((eq (char-after) ??)
1143 ;; A question mark. Can't be a label, so stop
1144 ;; looking for more : and ?.
1145 (setq c-maybe-labelp nil
1146 skip-chars (substring c-stmt-delim-chars 0 -2)))
1147 ((memq (char-after) '(?# ?\n ?\r)) ; A virtual semicolon?
1148 (if (and (eq (char-before) ?\\) (memq (char-after) '(?\n ?\r)))
1149 (backward-char))
1150 (skip-chars-backward " \t" from)
1151 (if (c-at-vsemi-p)
1152 (throw 'done (point))
1153 (forward-line)))
1154 (t (throw 'done (point)))))
1155 ;; In trailing space after an as yet undetected virtual semicolon?
1156 (c-backward-syntactic-ws from)
1157 (if (and (< (point) to)
1158 (c-at-vsemi-p))
1159 (point)
1160 nil)))))
1161
1162 (defun c-at-statement-start-p ()
1163 "Return non-nil if the point is at the first token in a statement
1164 or somewhere in the syntactic whitespace before it.
1165
1166 A \"statement\" here is not restricted to those inside code blocks.
1167 Any kind of declaration-like construct that occur outside function
1168 bodies is also considered a \"statement\".
1169
1170 Note that this function might do hidden buffer changes. See the
1171 comment at the start of cc-engine.el for more info."
1172
1173 (save-excursion
1174 (let ((end (point))
1175 c-maybe-labelp)
1176 (c-syntactic-skip-backward (substring c-stmt-delim-chars 1) nil t)
1177 (or (bobp)
1178 (eq (char-before) ?})
1179 (and (eq (char-before) ?{)
1180 (not (and c-special-brace-lists
1181 (progn (backward-char)
1182 (c-looking-at-special-brace-list)))))
1183 (c-crosses-statement-barrier-p (point) end)))))
1184
1185 (defun c-at-expression-start-p ()
1186 "Return non-nil if the point is at the first token in an expression or
1187 statement, or somewhere in the syntactic whitespace before it.
1188
1189 An \"expression\" here is a bit different from the normal language
1190 grammar sense: It's any sequence of expression tokens except commas,
1191 unless they are enclosed inside parentheses of some kind. Also, an
1192 expression never continues past an enclosing parenthesis, but it might
1193 contain parenthesis pairs of any sort except braces.
1194
1195 Since expressions never cross statement boundaries, this function also
1196 recognizes statement beginnings, just like `c-at-statement-start-p'.
1197
1198 Note that this function might do hidden buffer changes. See the
1199 comment at the start of cc-engine.el for more info."
1200
1201 (save-excursion
1202 (let ((end (point))
1203 (c-stmt-delim-chars c-stmt-delim-chars-with-comma)
1204 c-maybe-labelp)
1205 (c-syntactic-skip-backward (substring c-stmt-delim-chars 1) nil t)
1206 (or (bobp)
1207 (memq (char-before) '(?{ ?}))
1208 (save-excursion (backward-char)
1209 (looking-at "\\s("))
1210 (c-crosses-statement-barrier-p (point) end)))))
1211
1212 \f
1213 ;; A set of functions that covers various idiosyncrasies in
1214 ;; implementations of `forward-comment'.
1215
1216 ;; Note: Some emacsen considers incorrectly that any line comment
1217 ;; ending with a backslash continues to the next line. I can't think
1218 ;; of any way to work around that in a reliable way without changing
1219 ;; the buffer, though. Suggestions welcome. ;) (No, temporarily
1220 ;; changing the syntax for backslash doesn't work since we must treat
1221 ;; escapes in string literals correctly.)
1222
1223 (defun c-forward-single-comment ()
1224 "Move forward past whitespace and the closest following comment, if any.
1225 Return t if a comment was found, nil otherwise. In either case, the
1226 point is moved past the following whitespace. Line continuations,
1227 i.e. a backslashes followed by line breaks, are treated as whitespace.
1228 The line breaks that end line comments are considered to be the
1229 comment enders, so the point will be put on the beginning of the next
1230 line if it moved past a line comment.
1231
1232 This function does not do any hidden buffer changes."
1233
1234 (let ((start (point)))
1235 (when (looking-at "\\([ \t\n\r\f\v]\\|\\\\[\n\r]\\)+")
1236 (goto-char (match-end 0)))
1237
1238 (when (forward-comment 1)
1239 (if (eobp)
1240 ;; Some emacsen (e.g. XEmacs 21) return t when moving
1241 ;; forwards at eob.
1242 nil
1243
1244 ;; Emacs includes the ending newline in a b-style (c++)
1245 ;; comment, but XEmacs doesn't. We depend on the Emacs
1246 ;; behavior (which also is symmetric).
1247 (if (and (eolp) (elt (parse-partial-sexp start (point)) 7))
1248 (condition-case nil (forward-char 1)))
1249
1250 t))))
1251
1252 (defsubst c-forward-comments ()
1253 "Move forward past all following whitespace and comments.
1254 Line continuations, i.e. a backslashes followed by line breaks, are
1255 treated as whitespace.
1256
1257 Note that this function might do hidden buffer changes. See the
1258 comment at the start of cc-engine.el for more info."
1259
1260 (while (or
1261 ;; If forward-comment in at least XEmacs 21 is given a large
1262 ;; positive value, it'll loop all the way through if it hits
1263 ;; eob.
1264 (and (forward-comment 5)
1265 ;; Some emacsen (e.g. XEmacs 21) return t when moving
1266 ;; forwards at eob.
1267 (not (eobp)))
1268
1269 (when (looking-at "\\\\[\n\r]")
1270 (forward-char 2)
1271 t))))
1272
1273 (defun c-backward-single-comment ()
1274 "Move backward past whitespace and the closest preceding comment, if any.
1275 Return t if a comment was found, nil otherwise. In either case, the
1276 point is moved past the preceding whitespace. Line continuations,
1277 i.e. a backslashes followed by line breaks, are treated as whitespace.
1278 The line breaks that end line comments are considered to be the
1279 comment enders, so the point cannot be at the end of the same line to
1280 move over a line comment.
1281
1282 This function does not do any hidden buffer changes."
1283
1284 (let ((start (point)))
1285 ;; When we got newline terminated comments, forward-comment in all
1286 ;; supported emacsen so far will stop at eol of each line not
1287 ;; ending with a comment when moving backwards. This corrects for
1288 ;; that, and at the same time handles line continuations.
1289 (while (progn
1290 (skip-chars-backward " \t\n\r\f\v")
1291 (and (looking-at "[\n\r]")
1292 (eq (char-before) ?\\)))
1293 (backward-char))
1294
1295 (if (bobp)
1296 ;; Some emacsen (e.g. Emacs 19.34) return t when moving
1297 ;; backwards at bob.
1298 nil
1299
1300 ;; Leave point after the closest following newline if we've
1301 ;; backed up over any above, since forward-comment won't move
1302 ;; backward over a line comment if point is at the end of the
1303 ;; same line.
1304 (re-search-forward "\\=\\s *[\n\r]" start t)
1305
1306 (if (if (forward-comment -1)
1307 (if (eolp)
1308 ;; If forward-comment above succeeded and we're at eol
1309 ;; then the newline we moved over above didn't end a
1310 ;; line comment, so we give it another go.
1311 (forward-comment -1)
1312 t))
1313
1314 ;; Emacs <= 20 and XEmacs move back over the closer of a
1315 ;; block comment that lacks an opener.
1316 (if (looking-at "\\*/")
1317 (progn (forward-char 2) nil)
1318 t)))))
1319
1320 (defsubst c-backward-comments ()
1321 "Move backward past all preceding whitespace and comments.
1322 Line continuations, i.e. a backslashes followed by line breaks, are
1323 treated as whitespace. The line breaks that end line comments are
1324 considered to be the comment enders, so the point cannot be at the end
1325 of the same line to move over a line comment. Unlike
1326 c-backward-syntactic-ws, this function doesn't move back over
1327 preprocessor directives.
1328
1329 Note that this function might do hidden buffer changes. See the
1330 comment at the start of cc-engine.el for more info."
1331
1332 (let ((start (point)))
1333 (while (and
1334 ;; `forward-comment' in some emacsen (e.g. XEmacs 21.4)
1335 ;; return t when moving backwards at bob.
1336 (not (bobp))
1337
1338 (if (forward-comment -1)
1339 (if (looking-at "\\*/")
1340 ;; Emacs <= 20 and XEmacs move back over the
1341 ;; closer of a block comment that lacks an opener.
1342 (progn (forward-char 2) nil)
1343 t)
1344
1345 ;; XEmacs treats line continuations as whitespace but
1346 ;; only in the backward direction, which seems a bit
1347 ;; odd. Anyway, this is necessary for Emacs.
1348 (when (and (looking-at "[\n\r]")
1349 (eq (char-before) ?\\)
1350 (< (point) start))
1351 (backward-char)
1352 t))))))
1353
1354 \f
1355 ;; Tools for skipping over syntactic whitespace.
1356
1357 ;; The following functions use text properties to cache searches over
1358 ;; large regions of syntactic whitespace. It works as follows:
1359 ;;
1360 ;; o If a syntactic whitespace region contains anything but simple
1361 ;; whitespace (i.e. space, tab and line breaks), the text property
1362 ;; `c-in-sws' is put over it. At places where we have stopped
1363 ;; within that region there's also a `c-is-sws' text property.
1364 ;; That since there typically are nested whitespace inside that
1365 ;; must be handled separately, e.g. whitespace inside a comment or
1366 ;; cpp directive. Thus, from one point with `c-is-sws' it's safe
1367 ;; to jump to another point with that property within the same
1368 ;; `c-in-sws' region. It can be likened to a ladder where
1369 ;; `c-in-sws' marks the bars and `c-is-sws' the rungs.
1370 ;;
1371 ;; o The `c-is-sws' property is put on the simple whitespace chars at
1372 ;; a "rung position" and also maybe on the first following char.
1373 ;; As many characters as can be conveniently found in this range
1374 ;; are marked, but no assumption can be made that the whole range
1375 ;; is marked (it could be clobbered by later changes, for
1376 ;; instance).
1377 ;;
1378 ;; Note that some part of the beginning of a sequence of simple
1379 ;; whitespace might be part of the end of a preceding line comment
1380 ;; or cpp directive and must not be considered part of the "rung".
1381 ;; Such whitespace is some amount of horizontal whitespace followed
1382 ;; by a newline. In the case of cpp directives it could also be
1383 ;; two newlines with horizontal whitespace between them.
1384 ;;
1385 ;; The reason to include the first following char is to cope with
1386 ;; "rung positions" that doesn't have any ordinary whitespace. If
1387 ;; `c-is-sws' is put on a token character it does not have
1388 ;; `c-in-sws' set simultaneously. That's the only case when that
1389 ;; can occur, and the reason for not extending the `c-in-sws'
1390 ;; region to cover it is that the `c-in-sws' region could then be
1391 ;; accidentally merged with a following one if the token is only
1392 ;; one character long.
1393 ;;
1394 ;; o On buffer changes the `c-in-sws' and `c-is-sws' properties are
1395 ;; removed in the changed region. If the change was inside
1396 ;; syntactic whitespace that means that the "ladder" is broken, but
1397 ;; a later call to `c-forward-sws' or `c-backward-sws' will use the
1398 ;; parts on either side and use an ordinary search only to "repair"
1399 ;; the gap.
1400 ;;
1401 ;; Special care needs to be taken if a region is removed: If there
1402 ;; are `c-in-sws' on both sides of it which do not connect inside
1403 ;; the region then they can't be joined. If e.g. a marked macro is
1404 ;; broken, syntactic whitespace inside the new text might be
1405 ;; marked. If those marks would become connected with the old
1406 ;; `c-in-sws' range around the macro then we could get a ladder
1407 ;; with one end outside the macro and the other at some whitespace
1408 ;; within it.
1409 ;;
1410 ;; The main motivation for this system is to increase the speed in
1411 ;; skipping over the large whitespace regions that can occur at the
1412 ;; top level in e.g. header files that contain a lot of comments and
1413 ;; cpp directives. For small comments inside code it's probably
1414 ;; slower than using `forward-comment' straightforwardly, but speed is
1415 ;; not a significant factor there anyway.
1416
1417 ; (defface c-debug-is-sws-face
1418 ; '((t (:background "GreenYellow")))
1419 ; "Debug face to mark the `c-is-sws' property.")
1420 ; (defface c-debug-in-sws-face
1421 ; '((t (:underline t)))
1422 ; "Debug face to mark the `c-in-sws' property.")
1423
1424 ; (defun c-debug-put-sws-faces ()
1425 ; ;; Put the sws debug faces on all the `c-is-sws' and `c-in-sws'
1426 ; ;; properties in the buffer.
1427 ; (interactive)
1428 ; (save-excursion
1429 ; (c-save-buffer-state (in-face)
1430 ; (goto-char (point-min))
1431 ; (setq in-face (if (get-text-property (point) 'c-is-sws)
1432 ; (point)))
1433 ; (while (progn
1434 ; (goto-char (next-single-property-change
1435 ; (point) 'c-is-sws nil (point-max)))
1436 ; (if in-face
1437 ; (progn
1438 ; (c-debug-add-face in-face (point) 'c-debug-is-sws-face)
1439 ; (setq in-face nil))
1440 ; (setq in-face (point)))
1441 ; (not (eobp))))
1442 ; (goto-char (point-min))
1443 ; (setq in-face (if (get-text-property (point) 'c-in-sws)
1444 ; (point)))
1445 ; (while (progn
1446 ; (goto-char (next-single-property-change
1447 ; (point) 'c-in-sws nil (point-max)))
1448 ; (if in-face
1449 ; (progn
1450 ; (c-debug-add-face in-face (point) 'c-debug-in-sws-face)
1451 ; (setq in-face nil))
1452 ; (setq in-face (point)))
1453 ; (not (eobp)))))))
1454
1455 (defmacro c-debug-sws-msg (&rest args)
1456 ;;`(message ,@args)
1457 )
1458
1459 (defmacro c-put-is-sws (beg end)
1460 ;; This macro does a hidden buffer change.
1461 `(let ((beg ,beg) (end ,end))
1462 (put-text-property beg end 'c-is-sws t)
1463 ,@(when (facep 'c-debug-is-sws-face)
1464 `((c-debug-add-face beg end 'c-debug-is-sws-face)))))
1465
1466 (defmacro c-put-in-sws (beg end)
1467 ;; This macro does a hidden buffer change.
1468 `(let ((beg ,beg) (end ,end))
1469 (put-text-property beg end 'c-in-sws t)
1470 ,@(when (facep 'c-debug-is-sws-face)
1471 `((c-debug-add-face beg end 'c-debug-in-sws-face)))))
1472
1473 (defmacro c-remove-is-sws (beg end)
1474 ;; This macro does a hidden buffer change.
1475 `(let ((beg ,beg) (end ,end))
1476 (remove-text-properties beg end '(c-is-sws nil))
1477 ,@(when (facep 'c-debug-is-sws-face)
1478 `((c-debug-remove-face beg end 'c-debug-is-sws-face)))))
1479
1480 (defmacro c-remove-in-sws (beg end)
1481 ;; This macro does a hidden buffer change.
1482 `(let ((beg ,beg) (end ,end))
1483 (remove-text-properties beg end '(c-in-sws nil))
1484 ,@(when (facep 'c-debug-is-sws-face)
1485 `((c-debug-remove-face beg end 'c-debug-in-sws-face)))))
1486
1487 (defmacro c-remove-is-and-in-sws (beg end)
1488 ;; This macro does a hidden buffer change.
1489 `(let ((beg ,beg) (end ,end))
1490 (remove-text-properties beg end '(c-is-sws nil c-in-sws nil))
1491 ,@(when (facep 'c-debug-is-sws-face)
1492 `((c-debug-remove-face beg end 'c-debug-is-sws-face)
1493 (c-debug-remove-face beg end 'c-debug-in-sws-face)))))
1494
1495 (defsubst c-invalidate-sws-region-after (beg end)
1496 ;; Called from `after-change-functions'. Note that if
1497 ;; `c-forward-sws' or `c-backward-sws' are used outside
1498 ;; `c-save-buffer-state' or similar then this will remove the cache
1499 ;; properties right after they're added.
1500 ;;
1501 ;; This function does hidden buffer changes.
1502
1503 (save-excursion
1504 ;; Adjust the end to remove the properties in any following simple
1505 ;; ws up to and including the next line break, if there is any
1506 ;; after the changed region. This is necessary e.g. when a rung
1507 ;; marked empty line is converted to a line comment by inserting
1508 ;; "//" before the line break. In that case the line break would
1509 ;; keep the rung mark which could make a later `c-backward-sws'
1510 ;; move into the line comment instead of over it.
1511 (goto-char end)
1512 (skip-chars-forward " \t\f\v")
1513 (when (and (eolp) (not (eobp)))
1514 (setq end (1+ (point)))))
1515
1516 (when (and (= beg end)
1517 (get-text-property beg 'c-in-sws)
1518 (> beg (point-min))
1519 (get-text-property (1- beg) 'c-in-sws))
1520 ;; Ensure that an `c-in-sws' range gets broken. Note that it isn't
1521 ;; safe to keep a range that was continuous before the change. E.g:
1522 ;;
1523 ;; #define foo
1524 ;; \
1525 ;; bar
1526 ;;
1527 ;; There can be a "ladder" between "#" and "b". Now, if the newline
1528 ;; after "foo" is removed then "bar" will become part of the cpp
1529 ;; directive instead of a syntactically relevant token. In that
1530 ;; case there's no longer syntactic ws from "#" to "b".
1531 (setq beg (1- beg)))
1532
1533 (c-debug-sws-msg "c-invalidate-sws-region-after [%s..%s]" beg end)
1534 (c-remove-is-and-in-sws beg end))
1535
1536 (defun c-forward-sws ()
1537 ;; Used by `c-forward-syntactic-ws' to implement the unbounded search.
1538 ;;
1539 ;; This function might do hidden buffer changes.
1540
1541 (let (;; `rung-pos' is set to a position as early as possible in the
1542 ;; unmarked part of the simple ws region.
1543 (rung-pos (point)) next-rung-pos rung-end-pos last-put-in-sws-pos
1544 rung-is-marked next-rung-is-marked simple-ws-end
1545 ;; `safe-start' is set when it's safe to cache the start position.
1546 ;; It's not set if we've initially skipped over comments and line
1547 ;; continuations since we might have gone out through the end of a
1548 ;; macro then. This provision makes `c-forward-sws' not populate the
1549 ;; cache in the majority of cases, but otoh is `c-backward-sws' by far
1550 ;; more common.
1551 safe-start)
1552
1553 ;; Skip simple ws and do a quick check on the following character to see
1554 ;; if it's anything that can't start syntactic ws, so we can bail out
1555 ;; early in the majority of cases when there just are a few ws chars.
1556 (skip-chars-forward " \t\n\r\f\v")
1557 (when (looking-at c-syntactic-ws-start)
1558
1559 (setq rung-end-pos (min (1+ (point)) (point-max)))
1560 (if (setq rung-is-marked (text-property-any rung-pos rung-end-pos
1561 'c-is-sws t))
1562 ;; Find the last rung position to avoid setting properties in all
1563 ;; the cases when the marked rung is complete.
1564 ;; (`next-single-property-change' is certain to move at least one
1565 ;; step forward.)
1566 (setq rung-pos (1- (next-single-property-change
1567 rung-is-marked 'c-is-sws nil rung-end-pos)))
1568 ;; Got no marked rung here. Since the simple ws might have started
1569 ;; inside a line comment or cpp directive we must set `rung-pos' as
1570 ;; high as possible.
1571 (setq rung-pos (point)))
1572
1573 (while
1574 (progn
1575 (while
1576 (when (and rung-is-marked
1577 (get-text-property (point) 'c-in-sws))
1578
1579 ;; The following search is the main reason that `c-in-sws'
1580 ;; and `c-is-sws' aren't combined to one property.
1581 (goto-char (next-single-property-change
1582 (point) 'c-in-sws nil (point-max)))
1583 (unless (get-text-property (point) 'c-is-sws)
1584 ;; If the `c-in-sws' region extended past the last
1585 ;; `c-is-sws' char we have to go back a bit.
1586 (or (get-text-property (1- (point)) 'c-is-sws)
1587 (goto-char (previous-single-property-change
1588 (point) 'c-is-sws)))
1589 (backward-char))
1590
1591 (c-debug-sws-msg
1592 "c-forward-sws cached move %s -> %s (max %s)"
1593 rung-pos (point) (point-max))
1594
1595 (setq rung-pos (point))
1596 (and (> (skip-chars-forward " \t\n\r\f\v") 0)
1597 (not (eobp))))
1598
1599 ;; We'll loop here if there is simple ws after the last rung.
1600 ;; That means that there's been some change in it and it's
1601 ;; possible that we've stepped into another ladder, so extend
1602 ;; the previous one to join with it if there is one, and try to
1603 ;; use the cache again.
1604 (c-debug-sws-msg
1605 "c-forward-sws extending rung with [%s..%s] (max %s)"
1606 (1+ rung-pos) (1+ (point)) (point-max))
1607 (unless (get-text-property (point) 'c-is-sws)
1608 ;; Remove any `c-in-sws' property from the last char of
1609 ;; the rung before we mark it with `c-is-sws', so that we
1610 ;; won't connect with the remains of a broken "ladder".
1611 (c-remove-in-sws (point) (1+ (point))))
1612 (c-put-is-sws (1+ rung-pos)
1613 (1+ (point)))
1614 (c-put-in-sws rung-pos
1615 (setq rung-pos (point)
1616 last-put-in-sws-pos rung-pos)))
1617
1618 (setq simple-ws-end (point))
1619 (c-forward-comments)
1620
1621 (cond
1622 ((/= (point) simple-ws-end)
1623 ;; Skipped over comments. Don't cache at eob in case the buffer
1624 ;; is narrowed.
1625 (not (eobp)))
1626
1627 ((save-excursion
1628 (and c-opt-cpp-prefix
1629 (looking-at c-opt-cpp-start)
1630 (progn (skip-chars-backward " \t")
1631 (bolp))
1632 (or (bobp)
1633 (progn (backward-char)
1634 (not (eq (char-before) ?\\))))))
1635 ;; Skip a preprocessor directive.
1636 (end-of-line)
1637 (while (and (eq (char-before) ?\\)
1638 (= (forward-line 1) 0))
1639 (end-of-line))
1640 (forward-line 1)
1641 (setq safe-start t)
1642 ;; Don't cache at eob in case the buffer is narrowed.
1643 (not (eobp)))))
1644
1645 ;; We've searched over a piece of non-white syntactic ws. See if this
1646 ;; can be cached.
1647 (setq next-rung-pos (point))
1648 (skip-chars-forward " \t\n\r\f\v")
1649 (setq rung-end-pos (min (1+ (point)) (point-max)))
1650
1651 (if (or
1652 ;; Cache if we haven't skipped comments only, and if we started
1653 ;; either from a marked rung or from a completely uncached
1654 ;; position.
1655 (and safe-start
1656 (or rung-is-marked
1657 (not (get-text-property simple-ws-end 'c-in-sws))))
1658
1659 ;; See if there's a marked rung in the encountered simple ws. If
1660 ;; so then we can cache, unless `safe-start' is nil. Even then
1661 ;; we need to do this to check if the cache can be used for the
1662 ;; next step.
1663 (and (setq next-rung-is-marked
1664 (text-property-any next-rung-pos rung-end-pos
1665 'c-is-sws t))
1666 safe-start))
1667
1668 (progn
1669 (c-debug-sws-msg
1670 "c-forward-sws caching [%s..%s] - [%s..%s] (max %s)"
1671 rung-pos (1+ simple-ws-end) next-rung-pos rung-end-pos
1672 (point-max))
1673
1674 ;; Remove the properties for any nested ws that might be cached.
1675 ;; Only necessary for `c-is-sws' since `c-in-sws' will be set
1676 ;; anyway.
1677 (c-remove-is-sws (1+ simple-ws-end) next-rung-pos)
1678 (unless (and rung-is-marked (= rung-pos simple-ws-end))
1679 (c-put-is-sws rung-pos
1680 (1+ simple-ws-end))
1681 (setq rung-is-marked t))
1682 (c-put-in-sws rung-pos
1683 (setq rung-pos (point)
1684 last-put-in-sws-pos rung-pos))
1685 (unless (get-text-property (1- rung-end-pos) 'c-is-sws)
1686 ;; Remove any `c-in-sws' property from the last char of
1687 ;; the rung before we mark it with `c-is-sws', so that we
1688 ;; won't connect with the remains of a broken "ladder".
1689 (c-remove-in-sws (1- rung-end-pos) rung-end-pos))
1690 (c-put-is-sws next-rung-pos
1691 rung-end-pos))
1692
1693 (c-debug-sws-msg
1694 "c-forward-sws not caching [%s..%s] - [%s..%s] (max %s)"
1695 rung-pos (1+ simple-ws-end) next-rung-pos rung-end-pos
1696 (point-max))
1697
1698 ;; Set `rung-pos' for the next rung. It's the same thing here as
1699 ;; initially, except that the rung position is set as early as
1700 ;; possible since we can't be in the ending ws of a line comment or
1701 ;; cpp directive now.
1702 (if (setq rung-is-marked next-rung-is-marked)
1703 (setq rung-pos (1- (next-single-property-change
1704 rung-is-marked 'c-is-sws nil rung-end-pos)))
1705 (setq rung-pos next-rung-pos))
1706 (setq safe-start t)))
1707
1708 ;; Make sure that the newly marked `c-in-sws' region doesn't connect to
1709 ;; another one after the point (which might occur when editing inside a
1710 ;; comment or macro).
1711 (when (eq last-put-in-sws-pos (point))
1712 (cond ((< last-put-in-sws-pos (point-max))
1713 (c-debug-sws-msg
1714 "c-forward-sws clearing at %s for cache separation"
1715 last-put-in-sws-pos)
1716 (c-remove-in-sws last-put-in-sws-pos
1717 (1+ last-put-in-sws-pos)))
1718 (t
1719 ;; If at eob we have to clear the last character before the end
1720 ;; instead since the buffer might be narrowed and there might
1721 ;; be a `c-in-sws' after (point-max). In this case it's
1722 ;; necessary to clear both properties.
1723 (c-debug-sws-msg
1724 "c-forward-sws clearing thoroughly at %s for cache separation"
1725 (1- last-put-in-sws-pos))
1726 (c-remove-is-and-in-sws (1- last-put-in-sws-pos)
1727 last-put-in-sws-pos))))
1728 )))
1729
1730 (defun c-backward-sws ()
1731 ;; Used by `c-backward-syntactic-ws' to implement the unbounded search.
1732 ;;
1733 ;; This function might do hidden buffer changes.
1734
1735 (let (;; `rung-pos' is set to a position as late as possible in the unmarked
1736 ;; part of the simple ws region.
1737 (rung-pos (point)) next-rung-pos last-put-in-sws-pos
1738 rung-is-marked simple-ws-beg cmt-skip-pos)
1739
1740 ;; Skip simple horizontal ws and do a quick check on the preceding
1741 ;; character to see if it's anying that can't end syntactic ws, so we can
1742 ;; bail out early in the majority of cases when there just are a few ws
1743 ;; chars. Newlines are complicated in the backward direction, so we can't
1744 ;; skip over them.
1745 (skip-chars-backward " \t\f")
1746 (when (and (not (bobp))
1747 (save-excursion
1748 (backward-char)
1749 (looking-at c-syntactic-ws-end)))
1750
1751 ;; Try to find a rung position in the simple ws preceding point, so that
1752 ;; we can get a cache hit even if the last bit of the simple ws has
1753 ;; changed recently.
1754 (setq simple-ws-beg (point))
1755 (skip-chars-backward " \t\n\r\f\v")
1756 (if (setq rung-is-marked (text-property-any
1757 (point) (min (1+ rung-pos) (point-max))
1758 'c-is-sws t))
1759 ;; `rung-pos' will be the earliest marked position, which means that
1760 ;; there might be later unmarked parts in the simple ws region.
1761 ;; It's not worth the effort to fix that; the last part of the
1762 ;; simple ws is also typically edited often, so it could be wasted.
1763 (goto-char (setq rung-pos rung-is-marked))
1764 (goto-char simple-ws-beg))
1765
1766 (while
1767 (progn
1768 (while
1769 (when (and rung-is-marked
1770 (not (bobp))
1771 (get-text-property (1- (point)) 'c-in-sws))
1772
1773 ;; The following search is the main reason that `c-in-sws'
1774 ;; and `c-is-sws' aren't combined to one property.
1775 (goto-char (previous-single-property-change
1776 (point) 'c-in-sws nil (point-min)))
1777 (unless (get-text-property (point) 'c-is-sws)
1778 ;; If the `c-in-sws' region extended past the first
1779 ;; `c-is-sws' char we have to go forward a bit.
1780 (goto-char (next-single-property-change
1781 (point) 'c-is-sws)))
1782
1783 (c-debug-sws-msg
1784 "c-backward-sws cached move %s <- %s (min %s)"
1785 (point) rung-pos (point-min))
1786
1787 (setq rung-pos (point))
1788 (if (and (< (min (skip-chars-backward " \t\f\v")
1789 (progn
1790 (setq simple-ws-beg (point))
1791 (skip-chars-backward " \t\n\r\f\v")))
1792 0)
1793 (setq rung-is-marked
1794 (text-property-any (point) rung-pos
1795 'c-is-sws t)))
1796 t
1797 (goto-char simple-ws-beg)
1798 nil))
1799
1800 ;; We'll loop here if there is simple ws before the first rung.
1801 ;; That means that there's been some change in it and it's
1802 ;; possible that we've stepped into another ladder, so extend
1803 ;; the previous one to join with it if there is one, and try to
1804 ;; use the cache again.
1805 (c-debug-sws-msg
1806 "c-backward-sws extending rung with [%s..%s] (min %s)"
1807 rung-is-marked rung-pos (point-min))
1808 (unless (get-text-property (1- rung-pos) 'c-is-sws)
1809 ;; Remove any `c-in-sws' property from the last char of
1810 ;; the rung before we mark it with `c-is-sws', so that we
1811 ;; won't connect with the remains of a broken "ladder".
1812 (c-remove-in-sws (1- rung-pos) rung-pos))
1813 (c-put-is-sws rung-is-marked
1814 rung-pos)
1815 (c-put-in-sws rung-is-marked
1816 (1- rung-pos))
1817 (setq rung-pos rung-is-marked
1818 last-put-in-sws-pos rung-pos))
1819
1820 (c-backward-comments)
1821 (setq cmt-skip-pos (point))
1822
1823 (cond
1824 ((and c-opt-cpp-prefix
1825 (/= cmt-skip-pos simple-ws-beg)
1826 (c-beginning-of-macro))
1827 ;; Inside a cpp directive. See if it should be skipped over.
1828 (let ((cpp-beg (point)))
1829
1830 ;; Move back over all line continuations in the region skipped
1831 ;; over by `c-backward-comments'. If we go past it then we
1832 ;; started inside the cpp directive.
1833 (goto-char simple-ws-beg)
1834 (beginning-of-line)
1835 (while (and (> (point) cmt-skip-pos)
1836 (progn (backward-char)
1837 (eq (char-before) ?\\)))
1838 (beginning-of-line))
1839
1840 (if (< (point) cmt-skip-pos)
1841 ;; Don't move past the cpp directive if we began inside
1842 ;; it. Note that the position at the end of the last line
1843 ;; of the macro is also considered to be within it.
1844 (progn (goto-char cmt-skip-pos)
1845 nil)
1846
1847 ;; It's worthwhile to spend a little bit of effort on finding
1848 ;; the end of the macro, to get a good `simple-ws-beg'
1849 ;; position for the cache. Note that `c-backward-comments'
1850 ;; could have stepped over some comments before going into
1851 ;; the macro, and then `simple-ws-beg' must be kept on the
1852 ;; same side of those comments.
1853 (goto-char simple-ws-beg)
1854 (skip-chars-backward " \t\n\r\f\v")
1855 (if (eq (char-before) ?\\)
1856 (forward-char))
1857 (forward-line 1)
1858 (if (< (point) simple-ws-beg)
1859 ;; Might happen if comments after the macro were skipped
1860 ;; over.
1861 (setq simple-ws-beg (point)))
1862
1863 (goto-char cpp-beg)
1864 t)))
1865
1866 ((/= (save-excursion
1867 (skip-chars-forward " \t\n\r\f\v" simple-ws-beg)
1868 (setq next-rung-pos (point)))
1869 simple-ws-beg)
1870 ;; Skipped over comments. Must put point at the end of
1871 ;; the simple ws at point since we might be after a line
1872 ;; comment or cpp directive that's been partially
1873 ;; narrowed out, and we can't risk marking the simple ws
1874 ;; at the end of it.
1875 (goto-char next-rung-pos)
1876 t)))
1877
1878 ;; We've searched over a piece of non-white syntactic ws. See if this
1879 ;; can be cached.
1880 (setq next-rung-pos (point))
1881 (skip-chars-backward " \t\f\v")
1882
1883 (if (or
1884 ;; Cache if we started either from a marked rung or from a
1885 ;; completely uncached position.
1886 rung-is-marked
1887 (not (get-text-property (1- simple-ws-beg) 'c-in-sws))
1888
1889 ;; Cache if there's a marked rung in the encountered simple ws.
1890 (save-excursion
1891 (skip-chars-backward " \t\n\r\f\v")
1892 (text-property-any (point) (min (1+ next-rung-pos) (point-max))
1893 'c-is-sws t)))
1894
1895 (progn
1896 (c-debug-sws-msg
1897 "c-backward-sws caching [%s..%s] - [%s..%s] (min %s)"
1898 (point) (1+ next-rung-pos)
1899 simple-ws-beg (min (1+ rung-pos) (point-max))
1900 (point-min))
1901
1902 ;; Remove the properties for any nested ws that might be cached.
1903 ;; Only necessary for `c-is-sws' since `c-in-sws' will be set
1904 ;; anyway.
1905 (c-remove-is-sws (1+ next-rung-pos) simple-ws-beg)
1906 (unless (and rung-is-marked (= simple-ws-beg rung-pos))
1907 (let ((rung-end-pos (min (1+ rung-pos) (point-max))))
1908 (unless (get-text-property (1- rung-end-pos) 'c-is-sws)
1909 ;; Remove any `c-in-sws' property from the last char of
1910 ;; the rung before we mark it with `c-is-sws', so that we
1911 ;; won't connect with the remains of a broken "ladder".
1912 (c-remove-in-sws (1- rung-end-pos) rung-end-pos))
1913 (c-put-is-sws simple-ws-beg
1914 rung-end-pos)
1915 (setq rung-is-marked t)))
1916 (c-put-in-sws (setq simple-ws-beg (point)
1917 last-put-in-sws-pos simple-ws-beg)
1918 rung-pos)
1919 (c-put-is-sws (setq rung-pos simple-ws-beg)
1920 (1+ next-rung-pos)))
1921
1922 (c-debug-sws-msg
1923 "c-backward-sws not caching [%s..%s] - [%s..%s] (min %s)"
1924 (point) (1+ next-rung-pos)
1925 simple-ws-beg (min (1+ rung-pos) (point-max))
1926 (point-min))
1927 (setq rung-pos next-rung-pos
1928 simple-ws-beg (point))
1929 ))
1930
1931 ;; Make sure that the newly marked `c-in-sws' region doesn't connect to
1932 ;; another one before the point (which might occur when editing inside a
1933 ;; comment or macro).
1934 (when (eq last-put-in-sws-pos (point))
1935 (cond ((< (point-min) last-put-in-sws-pos)
1936 (c-debug-sws-msg
1937 "c-backward-sws clearing at %s for cache separation"
1938 (1- last-put-in-sws-pos))
1939 (c-remove-in-sws (1- last-put-in-sws-pos)
1940 last-put-in-sws-pos))
1941 ((> (point-min) 1)
1942 ;; If at bob and the buffer is narrowed, we have to clear the
1943 ;; character we're standing on instead since there might be a
1944 ;; `c-in-sws' before (point-min). In this case it's necessary
1945 ;; to clear both properties.
1946 (c-debug-sws-msg
1947 "c-backward-sws clearing thoroughly at %s for cache separation"
1948 last-put-in-sws-pos)
1949 (c-remove-is-and-in-sws last-put-in-sws-pos
1950 (1+ last-put-in-sws-pos)))))
1951 )))
1952
1953 \f
1954 ;; Other whitespace tools
1955 (defun c-partial-ws-p (beg end)
1956 ;; Is the region (beg end) WS, and is there WS (or BOB/EOB) next to the
1957 ;; region? This is a "heuristic" function. .....
1958 ;;
1959 ;; The motivation for the second bit is to check whether removing this
1960 ;; region would coalesce two symbols.
1961 ;;
1962 ;; FIXME!!! This function doesn't check virtual semicolons in any way. Be
1963 ;; careful about using this function for, e.g. AWK. (2007/3/7)
1964 (save-excursion
1965 (let ((end+1 (min (1+ end) (point-max))))
1966 (or (progn (goto-char (max (point-min) (1- beg)))
1967 (c-skip-ws-forward end)
1968 (eq (point) end))
1969 (progn (goto-char beg)
1970 (c-skip-ws-forward end+1)
1971 (eq (point) end+1))))))
1972 \f
1973 ;; A system for finding noteworthy parens before the point.
1974
1975 (defconst c-state-cache-too-far 5000)
1976 ;; A maximum comfortable scanning distance, e.g. between
1977 ;; `c-state-cache-good-pos' and "HERE" (where we call c-parse-state). When
1978 ;; this distance is exceeded, we take "emergency meausures", e.g. by clearing
1979 ;; the cache and starting again from point-min or a beginning of defun. This
1980 ;; value can be tuned for efficiency or set to a lower value for testing.
1981
1982 (defvar c-state-cache nil)
1983 (make-variable-buffer-local 'c-state-cache)
1984 ;; The state cache used by `c-parse-state' to cut down the amount of
1985 ;; searching. It's the result from some earlier `c-parse-state' call. See
1986 ;; `c-parse-state''s doc string for details of its structure.
1987 ;;
1988 ;; The use of the cached info is more effective if the next
1989 ;; `c-parse-state' call is on a line close by the one the cached state
1990 ;; was made at; the cache can actually slow down a little if the
1991 ;; cached state was made very far back in the buffer. The cache is
1992 ;; most effective if `c-parse-state' is used on each line while moving
1993 ;; forward.
1994
1995 (defvar c-state-cache-good-pos 1)
1996 (make-variable-buffer-local 'c-state-cache-good-pos)
1997 ;; This is a position where `c-state-cache' is known to be correct, or
1998 ;; nil (see below). It's a position inside one of the recorded unclosed
1999 ;; parens or the top level, but not further nested inside any literal or
2000 ;; subparen that is closed before the last recorded position.
2001 ;;
2002 ;; The exact position is chosen to try to be close to yet earlier than
2003 ;; the position where `c-state-cache' will be called next. Right now
2004 ;; the heuristic is to set it to the position after the last found
2005 ;; closing paren (of any type) before the line on which
2006 ;; `c-parse-state' was called. That is chosen primarily to work well
2007 ;; with refontification of the current line.
2008 ;;
2009 ;; 2009-07-28: When `c-state-point-min' and the last position where
2010 ;; `c-parse-state' or for which `c-invalidate-state-cache' was called, are
2011 ;; both in the same literal, there is no such "good position", and
2012 ;; c-state-cache-good-pos is then nil. This is the ONLY circumstance in which
2013 ;; it can be nil. In this case, `c-state-point-min-literal' will be non-nil.
2014 ;;
2015 ;; 2009-06-12: In a brace desert, c-state-cache-good-pos may also be in
2016 ;; the middle of the desert, as long as it is not within a brace pair
2017 ;; recorded in `c-state-cache' or a paren/bracket pair.
2018
2019
2020 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2021 ;; We maintain a simple cache of positions which aren't in a literal, so as to
2022 ;; speed up testing for non-literality.
2023 (defconst c-state-nonlit-pos-interval 10000)
2024 ;; The approximate interval between entries in `c-state-nonlit-pos-cache'.
2025
2026 (defvar c-state-nonlit-pos-cache nil)
2027 (make-variable-buffer-local 'c-state-nonlit-pos-cache)
2028 ;; A list of buffer positions which are known not to be in a literal. This is
2029 ;; ordered with higher positions at the front of the list. Only those which
2030 ;; are less than `c-state-nonlit-pos-cache-limit' are valid.
2031
2032 (defvar c-state-nonlit-pos-cache-limit 1)
2033 (make-variable-buffer-local 'c-state-nonlit-pos-cache-limit)
2034 ;; An upper limit on valid entries in `c-state-nonlit-pos-cache'. This is
2035 ;; reduced by buffer changes, and increased by invocations of
2036 ;; `c-state-literal-at'.
2037
2038 (defsubst c-state-pp-to-literal (from to)
2039 ;; Do a parse-partial-sexp from FROM to TO, returning the bounds of any
2040 ;; literal at TO as a cons, otherwise NIL.
2041 ;; FROM must not be in a literal, and the buffer should already be wide
2042 ;; enough.
2043 (save-excursion
2044 (let ((s (parse-partial-sexp from to)))
2045 (when (or (nth 3 s) (nth 4 s)) ; in a string or comment
2046 (parse-partial-sexp (point) (point-max)
2047 nil ; TARGETDEPTH
2048 nil ; STOPBEFORE
2049 s ; OLDSTATE
2050 'syntax-table) ; stop at end of literal
2051 (cons (nth 8 s) (point))))))
2052
2053 (defun c-state-literal-at (here)
2054 ;; If position HERE is inside a literal, return (START . END), the
2055 ;; boundaries of the literal (which may be outside the accessible bit of the
2056 ;; buffer). Otherwise, return nil.
2057 ;;
2058 ;; This function is almost the same as `c-literal-limits'. It differs in
2059 ;; that it is a lower level function, and that it rigourously follows the
2060 ;; syntax from BOB, whereas `c-literal-limits' uses a "local" safe position.
2061 (save-restriction
2062 (widen)
2063 (save-excursion
2064 (let ((c c-state-nonlit-pos-cache)
2065 pos npos lit)
2066 ;; Trim the cache to take account of buffer changes.
2067 (while (and c (> (car c) c-state-nonlit-pos-cache-limit))
2068 (setq c (cdr c)))
2069 (setq c-state-nonlit-pos-cache c)
2070
2071 (while (and c (> (car c) here))
2072 (setq c (cdr c)))
2073 (setq pos (or (car c) (point-min)))
2074
2075 (while (<= (setq npos (+ pos c-state-nonlit-pos-interval))
2076 here)
2077 (setq lit (c-state-pp-to-literal pos npos))
2078 (setq pos (or (cdr lit) npos)) ; end of literal containing npos.
2079 (setq c-state-nonlit-pos-cache (cons pos c-state-nonlit-pos-cache)))
2080
2081 (if (> pos c-state-nonlit-pos-cache-limit)
2082 (setq c-state-nonlit-pos-cache-limit pos))
2083 (if (< pos here)
2084 (setq lit (c-state-pp-to-literal pos here)))
2085 lit))))
2086
2087 (defsubst c-state-lit-beg (pos)
2088 ;; Return the start of the literal containing POS, or POS itself.
2089 (or (car (c-state-literal-at pos))
2090 pos))
2091
2092 (defsubst c-state-cache-non-literal-place (pos state)
2093 ;; Return a position outside of a string/comment at or before POS.
2094 ;; STATE is the parse-partial-sexp state at POS.
2095 (if (or (nth 3 state) ; in a string?
2096 (nth 4 state)) ; in a comment?
2097 (nth 8 state)
2098 pos))
2099
2100
2101 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2102 ;; Stuff to do with point-min, and coping with any literal there.
2103 (defvar c-state-point-min 1)
2104 (make-variable-buffer-local 'c-state-point-min)
2105 ;; This is (point-min) when `c-state-cache' was last calculated. A change of
2106 ;; narrowing is likely to affect the parens that are visible before the point.
2107
2108 (defvar c-state-point-min-lit-type nil)
2109 (make-variable-buffer-local 'c-state-point-min-lit-type)
2110 (defvar c-state-point-min-lit-start nil)
2111 (make-variable-buffer-local 'c-state-point-min-lit-start)
2112 ;; These two variables define the literal, if any, containing point-min.
2113 ;; Their values are, respectively, 'string, c, or c++, and the start of the
2114 ;; literal. If there's no literal there, they're both nil.
2115
2116 (defvar c-state-min-scan-pos 1)
2117 (make-variable-buffer-local 'c-state-min-scan-pos)
2118 ;; This is the earliest buffer-pos from which scanning can be done. It is
2119 ;; either the end of the literal containing point-min, or point-min itself.
2120 ;; It becomes nil if the buffer is changed earlier than this point.
2121 (defun c-state-get-min-scan-pos ()
2122 ;; Return the lowest valid scanning pos. This will be the end of the
2123 ;; literal enclosing point-min, or point-min itself.
2124 (or c-state-min-scan-pos
2125 (save-restriction
2126 (save-excursion
2127 (widen)
2128 (goto-char c-state-point-min-lit-start)
2129 (if (eq c-state-point-min-lit-type 'string)
2130 (forward-sexp)
2131 (forward-comment 1))
2132 (setq c-state-min-scan-pos (point))))))
2133
2134 (defun c-state-mark-point-min-literal ()
2135 ;; Determine the properties of any literal containing POINT-MIN, setting the
2136 ;; variables `c-state-point-min-lit-type', `c-state-point-min-lit-start',
2137 ;; and `c-state-min-scan-pos' accordingly. The return value is meaningless.
2138 (let ((p-min (point-min))
2139 lit)
2140 (save-restriction
2141 (widen)
2142 (setq lit (c-state-literal-at p-min))
2143 (if lit
2144 (setq c-state-point-min-lit-type
2145 (save-excursion
2146 (goto-char (car lit))
2147 (cond
2148 ((looking-at c-block-comment-start-regexp) 'c)
2149 ((looking-at c-line-comment-starter) 'c++)
2150 (t 'string)))
2151 c-state-point-min-lit-start (car lit)
2152 c-state-min-scan-pos (cdr lit))
2153 (setq c-state-point-min-lit-type nil
2154 c-state-point-min-lit-start nil
2155 c-state-min-scan-pos p-min)))))
2156
2157
2158 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2159 ;; A variable which signals a brace dessert - helpful for reducing the number
2160 ;; of fruitless backward scans.
2161 (defvar c-state-brace-pair-desert nil)
2162 (make-variable-buffer-local 'c-state-brace-pair-desert)
2163 ;; Used only in `c-append-lower-brace-pair-to-state-cache'. It is set when an
2164 ;; that defun has searched backwards for a brace pair and not found one. Its
2165 ;; value is either nil or a cons (PA . FROM), where PA is the position of the
2166 ;; enclosing opening paren/brace/bracket which bounds the backwards search (or
2167 ;; nil when at top level) and FROM is where the backward search started. It
2168 ;; is reset to nil in `c-invalidate-state-cache'.
2169
2170
2171 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2172 ;; Lowish level functions/macros which work directly on `c-state-cache', or a
2173 ;; list of like structure.
2174 (defmacro c-state-cache-top-lparen (&optional cache)
2175 ;; Return the address of the top left brace/bracket/paren recorded in CACHE
2176 ;; (default `c-state-cache') (or nil).
2177 (let ((cash (or cache 'c-state-cache)))
2178 `(if (consp (car ,cash))
2179 (caar ,cash)
2180 (car ,cash))))
2181
2182 (defmacro c-state-cache-top-paren (&optional cache)
2183 ;; Return the address of the latest brace/bracket/paren (whether left or
2184 ;; right) recorded in CACHE (default `c-state-cache') or nil.
2185 (let ((cash (or cache 'c-state-cache)))
2186 `(if (consp (car ,cash))
2187 (cdar ,cash)
2188 (car ,cash))))
2189
2190 (defmacro c-state-cache-after-top-paren (&optional cache)
2191 ;; Return the position just after the latest brace/bracket/paren (whether
2192 ;; left or right) recorded in CACHE (default `c-state-cache') or nil.
2193 (let ((cash (or cache 'c-state-cache)))
2194 `(if (consp (car ,cash))
2195 (cdar ,cash)
2196 (and (car ,cash)
2197 (1+ (car ,cash))))))
2198
2199 (defun c-get-cache-scan-pos (here)
2200 ;; From the state-cache, determine the buffer position from which we might
2201 ;; scan forward to HERE to update this cache. This position will be just
2202 ;; after a paren/brace/bracket recorded in the cache, if possible, otherwise
2203 ;; return the earliest position in the accessible region which isn't within
2204 ;; a literal. If the visible portion of the buffer is entirely within a
2205 ;; literal, return NIL.
2206 (let ((c c-state-cache) elt)
2207 ;(while (>= (or (c-state-cache-top-lparen c) 1) here)
2208 (while (and c
2209 (>= (c-state-cache-top-lparen c) here))
2210 (setq c (cdr c)))
2211
2212 (setq elt (car c))
2213 (cond
2214 ((consp elt)
2215 (if (> (cdr elt) here)
2216 (1+ (car elt))
2217 (cdr elt)))
2218 (elt (1+ elt))
2219 ((<= (c-state-get-min-scan-pos) here)
2220 (c-state-get-min-scan-pos))
2221 (t nil))))
2222
2223 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2224 ;; Variables which keep track of preprocessor constructs.
2225 (defvar c-state-old-cpp-beg nil)
2226 (make-variable-buffer-local 'c-state-old-cpp-beg)
2227 (defvar c-state-old-cpp-end nil)
2228 (make-variable-buffer-local 'c-state-old-cpp-end)
2229 ;; These are the limits of the macro containing point at the previous call of
2230 ;; `c-parse-state', or nil.
2231
2232 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2233 ;; Defuns which analyse the buffer, yet don't change `c-state-cache'.
2234 (defun c-get-fallback-scan-pos (here)
2235 ;; Return a start position for building `c-state-cache' from
2236 ;; scratch. This will be at the top level, 2 defuns back.
2237 (save-excursion
2238 ;; Go back 2 bods, but ignore any bogus positions returned by
2239 ;; beginning-of-defun (i.e. open paren in column zero).
2240 (goto-char here)
2241 (let ((cnt 2))
2242 (while (not (or (bobp) (zerop cnt)))
2243 (c-beginning-of-defun-1) ; Pure elisp BOD.
2244 (if (eq (char-after) ?\{)
2245 (setq cnt (1- cnt)))))
2246 (point)))
2247
2248 (defun c-state-balance-parens-backwards (here- here+ top)
2249 ;; Return the position of the opening paren/brace/bracket before HERE- which
2250 ;; matches the outermost close p/b/b between HERE+ and TOP. Except when
2251 ;; there's a macro, HERE- and HERE+ are the same. Like this:
2252 ;;
2253 ;; ............................................
2254 ;; | |
2255 ;; ( [ ( .........#macro.. ) ( ) ] )
2256 ;; ^ ^ ^ ^
2257 ;; | | | |
2258 ;; return HERE- HERE+ TOP
2259 ;;
2260 ;; If there aren't enough opening paren/brace/brackets, return the position
2261 ;; of the outermost one found, or HERE- if there are none. If there are no
2262 ;; closeing p/b/bs between HERE+ and TOP, return HERE-. HERE-/+ and TOP
2263 ;; must not be inside literals. Only the accessible portion of the buffer
2264 ;; will be scanned.
2265
2266 ;; PART 1: scan from `here+' up to `top', accumulating ")"s which enclose
2267 ;; `here'. Go round the next loop each time we pass over such a ")". These
2268 ;; probably match "("s before `here-'.
2269 (let (pos pa ren+1 lonely-rens)
2270 (save-excursion
2271 (save-restriction
2272 (narrow-to-region (point-min) top) ; This can move point, sometimes.
2273 (setq pos here+)
2274 (c-safe
2275 (while
2276 (setq ren+1 (scan-lists pos 1 1)) ; might signal
2277 (setq lonely-rens (cons ren+1 lonely-rens)
2278 pos ren+1)))))
2279
2280 ;; PART 2: Scan back before `here-' searching for the "("s
2281 ;; matching/mismatching the ")"s found above. We only need to direct the
2282 ;; caller to scan when we've encountered unmatched right parens.
2283 (setq pos here-)
2284 (when lonely-rens
2285 (c-safe
2286 (while
2287 (and lonely-rens ; actual values aren't used.
2288 (setq pa (scan-lists pos -1 1)))
2289 (setq pos pa)
2290 (setq lonely-rens (cdr lonely-rens)))))
2291 pos))
2292
2293 (defun c-parse-state-get-strategy (here good-pos)
2294 ;; Determine the scanning strategy for adjusting `c-parse-state', attempting
2295 ;; to minimise the amount of scanning. HERE is the pertinent position in
2296 ;; the buffer, GOOD-POS is a position where `c-state-cache' (possibly with
2297 ;; its head trimmed) is known to be good, or nil if there is no such
2298 ;; position.
2299 ;;
2300 ;; The return value is a list, one of the following:
2301 ;;
2302 ;; o - ('forward CACHE-POS START-POINT) - scan forward from START-POINT,
2303 ;; which is not less than CACHE-POS.
2304 ;; o - ('backward CACHE-POS nil) - scan backwards (from HERE).
2305 ;; o - ('BOD nil START-POINT) - scan forwards from START-POINT, which is at the
2306 ;; top level.
2307 ;; o - ('IN-LIT nil nil) - point is inside the literal containing point-min.
2308 ;; , where CACHE-POS is the highest position recorded in `c-state-cache' at
2309 ;; or below HERE.
2310 (let ((cache-pos (c-get-cache-scan-pos here)) ; highest position below HERE in cache (or 1)
2311 BOD-pos ; position of 2nd BOD before HERE.
2312 strategy ; 'forward, 'backward, 'BOD, or 'IN-LIT.
2313 start-point
2314 how-far) ; putative scanning distance.
2315 (setq good-pos (or good-pos (c-state-get-min-scan-pos)))
2316 (cond
2317 ((< here (c-state-get-min-scan-pos))
2318 (setq strategy 'IN-LIT
2319 start-point nil
2320 cache-pos nil
2321 how-far 0))
2322 ((<= good-pos here)
2323 (setq strategy 'forward
2324 start-point (max good-pos cache-pos)
2325 how-far (- here start-point)))
2326 ((< (- good-pos here) (- here cache-pos)) ; FIXME!!! ; apply some sort of weighting.
2327 (setq strategy 'backward
2328 how-far (- good-pos here)))
2329 (t
2330 (setq strategy 'forward
2331 how-far (- here cache-pos)
2332 start-point cache-pos)))
2333
2334 ;; Might we be better off starting from the top level, two defuns back,
2335 ;; instead?
2336 (when (> how-far c-state-cache-too-far)
2337 (setq BOD-pos (c-get-fallback-scan-pos here)) ; somewhat EXPENSIVE!!!
2338 (if (< (- here BOD-pos) how-far)
2339 (setq strategy 'BOD
2340 start-point BOD-pos)))
2341
2342 (list
2343 strategy
2344 (and (memq strategy '(forward backward)) cache-pos)
2345 (and (memq strategy '(forward BOD)) start-point))))
2346
2347
2348 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2349 ;; Routines which change `c-state-cache' and associated values.
2350 (defun c-renarrow-state-cache ()
2351 ;; The region (more precisely, point-min) has changed since we
2352 ;; calculated `c-state-cache'. Amend `c-state-cache' accordingly.
2353 (if (< (point-min) c-state-point-min)
2354 ;; If point-min has MOVED BACKWARDS then we drop the state completely.
2355 ;; It would be possible to do a better job here and recalculate the top
2356 ;; only.
2357 (progn
2358 (c-state-mark-point-min-literal)
2359 (setq c-state-cache nil
2360 c-state-cache-good-pos c-state-min-scan-pos
2361 c-state-brace-pair-desert nil))
2362
2363 ;; point-min has MOVED FORWARD.
2364
2365 ;; Is the new point-min inside a (different) literal?
2366 (unless (and c-state-point-min-lit-start ; at prev. point-min
2367 (< (point-min) (c-state-get-min-scan-pos)))
2368 (c-state-mark-point-min-literal))
2369
2370 ;; Cut off a bit of the tail from `c-state-cache'.
2371 (let ((ptr (cons nil c-state-cache))
2372 pa)
2373 (while (and (setq pa (c-state-cache-top-lparen (cdr ptr)))
2374 (>= pa (point-min)))
2375 (setq ptr (cdr ptr)))
2376
2377 (when (consp ptr)
2378 (if (eq (cdr ptr) c-state-cache)
2379 (setq c-state-cache nil
2380 c-state-cache-good-pos c-state-min-scan-pos)
2381 (setcdr ptr nil)
2382 (setq c-state-cache-good-pos (1+ (c-state-cache-top-lparen))))
2383 )))
2384
2385 (setq c-state-point-min (point-min)))
2386
2387 (defun c-append-lower-brace-pair-to-state-cache (from &optional upper-lim)
2388 ;; If there is a brace pair preceding FROM in the buffer (not necessarily
2389 ;; immediately preceding), push a cons onto `c-state-cache' to represent it.
2390 ;; FROM must not be inside a literal. If UPPER-LIM is non-nil, we append
2391 ;; the highest brace pair whose "}" is below UPPER-LIM.
2392 ;;
2393 ;; Return non-nil when this has been done.
2394 ;;
2395 ;; This routine should be fast. Since it can get called a LOT, we maintain
2396 ;; `c-state-brace-pair-desert', a small cache of "failures", such that we
2397 ;; reduce the time wasted in repeated fruitless searches in brace deserts.
2398 (save-excursion
2399 (save-restriction
2400 (let ((bra from) ce ; Positions of "{" and "}".
2401 new-cons
2402 (cache-pos (c-state-cache-top-lparen)) ; might be nil.
2403 (macro-start-or-from
2404 (progn (goto-char from)
2405 (c-beginning-of-macro)
2406 (point))))
2407 (or upper-lim (setq upper-lim from))
2408
2409 ;; If we're essentially repeating a fruitless search, just give up.
2410 (unless (and c-state-brace-pair-desert
2411 (eq cache-pos (car c-state-brace-pair-desert))
2412 (<= from (cdr c-state-brace-pair-desert)))
2413 ;; Only search what we absolutely need to:
2414 (if (and c-state-brace-pair-desert
2415 (> from (cdr c-state-brace-pair-desert)))
2416 (narrow-to-region (cdr c-state-brace-pair-desert) (point-max)))
2417
2418 ;; In the next pair of nested loops, the inner one moves back past a
2419 ;; pair of (mis-)matching parens or brackets; the outer one moves
2420 ;; back over a sequence of unmatched close brace/paren/bracket each
2421 ;; time round.
2422 (while
2423 (progn
2424 (c-safe
2425 (while
2426 (and (setq ce (scan-lists bra -1 -1)) ; back past )/]/}; might signal
2427 (setq bra (scan-lists ce -1 1)) ; back past (/[/{; might signal
2428 (or (> ce upper-lim)
2429 (not (eq (char-after bra) ?\{))
2430 (and (goto-char bra)
2431 (c-beginning-of-macro)
2432 (< (point) macro-start-or-from))))))
2433 (and ce (< ce bra)))
2434 (setq bra ce)) ; If we just backed over an unbalanced closing
2435 ; brace, ignore it.
2436
2437 (if (and ce (< bra ce) (eq (char-after bra) ?\{))
2438 ;; We've found the desired brace-pair.
2439 (progn
2440 (setq new-cons (cons bra (1+ ce)))
2441 (cond
2442 ((consp (car c-state-cache))
2443 (setcar c-state-cache new-cons))
2444 ((and (numberp (car c-state-cache)) ; probably never happens
2445 (< ce (car c-state-cache)))
2446 (setcdr c-state-cache
2447 (cons new-cons (cdr c-state-cache))))
2448 (t (setq c-state-cache (cons new-cons c-state-cache)))))
2449
2450 ;; We haven't found a brace pair. Record this.
2451 (setq c-state-brace-pair-desert (cons cache-pos from))))))))
2452
2453 (defsubst c-state-push-any-brace-pair (bra+1 macro-start-or-here)
2454 ;; If BRA+1 is nil, do nothing. Otherwise, BRA+1 is the buffer position
2455 ;; following a {, and that brace has a (mis-)matching } (or ]), and we
2456 ;; "push" "a" brace pair onto `c-state-cache'.
2457 ;;
2458 ;; Here "push" means overwrite the top element if it's itself a brace-pair,
2459 ;; otherwise push it normally.
2460 ;;
2461 ;; The brace pair we push is normally the one surrounding BRA+1, but if the
2462 ;; latter is inside a macro, not being a macro containing
2463 ;; MACRO-START-OR-HERE, we scan backwards through the buffer for a non-macro
2464 ;; base pair. This latter case is assumed to be rare.
2465 ;;
2466 ;; Note: POINT is not preserved in this routine.
2467 (if bra+1
2468 (if (or (> bra+1 macro-start-or-here)
2469 (progn (goto-char bra+1)
2470 (not (c-beginning-of-macro))))
2471 (setq c-state-cache
2472 (cons (cons (1- bra+1)
2473 (scan-lists bra+1 1 1))
2474 (if (consp (car c-state-cache))
2475 (cdr c-state-cache)
2476 c-state-cache)))
2477 ;; N.B. This defsubst codes one method for the simple, normal case,
2478 ;; and a more sophisticated, slower way for the general case. Don't
2479 ;; eliminate this defsubst - it's a speed optimisation.
2480 (c-append-lower-brace-pair-to-state-cache (1- bra+1)))))
2481
2482 (defun c-append-to-state-cache (from)
2483 ;; Scan the buffer from FROM to (point-max), adding elements into
2484 ;; `c-state-cache' for braces etc. Return a candidate for
2485 ;; `c-state-cache-good-pos'.
2486 ;;
2487 ;; FROM must be after the latest brace/paren/bracket in `c-state-cache', if
2488 ;; any. Typically, it is immediately after it. It must not be inside a
2489 ;; literal.
2490 (let ((here-bol (c-point 'bol (point-max)))
2491 (macro-start-or-here
2492 (save-excursion (goto-char (point-max))
2493 (if (c-beginning-of-macro)
2494 (point)
2495 (point-max))))
2496 pa+1 ; pos just after an opening PAren (or brace).
2497 (ren+1 from) ; usually a pos just after an closing paREN etc.
2498 ; Is actually the pos. to scan for a (/{/[ from,
2499 ; which sometimes is after a silly )/}/].
2500 paren+1 ; Pos after some opening or closing paren.
2501 paren+1s ; A list of `paren+1's; used to determine a
2502 ; good-pos.
2503 bra+1 ce+1 ; just after L/R bra-ces.
2504 bra+1s ; list of OLD values of bra+1.
2505 mstart) ; start of a macro.
2506
2507 (save-excursion
2508 ;; Each time round the following loop, we enter a succesively deeper
2509 ;; level of brace/paren nesting. (Except sometimes we "continue at
2510 ;; the existing level".) `pa+1' is a pos inside an opening
2511 ;; brace/paren/bracket, usually just after it.
2512 (while
2513 (progn
2514 ;; Each time round the next loop moves forward over an opening then
2515 ;; a closing brace/bracket/paren. This loop is white hot, so it
2516 ;; plays ugly tricks to go fast. DON'T PUT ANYTHING INTO THIS
2517 ;; LOOP WHICH ISN'T ABSOLUTELY NECESSARY!!! It terminates when a
2518 ;; call of `scan-lists' signals an error, which happens when there
2519 ;; are no more b/b/p's to scan.
2520 (c-safe
2521 (while t
2522 (setq pa+1 (scan-lists ren+1 1 -1) ; Into (/{/[; might signal
2523 paren+1s (cons pa+1 paren+1s))
2524 (setq ren+1 (scan-lists pa+1 1 1)) ; Out of )/}/]; might signal
2525 (if (and (eq (char-before pa+1) ?{)) ; Check for a macro later.
2526 (setq bra+1 pa+1))
2527 (setcar paren+1s ren+1)))
2528
2529 (if (and pa+1 (> pa+1 ren+1))
2530 ;; We've just entered a deeper nesting level.
2531 (progn
2532 ;; Insert the brace pair (if present) and the single open
2533 ;; paren/brace/bracket into `c-state-cache' It cannot be
2534 ;; inside a macro, except one around point, because of what
2535 ;; `c-neutralize-syntax-in-CPP' has done.
2536 (c-state-push-any-brace-pair bra+1 macro-start-or-here)
2537 ;; Insert the opening brace/bracket/paren position.
2538 (setq c-state-cache (cons (1- pa+1) c-state-cache))
2539 ;; Clear admin stuff for the next more nested part of the scan.
2540 (setq ren+1 pa+1 pa+1 nil bra+1 nil bra+1s nil)
2541 t) ; Carry on the loop
2542
2543 ;; All open p/b/b's at this nesting level, if any, have probably
2544 ;; been closed by matching/mismatching ones. We're probably
2545 ;; finished - we just need to check for having found an
2546 ;; unmatched )/}/], which we ignore. Such a )/}/] can't be in a
2547 ;; macro, due the action of `c-neutralize-syntax-in-CPP'.
2548 (c-safe (setq ren+1 (scan-lists ren+1 1 1)))))) ; acts as loop control.
2549
2550 ;; Record the final, innermost, brace-pair if there is one.
2551 (c-state-push-any-brace-pair bra+1 macro-start-or-here)
2552
2553 ;; Determine a good pos
2554 (while (and (setq paren+1 (car paren+1s))
2555 (> (if (> paren+1 macro-start-or-here)
2556 paren+1
2557 (goto-char paren+1)
2558 (setq mstart (and (c-beginning-of-macro)
2559 (point)))
2560 (or mstart paren+1))
2561 here-bol))
2562 (setq paren+1s (cdr paren+1s)))
2563 (cond
2564 ((and paren+1 mstart)
2565 (min paren+1 mstart))
2566 (paren+1)
2567 (t from)))))
2568
2569 (defun c-remove-stale-state-cache (good-pos pps-point)
2570 ;; Remove stale entries from the `c-cache-state', i.e. those which will
2571 ;; not be in it when it is amended for position (point-max).
2572 ;; Additionally, the "outermost" open-brace entry before (point-max)
2573 ;; will be converted to a cons if the matching close-brace is scanned.
2574 ;;
2575 ;; GOOD-POS is a "maximal" "safe position" - there must be no open
2576 ;; parens/braces/brackets between GOOD-POS and (point-max).
2577 ;;
2578 ;; As a second thing, calculate the result of parse-partial-sexp at
2579 ;; PPS-POINT, w.r.t. GOOD-POS. The motivation here is that
2580 ;; `c-state-cache-good-pos' may become PPS-POINT, but the caller may need to
2581 ;; adjust it to get outside a string/comment. (Sorry about this! The code
2582 ;; needs to be FAST).
2583 ;;
2584 ;; Return a list (GOOD-POS SCAN-BACK-POS PPS-STATE), where
2585 ;; o - GOOD-POS is a position where the new value `c-state-cache' is known
2586 ;; to be good (we aim for this to be as high as possible);
2587 ;; o - SCAN-BACK-POS, if not nil, indicates there may be a brace pair
2588 ;; preceding POS which needs to be recorded in `c-state-cache'. It is a
2589 ;; position to scan backwards from.
2590 ;; o - PPS-STATE is the parse-partial-sexp state at PPS-POINT.
2591 (save-restriction
2592 (narrow-to-region 1 (point-max))
2593 (save-excursion
2594 (let* ((in-macro-start ; start of macro containing (point-max) or nil.
2595 (save-excursion
2596 (goto-char (point-max))
2597 (and (c-beginning-of-macro)
2598 (point))))
2599 (good-pos-actual-macro-start ; Start of macro containing good-pos
2600 ; or nil
2601 (and (< good-pos (point-max))
2602 (save-excursion
2603 (goto-char good-pos)
2604 (and (c-beginning-of-macro)
2605 (point)))))
2606 (good-pos-actual-macro-end ; End of this macro, (maybe
2607 ; (point-max)), or nil.
2608 (and good-pos-actual-macro-start
2609 (save-excursion
2610 (goto-char good-pos-actual-macro-start)
2611 (c-end-of-macro)
2612 (point))))
2613 pps-state ; Will be 9 or 10 elements long.
2614 pos
2615 upper-lim ; ,beyond which `c-state-cache' entries are removed
2616 scan-back-pos
2617 pair-beg pps-point-state target-depth)
2618
2619 ;; Remove entries beyond (point-max). Also remove any entries inside
2620 ;; a macro, unless (point-max) is in the same macro.
2621 (setq upper-lim
2622 (if (or (null c-state-old-cpp-beg)
2623 (and (> (point-max) c-state-old-cpp-beg)
2624 (< (point-max) c-state-old-cpp-end)))
2625 (point-max)
2626 (min (point-max) c-state-old-cpp-beg)))
2627 (while (and c-state-cache (>= (c-state-cache-top-lparen) upper-lim))
2628 (setq c-state-cache (cdr c-state-cache)))
2629 ;; If `upper-lim' is inside the last recorded brace pair, remove its
2630 ;; RBrace and indicate we'll need to search backwards for a previous
2631 ;; brace pair.
2632 (when (and c-state-cache
2633 (consp (car c-state-cache))
2634 (> (cdar c-state-cache) upper-lim))
2635 (setcar c-state-cache (caar c-state-cache))
2636 (setq scan-back-pos (car c-state-cache)))
2637
2638 ;; The next loop jumps forward out of a nested level of parens each
2639 ;; time round; the corresponding elements in `c-state-cache' are
2640 ;; removed. `pos' is just after the brace-pair or the open paren at
2641 ;; (car c-state-cache). There can be no open parens/braces/brackets
2642 ;; between `good-pos'/`good-pos-actual-macro-start' and (point-max),
2643 ;; due to the interface spec to this function.
2644 (setq pos (if (and good-pos-actual-macro-end
2645 (not (eq good-pos-actual-macro-start
2646 in-macro-start)))
2647 (1+ good-pos-actual-macro-end) ; get outside the macro as
2648 ; marked by a `category' text property.
2649 good-pos))
2650 (goto-char pos)
2651 (while (and c-state-cache
2652 (< (point) (point-max)))
2653 (cond
2654 ((null pps-state) ; first time through
2655 (setq target-depth -1))
2656 ((eq (car pps-state) target-depth) ; found closing ),},]
2657 (setq target-depth (1- (car pps-state))))
2658 ;; Do nothing when we've merely reached pps-point.
2659 )
2660
2661 ;; Scan!
2662 (setq pps-state
2663 (parse-partial-sexp
2664 (point) (if (< (point) pps-point) pps-point (point-max))
2665 target-depth
2666 nil pps-state))
2667
2668 (if (= (point) pps-point)
2669 (setq pps-point-state pps-state))
2670
2671 (when (eq (car pps-state) target-depth)
2672 (setq pos (point)) ; POS is now just after an R-paren/brace.
2673 (cond
2674 ((and (consp (car c-state-cache))
2675 (eq (point) (cdar c-state-cache)))
2676 ;; We've just moved out of the paren pair containing the brace-pair
2677 ;; at (car c-state-cache). `pair-beg' is where the open paren is,
2678 ;; and is potentially where the open brace of a cons in
2679 ;; c-state-cache will be.
2680 (setq pair-beg (car-safe (cdr c-state-cache))
2681 c-state-cache (cdr-safe (cdr c-state-cache)))) ; remove {}pair + containing Lparen.
2682 ((numberp (car c-state-cache))
2683 (setq pair-beg (car c-state-cache)
2684 c-state-cache (cdr c-state-cache))) ; remove this
2685 ; containing Lparen
2686 ((numberp (cadr c-state-cache))
2687 (setq pair-beg (cadr c-state-cache)
2688 c-state-cache (cddr c-state-cache))) ; Remove a paren pair
2689 ; together with enclosed brace pair.
2690 ;; (t nil) ; Ignore an unmated Rparen.
2691 )))
2692
2693 (if (< (point) pps-point)
2694 (setq pps-state (parse-partial-sexp (point) pps-point
2695 nil nil ; TARGETDEPTH, STOPBEFORE
2696 pps-state)))
2697
2698 ;; If the last paren pair we moved out of was actually a brace pair,
2699 ;; insert it into `c-state-cache'.
2700 (when (and pair-beg (eq (char-after pair-beg) ?{))
2701 (if (consp (car-safe c-state-cache))
2702 (setq c-state-cache (cdr c-state-cache)))
2703 (setq c-state-cache (cons (cons pair-beg pos)
2704 c-state-cache)))
2705
2706 (list pos scan-back-pos pps-state)))))
2707
2708 (defun c-remove-stale-state-cache-backwards (here cache-pos)
2709 ;; Strip stale elements of `c-state-cache' by moving backwards through the
2710 ;; buffer, and inform the caller of the scenario detected.
2711 ;;
2712 ;; HERE is the position we're setting `c-state-cache' for.
2713 ;; CACHE-POS is just after the latest recorded position in `c-state-cache'
2714 ;; before HERE, or a position at or near point-min which isn't in a
2715 ;; literal.
2716 ;;
2717 ;; This function must only be called only when (> `c-state-cache-good-pos'
2718 ;; HERE). Usually the gap between CACHE-POS and HERE is large. It is thus
2719 ;; optimised to eliminate (or minimise) scanning between these two
2720 ;; positions.
2721 ;;
2722 ;; Return a three element list (GOOD-POS SCAN-BACK-POS FWD-FLAG), where:
2723 ;; o - GOOD-POS is a "good position", where `c-state-cache' is valid, or
2724 ;; could become so after missing elements are inserted into
2725 ;; `c-state-cache'. This is JUST AFTER an opening or closing
2726 ;; brace/paren/bracket which is already in `c-state-cache' or just before
2727 ;; one otherwise. exceptionally (when there's no such b/p/b handy) the BOL
2728 ;; before `here''s line, or the start of the literal containing it.
2729 ;; o - SCAN-BACK-POS, if non-nil, indicates there may be a brace pair
2730 ;; preceding POS which isn't recorded in `c-state-cache'. It is a position
2731 ;; to scan backwards from.
2732 ;; o - FWD-FLAG, if non-nil, indicates there may be parens/braces between
2733 ;; POS and HERE which aren't recorded in `c-state-cache'.
2734 ;;
2735 ;; The comments in this defun use "paren" to mean parenthesis or square
2736 ;; bracket (as contrasted with a brace), and "(" and ")" likewise.
2737 ;;
2738 ;; . {..} (..) (..) ( .. { } ) (...) ( .... . ..)
2739 ;; | | | | | |
2740 ;; CP E here D C good
2741 (let ((pos c-state-cache-good-pos)
2742 pa ren ; positions of "(" and ")"
2743 dropped-cons ; whether the last element dropped from `c-state-cache'
2744 ; was a cons (representing a brace-pair)
2745 good-pos ; see above.
2746 lit ; (START . END) of a literal containing some point.
2747 here-lit-start here-lit-end ; bounds of literal containing `here'
2748 ; or `here' itself.
2749 here- here+ ; start/end of macro around HERE, or HERE
2750 (here-bol (c-point 'bol here))
2751 (too-far-back (max (- here c-state-cache-too-far) 1)))
2752
2753 ;; Remove completely irrelevant entries from `c-state-cache'.
2754 (while (and c-state-cache
2755 (>= (setq pa (c-state-cache-top-lparen)) here))
2756 (setq dropped-cons (consp (car c-state-cache)))
2757 (setq c-state-cache (cdr c-state-cache))
2758 (setq pos pa))
2759 ;; At this stage, (> pos here);
2760 ;; (< (c-state-cache-top-lparen) here) (or is nil).
2761
2762 (cond
2763 ((and (consp (car c-state-cache))
2764 (> (cdar c-state-cache) here))
2765 ;; CASE 1: The top of the cache is a brace pair which now encloses
2766 ;; `here'. As good-pos, return the address. of the "{". Since we've no
2767 ;; knowledge of what's inside these braces, we have no alternative but
2768 ;; to direct the caller to scan the buffer from the opening brace.
2769 (setq pos (caar c-state-cache))
2770 (setcar c-state-cache pos)
2771 (list (1+ pos) pos t)) ; return value. We've just converted a brace pair
2772 ; entry into a { entry, so the caller needs to
2773 ; search for a brace pair before the {.
2774
2775 ;; `here' might be inside a literal. Check for this.
2776 ((progn
2777 (setq lit (c-state-literal-at here)
2778 here-lit-start (or (car lit) here)
2779 here-lit-end (or (cdr lit) here))
2780 ;; Has `here' just "newly entered" a macro?
2781 (save-excursion
2782 (goto-char here-lit-start)
2783 (if (and (c-beginning-of-macro)
2784 (or (null c-state-old-cpp-beg)
2785 (not (= (point) c-state-old-cpp-beg))))
2786 (progn
2787 (setq here- (point))
2788 (c-end-of-macro)
2789 (setq here+ (point)))
2790 (setq here- here-lit-start
2791 here+ here-lit-end)))
2792
2793 ;; `here' might be nested inside any depth of parens (or brackets but
2794 ;; not braces). Scan backwards to find the outermost such opening
2795 ;; paren, if there is one. This will be the scan position to return.
2796 (save-restriction
2797 (narrow-to-region cache-pos (point-max))
2798 (setq pos (c-state-balance-parens-backwards here- here+ pos)))
2799 nil)) ; for the cond
2800
2801 ((< pos here-lit-start)
2802 ;; CASE 2: Address of outermost ( or [ which now encloses `here', but
2803 ;; didn't enclose the (previous) `c-state-cache-good-pos'. If there is
2804 ;; a brace pair preceding this, it will already be in `c-state-cache',
2805 ;; unless there was a brace pair after it, i.e. there'll only be one to
2806 ;; scan for if we've just deleted one.
2807 (list pos (and dropped-cons pos) t)) ; Return value.
2808
2809 ;; `here' isn't enclosed in a (previously unrecorded) bracket/paren.
2810 ;; Further forward scanning isn't needed, but we still need to find a
2811 ;; GOOD-POS. Step out of all enclosing "("s on HERE's line.
2812 ((progn
2813 (save-restriction
2814 (narrow-to-region here-bol (point-max))
2815 (setq pos here-lit-start)
2816 (c-safe (while (setq pa (scan-lists pos -1 1))
2817 (setq pos pa)))) ; might signal
2818 nil)) ; for the cond
2819
2820 ((setq ren (c-safe-scan-lists pos -1 -1 too-far-back))
2821 ;; CASE 3: After a }/)/] before `here''s BOL.
2822 (list (1+ ren) (and dropped-cons pos) nil)) ; Return value
2823
2824 (t
2825 ;; CASE 4; Best of a bad job: BOL before `here-bol', or beginning of
2826 ;; literal containing it.
2827 (setq good-pos (c-state-lit-beg (c-point 'bopl here-bol)))
2828 (list good-pos (and dropped-cons good-pos) nil)))))
2829
2830
2831 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2832 ;; Externally visible routines.
2833
2834 (defun c-state-cache-init ()
2835 (setq c-state-cache nil
2836 c-state-cache-good-pos 1
2837 c-state-nonlit-pos-cache nil
2838 c-state-nonlit-pos-cache-limit 1
2839 c-state-brace-pair-desert nil
2840 c-state-point-min 1
2841 c-state-point-min-lit-type nil
2842 c-state-point-min-lit-start nil
2843 c-state-min-scan-pos 1
2844 c-state-old-cpp-beg nil
2845 c-state-old-cpp-end nil)
2846 (c-state-mark-point-min-literal))
2847
2848 (defun c-invalidate-state-cache-1 (here)
2849 ;; Invalidate all info on `c-state-cache' that applies to the buffer at HERE
2850 ;; or higher and set `c-state-cache-good-pos' accordingly. The cache is
2851 ;; left in a consistent state.
2852 ;;
2853 ;; This is much like `c-whack-state-after', but it never changes a paren
2854 ;; pair element into an open paren element. Doing that would mean that the
2855 ;; new open paren wouldn't have the required preceding paren pair element.
2856 ;;
2857 ;; This function is called from c-after-change.
2858
2859 ;; The cache of non-literals:
2860 (if (< here c-state-nonlit-pos-cache-limit)
2861 (setq c-state-nonlit-pos-cache-limit here))
2862
2863 ;; `c-state-cache':
2864 ;; Case 1: if `here' is in a literal containing point-min, everything
2865 ;; becomes (or is already) nil.
2866 (if (or (null c-state-cache-good-pos)
2867 (< here (c-state-get-min-scan-pos)))
2868 (setq c-state-cache nil
2869 c-state-cache-good-pos nil
2870 c-state-min-scan-pos nil)
2871
2872 ;;; Truncate `c-state-cache' and set `c-state-cache-good-pos' to a value below
2873 ;;; `here'. To maintain its consistency, we may need to insert a new brace
2874 ;;; pair.
2875 (let ((here-bol (c-point 'bol here))
2876 too-high-pa ; recorded {/(/[ next above here, or nil.
2877 dropped-cons ; was the last removed element a brace pair?
2878 pa)
2879 ;; The easy bit - knock over-the-top bits off `c-state-cache'.
2880 (while (and c-state-cache
2881 (>= (setq pa (c-state-cache-top-paren)) here))
2882 (setq dropped-cons (consp (car c-state-cache))
2883 too-high-pa (c-state-cache-top-lparen)
2884 c-state-cache (cdr c-state-cache)))
2885
2886 ;; Do we need to add in an earlier brace pair, having lopped one off?
2887 (if (and dropped-cons
2888 (< too-high-pa (+ here c-state-cache-too-far)))
2889 (c-append-lower-brace-pair-to-state-cache too-high-pa here-bol))
2890 (setq c-state-cache-good-pos (or (c-state-cache-after-top-paren)
2891 (c-state-get-min-scan-pos)))))
2892
2893 ;; The brace-pair desert marker:
2894 (when (car c-state-brace-pair-desert)
2895 (if (< here (car c-state-brace-pair-desert))
2896 (setq c-state-brace-pair-desert nil)
2897 (if (< here (cdr c-state-brace-pair-desert))
2898 (setcdr c-state-brace-pair-desert here)))))
2899
2900 (defun c-parse-state-1 ()
2901 ;; Find and record all noteworthy parens between some good point earlier in
2902 ;; the file and point. That good point is at least the beginning of the
2903 ;; top-level construct we are in, or the beginning of the preceding
2904 ;; top-level construct if we aren't in one.
2905 ;;
2906 ;; The returned value is a list of the noteworthy parens with the last one
2907 ;; first. If an element in the list is an integer, it's the position of an
2908 ;; open paren (of any type) which has not been closed before the point. If
2909 ;; an element is a cons, it gives the position of a closed BRACE paren
2910 ;; pair[*]; the car is the start brace position and the cdr is the position
2911 ;; following the closing brace. Only the last closed brace paren pair
2912 ;; before each open paren and before the point is recorded, and thus the
2913 ;; state never contains two cons elements in succession. When a close brace
2914 ;; has no matching open brace (e.g., the matching brace is outside the
2915 ;; visible region), it is not represented in the returned value.
2916 ;;
2917 ;; [*] N.B. The close "brace" might be a mismatching close bracket or paren.
2918 ;; This defun explicitly treats mismatching parens/braces/brackets as
2919 ;; matching. It is the open brace which makes it a "brace" pair.
2920 ;;
2921 ;; If POINT is within a macro, open parens and brace pairs within
2922 ;; THIS macro MIGHT be recorded. This depends on whether their
2923 ;; syntactic properties have been suppressed by
2924 ;; `c-neutralize-syntax-in-CPP'. This might need fixing (2008-12-11).
2925 ;;
2926 ;; Currently no characters which are given paren syntax with the
2927 ;; syntax-table property are recorded, i.e. angle bracket arglist
2928 ;; parens are never present here. Note that this might change.
2929 ;;
2930 ;; BUG: This function doesn't cope entirely well with unbalanced
2931 ;; parens in macros. (2008-12-11: this has probably been resolved
2932 ;; by the function `c-neutralize-syntax-in-CPP'.) E.g. in the
2933 ;; following case the brace before the macro isn't balanced with the
2934 ;; one after it:
2935 ;;
2936 ;; {
2937 ;; #define X {
2938 ;; }
2939 ;;
2940 ;; Note to maintainers: this function DOES get called with point
2941 ;; within comments and strings, so don't assume it doesn't!
2942 ;;
2943 ;; This function might do hidden buffer changes.
2944 (let* ((here (point))
2945 (here-bopl (c-point 'bopl))
2946 strategy ; 'forward, 'backward etc..
2947 ;; Candidate positions to start scanning from:
2948 cache-pos ; highest position below HERE already existing in
2949 ; cache (or 1).
2950 good-pos
2951 start-point
2952 bopl-state
2953 res
2954 scan-backward-pos scan-forward-p) ; used for 'backward.
2955 ;; If POINT-MIN has changed, adjust the cache
2956 (unless (= (point-min) c-state-point-min)
2957 (c-renarrow-state-cache))
2958
2959 ;; Strategy?
2960 (setq res (c-parse-state-get-strategy here c-state-cache-good-pos)
2961 strategy (car res)
2962 cache-pos (cadr res)
2963 start-point (nth 2 res))
2964
2965 (when (eq strategy 'BOD)
2966 (setq c-state-cache nil
2967 c-state-cache-good-pos start-point))
2968
2969 ;; SCAN!
2970 (save-restriction
2971 (cond
2972 ((memq strategy '(forward BOD))
2973 (narrow-to-region (point-min) here)
2974 (setq res (c-remove-stale-state-cache start-point here-bopl))
2975 (setq cache-pos (car res)
2976 scan-backward-pos (cadr res)
2977 bopl-state (car (cddr res))) ; will be nil if (< here-bopl
2978 ; start-point)
2979 (if scan-backward-pos
2980 (c-append-lower-brace-pair-to-state-cache scan-backward-pos))
2981 (setq good-pos
2982 (c-append-to-state-cache cache-pos))
2983 (setq c-state-cache-good-pos
2984 (if (and bopl-state
2985 (< good-pos (- here c-state-cache-too-far)))
2986 (c-state-cache-non-literal-place here-bopl bopl-state)
2987 good-pos)))
2988
2989 ((eq strategy 'backward)
2990 (setq res (c-remove-stale-state-cache-backwards here cache-pos)
2991 good-pos (car res)
2992 scan-backward-pos (cadr res)
2993 scan-forward-p (car (cddr res)))
2994 (if scan-backward-pos
2995 (c-append-lower-brace-pair-to-state-cache
2996 scan-backward-pos))
2997 (setq c-state-cache-good-pos
2998 (if scan-forward-p
2999 (progn (narrow-to-region (point-min) here)
3000 (c-append-to-state-cache good-pos))
3001
3002 (c-get-cache-scan-pos good-pos))))
3003
3004 (t ; (eq strategy 'IN-LIT)
3005 (setq c-state-cache nil
3006 c-state-cache-good-pos nil)))))
3007
3008 c-state-cache)
3009
3010 (defun c-invalidate-state-cache (here)
3011 ;; This is a wrapper over `c-invalidate-state-cache-1'.
3012 ;;
3013 ;; It suppresses the syntactic effect of the < and > (template) brackets and
3014 ;; of all parens in preprocessor constructs, except for any such construct
3015 ;; containing point. We can then call `c-invalidate-state-cache-1' without
3016 ;; worrying further about macros and template delimiters.
3017 (c-with-<->-as-parens-suppressed
3018 (if (and c-state-old-cpp-beg
3019 (< c-state-old-cpp-beg here))
3020 (c-with-all-but-one-cpps-commented-out
3021 c-state-old-cpp-beg
3022 (min c-state-old-cpp-end here)
3023 (c-invalidate-state-cache-1 here))
3024 (c-with-cpps-commented-out
3025 (c-invalidate-state-cache-1 here)))))
3026
3027 (defun c-parse-state ()
3028 ;; This is a wrapper over `c-parse-state-1'. See that function for a
3029 ;; description of the functionality and return value.
3030 ;;
3031 ;; It suppresses the syntactic effect of the < and > (template) brackets and
3032 ;; of all parens in preprocessor constructs, except for any such construct
3033 ;; containing point. We can then call `c-parse-state-1' without worrying
3034 ;; further about macros and template delimiters.
3035 (let (here-cpp-beg here-cpp-end)
3036 (save-excursion
3037 (when (c-beginning-of-macro)
3038 (setq here-cpp-beg (point))
3039 (unless
3040 (> (setq here-cpp-end (c-syntactic-end-of-macro))
3041 here-cpp-beg)
3042 (setq here-cpp-beg nil here-cpp-end nil))))
3043 ;; FIXME!!! Put in a `condition-case' here to protect the integrity of the
3044 ;; subsystem.
3045 (prog1
3046 (c-with-<->-as-parens-suppressed
3047 (if (and here-cpp-beg (> here-cpp-end here-cpp-beg))
3048 (c-with-all-but-one-cpps-commented-out
3049 here-cpp-beg here-cpp-end
3050 (c-parse-state-1))
3051 (c-with-cpps-commented-out
3052 (c-parse-state-1))))
3053 (setq c-state-old-cpp-beg (and here-cpp-beg (copy-marker here-cpp-beg t))
3054 c-state-old-cpp-end (and here-cpp-end (copy-marker here-cpp-end t)))
3055 )))
3056
3057 ;; Debug tool to catch cache inconsistencies. This is called from
3058 ;; 000tests.el.
3059 (defvar c-debug-parse-state nil)
3060 (unless (fboundp 'c-real-parse-state)
3061 (fset 'c-real-parse-state (symbol-function 'c-parse-state)))
3062 (cc-bytecomp-defun c-real-parse-state)
3063 (defun c-debug-parse-state ()
3064 (let ((here (point)) (res1 (c-real-parse-state)) res2)
3065 (let ((c-state-cache nil)
3066 (c-state-cache-good-pos 1)
3067 (c-state-nonlit-pos-cache nil)
3068 (c-state-nonlit-pos-cache-limit 1)
3069 (c-state-brace-pair-desert nil)
3070 (c-state-point-min 1)
3071 (c-state-point-min-lit-type nil)
3072 (c-state-point-min-lit-start nil)
3073 (c-state-min-scan-pos 1)
3074 (c-state-old-cpp-beg nil)
3075 (c-state-old-cpp-end nil))
3076 (setq res2 (c-real-parse-state)))
3077 (unless (equal res1 res2)
3078 ;; The cache can actually go further back due to the ad-hoc way
3079 ;; the first paren is found, so try to whack off a bit of its
3080 ;; start before complaining.
3081 (save-excursion
3082 (goto-char (or (c-least-enclosing-brace res2) (point)))
3083 (c-beginning-of-defun-1)
3084 (while (not (or (bobp) (eq (char-after) ?{)))
3085 (c-beginning-of-defun-1))
3086 (unless (equal (c-whack-state-before (point) res1) res2)
3087 (message (concat "c-parse-state inconsistency at %s: "
3088 "using cache: %s, from scratch: %s")
3089 here res1 res2))))
3090 res1))
3091
3092 (defun c-toggle-parse-state-debug (&optional arg)
3093 (interactive "P")
3094 (setq c-debug-parse-state (c-calculate-state arg c-debug-parse-state))
3095 (fset 'c-parse-state (symbol-function (if c-debug-parse-state
3096 'c-debug-parse-state
3097 'c-real-parse-state)))
3098 (c-keep-region-active))
3099 (when c-debug-parse-state
3100 (c-toggle-parse-state-debug 1))
3101
3102 \f
3103 (defun c-whack-state-before (bufpos paren-state)
3104 ;; Whack off any state information from PAREN-STATE which lies
3105 ;; before BUFPOS. Not destructive on PAREN-STATE.
3106 (let* ((newstate (list nil))
3107 (ptr newstate)
3108 car)
3109 (while paren-state
3110 (setq car (car paren-state)
3111 paren-state (cdr paren-state))
3112 (if (< (if (consp car) (car car) car) bufpos)
3113 (setq paren-state nil)
3114 (setcdr ptr (list car))
3115 (setq ptr (cdr ptr))))
3116 (cdr newstate)))
3117
3118 (defun c-whack-state-after (bufpos paren-state)
3119 ;; Whack off any state information from PAREN-STATE which lies at or
3120 ;; after BUFPOS. Not destructive on PAREN-STATE.
3121 (catch 'done
3122 (while paren-state
3123 (let ((car (car paren-state)))
3124 (if (consp car)
3125 ;; just check the car, because in a balanced brace
3126 ;; expression, it must be impossible for the corresponding
3127 ;; close brace to be before point, but the open brace to
3128 ;; be after.
3129 (if (<= bufpos (car car))
3130 nil ; whack it off
3131 (if (< bufpos (cdr car))
3132 ;; its possible that the open brace is before
3133 ;; bufpos, but the close brace is after. In that
3134 ;; case, convert this to a non-cons element. The
3135 ;; rest of the state is before bufpos, so we're
3136 ;; done.
3137 (throw 'done (cons (car car) (cdr paren-state)))
3138 ;; we know that both the open and close braces are
3139 ;; before bufpos, so we also know that everything else
3140 ;; on state is before bufpos.
3141 (throw 'done paren-state)))
3142 (if (<= bufpos car)
3143 nil ; whack it off
3144 ;; it's before bufpos, so everything else should too.
3145 (throw 'done paren-state)))
3146 (setq paren-state (cdr paren-state)))
3147 nil)))
3148
3149 (defun c-most-enclosing-brace (paren-state &optional bufpos)
3150 ;; Return the bufpos of the innermost enclosing open paren before
3151 ;; bufpos, or nil if none was found.
3152 (let (enclosingp)
3153 (or bufpos (setq bufpos 134217727))
3154 (while paren-state
3155 (setq enclosingp (car paren-state)
3156 paren-state (cdr paren-state))
3157 (if (or (consp enclosingp)
3158 (>= enclosingp bufpos))
3159 (setq enclosingp nil)
3160 (setq paren-state nil)))
3161 enclosingp))
3162
3163 (defun c-least-enclosing-brace (paren-state)
3164 ;; Return the bufpos of the outermost enclosing open paren, or nil
3165 ;; if none was found.
3166 (let (pos elem)
3167 (while paren-state
3168 (setq elem (car paren-state)
3169 paren-state (cdr paren-state))
3170 (if (integerp elem)
3171 (setq pos elem)))
3172 pos))
3173
3174 (defun c-safe-position (bufpos paren-state)
3175 ;; Return the closest "safe" position recorded on PAREN-STATE that
3176 ;; is higher up than BUFPOS. Return nil if PAREN-STATE doesn't
3177 ;; contain any. Return nil if BUFPOS is nil, which is useful to
3178 ;; find the closest limit before a given limit that might be nil.
3179 ;;
3180 ;; A "safe" position is a position at or after a recorded open
3181 ;; paren, or after a recorded close paren. The returned position is
3182 ;; thus either the first position after a close brace, or the first
3183 ;; position after an enclosing paren, or at the enclosing paren in
3184 ;; case BUFPOS is immediately after it.
3185 (when bufpos
3186 (let (elem)
3187 (catch 'done
3188 (while paren-state
3189 (setq elem (car paren-state))
3190 (if (consp elem)
3191 (cond ((< (cdr elem) bufpos)
3192 (throw 'done (cdr elem)))
3193 ((< (car elem) bufpos)
3194 ;; See below.
3195 (throw 'done (min (1+ (car elem)) bufpos))))
3196 (if (< elem bufpos)
3197 ;; elem is the position at and not after the opening paren, so
3198 ;; we can go forward one more step unless it's equal to
3199 ;; bufpos. This is useful in some cases avoid an extra paren
3200 ;; level between the safe position and bufpos.
3201 (throw 'done (min (1+ elem) bufpos))))
3202 (setq paren-state (cdr paren-state)))))))
3203
3204 (defun c-beginning-of-syntax ()
3205 ;; This is used for `font-lock-beginning-of-syntax-function'. It
3206 ;; goes to the closest previous point that is known to be outside
3207 ;; any string literal or comment. `c-state-cache' is used if it has
3208 ;; a position in the vicinity.
3209 (let* ((paren-state c-state-cache)
3210 elem
3211
3212 (pos (catch 'done
3213 ;; Note: Similar code in `c-safe-position'. The
3214 ;; difference is that we accept a safe position at
3215 ;; the point and don't bother to go forward past open
3216 ;; parens.
3217 (while paren-state
3218 (setq elem (car paren-state))
3219 (if (consp elem)
3220 (cond ((<= (cdr elem) (point))
3221 (throw 'done (cdr elem)))
3222 ((<= (car elem) (point))
3223 (throw 'done (car elem))))
3224 (if (<= elem (point))
3225 (throw 'done elem)))
3226 (setq paren-state (cdr paren-state)))
3227 (point-min))))
3228
3229 (if (> pos (- (point) 4000))
3230 (goto-char pos)
3231 ;; The position is far back. Try `c-beginning-of-defun-1'
3232 ;; (although we can't be entirely sure it will go to a position
3233 ;; outside a comment or string in current emacsen). FIXME:
3234 ;; Consult `syntax-ppss' here.
3235 (c-beginning-of-defun-1)
3236 (if (< (point) pos)
3237 (goto-char pos)))))
3238
3239 \f
3240 ;; Tools for scanning identifiers and other tokens.
3241
3242 (defun c-on-identifier ()
3243 "Return non-nil if the point is on or directly after an identifier.
3244 Keywords are recognized and not considered identifiers. If an
3245 identifier is detected, the returned value is its starting position.
3246 If an identifier ends at the point and another begins at it \(can only
3247 happen in Pike) then the point for the preceding one is returned.
3248
3249 Note that this function might do hidden buffer changes. See the
3250 comment at the start of cc-engine.el for more info."
3251
3252 ;; FIXME: Shouldn't this function handle "operator" in C++?
3253
3254 (save-excursion
3255 (skip-syntax-backward "w_")
3256
3257 (or
3258
3259 ;; Check for a normal (non-keyword) identifier.
3260 (and (looking-at c-symbol-start)
3261 (not (looking-at c-keywords-regexp))
3262 (point))
3263
3264 (when (c-major-mode-is 'pike-mode)
3265 ;; Handle the `<operator> syntax in Pike.
3266 (let ((pos (point)))
3267 (skip-chars-backward "-!%&*+/<=>^|~[]()")
3268 (and (if (< (skip-chars-backward "`") 0)
3269 t
3270 (goto-char pos)
3271 (eq (char-after) ?\`))
3272 (looking-at c-symbol-key)
3273 (>= (match-end 0) pos)
3274 (point))))
3275
3276 ;; Handle the "operator +" syntax in C++.
3277 (when (and c-overloadable-operators-regexp
3278 (= (c-backward-token-2 0) 0))
3279
3280 (cond ((and (looking-at c-overloadable-operators-regexp)
3281 (or (not c-opt-op-identifier-prefix)
3282 (and (= (c-backward-token-2 1) 0)
3283 (looking-at c-opt-op-identifier-prefix))))
3284 (point))
3285
3286 ((save-excursion
3287 (and c-opt-op-identifier-prefix
3288 (looking-at c-opt-op-identifier-prefix)
3289 (= (c-forward-token-2 1) 0)
3290 (looking-at c-overloadable-operators-regexp)))
3291 (point))))
3292
3293 )))
3294
3295 (defsubst c-simple-skip-symbol-backward ()
3296 ;; If the point is at the end of a symbol then skip backward to the
3297 ;; beginning of it. Don't move otherwise. Return non-nil if point
3298 ;; moved.
3299 ;;
3300 ;; This function might do hidden buffer changes.
3301 (or (< (skip-syntax-backward "w_") 0)
3302 (and (c-major-mode-is 'pike-mode)
3303 ;; Handle the `<operator> syntax in Pike.
3304 (let ((pos (point)))
3305 (if (and (< (skip-chars-backward "-!%&*+/<=>^|~[]()") 0)
3306 (< (skip-chars-backward "`") 0)
3307 (looking-at c-symbol-key)
3308 (>= (match-end 0) pos))
3309 t
3310 (goto-char pos)
3311 nil)))))
3312
3313 (defun c-beginning-of-current-token (&optional back-limit)
3314 ;; Move to the beginning of the current token. Do not move if not
3315 ;; in the middle of one. BACK-LIMIT may be used to bound the
3316 ;; backward search; if given it's assumed to be at the boundary
3317 ;; between two tokens. Return non-nil if the point is moved, nil
3318 ;; otherwise.
3319 ;;
3320 ;; This function might do hidden buffer changes.
3321 (let ((start (point)))
3322 (if (looking-at "\\w\\|\\s_")
3323 (skip-syntax-backward "w_" back-limit)
3324 (when (< (skip-syntax-backward ".()" back-limit) 0)
3325 (while (let ((pos (or (and (looking-at c-nonsymbol-token-regexp)
3326 (match-end 0))
3327 ;; `c-nonsymbol-token-regexp' should always match
3328 ;; since we've skipped backward over punctuator
3329 ;; or paren syntax, but consume one char in case
3330 ;; it doesn't so that we don't leave point before
3331 ;; some earlier incorrect token.
3332 (1+ (point)))))
3333 (if (<= pos start)
3334 (goto-char pos))))))
3335 (< (point) start)))
3336
3337 (defun c-end-of-current-token (&optional back-limit)
3338 ;; Move to the end of the current token. Do not move if not in the
3339 ;; middle of one. BACK-LIMIT may be used to bound the backward
3340 ;; search; if given it's assumed to be at the boundary between two
3341 ;; tokens. Return non-nil if the point is moved, nil otherwise.
3342 ;;
3343 ;; This function might do hidden buffer changes.
3344 (let ((start (point)))
3345 (cond ((< (skip-syntax-backward "w_" (1- start)) 0)
3346 (skip-syntax-forward "w_"))
3347 ((< (skip-syntax-backward ".()" back-limit) 0)
3348 (while (progn
3349 (if (looking-at c-nonsymbol-token-regexp)
3350 (goto-char (match-end 0))
3351 ;; `c-nonsymbol-token-regexp' should always match since
3352 ;; we've skipped backward over punctuator or paren
3353 ;; syntax, but move forward in case it doesn't so that
3354 ;; we don't leave point earlier than we started with.
3355 (forward-char))
3356 (< (point) start)))))
3357 (> (point) start)))
3358
3359 (defconst c-jump-syntax-balanced
3360 (if (memq 'gen-string-delim c-emacs-features)
3361 "\\w\\|\\s_\\|\\s\(\\|\\s\)\\|\\s\"\\|\\s|"
3362 "\\w\\|\\s_\\|\\s\(\\|\\s\)\\|\\s\""))
3363
3364 (defconst c-jump-syntax-unbalanced
3365 (if (memq 'gen-string-delim c-emacs-features)
3366 "\\w\\|\\s_\\|\\s\"\\|\\s|"
3367 "\\w\\|\\s_\\|\\s\""))
3368
3369 (defun c-forward-token-2 (&optional count balanced limit)
3370 "Move forward by tokens.
3371 A token is defined as all symbols and identifiers which aren't
3372 syntactic whitespace \(note that multicharacter tokens like \"==\" are
3373 treated properly). Point is always either left at the beginning of a
3374 token or not moved at all. COUNT specifies the number of tokens to
3375 move; a negative COUNT moves in the opposite direction. A COUNT of 0
3376 moves to the next token beginning only if not already at one. If
3377 BALANCED is true, move over balanced parens, otherwise move into them.
3378 Also, if BALANCED is true, never move out of an enclosing paren.
3379
3380 LIMIT sets the limit for the movement and defaults to the point limit.
3381 The case when LIMIT is set in the middle of a token, comment or macro
3382 is handled correctly, i.e. the point won't be left there.
3383
3384 Return the number of tokens left to move \(positive or negative). If
3385 BALANCED is true, a move over a balanced paren counts as one. Note
3386 that if COUNT is 0 and no appropriate token beginning is found, 1 will
3387 be returned. Thus, a return value of 0 guarantees that point is at
3388 the requested position and a return value less \(without signs) than
3389 COUNT guarantees that point is at the beginning of some token.
3390
3391 Note that this function might do hidden buffer changes. See the
3392 comment at the start of cc-engine.el for more info."
3393
3394 (or count (setq count 1))
3395 (if (< count 0)
3396 (- (c-backward-token-2 (- count) balanced limit))
3397
3398 (let ((jump-syntax (if balanced
3399 c-jump-syntax-balanced
3400 c-jump-syntax-unbalanced))
3401 (last (point))
3402 (prev (point)))
3403
3404 (if (zerop count)
3405 ;; If count is zero we should jump if in the middle of a token.
3406 (c-end-of-current-token))
3407
3408 (save-restriction
3409 (if limit (narrow-to-region (point-min) limit))
3410 (if (/= (point)
3411 (progn (c-forward-syntactic-ws) (point)))
3412 ;; Skip whitespace. Count this as a move if we did in
3413 ;; fact move.
3414 (setq count (max (1- count) 0)))
3415
3416 (if (eobp)
3417 ;; Moved out of bounds. Make sure the returned count isn't zero.
3418 (progn
3419 (if (zerop count) (setq count 1))
3420 (goto-char last))
3421
3422 ;; Use `condition-case' to avoid having the limit tests
3423 ;; inside the loop.
3424 (condition-case nil
3425 (while (and
3426 (> count 0)
3427 (progn
3428 (setq last (point))
3429 (cond ((looking-at jump-syntax)
3430 (goto-char (scan-sexps (point) 1))
3431 t)
3432 ((looking-at c-nonsymbol-token-regexp)
3433 (goto-char (match-end 0))
3434 t)
3435 ;; `c-nonsymbol-token-regexp' above should always
3436 ;; match if there are correct tokens. Try to
3437 ;; widen to see if the limit was set in the
3438 ;; middle of one, else fall back to treating
3439 ;; the offending thing as a one character token.
3440 ((and limit
3441 (save-restriction
3442 (widen)
3443 (looking-at c-nonsymbol-token-regexp)))
3444 nil)
3445 (t
3446 (forward-char)
3447 t))))
3448 (c-forward-syntactic-ws)
3449 (setq prev last
3450 count (1- count)))
3451 (error (goto-char last)))
3452
3453 (when (eobp)
3454 (goto-char prev)
3455 (setq count (1+ count)))))
3456
3457 count)))
3458
3459 (defun c-backward-token-2 (&optional count balanced limit)
3460 "Move backward by tokens.
3461 See `c-forward-token-2' for details."
3462
3463 (or count (setq count 1))
3464 (if (< count 0)
3465 (- (c-forward-token-2 (- count) balanced limit))
3466
3467 (or limit (setq limit (point-min)))
3468 (let ((jump-syntax (if balanced
3469 c-jump-syntax-balanced
3470 c-jump-syntax-unbalanced))
3471 (last (point)))
3472
3473 (if (zerop count)
3474 ;; The count is zero so try to skip to the beginning of the
3475 ;; current token.
3476 (if (> (point)
3477 (progn (c-beginning-of-current-token) (point)))
3478 (if (< (point) limit)
3479 ;; The limit is inside the same token, so return 1.
3480 (setq count 1))
3481
3482 ;; We're not in the middle of a token. If there's
3483 ;; whitespace after the point then we must move backward,
3484 ;; so set count to 1 in that case.
3485 (and (looking-at c-syntactic-ws-start)
3486 ;; If we're looking at a '#' that might start a cpp
3487 ;; directive then we have to do a more elaborate check.
3488 (or (/= (char-after) ?#)
3489 (not c-opt-cpp-prefix)
3490 (save-excursion
3491 (and (= (point)
3492 (progn (beginning-of-line)
3493 (looking-at "[ \t]*")
3494 (match-end 0)))
3495 (or (bobp)
3496 (progn (backward-char)
3497 (not (eq (char-before) ?\\)))))))
3498 (setq count 1))))
3499
3500 ;; Use `condition-case' to avoid having to check for buffer
3501 ;; limits in `backward-char', `scan-sexps' and `goto-char' below.
3502 (condition-case nil
3503 (while (and
3504 (> count 0)
3505 (progn
3506 (c-backward-syntactic-ws)
3507 (backward-char)
3508 (if (looking-at jump-syntax)
3509 (goto-char (scan-sexps (1+ (point)) -1))
3510 ;; This can be very inefficient if there's a long
3511 ;; sequence of operator tokens without any separation.
3512 ;; That doesn't happen in practice, anyway.
3513 (c-beginning-of-current-token))
3514 (>= (point) limit)))
3515 (setq last (point)
3516 count (1- count)))
3517 (error (goto-char last)))
3518
3519 (if (< (point) limit)
3520 (goto-char last))
3521
3522 count)))
3523
3524 (defun c-forward-token-1 (&optional count balanced limit)
3525 "Like `c-forward-token-2' but doesn't treat multicharacter operator
3526 tokens like \"==\" as single tokens, i.e. all sequences of symbol
3527 characters are jumped over character by character. This function is
3528 for compatibility only; it's only a wrapper over `c-forward-token-2'."
3529 (let ((c-nonsymbol-token-regexp "\\s.\\|\\s\(\\|\\s\)"))
3530 (c-forward-token-2 count balanced limit)))
3531
3532 (defun c-backward-token-1 (&optional count balanced limit)
3533 "Like `c-backward-token-2' but doesn't treat multicharacter operator
3534 tokens like \"==\" as single tokens, i.e. all sequences of symbol
3535 characters are jumped over character by character. This function is
3536 for compatibility only; it's only a wrapper over `c-backward-token-2'."
3537 (let ((c-nonsymbol-token-regexp "\\s.\\|\\s\(\\|\\s\)"))
3538 (c-backward-token-2 count balanced limit)))
3539
3540 \f
3541 ;; Tools for doing searches restricted to syntactically relevant text.
3542
3543 (defun c-syntactic-re-search-forward (regexp &optional bound noerror
3544 paren-level not-inside-token
3545 lookbehind-submatch)
3546 "Like `re-search-forward', but only report matches that are found
3547 in syntactically significant text. I.e. matches in comments, macros
3548 or string literals are ignored. The start point is assumed to be
3549 outside any comment, macro or string literal, or else the content of
3550 that region is taken as syntactically significant text.
3551
3552 If PAREN-LEVEL is non-nil, an additional restriction is added to
3553 ignore matches in nested paren sexps. The search will also not go
3554 outside the current list sexp, which has the effect that if the point
3555 should be moved to BOUND when no match is found \(i.e. NOERROR is
3556 neither nil nor t), then it will be at the closing paren if the end of
3557 the current list sexp is encountered first.
3558
3559 If NOT-INSIDE-TOKEN is non-nil, matches in the middle of tokens are
3560 ignored. Things like multicharacter operators and special symbols
3561 \(e.g. \"`()\" in Pike) are handled but currently not floating point
3562 constants.
3563
3564 If LOOKBEHIND-SUBMATCH is non-nil, it's taken as a number of a
3565 subexpression in REGEXP. The end of that submatch is used as the
3566 position to check for syntactic significance. If LOOKBEHIND-SUBMATCH
3567 isn't used or if that subexpression didn't match then the start
3568 position of the whole match is used instead. The \"look behind\"
3569 subexpression is never tested before the starting position, so it
3570 might be a good idea to include \\=\\= as a match alternative in it.
3571
3572 Optimization note: Matches might be missed if the \"look behind\"
3573 subexpression can match the end of nonwhite syntactic whitespace,
3574 i.e. the end of comments or cpp directives. This since the function
3575 skips over such things before resuming the search. It's on the other
3576 hand not safe to assume that the \"look behind\" subexpression never
3577 matches syntactic whitespace.
3578
3579 Bug: Unbalanced parens inside cpp directives are currently not handled
3580 correctly \(i.e. they don't get ignored as they should) when
3581 PAREN-LEVEL is set.
3582
3583 Note that this function might do hidden buffer changes. See the
3584 comment at the start of cc-engine.el for more info."
3585
3586 (or bound (setq bound (point-max)))
3587 (if paren-level (setq paren-level -1))
3588
3589 ;;(message "c-syntactic-re-search-forward %s %s %S" (point) bound regexp)
3590
3591 (let ((start (point))
3592 tmp
3593 ;; Start position for the last search.
3594 search-pos
3595 ;; The `parse-partial-sexp' state between the start position
3596 ;; and the point.
3597 state
3598 ;; The current position after the last state update. The next
3599 ;; `parse-partial-sexp' continues from here.
3600 (state-pos (point))
3601 ;; The position at which to check the state and the state
3602 ;; there. This is separate from `state-pos' since we might
3603 ;; need to back up before doing the next search round.
3604 check-pos check-state
3605 ;; Last position known to end a token.
3606 (last-token-end-pos (point-min))
3607 ;; Set when a valid match is found.
3608 found)
3609
3610 (condition-case err
3611 (while
3612 (and
3613 (progn
3614 (setq search-pos (point))
3615 (re-search-forward regexp bound noerror))
3616
3617 (progn
3618 (setq state (parse-partial-sexp
3619 state-pos (match-beginning 0) paren-level nil state)
3620 state-pos (point))
3621 (if (setq check-pos (and lookbehind-submatch
3622 (or (not paren-level)
3623 (>= (car state) 0))
3624 (match-end lookbehind-submatch)))
3625 (setq check-state (parse-partial-sexp
3626 state-pos check-pos paren-level nil state))
3627 (setq check-pos state-pos
3628 check-state state))
3629
3630 ;; NOTE: If we got a look behind subexpression and get
3631 ;; an insignificant match in something that isn't
3632 ;; syntactic whitespace (i.e. strings or in nested
3633 ;; parentheses), then we can never skip more than a
3634 ;; single character from the match start position
3635 ;; (i.e. `state-pos' here) before continuing the
3636 ;; search. That since the look behind subexpression
3637 ;; might match the end of the insignificant region in
3638 ;; the next search.
3639
3640 (cond
3641 ((elt check-state 7)
3642 ;; Match inside a line comment. Skip to eol. Use
3643 ;; `re-search-forward' instead of `skip-chars-forward' to get
3644 ;; the right bound behavior.
3645 (re-search-forward "[\n\r]" bound noerror))
3646
3647 ((elt check-state 4)
3648 ;; Match inside a block comment. Skip to the '*/'.
3649 (search-forward "*/" bound noerror))
3650
3651 ((and (not (elt check-state 5))
3652 (eq (char-before check-pos) ?/)
3653 (not (c-get-char-property (1- check-pos) 'syntax-table))
3654 (memq (char-after check-pos) '(?/ ?*)))
3655 ;; Match in the middle of the opener of a block or line
3656 ;; comment.
3657 (if (= (char-after check-pos) ?/)
3658 (re-search-forward "[\n\r]" bound noerror)
3659 (search-forward "*/" bound noerror)))
3660
3661 ;; The last `parse-partial-sexp' above might have
3662 ;; stopped short of the real check position if the end
3663 ;; of the current sexp was encountered in paren-level
3664 ;; mode. The checks above are always false in that
3665 ;; case, and since they can do better skipping in
3666 ;; lookbehind-submatch mode, we do them before
3667 ;; checking the paren level.
3668
3669 ((and paren-level
3670 (/= (setq tmp (car check-state)) 0))
3671 ;; Check the paren level first since we're short of the
3672 ;; syntactic checking position if the end of the
3673 ;; current sexp was encountered by `parse-partial-sexp'.
3674 (if (> tmp 0)
3675
3676 ;; Inside a nested paren sexp.
3677 (if lookbehind-submatch
3678 ;; See the NOTE above.
3679 (progn (goto-char state-pos) t)
3680 ;; Skip out of the paren quickly.
3681 (setq state (parse-partial-sexp state-pos bound 0 nil state)
3682 state-pos (point)))
3683
3684 ;; Have exited the current paren sexp.
3685 (if noerror
3686 (progn
3687 ;; The last `parse-partial-sexp' call above
3688 ;; has left us just after the closing paren
3689 ;; in this case, so we can modify the bound
3690 ;; to leave the point at the right position
3691 ;; upon return.
3692 (setq bound (1- (point)))
3693 nil)
3694 (signal 'search-failed (list regexp)))))
3695
3696 ((setq tmp (elt check-state 3))
3697 ;; Match inside a string.
3698 (if (or lookbehind-submatch
3699 (not (integerp tmp)))
3700 ;; See the NOTE above.
3701 (progn (goto-char state-pos) t)
3702 ;; Skip to the end of the string before continuing.
3703 (let ((ender (make-string 1 tmp)) (continue t))
3704 (while (if (search-forward ender bound noerror)
3705 (progn
3706 (setq state (parse-partial-sexp
3707 state-pos (point) nil nil state)
3708 state-pos (point))
3709 (elt state 3))
3710 (setq continue nil)))
3711 continue)))
3712
3713 ((save-excursion
3714 (save-match-data
3715 (c-beginning-of-macro start)))
3716 ;; Match inside a macro. Skip to the end of it.
3717 (c-end-of-macro)
3718 (cond ((<= (point) bound) t)
3719 (noerror nil)
3720 (t (signal 'search-failed (list regexp)))))
3721
3722 ((and not-inside-token
3723 (or (< check-pos last-token-end-pos)
3724 (< check-pos
3725 (save-excursion
3726 (goto-char check-pos)
3727 (save-match-data
3728 (c-end-of-current-token last-token-end-pos))
3729 (setq last-token-end-pos (point))))))
3730 ;; Inside a token.
3731 (if lookbehind-submatch
3732 ;; See the NOTE above.
3733 (goto-char state-pos)
3734 (goto-char (min last-token-end-pos bound))))
3735
3736 (t
3737 ;; A real match.
3738 (setq found t)
3739 nil)))
3740
3741 ;; Should loop to search again, but take care to avoid
3742 ;; looping on the same spot.
3743 (or (/= search-pos (point))
3744 (if (= (point) bound)
3745 (if noerror
3746 nil
3747 (signal 'search-failed (list regexp)))
3748 (forward-char)
3749 t))))
3750
3751 (error
3752 (goto-char start)
3753 (signal (car err) (cdr err))))
3754
3755 ;;(message "c-syntactic-re-search-forward done %s" (or (match-end 0) (point)))
3756
3757 (if found
3758 (progn
3759 (goto-char (match-end 0))
3760 (match-end 0))
3761
3762 ;; Search failed. Set point as appropriate.
3763 (if (eq noerror t)
3764 (goto-char start)
3765 (goto-char bound))
3766 nil)))
3767
3768 (defvar safe-pos-list) ; bound in c-syntactic-skip-backward
3769
3770 (defsubst c-ssb-lit-begin ()
3771 ;; Return the start of the literal point is in, or nil.
3772 ;; We read and write the variables `safe-pos', `safe-pos-list', `state'
3773 ;; bound in the caller.
3774
3775 ;; Use `parse-partial-sexp' from a safe position down to the point to check
3776 ;; if it's outside comments and strings.
3777 (save-excursion
3778 (let ((pos (point)) safe-pos state pps-end-pos)
3779 ;; Pick a safe position as close to the point as possible.
3780 ;;
3781 ;; FIXME: Consult `syntax-ppss' here if our cache doesn't give a good
3782 ;; position.
3783
3784 (while (and safe-pos-list
3785 (> (car safe-pos-list) (point)))
3786 (setq safe-pos-list (cdr safe-pos-list)))
3787 (unless (setq safe-pos (car-safe safe-pos-list))
3788 (setq safe-pos (max (or (c-safe-position
3789 (point) (or c-state-cache
3790 (c-parse-state)))
3791 0)
3792 (point-min))
3793 safe-pos-list (list safe-pos)))
3794
3795 ;; Cache positions along the way to use if we have to back up more. We
3796 ;; cache every closing paren on the same level. If the paren cache is
3797 ;; relevant in this region then we're typically already on the same
3798 ;; level as the target position. Note that we might cache positions
3799 ;; after opening parens in case safe-pos is in a nested list. That's
3800 ;; both uncommon and harmless.
3801 (while (progn
3802 (setq state (parse-partial-sexp
3803 safe-pos pos 0))
3804 (< (point) pos))
3805 (setq safe-pos (point)
3806 safe-pos-list (cons safe-pos safe-pos-list)))
3807
3808 ;; If the state contains the start of the containing sexp we cache that
3809 ;; position too, so that parse-partial-sexp in the next run has a bigger
3810 ;; chance of starting at the same level as the target position and thus
3811 ;; will get more good safe positions into the list.
3812 (if (elt state 1)
3813 (setq safe-pos (1+ (elt state 1))
3814 safe-pos-list (cons safe-pos safe-pos-list)))
3815
3816 (if (or (elt state 3) (elt state 4))
3817 ;; Inside string or comment. Continue search at the
3818 ;; beginning of it.
3819 (elt state 8)))))
3820
3821 (defun c-syntactic-skip-backward (skip-chars &optional limit paren-level)
3822 "Like `skip-chars-backward' but only look at syntactically relevant chars,
3823 i.e. don't stop at positions inside syntactic whitespace or string
3824 literals. Preprocessor directives are also ignored, with the exception
3825 of the one that the point starts within, if any. If LIMIT is given,
3826 it's assumed to be at a syntactically relevant position.
3827
3828 If PAREN-LEVEL is non-nil, the function won't stop in nested paren
3829 sexps, and the search will also not go outside the current paren sexp.
3830 However, if LIMIT or the buffer limit is reached inside a nested paren
3831 then the point will be left at the limit.
3832
3833 Non-nil is returned if the point moved, nil otherwise.
3834
3835 Note that this function might do hidden buffer changes. See the
3836 comment at the start of cc-engine.el for more info."
3837
3838 (let ((start (point))
3839 state-2
3840 ;; A list of syntactically relevant positions in descending
3841 ;; order. It's used to avoid scanning repeatedly over
3842 ;; potentially large regions with `parse-partial-sexp' to verify
3843 ;; each position. Used in `c-ssb-lit-begin'
3844 safe-pos-list
3845 ;; The result from `c-beginning-of-macro' at the start position or the
3846 ;; start position itself if it isn't within a macro. Evaluated on
3847 ;; demand.
3848 start-macro-beg
3849 ;; The earliest position after the current one with the same paren
3850 ;; level. Used only when `paren-level' is set.
3851 lit-beg
3852 (paren-level-pos (point)))
3853
3854 (while
3855 (progn
3856 ;; The next loop "tries" to find the end point each time round,
3857 ;; loops when it hasn't succeeded.
3858 (while
3859 (and
3860 (< (skip-chars-backward skip-chars limit) 0)
3861
3862 (let ((pos (point)) state-2 pps-end-pos)
3863
3864 (cond
3865 ;; Don't stop inside a literal
3866 ((setq lit-beg (c-ssb-lit-begin))
3867 (goto-char lit-beg)
3868 t)
3869
3870 ((and paren-level
3871 (save-excursion
3872 (setq state-2 (parse-partial-sexp
3873 pos paren-level-pos -1)
3874 pps-end-pos (point))
3875 (/= (car state-2) 0)))
3876 ;; Not at the right level.
3877
3878 (if (and (< (car state-2) 0)
3879 ;; We stop above if we go out of a paren.
3880 ;; Now check whether it precedes or is
3881 ;; nested in the starting sexp.
3882 (save-excursion
3883 (setq state-2
3884 (parse-partial-sexp
3885 pps-end-pos paren-level-pos
3886 nil nil state-2))
3887 (< (car state-2) 0)))
3888
3889 ;; We've stopped short of the starting position
3890 ;; so the hit was inside a nested list. Go up
3891 ;; until we are at the right level.
3892 (condition-case nil
3893 (progn
3894 (goto-char (scan-lists pos -1
3895 (- (car state-2))))
3896 (setq paren-level-pos (point))
3897 (if (and limit (>= limit paren-level-pos))
3898 (progn
3899 (goto-char limit)
3900 nil)
3901 t))
3902 (error
3903 (goto-char (or limit (point-min)))
3904 nil))
3905
3906 ;; The hit was outside the list at the start
3907 ;; position. Go to the start of the list and exit.
3908 (goto-char (1+ (elt state-2 1)))
3909 nil))
3910
3911 ((c-beginning-of-macro limit)
3912 ;; Inside a macro.
3913 (if (< (point)
3914 (or start-macro-beg
3915 (setq start-macro-beg
3916 (save-excursion
3917 (goto-char start)
3918 (c-beginning-of-macro limit)
3919 (point)))))
3920 t
3921
3922 ;; It's inside the same macro we started in so it's
3923 ;; a relevant match.
3924 (goto-char pos)
3925 nil))))))
3926
3927 (> (point)
3928 (progn
3929 ;; Skip syntactic ws afterwards so that we don't stop at the
3930 ;; end of a comment if `skip-chars' is something like "^/".
3931 (c-backward-syntactic-ws)
3932 (point)))))
3933
3934 ;; We might want to extend this with more useful return values in
3935 ;; the future.
3936 (/= (point) start)))
3937
3938 ;; The following is an alternative implementation of
3939 ;; `c-syntactic-skip-backward' that uses backward movement to keep
3940 ;; track of the syntactic context. It turned out to be generally
3941 ;; slower than the one above which uses forward checks from earlier
3942 ;; safe positions.
3943 ;;
3944 ;;(defconst c-ssb-stop-re
3945 ;; ;; The regexp matching chars `c-syntactic-skip-backward' needs to
3946 ;; ;; stop at to avoid going into comments and literals.
3947 ;; (concat
3948 ;; ;; Match comment end syntax and string literal syntax. Also match
3949 ;; ;; '/' for block comment endings (not covered by comment end
3950 ;; ;; syntax).
3951 ;; "\\s>\\|/\\|\\s\""
3952 ;; (if (memq 'gen-string-delim c-emacs-features)
3953 ;; "\\|\\s|"
3954 ;; "")
3955 ;; (if (memq 'gen-comment-delim c-emacs-features)
3956 ;; "\\|\\s!"
3957 ;; "")))
3958 ;;
3959 ;;(defconst c-ssb-stop-paren-re
3960 ;; ;; Like `c-ssb-stop-re' but also stops at paren chars.
3961 ;; (concat c-ssb-stop-re "\\|\\s(\\|\\s)"))
3962 ;;
3963 ;;(defconst c-ssb-sexp-end-re
3964 ;; ;; Regexp matching the ending syntax of a complex sexp.
3965 ;; (concat c-string-limit-regexp "\\|\\s)"))
3966 ;;
3967 ;;(defun c-syntactic-skip-backward (skip-chars &optional limit paren-level)
3968 ;; "Like `skip-chars-backward' but only look at syntactically relevant chars,
3969 ;;i.e. don't stop at positions inside syntactic whitespace or string
3970 ;;literals. Preprocessor directives are also ignored. However, if the
3971 ;;point is within a comment, string literal or preprocessor directory to
3972 ;;begin with, its contents is treated as syntactically relevant chars.
3973 ;;If LIMIT is given, it limits the backward search and the point will be
3974 ;;left there if no earlier position is found.
3975 ;;
3976 ;;If PAREN-LEVEL is non-nil, the function won't stop in nested paren
3977 ;;sexps, and the search will also not go outside the current paren sexp.
3978 ;;However, if LIMIT or the buffer limit is reached inside a nested paren
3979 ;;then the point will be left at the limit.
3980 ;;
3981 ;;Non-nil is returned if the point moved, nil otherwise.
3982 ;;
3983 ;;Note that this function might do hidden buffer changes. See the
3984 ;;comment at the start of cc-engine.el for more info."
3985 ;;
3986 ;; (save-restriction
3987 ;; (when limit
3988 ;; (narrow-to-region limit (point-max)))
3989 ;;
3990 ;; (let ((start (point)))
3991 ;; (catch 'done
3992 ;; (while (let ((last-pos (point))
3993 ;; (stop-pos (progn
3994 ;; (skip-chars-backward skip-chars)
3995 ;; (point))))
3996 ;;
3997 ;; ;; Skip back over the same region as
3998 ;; ;; `skip-chars-backward' above, but keep to
3999 ;; ;; syntactically relevant positions.
4000 ;; (goto-char last-pos)
4001 ;; (while (and
4002 ;; ;; `re-search-backward' with a single char regexp
4003 ;; ;; should be fast.
4004 ;; (re-search-backward
4005 ;; (if paren-level c-ssb-stop-paren-re c-ssb-stop-re)
4006 ;; stop-pos 'move)
4007 ;;
4008 ;; (progn
4009 ;; (cond
4010 ;; ((looking-at "\\s(")
4011 ;; ;; `paren-level' is set and we've found the
4012 ;; ;; start of the containing paren.
4013 ;; (forward-char)
4014 ;; (throw 'done t))
4015 ;;
4016 ;; ((looking-at c-ssb-sexp-end-re)
4017 ;; ;; We're at the end of a string literal or paren
4018 ;; ;; sexp (if `paren-level' is set).
4019 ;; (forward-char)
4020 ;; (condition-case nil
4021 ;; (c-backward-sexp)
4022 ;; (error
4023 ;; (goto-char limit)
4024 ;; (throw 'done t))))
4025 ;;
4026 ;; (t
4027 ;; (forward-char)
4028 ;; ;; At the end of some syntactic ws or possibly
4029 ;; ;; after a plain '/' operator.
4030 ;; (let ((pos (point)))
4031 ;; (c-backward-syntactic-ws)
4032 ;; (if (= pos (point))
4033 ;; ;; Was a plain '/' operator. Go past it.
4034 ;; (backward-char)))))
4035 ;;
4036 ;; (> (point) stop-pos))))
4037 ;;
4038 ;; ;; Now the point is either at `stop-pos' or at some
4039 ;; ;; position further back if `stop-pos' was at a
4040 ;; ;; syntactically irrelevant place.
4041 ;;
4042 ;; ;; Skip additional syntactic ws so that we don't stop
4043 ;; ;; at the end of a comment if `skip-chars' is
4044 ;; ;; something like "^/".
4045 ;; (c-backward-syntactic-ws)
4046 ;;
4047 ;; (< (point) stop-pos))))
4048 ;;
4049 ;; ;; We might want to extend this with more useful return values
4050 ;; ;; in the future.
4051 ;; (/= (point) start))))
4052
4053 \f
4054 ;; Tools for handling comments and string literals.
4055
4056 (defun c-slow-in-literal (&optional lim detect-cpp)
4057 "Return the type of literal point is in, if any.
4058 The return value is `c' if in a C-style comment, `c++' if in a C++
4059 style comment, `string' if in a string literal, `pound' if DETECT-CPP
4060 is non-nil and in a preprocessor line, or nil if somewhere else.
4061 Optional LIM is used as the backward limit of the search. If omitted,
4062 or nil, `c-beginning-of-defun' is used.
4063
4064 The last point calculated is cached if the cache is enabled, i.e. if
4065 `c-in-literal-cache' is bound to a two element vector.
4066
4067 Note that this function might do hidden buffer changes. See the
4068 comment at the start of cc-engine.el for more info."
4069
4070 (if (and (vectorp c-in-literal-cache)
4071 (= (point) (aref c-in-literal-cache 0)))
4072 (aref c-in-literal-cache 1)
4073 (let ((rtn (save-excursion
4074 (let* ((pos (point))
4075 (lim (or lim (progn
4076 (c-beginning-of-syntax)
4077 (point))))
4078 (state (parse-partial-sexp lim pos)))
4079 (cond
4080 ((elt state 3) 'string)
4081 ((elt state 4) (if (elt state 7) 'c++ 'c))
4082 ((and detect-cpp (c-beginning-of-macro lim)) 'pound)
4083 (t nil))))))
4084 ;; cache this result if the cache is enabled
4085 (if (not c-in-literal-cache)
4086 (setq c-in-literal-cache (vector (point) rtn)))
4087 rtn)))
4088
4089 ;; XEmacs has a built-in function that should make this much quicker.
4090 ;; I don't think we even need the cache, which makes our lives more
4091 ;; complicated anyway. In this case, lim is only used to detect
4092 ;; cpp directives.
4093 ;;
4094 ;; Note that there is a bug in Xemacs's buffer-syntactic-context when used in
4095 ;; conjunction with syntax-table-properties. The bug is present in, e.g.,
4096 ;; Xemacs 21.4.4. It manifested itself thus:
4097 ;;
4098 ;; Starting with an empty AWK Mode buffer, type
4099 ;; /regexp/ {<C-j>
4100 ;; Point gets wrongly left at column 0, rather than being indented to tab-width.
4101 ;;
4102 ;; AWK Mode is designed such that when the first / is typed, it gets the
4103 ;; syntax-table property "string fence". When the second / is typed, BOTH /s
4104 ;; are given the s-t property "string". However, buffer-syntactic-context
4105 ;; fails to take account of the change of the s-t property on the opening / to
4106 ;; "string", and reports that the { is within a string started by the second /.
4107 ;;
4108 ;; The workaround for this is for the AWK Mode initialisation to switch the
4109 ;; defalias for c-in-literal to c-slow-in-literal. This will slow down other
4110 ;; cc-modes in Xemacs whenever an awk-buffer has been initialised.
4111 ;;
4112 ;; (Alan Mackenzie, 2003/4/30).
4113
4114 (defun c-fast-in-literal (&optional lim detect-cpp)
4115 ;; This function might do hidden buffer changes.
4116 (let ((context (buffer-syntactic-context)))
4117 (cond
4118 ((eq context 'string) 'string)
4119 ((eq context 'comment) 'c++)
4120 ((eq context 'block-comment) 'c)
4121 ((and detect-cpp (save-excursion (c-beginning-of-macro lim))) 'pound))))
4122
4123 (defalias 'c-in-literal
4124 (if (fboundp 'buffer-syntactic-context)
4125 'c-fast-in-literal ; XEmacs
4126 'c-slow-in-literal)) ; GNU Emacs
4127
4128 ;; The defalias above isn't enough to shut up the byte compiler.
4129 (cc-bytecomp-defun c-in-literal)
4130
4131 (defun c-literal-limits (&optional lim near not-in-delimiter)
4132 "Return a cons of the beginning and end positions of the comment or
4133 string surrounding point (including both delimiters), or nil if point
4134 isn't in one. If LIM is non-nil, it's used as the \"safe\" position
4135 to start parsing from. If NEAR is non-nil, then the limits of any
4136 literal next to point is returned. \"Next to\" means there's only
4137 spaces and tabs between point and the literal. The search for such a
4138 literal is done first in forward direction. If NOT-IN-DELIMITER is
4139 non-nil, the case when point is inside a starting delimiter won't be
4140 recognized. This only has effect for comments which have starting
4141 delimiters with more than one character.
4142
4143 Note that this function might do hidden buffer changes. See the
4144 comment at the start of cc-engine.el for more info."
4145
4146 (save-excursion
4147 (let* ((pos (point))
4148 (lim (or lim (progn
4149 (c-beginning-of-syntax)
4150 (point))))
4151 (state (parse-partial-sexp lim pos)))
4152
4153 (cond ((elt state 3) ; String.
4154 (goto-char (elt state 8))
4155 (cons (point) (or (c-safe (c-forward-sexp 1) (point))
4156 (point-max))))
4157
4158 ((elt state 4) ; Comment.
4159 (goto-char (elt state 8))
4160 (cons (point) (progn (c-forward-single-comment) (point))))
4161
4162 ((and (not not-in-delimiter)
4163 (not (elt state 5))
4164 (eq (char-before) ?/)
4165 (looking-at "[/*]"))
4166 ;; We're standing in a comment starter.
4167 (backward-char 1)
4168 (cons (point) (progn (c-forward-single-comment) (point))))
4169
4170 (near
4171 (goto-char pos)
4172
4173 ;; Search forward for a literal.
4174 (skip-chars-forward " \t")
4175
4176 (cond
4177 ((looking-at c-string-limit-regexp) ; String.
4178 (cons (point) (or (c-safe (c-forward-sexp 1) (point))
4179 (point-max))))
4180
4181 ((looking-at c-comment-start-regexp) ; Line or block comment.
4182 (cons (point) (progn (c-forward-single-comment) (point))))
4183
4184 (t
4185 ;; Search backward.
4186 (skip-chars-backward " \t")
4187
4188 (let ((end (point)) beg)
4189 (cond
4190 ((save-excursion
4191 (< (skip-syntax-backward c-string-syntax) 0)) ; String.
4192 (setq beg (c-safe (c-backward-sexp 1) (point))))
4193
4194 ((and (c-safe (forward-char -2) t)
4195 (looking-at "*/"))
4196 ;; Block comment. Due to the nature of line
4197 ;; comments, they will always be covered by the
4198 ;; normal case above.
4199 (goto-char end)
4200 (c-backward-single-comment)
4201 ;; If LIM is bogus, beg will be bogus.
4202 (setq beg (point))))
4203
4204 (if beg (cons beg end))))))
4205 ))))
4206
4207 ;; In case external callers use this; it did have a docstring.
4208 (defalias 'c-literal-limits-fast 'c-literal-limits)
4209
4210 (defun c-collect-line-comments (range)
4211 "If the argument is a cons of two buffer positions (such as returned by
4212 `c-literal-limits'), and that range contains a C++ style line comment,
4213 then an extended range is returned that contains all adjacent line
4214 comments (i.e. all comments that starts in the same column with no
4215 empty lines or non-whitespace characters between them). Otherwise the
4216 argument is returned.
4217
4218 Note that this function might do hidden buffer changes. See the
4219 comment at the start of cc-engine.el for more info."
4220
4221 (save-excursion
4222 (condition-case nil
4223 (if (and (consp range) (progn
4224 (goto-char (car range))
4225 (looking-at c-line-comment-starter)))
4226 (let ((col (current-column))
4227 (beg (point))
4228 (bopl (c-point 'bopl))
4229 (end (cdr range)))
4230 ;; Got to take care in the backward direction to handle
4231 ;; comments which are preceded by code.
4232 (while (and (c-backward-single-comment)
4233 (>= (point) bopl)
4234 (looking-at c-line-comment-starter)
4235 (= col (current-column)))
4236 (setq beg (point)
4237 bopl (c-point 'bopl)))
4238 (goto-char end)
4239 (while (and (progn (skip-chars-forward " \t")
4240 (looking-at c-line-comment-starter))
4241 (= col (current-column))
4242 (prog1 (zerop (forward-line 1))
4243 (setq end (point)))))
4244 (cons beg end))
4245 range)
4246 (error range))))
4247
4248 (defun c-literal-type (range)
4249 "Convenience function that given the result of `c-literal-limits',
4250 returns nil or the type of literal that the range surrounds, one
4251 of the symbols 'c, 'c++ or 'string. It's much faster than using
4252 `c-in-literal' and is intended to be used when you need both the
4253 type of a literal and its limits.
4254
4255 Note that this function might do hidden buffer changes. See the
4256 comment at the start of cc-engine.el for more info."
4257
4258 (if (consp range)
4259 (save-excursion
4260 (goto-char (car range))
4261 (cond ((looking-at c-string-limit-regexp) 'string)
4262 ((or (looking-at "//") ; c++ line comment
4263 (and (looking-at "\\s<") ; comment starter
4264 (looking-at "#"))) ; awk comment.
4265 'c++)
4266 (t 'c))) ; Assuming the range is valid.
4267 range))
4268
4269 \f
4270 ;; `c-find-decl-spots' and accompanying stuff.
4271
4272 ;; Variables used in `c-find-decl-spots' to cache the search done for
4273 ;; the first declaration in the last call. When that function starts,
4274 ;; it needs to back up over syntactic whitespace to look at the last
4275 ;; token before the region being searched. That can sometimes cause
4276 ;; moves back and forth over a quite large region of comments and
4277 ;; macros, which would be repeated for each changed character when
4278 ;; we're called during fontification, since font-lock refontifies the
4279 ;; current line for each change. Thus it's worthwhile to cache the
4280 ;; first match.
4281 ;;
4282 ;; `c-find-decl-syntactic-pos' is a syntactically relevant position in
4283 ;; the syntactic whitespace less or equal to some start position.
4284 ;; There's no cached value if it's nil.
4285 ;;
4286 ;; `c-find-decl-match-pos' is the match position if
4287 ;; `c-find-decl-prefix-search' matched before the syntactic whitespace
4288 ;; at `c-find-decl-syntactic-pos', or nil if there's no such match.
4289 (defvar c-find-decl-syntactic-pos nil)
4290 (make-variable-buffer-local 'c-find-decl-syntactic-pos)
4291 (defvar c-find-decl-match-pos nil)
4292 (make-variable-buffer-local 'c-find-decl-match-pos)
4293
4294 (defsubst c-invalidate-find-decl-cache (change-min-pos)
4295 (and c-find-decl-syntactic-pos
4296 (< change-min-pos c-find-decl-syntactic-pos)
4297 (setq c-find-decl-syntactic-pos nil)))
4298
4299 ; (defface c-debug-decl-spot-face
4300 ; '((t (:background "Turquoise")))
4301 ; "Debug face to mark the spots where `c-find-decl-spots' stopped.")
4302 ; (defface c-debug-decl-sws-face
4303 ; '((t (:background "Khaki")))
4304 ; "Debug face to mark the syntactic whitespace between the declaration
4305 ; spots and the preceding token end.")
4306
4307 (defmacro c-debug-put-decl-spot-faces (match-pos decl-pos)
4308 (when (facep 'c-debug-decl-spot-face)
4309 `(c-save-buffer-state ((match-pos ,match-pos) (decl-pos ,decl-pos))
4310 (c-debug-add-face (max match-pos (point-min)) decl-pos
4311 'c-debug-decl-sws-face)
4312 (c-debug-add-face decl-pos (min (1+ decl-pos) (point-max))
4313 'c-debug-decl-spot-face))))
4314 (defmacro c-debug-remove-decl-spot-faces (beg end)
4315 (when (facep 'c-debug-decl-spot-face)
4316 `(c-save-buffer-state ()
4317 (c-debug-remove-face ,beg ,end 'c-debug-decl-spot-face)
4318 (c-debug-remove-face ,beg ,end 'c-debug-decl-sws-face))))
4319
4320 (defmacro c-find-decl-prefix-search ()
4321 ;; Macro used inside `c-find-decl-spots'. It ought to be a defun,
4322 ;; but it contains lots of free variables that refer to things
4323 ;; inside `c-find-decl-spots'. The point is left at `cfd-match-pos'
4324 ;; if there is a match, otherwise at `cfd-limit'.
4325 ;;
4326 ;; This macro might do hidden buffer changes.
4327
4328 '(progn
4329 ;; Find the next property match position if we haven't got one already.
4330 (unless cfd-prop-match
4331 (save-excursion
4332 (while (progn
4333 (goto-char (next-single-property-change
4334 (point) 'c-type nil cfd-limit))
4335 (and (< (point) cfd-limit)
4336 (not (eq (c-get-char-property (1- (point)) 'c-type)
4337 'c-decl-end)))))
4338 (setq cfd-prop-match (point))))
4339
4340 ;; Find the next `c-decl-prefix-or-start-re' match if we haven't
4341 ;; got one already.
4342 (unless cfd-re-match
4343
4344 (if (> cfd-re-match-end (point))
4345 (goto-char cfd-re-match-end))
4346
4347 (while (if (setq cfd-re-match-end
4348 (re-search-forward c-decl-prefix-or-start-re
4349 cfd-limit 'move))
4350
4351 ;; Match. Check if it's inside a comment or string literal.
4352 (c-got-face-at
4353 (if (setq cfd-re-match (match-end 1))
4354 ;; Matched the end of a token preceding a decl spot.
4355 (progn
4356 (goto-char cfd-re-match)
4357 (1- cfd-re-match))
4358 ;; Matched a token that start a decl spot.
4359 (goto-char (match-beginning 0))
4360 (point))
4361 c-literal-faces)
4362
4363 ;; No match. Finish up and exit the loop.
4364 (setq cfd-re-match cfd-limit)
4365 nil)
4366
4367 ;; Skip out of comments and string literals.
4368 (while (progn
4369 (goto-char (next-single-property-change
4370 (point) 'face nil cfd-limit))
4371 (and (< (point) cfd-limit)
4372 (c-got-face-at (point) c-literal-faces)))))
4373
4374 ;; If we matched at the decl start, we have to back up over the
4375 ;; preceding syntactic ws to set `cfd-match-pos' and to catch
4376 ;; any decl spots in the syntactic ws.
4377 (unless cfd-re-match
4378 (c-backward-syntactic-ws)
4379 (setq cfd-re-match (point))))
4380
4381 ;; Choose whichever match is closer to the start.
4382 (if (< cfd-re-match cfd-prop-match)
4383 (setq cfd-match-pos cfd-re-match
4384 cfd-re-match nil)
4385 (setq cfd-match-pos cfd-prop-match
4386 cfd-prop-match nil))
4387
4388 (goto-char cfd-match-pos)
4389
4390 (when (< cfd-match-pos cfd-limit)
4391 ;; Skip forward past comments only so we don't skip macros.
4392 (c-forward-comments)
4393 ;; Set the position to continue at. We can avoid going over
4394 ;; the comments skipped above a second time, but it's possible
4395 ;; that the comment skipping has taken us past `cfd-prop-match'
4396 ;; since the property might be used inside comments.
4397 (setq cfd-continue-pos (if cfd-prop-match
4398 (min cfd-prop-match (point))
4399 (point))))))
4400
4401 (defun c-find-decl-spots (cfd-limit cfd-decl-re cfd-face-checklist cfd-fun)
4402 ;; Call CFD-FUN for each possible spot for a declaration, cast or
4403 ;; label from the point to CFD-LIMIT.
4404 ;;
4405 ;; CFD-FUN is called with point at the start of the spot. It's
4406 ;; passed two arguments: The first is the end position of the token
4407 ;; preceding the spot, or 0 for the implicit match at bob. The
4408 ;; second is a flag that is t when the match is inside a macro. If
4409 ;; CFD-FUN adds `c-decl-end' properties somewhere below the current
4410 ;; spot, it should return non-nil to ensure that the next search
4411 ;; will find them.
4412 ;;
4413 ;; Such a spot is:
4414 ;; o The first token after bob.
4415 ;; o The first token after the end of submatch 1 in
4416 ;; `c-decl-prefix-or-start-re' when that submatch matches.
4417 ;; o The start of each `c-decl-prefix-or-start-re' match when
4418 ;; submatch 1 doesn't match.
4419 ;; o The first token after the end of each occurrence of the
4420 ;; `c-type' text property with the value `c-decl-end', provided
4421 ;; `c-type-decl-end-used' is set.
4422 ;;
4423 ;; Only a spot that match CFD-DECL-RE and whose face is in the
4424 ;; CFD-FACE-CHECKLIST list causes CFD-FUN to be called. The face
4425 ;; check is disabled if CFD-FACE-CHECKLIST is nil.
4426 ;;
4427 ;; If the match is inside a macro then the buffer is narrowed to the
4428 ;; end of it, so that CFD-FUN can investigate the following tokens
4429 ;; without matching something that begins inside a macro and ends
4430 ;; outside it. It's to avoid this work that the CFD-DECL-RE and
4431 ;; CFD-FACE-CHECKLIST checks exist.
4432 ;;
4433 ;; The spots are visited approximately in order from top to bottom.
4434 ;; It's however the positions where `c-decl-prefix-or-start-re'
4435 ;; matches and where `c-decl-end' properties are found that are in
4436 ;; order. Since the spots often are at the following token, they
4437 ;; might be visited out of order insofar as more spots are reported
4438 ;; later on within the syntactic whitespace between the match
4439 ;; positions and their spots.
4440 ;;
4441 ;; It's assumed that comments and strings are fontified in the
4442 ;; searched range.
4443 ;;
4444 ;; This is mainly used in fontification, and so has an elaborate
4445 ;; cache to handle repeated calls from the same start position; see
4446 ;; the variables above.
4447 ;;
4448 ;; All variables in this function begin with `cfd-' to avoid name
4449 ;; collision with the (dynamically bound) variables used in CFD-FUN.
4450 ;;
4451 ;; This function might do hidden buffer changes.
4452
4453 (let ((cfd-start-pos (point))
4454 (cfd-buffer-end (point-max))
4455 ;; The end of the token preceding the decl spot last found
4456 ;; with `c-decl-prefix-or-start-re'. `cfd-limit' if there's
4457 ;; no match.
4458 cfd-re-match
4459 ;; The end position of the last `c-decl-prefix-or-start-re'
4460 ;; match. If this is greater than `cfd-continue-pos', the
4461 ;; next regexp search is started here instead.
4462 (cfd-re-match-end (point-min))
4463 ;; The end of the last `c-decl-end' found by
4464 ;; `c-find-decl-prefix-search'. `cfd-limit' if there's no
4465 ;; match. If searching for the property isn't needed then we
4466 ;; disable it by setting it to `cfd-limit' directly.
4467 (cfd-prop-match (unless c-type-decl-end-used cfd-limit))
4468 ;; The end of the token preceding the decl spot last found by
4469 ;; `c-find-decl-prefix-search'. 0 for the implicit match at
4470 ;; bob. `cfd-limit' if there's no match. In other words,
4471 ;; this is the minimum of `cfd-re-match' and `cfd-prop-match'.
4472 (cfd-match-pos cfd-limit)
4473 ;; The position to continue searching at.
4474 cfd-continue-pos
4475 ;; The position of the last "real" token we've stopped at.
4476 ;; This can be greater than `cfd-continue-pos' when we get
4477 ;; hits inside macros or at `c-decl-end' positions inside
4478 ;; comments.
4479 (cfd-token-pos 0)
4480 ;; The end position of the last entered macro.
4481 (cfd-macro-end 0))
4482
4483 ;; Initialize by finding a syntactically relevant start position
4484 ;; before the point, and do the first `c-decl-prefix-or-start-re'
4485 ;; search unless we're at bob.
4486
4487 (let (start-in-literal start-in-macro syntactic-pos)
4488 ;; Must back up a bit since we look for the end of the previous
4489 ;; statement or declaration, which is earlier than the first
4490 ;; returned match.
4491
4492 (cond
4493 ;; First we need to move to a syntactically relevant position.
4494 ;; Begin by backing out of comment or string literals.
4495 ((and
4496 (when (c-got-face-at (point) c-literal-faces)
4497 ;; Try to use the faces to back up to the start of the
4498 ;; literal. FIXME: What if the point is on a declaration
4499 ;; inside a comment?
4500 (while (and (not (bobp))
4501 (c-got-face-at (1- (point)) c-literal-faces))
4502 (goto-char (previous-single-property-change
4503 (point) 'face nil (point-min))))
4504
4505 ;; XEmacs doesn't fontify the quotes surrounding string
4506 ;; literals.
4507 (and (featurep 'xemacs)
4508 (eq (get-text-property (point) 'face)
4509 'font-lock-string-face)
4510 (not (bobp))
4511 (progn (backward-char)
4512 (not (looking-at c-string-limit-regexp)))
4513 (forward-char))
4514
4515 ;; Don't trust the literal to contain only literal faces
4516 ;; (the font lock package might not have fontified the
4517 ;; start of it at all, for instance) so check that we have
4518 ;; arrived at something that looks like a start or else
4519 ;; resort to `c-literal-limits'.
4520 (unless (looking-at c-literal-start-regexp)
4521 (let ((range (c-literal-limits)))
4522 (if range (goto-char (car range)))))
4523
4524 (setq start-in-literal (point)))
4525
4526 ;; The start is in a literal. If the limit is in the same
4527 ;; one we don't have to find a syntactic position etc. We
4528 ;; only check that if the limit is at or before bonl to save
4529 ;; time; it covers the by far most common case when font-lock
4530 ;; refontifies the current line only.
4531 (<= cfd-limit (c-point 'bonl cfd-start-pos))
4532 (save-excursion
4533 (goto-char cfd-start-pos)
4534 (while (progn
4535 (goto-char (next-single-property-change
4536 (point) 'face nil cfd-limit))
4537 (and (< (point) cfd-limit)
4538 (c-got-face-at (point) c-literal-faces))))
4539 (= (point) cfd-limit)))
4540
4541 ;; Completely inside a literal. Set up variables to trig the
4542 ;; (< cfd-continue-pos cfd-start-pos) case below and it'll
4543 ;; find a suitable start position.
4544 (setq cfd-continue-pos start-in-literal))
4545
4546 ;; Check if the region might be completely inside a macro, to
4547 ;; optimize that like the completely-inside-literal above.
4548 ((save-excursion
4549 (and (= (forward-line 1) 0)
4550 (bolp) ; forward-line has funny behavior at eob.
4551 (>= (point) cfd-limit)
4552 (progn (backward-char)
4553 (eq (char-before) ?\\))))
4554 ;; (Maybe) completely inside a macro. Only need to trig the
4555 ;; (< cfd-continue-pos cfd-start-pos) case below to make it
4556 ;; set things up.
4557 (setq cfd-continue-pos (1- cfd-start-pos)
4558 start-in-macro t))
4559
4560 (t
4561 ;; Back out of any macro so we don't miss any declaration
4562 ;; that could follow after it.
4563 (when (c-beginning-of-macro)
4564 (setq start-in-macro t))
4565
4566 ;; Now we're at a proper syntactically relevant position so we
4567 ;; can use the cache. But first clear it if it applied
4568 ;; further down.
4569 (c-invalidate-find-decl-cache cfd-start-pos)
4570
4571 (setq syntactic-pos (point))
4572 (unless (eq syntactic-pos c-find-decl-syntactic-pos)
4573 ;; Don't have to do this if the cache is relevant here,
4574 ;; typically if the same line is refontified again. If
4575 ;; we're just some syntactic whitespace further down we can
4576 ;; still use the cache to limit the skipping.
4577 (c-backward-syntactic-ws c-find-decl-syntactic-pos))
4578
4579 ;; If we hit `c-find-decl-syntactic-pos' and
4580 ;; `c-find-decl-match-pos' is set then we install the cached
4581 ;; values. If we hit `c-find-decl-syntactic-pos' and
4582 ;; `c-find-decl-match-pos' is nil then we know there's no decl
4583 ;; prefix in the whitespace before `c-find-decl-syntactic-pos'
4584 ;; and so we can continue the search from this point. If we
4585 ;; didn't hit `c-find-decl-syntactic-pos' then we're now in
4586 ;; the right spot to begin searching anyway.
4587 (if (and (eq (point) c-find-decl-syntactic-pos)
4588 c-find-decl-match-pos)
4589 (setq cfd-match-pos c-find-decl-match-pos
4590 cfd-continue-pos syntactic-pos)
4591
4592 (setq c-find-decl-syntactic-pos syntactic-pos)
4593
4594 (when (if (bobp)
4595 ;; Always consider bob a match to get the first
4596 ;; declaration in the file. Do this separately instead of
4597 ;; letting `c-decl-prefix-or-start-re' match bob, so that
4598 ;; regexp always can consume at least one character to
4599 ;; ensure that we won't get stuck in an infinite loop.
4600 (setq cfd-re-match 0)
4601 (backward-char)
4602 (c-beginning-of-current-token)
4603 (< (point) cfd-limit))
4604 ;; Do an initial search now. In the bob case above it's
4605 ;; only done to search for a `c-decl-end' spot.
4606 (c-find-decl-prefix-search))
4607
4608 (setq c-find-decl-match-pos (and (< cfd-match-pos cfd-start-pos)
4609 cfd-match-pos)))))
4610
4611 ;; Advance `cfd-continue-pos' if it's before the start position.
4612 ;; The closest continue position that might have effect at or
4613 ;; after the start depends on what we started in. This also
4614 ;; finds a suitable start position in the special cases when the
4615 ;; region is completely within a literal or macro.
4616 (when (and cfd-continue-pos (< cfd-continue-pos cfd-start-pos))
4617
4618 (cond
4619 (start-in-macro
4620 ;; If we're in a macro then it's the closest preceding token
4621 ;; in the macro. Check this before `start-in-literal',
4622 ;; since if we're inside a literal in a macro, the preceding
4623 ;; token is earlier than any `c-decl-end' spot inside the
4624 ;; literal (comment).
4625 (goto-char (or start-in-literal cfd-start-pos))
4626 ;; The only syntactic ws in macros are comments.
4627 (c-backward-comments)
4628 (backward-char)
4629 (c-beginning-of-current-token))
4630
4631 (start-in-literal
4632 ;; If we're in a comment it can only be the closest
4633 ;; preceding `c-decl-end' position within that comment, if
4634 ;; any. Go back to the beginning of such a property so that
4635 ;; `c-find-decl-prefix-search' will find the end of it.
4636 ;; (Can't stop at the end and install it directly on
4637 ;; `cfd-prop-match' since that variable might be cleared
4638 ;; after `cfd-fun' below.)
4639 ;;
4640 ;; Note that if the literal is a string then the property
4641 ;; search will simply skip to the beginning of it right
4642 ;; away.
4643 (if (not c-type-decl-end-used)
4644 (goto-char start-in-literal)
4645 (goto-char cfd-start-pos)
4646 (while (progn
4647 (goto-char (previous-single-property-change
4648 (point) 'c-type nil start-in-literal))
4649 (and (> (point) start-in-literal)
4650 (not (eq (c-get-char-property (point) 'c-type)
4651 'c-decl-end))))))
4652
4653 (when (= (point) start-in-literal)
4654 ;; Didn't find any property inside the comment, so we can
4655 ;; skip it entirely. (This won't skip past a string, but
4656 ;; that'll be handled quickly by the next
4657 ;; `c-find-decl-prefix-search' anyway.)
4658 (c-forward-single-comment)
4659 (if (> (point) cfd-limit)
4660 (goto-char cfd-limit))))
4661
4662 (t
4663 ;; If we started in normal code, the only match that might
4664 ;; apply before the start is what we already got in
4665 ;; `cfd-match-pos' so we can continue at the start position.
4666 ;; (Note that we don't get here if the first match is below
4667 ;; it.)
4668 (goto-char cfd-start-pos)))
4669
4670 ;; Delete found matches if they are before our new continue
4671 ;; position, so that `c-find-decl-prefix-search' won't back up
4672 ;; to them later on.
4673 (setq cfd-continue-pos (point))
4674 (when (and cfd-re-match (< cfd-re-match cfd-continue-pos))
4675 (setq cfd-re-match nil))
4676 (when (and cfd-prop-match (< cfd-prop-match cfd-continue-pos))
4677 (setq cfd-prop-match nil)))
4678
4679 (if syntactic-pos
4680 ;; This is the normal case and we got a proper syntactic
4681 ;; position. If there's a match then it's always outside
4682 ;; macros and comments, so advance to the next token and set
4683 ;; `cfd-token-pos'. The loop below will later go back using
4684 ;; `cfd-continue-pos' to fix declarations inside the
4685 ;; syntactic ws.
4686 (when (and cfd-match-pos (< cfd-match-pos syntactic-pos))
4687 (goto-char syntactic-pos)
4688 (c-forward-syntactic-ws)
4689 (and cfd-continue-pos
4690 (< cfd-continue-pos (point))
4691 (setq cfd-token-pos (point))))
4692
4693 ;; Have one of the special cases when the region is completely
4694 ;; within a literal or macro. `cfd-continue-pos' is set to a
4695 ;; good start position for the search, so do it.
4696 (c-find-decl-prefix-search)))
4697
4698 ;; Now loop. Round what? (ACM, 2006/7/5). We already got the first match.
4699
4700 (while (progn
4701 (while (and
4702 (< cfd-match-pos cfd-limit)
4703
4704 (or
4705 ;; Kludge to filter out matches on the "<" that
4706 ;; aren't open parens, for the sake of languages
4707 ;; that got `c-recognize-<>-arglists' set.
4708 (and (eq (char-before cfd-match-pos) ?<)
4709 (not (c-get-char-property (1- cfd-match-pos)
4710 'syntax-table)))
4711
4712 ;; If `cfd-continue-pos' is less or equal to
4713 ;; `cfd-token-pos', we've got a hit inside a macro
4714 ;; that's in the syntactic whitespace before the last
4715 ;; "real" declaration we've checked. If they're equal
4716 ;; we've arrived at the declaration a second time, so
4717 ;; there's nothing to do.
4718 (= cfd-continue-pos cfd-token-pos)
4719
4720 (progn
4721 ;; If `cfd-continue-pos' is less than `cfd-token-pos'
4722 ;; we're still searching for declarations embedded in
4723 ;; the syntactic whitespace. In that case we need
4724 ;; only to skip comments and not macros, since they
4725 ;; can't be nested, and that's already been done in
4726 ;; `c-find-decl-prefix-search'.
4727 (when (> cfd-continue-pos cfd-token-pos)
4728 (c-forward-syntactic-ws)
4729 (setq cfd-token-pos (point)))
4730
4731 ;; Continue if the following token fails the
4732 ;; CFD-DECL-RE and CFD-FACE-CHECKLIST checks.
4733 (when (or (>= (point) cfd-limit)
4734 (not (looking-at cfd-decl-re))
4735 (and cfd-face-checklist
4736 (not (c-got-face-at
4737 (point) cfd-face-checklist))))
4738 (goto-char cfd-continue-pos)
4739 t)))
4740
4741 (< (point) cfd-limit))
4742 (c-find-decl-prefix-search))
4743
4744 (< (point) cfd-limit))
4745
4746 (when (and
4747 (>= (point) cfd-start-pos)
4748
4749 (progn
4750 ;; Narrow to the end of the macro if we got a hit inside
4751 ;; one, to avoid recognizing things that start inside the
4752 ;; macro and end outside it.
4753 (when (> cfd-match-pos cfd-macro-end)
4754 ;; Not in the same macro as in the previous round.
4755 (save-excursion
4756 (goto-char cfd-match-pos)
4757 (setq cfd-macro-end
4758 (if (save-excursion (and (c-beginning-of-macro)
4759 (< (point) cfd-match-pos)))
4760 (progn (c-end-of-macro)
4761 (point))
4762 0))))
4763
4764 (if (zerop cfd-macro-end)
4765 t
4766 (if (> cfd-macro-end (point))
4767 (progn (narrow-to-region (point-min) cfd-macro-end)
4768 t)
4769 ;; The matched token was the last thing in the macro,
4770 ;; so the whole match is bogus.
4771 (setq cfd-macro-end 0)
4772 nil))))
4773
4774 (c-debug-put-decl-spot-faces cfd-match-pos (point))
4775 (if (funcall cfd-fun cfd-match-pos (/= cfd-macro-end 0))
4776 (setq cfd-prop-match nil))
4777
4778 (when (/= cfd-macro-end 0)
4779 ;; Restore limits if we did macro narrowment above.
4780 (narrow-to-region (point-min) cfd-buffer-end)))
4781
4782 (goto-char cfd-continue-pos)
4783 (if (= cfd-continue-pos cfd-limit)
4784 (setq cfd-match-pos cfd-limit)
4785 (c-find-decl-prefix-search)))))
4786
4787 \f
4788 ;; A cache for found types.
4789
4790 ;; Buffer local variable that contains an obarray with the types we've
4791 ;; found. If a declaration is recognized somewhere we record the
4792 ;; fully qualified identifier in it to recognize it as a type
4793 ;; elsewhere in the file too. This is not accurate since we do not
4794 ;; bother with the scoping rules of the languages, but in practice the
4795 ;; same name is seldom used as both a type and something else in a
4796 ;; file, and we only use this as a last resort in ambiguous cases (see
4797 ;; `c-forward-decl-or-cast-1').
4798 ;;
4799 ;; Not every type need be in this cache. However, things which have
4800 ;; ceased to be types must be removed from it.
4801 ;;
4802 ;; Template types in C++ are added here too but with the template
4803 ;; arglist replaced with "<>" in references or "<" for the one in the
4804 ;; primary type. E.g. the type "Foo<A,B>::Bar<C>" is stored as
4805 ;; "Foo<>::Bar<". This avoids storing very long strings (since C++
4806 ;; template specs can be fairly sized programs in themselves) and
4807 ;; improves the hit ratio (it's a type regardless of the template
4808 ;; args; it's just not the same type, but we're only interested in
4809 ;; recognizing types, not telling distinct types apart). Note that
4810 ;; template types in references are added here too; from the example
4811 ;; above there will also be an entry "Foo<".
4812 (defvar c-found-types nil)
4813 (make-variable-buffer-local 'c-found-types)
4814
4815 (defsubst c-clear-found-types ()
4816 ;; Clears `c-found-types'.
4817 (setq c-found-types (make-vector 53 0)))
4818
4819 (defun c-add-type (from to)
4820 ;; Add the given region as a type in `c-found-types'. If the region
4821 ;; doesn't match an existing type but there is a type which is equal
4822 ;; to the given one except that the last character is missing, then
4823 ;; the shorter type is removed. That's done to avoid adding all
4824 ;; prefixes of a type as it's being entered and font locked. This
4825 ;; doesn't cover cases like when characters are removed from a type
4826 ;; or added in the middle. We'd need the position of point when the
4827 ;; font locking is invoked to solve this well.
4828 ;;
4829 ;; This function might do hidden buffer changes.
4830 (let ((type (c-syntactic-content from to c-recognize-<>-arglists)))
4831 (unless (intern-soft type c-found-types)
4832 (unintern (substring type 0 -1) c-found-types)
4833 (intern type c-found-types))))
4834
4835 (defun c-unfind-type (name)
4836 ;; Remove the "NAME" from c-found-types, if present.
4837 (unintern name c-found-types))
4838
4839 (defsubst c-check-type (from to)
4840 ;; Return non-nil if the given region contains a type in
4841 ;; `c-found-types'.
4842 ;;
4843 ;; This function might do hidden buffer changes.
4844 (intern-soft (c-syntactic-content from to c-recognize-<>-arglists)
4845 c-found-types))
4846
4847 (defun c-list-found-types ()
4848 ;; Return all the types in `c-found-types' as a sorted list of
4849 ;; strings.
4850 (let (type-list)
4851 (mapatoms (lambda (type)
4852 (setq type-list (cons (symbol-name type)
4853 type-list)))
4854 c-found-types)
4855 (sort type-list 'string-lessp)))
4856
4857 ;; Shut up the byte compiler.
4858 (defvar c-maybe-stale-found-type)
4859
4860 (defun c-trim-found-types (beg end old-len)
4861 ;; An after change function which, in conjunction with the info in
4862 ;; c-maybe-stale-found-type (set in c-before-change), removes a type
4863 ;; from `c-found-types', should this type have become stale. For
4864 ;; example, this happens to "foo" when "foo \n bar();" becomes
4865 ;; "foo(); \n bar();". Such stale types, if not removed, foul up
4866 ;; the fontification.
4867 ;;
4868 ;; Have we, perhaps, added non-ws characters to the front/back of a found
4869 ;; type?
4870 (when (> end beg)
4871 (save-excursion
4872 (when (< end (point-max))
4873 (goto-char end)
4874 (if (and (c-beginning-of-current-token) ; only moves when we started in the middle
4875 (progn (goto-char end)
4876 (c-end-of-current-token)))
4877 (c-unfind-type (buffer-substring-no-properties
4878 end (point)))))
4879 (when (> beg (point-min))
4880 (goto-char beg)
4881 (if (and (c-end-of-current-token) ; only moves when we started in the middle
4882 (progn (goto-char beg)
4883 (c-beginning-of-current-token)))
4884 (c-unfind-type (buffer-substring-no-properties
4885 (point) beg))))))
4886
4887 (if c-maybe-stale-found-type ; e.g. (c-decl-id-start "foo" 97 107 " (* ooka) " "o")
4888 (cond
4889 ;; Changing the amount of (already existing) whitespace - don't do anything.
4890 ((and (c-partial-ws-p beg end)
4891 (or (= beg end) ; removal of WS
4892 (string-match "^[ \t\n\r\f\v]*$" (nth 5 c-maybe-stale-found-type)))))
4893
4894 ;; The syntactic relationship which defined a "found type" has been
4895 ;; destroyed.
4896 ((eq (car c-maybe-stale-found-type) 'c-decl-id-start)
4897 (c-unfind-type (cadr c-maybe-stale-found-type)))
4898 ;; ((eq (car c-maybe-stale-found-type) 'c-decl-type-start) FIXME!!!
4899 )))
4900
4901 \f
4902 ;; Setting and removing syntax properties on < and > in languages (C++
4903 ;; and Java) where they can be template/generic delimiters as well as
4904 ;; their normal meaning of "less/greater than".
4905
4906 ;; Normally, < and > have syntax 'punctuation'. When they are found to
4907 ;; be delimiters, they are marked as such with the category properties
4908 ;; c-<-as-paren-syntax, c->-as-paren-syntax respectively.
4909
4910 ;; STRATEGY:
4911 ;;
4912 ;; It is impossible to determine with certainty whether a <..> pair in
4913 ;; C++ is two comparison operators or is template delimiters, unless
4914 ;; one duplicates a lot of a C++ compiler. For example, the following
4915 ;; code fragment:
4916 ;;
4917 ;; foo (a < b, c > d) ;
4918 ;;
4919 ;; could be a function call with two integer parameters (each a
4920 ;; relational expression), or it could be a constructor for class foo
4921 ;; taking one parameter d of templated type "a < b, c >". They are
4922 ;; somewhat easier to distinguish in Java.
4923 ;;
4924 ;; The strategy now (2010-01) adopted is to mark and unmark < and
4925 ;; > IN MATCHING PAIRS ONLY. [Previously, they were marked
4926 ;; individually when their context so indicated. This gave rise to
4927 ;; intractible problems when one of a matching pair was deleted, or
4928 ;; pulled into a literal.]
4929 ;;
4930 ;; At each buffer change, the syntax-table properties are removed in a
4931 ;; before-change function and reapplied, when needed, in an
4932 ;; after-change function. It is far more important that the
4933 ;; properties get removed when they they are spurious than that they
4934 ;; be present when wanted.
4935 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
4936 (defun c-clear-<-pair-props (&optional pos)
4937 ;; POS (default point) is at a < character. If it is marked with
4938 ;; open paren syntax-table text property, remove the property,
4939 ;; together with the close paren property on the matching > (if
4940 ;; any).
4941 (save-excursion
4942 (if pos
4943 (goto-char pos)
4944 (setq pos (point)))
4945 (when (equal (c-get-char-property (point) 'syntax-table)
4946 c-<-as-paren-syntax)
4947 (with-syntax-table c-no-parens-syntax-table ; ignore unbalanced [,{,(,..
4948 (c-go-list-forward))
4949 (when (equal (c-get-char-property (1- (point)) 'syntax-table)
4950 c->-as-paren-syntax) ; should always be true.
4951 (c-clear-char-property (1- (point)) 'category))
4952 (c-clear-char-property pos 'category))))
4953
4954 (defun c-clear->-pair-props (&optional pos)
4955 ;; POS (default point) is at a > character. If it is marked with
4956 ;; close paren syntax-table property, remove the property, together
4957 ;; with the open paren property on the matching < (if any).
4958 (save-excursion
4959 (if pos
4960 (goto-char pos)
4961 (setq pos (point)))
4962 (when (equal (c-get-char-property (point) 'syntax-table)
4963 c->-as-paren-syntax)
4964 (with-syntax-table c-no-parens-syntax-table ; ignore unbalanced [,{,(,..
4965 (c-go-up-list-backward))
4966 (when (equal (c-get-char-property (point) 'syntax-table)
4967 c-<-as-paren-syntax) ; should always be true.
4968 (c-clear-char-property (point) 'category))
4969 (c-clear-char-property pos 'category))))
4970
4971 (defun c-clear-<>-pair-props (&optional pos)
4972 ;; POS (default point) is at a < or > character. If it has an
4973 ;; open/close paren syntax-table property, remove this property both
4974 ;; from the current character and its partner (which will also be
4975 ;; thusly marked).
4976 (cond
4977 ((eq (char-after) ?\<)
4978 (c-clear-<-pair-props pos))
4979 ((eq (char-after) ?\>)
4980 (c-clear->-pair-props pos))
4981 (t (c-benign-error
4982 "c-clear-<>-pair-props called from wrong position"))))
4983
4984 (defun c-clear-<-pair-props-if-match-after (lim &optional pos)
4985 ;; POS (default point) is at a < character. If it is both marked
4986 ;; with open/close paren syntax-table property, and has a matching >
4987 ;; (also marked) which is after LIM, remove the property both from
4988 ;; the current > and its partner.
4989 (save-excursion
4990 (if pos
4991 (goto-char pos)
4992 (setq pos (point)))
4993 (when (equal (c-get-char-property (point) 'syntax-table)
4994 c-<-as-paren-syntax)
4995 (with-syntax-table c-no-parens-syntax-table ; ignore unbalanced [,{,(,..
4996 (c-go-list-forward))
4997 (when (and (>= (point) lim)
4998 (equal (c-get-char-property (1- (point)) 'syntax-table)
4999 c->-as-paren-syntax)) ; should always be true.
5000 (c-unmark-<->-as-paren (1- (point)))
5001 (c-unmark-<->-as-paren pos)))))
5002
5003 (defun c-clear->-pair-props-if-match-before (lim &optional pos)
5004 ;; POS (default point) is at a > character. If it is both marked
5005 ;; with open/close paren syntax-table property, and has a matching <
5006 ;; (also marked) which is before LIM, remove the property both from
5007 ;; the current < and its partner.
5008 (save-excursion
5009 (if pos
5010 (goto-char pos)
5011 (setq pos (point)))
5012 (when (equal (c-get-char-property (point) 'syntax-table)
5013 c->-as-paren-syntax)
5014 (with-syntax-table c-no-parens-syntax-table ; ignore unbalanced [,{,(,..
5015 (c-go-up-list-backward))
5016 (when (and (<= (point) lim)
5017 (equal (c-get-char-property (point) 'syntax-table)
5018 c-<-as-paren-syntax)) ; should always be true.
5019 (c-unmark-<->-as-paren (point))
5020 (c-unmark-<->-as-paren pos)))))
5021
5022 (defun c-before-change-check-<>-operators (beg end)
5023 ;; Unmark certain pairs of "< .... >" which are currently marked as
5024 ;; template/generic delimiters. (This marking is via syntax-table
5025 ;; text properties).
5026 ;;
5027 ;; These pairs are those which are in the current "statement" (i.e.,
5028 ;; the region between the {, }, or ; before BEG and the one after
5029 ;; END), and which enclose any part of the interval (BEG END).
5030 ;;
5031 ;; Note that in C++ (?and Java), template/generic parens cannot
5032 ;; enclose a brace or semicolon, so we use these as bounds on the
5033 ;; region we must work on.
5034 ;;
5035 ;; This function is called from before-change-functions (via
5036 ;; c-get-state-before-change-functions). Thus the buffer is widened,
5037 ;; and point is undefined, both at entry and exit.
5038 ;;
5039 ;; FIXME!!! This routine ignores the possibility of macros entirely.
5040 ;; 2010-01-29.
5041 (save-excursion
5042 (let ((beg-lit-limits (progn (goto-char beg) (c-literal-limits)))
5043 (end-lit-limits (progn (goto-char end) (c-literal-limits))))
5044 ;; Locate the barrier before the changed region
5045 (goto-char (if beg-lit-limits (car beg-lit-limits) beg))
5046 (c-syntactic-skip-backward "^;{}" (max (- beg 2048) (point-min)))
5047
5048 ;; Remove the syntax-table properties from each pertinent <...> pair.
5049 ;; Firsly, the ones with the < before beg and > after beg.
5050 (while (c-search-forward-char-property 'category 'c-<-as-paren-syntax beg)
5051 (c-clear-<-pair-props-if-match-after beg (1- (point))))
5052
5053 ;; Locate the barrier after END.
5054 (goto-char (if end-lit-limits (cdr end-lit-limits) end))
5055 (c-syntactic-re-search-forward "[;{}]"
5056 (min (+ end 2048) (point-max)) 'end)
5057
5058 ;; Remove syntax-table properties from the remaining pertinent <...>
5059 ;; pairs, those with a > after end and < before end.
5060 (while (c-search-backward-char-property 'category 'c->-as-paren-syntax end)
5061 (c-clear->-pair-props-if-match-before end)))))
5062
5063
5064
5065 (defun c-after-change-check-<>-operators (beg end)
5066 ;; This is called from `after-change-functions' when
5067 ;; c-recognize-<>-arglists' is set. It ensures that no "<" or ">"
5068 ;; chars with paren syntax become part of another operator like "<<"
5069 ;; or ">=".
5070 ;;
5071 ;; This function might do hidden buffer changes.
5072
5073 (save-excursion
5074 (goto-char beg)
5075 (when (or (looking-at "[<>]")
5076 (< (skip-chars-backward "<>") 0))
5077
5078 (goto-char beg)
5079 (c-beginning-of-current-token)
5080 (when (and (< (point) beg)
5081 (looking-at c-<>-multichar-token-regexp)
5082 (< beg (setq beg (match-end 0))))
5083 (while (progn (skip-chars-forward "^<>" beg)
5084 (< (point) beg))
5085 (c-clear-<>-pair-props)
5086 (forward-char))))
5087
5088 (when (< beg end)
5089 (goto-char end)
5090 (when (or (looking-at "[<>]")
5091 (< (skip-chars-backward "<>") 0))
5092
5093 (goto-char end)
5094 (c-beginning-of-current-token)
5095 (when (and (< (point) end)
5096 (looking-at c-<>-multichar-token-regexp)
5097 (< end (setq end (match-end 0))))
5098 (while (progn (skip-chars-forward "^<>" end)
5099 (< (point) end))
5100 (c-clear-<>-pair-props)
5101 (forward-char)))))))
5102
5103
5104 \f
5105 ;; Handling of small scale constructs like types and names.
5106
5107 ;; Dynamically bound variable that instructs `c-forward-type' to also
5108 ;; treat possible types (i.e. those that it normally returns 'maybe or
5109 ;; 'found for) as actual types (and always return 'found for them).
5110 ;; This means that it records them in `c-record-type-identifiers' if
5111 ;; that is set, and that it adds them to `c-found-types'.
5112 (defvar c-promote-possible-types nil)
5113
5114 ;; Dynamically bound variable that instructs `c-forward-<>-arglist' to
5115 ;; mark up successfully parsed arglists with paren syntax properties on
5116 ;; the surrounding angle brackets and with `c-<>-arg-sep' in the
5117 ;; `c-type' property of each argument separating comma.
5118 ;;
5119 ;; Setting this variable also makes `c-forward-<>-arglist' recurse into
5120 ;; all arglists for side effects (i.e. recording types), otherwise it
5121 ;; exploits any existing paren syntax properties to quickly jump to the
5122 ;; end of already parsed arglists.
5123 ;;
5124 ;; Marking up the arglists is not the default since doing that correctly
5125 ;; depends on a proper value for `c-restricted-<>-arglists'.
5126 (defvar c-parse-and-markup-<>-arglists nil)
5127
5128 ;; Dynamically bound variable that instructs `c-forward-<>-arglist' to
5129 ;; not accept arglists that contain binary operators.
5130 ;;
5131 ;; This is primarily used to handle C++ template arglists. C++
5132 ;; disambiguates them by checking whether the preceding name is a
5133 ;; template or not. We can't do that, so we assume it is a template
5134 ;; if it can be parsed as one. That usually works well since
5135 ;; comparison expressions on the forms "a < b > c" or "a < b, c > d"
5136 ;; in almost all cases would be pointless.
5137 ;;
5138 ;; However, in function arglists, e.g. in "foo (a < b, c > d)", we
5139 ;; should let the comma separate the function arguments instead. And
5140 ;; in a context where the value of the expression is taken, e.g. in
5141 ;; "if (a < b || c > d)", it's probably not a template.
5142 (defvar c-restricted-<>-arglists nil)
5143
5144 ;; Dynamically bound variables that instructs
5145 ;; `c-forward-keyword-clause', `c-forward-<>-arglist',
5146 ;; `c-forward-name', `c-forward-type', `c-forward-decl-or-cast-1', and
5147 ;; `c-forward-label' to record the ranges of all the type and
5148 ;; reference identifiers they encounter. They will build lists on
5149 ;; these variables where each element is a cons of the buffer
5150 ;; positions surrounding each identifier. This recording is only
5151 ;; activated when `c-record-type-identifiers' is non-nil.
5152 ;;
5153 ;; All known types that can't be identifiers are recorded, and also
5154 ;; other possible types if `c-promote-possible-types' is set.
5155 ;; Recording is however disabled inside angle bracket arglists that
5156 ;; are encountered inside names and other angle bracket arglists.
5157 ;; Such occurrences are taken care of by `c-font-lock-<>-arglists'
5158 ;; instead.
5159 ;;
5160 ;; Only the names in C++ template style references (e.g. "tmpl" in
5161 ;; "tmpl<a,b>::foo") are recorded as references, other references
5162 ;; aren't handled here.
5163 ;;
5164 ;; `c-forward-label' records the label identifier(s) on
5165 ;; `c-record-ref-identifiers'.
5166 (defvar c-record-type-identifiers nil)
5167 (defvar c-record-ref-identifiers nil)
5168
5169 ;; This variable will receive a cons cell of the range of the last
5170 ;; single identifier symbol stepped over by `c-forward-name' if it's
5171 ;; successful. This is the range that should be put on one of the
5172 ;; record lists above by the caller. It's assigned nil if there's no
5173 ;; such symbol in the name.
5174 (defvar c-last-identifier-range nil)
5175
5176 (defmacro c-record-type-id (range)
5177 (if (eq (car-safe range) 'cons)
5178 ;; Always true.
5179 `(setq c-record-type-identifiers
5180 (cons ,range c-record-type-identifiers))
5181 `(let ((range ,range))
5182 (if range
5183 (setq c-record-type-identifiers
5184 (cons range c-record-type-identifiers))))))
5185
5186 (defmacro c-record-ref-id (range)
5187 (if (eq (car-safe range) 'cons)
5188 ;; Always true.
5189 `(setq c-record-ref-identifiers
5190 (cons ,range c-record-ref-identifiers))
5191 `(let ((range ,range))
5192 (if range
5193 (setq c-record-ref-identifiers
5194 (cons range c-record-ref-identifiers))))))
5195
5196 ;; Dynamically bound variable that instructs `c-forward-type' to
5197 ;; record the ranges of types that only are found. Behaves otherwise
5198 ;; like `c-record-type-identifiers'.
5199 (defvar c-record-found-types nil)
5200
5201 (defmacro c-forward-keyword-prefixed-id (type)
5202 ;; Used internally in `c-forward-keyword-clause' to move forward
5203 ;; over a type (if TYPE is 'type) or a name (otherwise) which
5204 ;; possibly is prefixed by keywords and their associated clauses.
5205 ;; Try with a type/name first to not trip up on those that begin
5206 ;; with a keyword. Return t if a known or found type is moved
5207 ;; over. The point is clobbered if nil is returned. If range
5208 ;; recording is enabled, the identifier is recorded on as a type
5209 ;; if TYPE is 'type or as a reference if TYPE is 'ref.
5210 ;;
5211 ;; This macro might do hidden buffer changes.
5212 `(let (res)
5213 (while (if (setq res ,(if (eq type 'type)
5214 `(c-forward-type)
5215 `(c-forward-name)))
5216 nil
5217 (and (looking-at c-keywords-regexp)
5218 (c-forward-keyword-clause 1))))
5219 (when (memq res '(t known found prefix))
5220 ,(when (eq type 'ref)
5221 `(when c-record-type-identifiers
5222 (c-record-ref-id c-last-identifier-range)))
5223 t)))
5224
5225 (defmacro c-forward-id-comma-list (type update-safe-pos)
5226 ;; Used internally in `c-forward-keyword-clause' to move forward
5227 ;; over a comma separated list of types or names using
5228 ;; `c-forward-keyword-prefixed-id'.
5229 ;;
5230 ;; This macro might do hidden buffer changes.
5231 `(while (and (progn
5232 ,(when update-safe-pos
5233 `(setq safe-pos (point)))
5234 (eq (char-after) ?,))
5235 (progn
5236 (forward-char)
5237 (c-forward-syntactic-ws)
5238 (c-forward-keyword-prefixed-id ,type)))))
5239
5240 (defun c-forward-keyword-clause (match)
5241 ;; Submatch MATCH in the current match data is assumed to surround a
5242 ;; token. If it's a keyword, move over it and any immediately
5243 ;; following clauses associated with it, stopping at the start of
5244 ;; the next token. t is returned in that case, otherwise the point
5245 ;; stays and nil is returned. The kind of clauses that are
5246 ;; recognized are those specified by `c-type-list-kwds',
5247 ;; `c-ref-list-kwds', `c-colon-type-list-kwds',
5248 ;; `c-paren-nontype-kwds', `c-paren-type-kwds', `c-<>-type-kwds',
5249 ;; and `c-<>-arglist-kwds'.
5250 ;;
5251 ;; This function records identifier ranges on
5252 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
5253 ;; `c-record-type-identifiers' is non-nil.
5254 ;;
5255 ;; Note that for `c-colon-type-list-kwds', which doesn't necessary
5256 ;; apply directly after the keyword, the type list is moved over
5257 ;; only when there is no unaccounted token before it (i.e. a token
5258 ;; that isn't moved over due to some other keyword list). The
5259 ;; identifier ranges in the list are still recorded if that should
5260 ;; be done, though.
5261 ;;
5262 ;; This function might do hidden buffer changes.
5263
5264 (let ((kwd-sym (c-keyword-sym (match-string match))) safe-pos pos
5265 ;; The call to `c-forward-<>-arglist' below is made after
5266 ;; `c-<>-sexp-kwds' keywords, so we're certain they actually
5267 ;; are angle bracket arglists and `c-restricted-<>-arglists'
5268 ;; should therefore be nil.
5269 (c-parse-and-markup-<>-arglists t)
5270 c-restricted-<>-arglists)
5271
5272 (when kwd-sym
5273 (goto-char (match-end match))
5274 (c-forward-syntactic-ws)
5275 (setq safe-pos (point))
5276
5277 (cond
5278 ((and (c-keyword-member kwd-sym 'c-type-list-kwds)
5279 (c-forward-keyword-prefixed-id type))
5280 ;; There's a type directly after a keyword in `c-type-list-kwds'.
5281 (c-forward-id-comma-list type t))
5282
5283 ((and (c-keyword-member kwd-sym 'c-ref-list-kwds)
5284 (c-forward-keyword-prefixed-id ref))
5285 ;; There's a name directly after a keyword in `c-ref-list-kwds'.
5286 (c-forward-id-comma-list ref t))
5287
5288 ((and (c-keyword-member kwd-sym 'c-paren-any-kwds)
5289 (eq (char-after) ?\())
5290 ;; There's an open paren after a keyword in `c-paren-any-kwds'.
5291
5292 (forward-char)
5293 (when (and (setq pos (c-up-list-forward))
5294 (eq (char-before pos) ?\)))
5295 (when (and c-record-type-identifiers
5296 (c-keyword-member kwd-sym 'c-paren-type-kwds))
5297 ;; Use `c-forward-type' on every identifier we can find
5298 ;; inside the paren, to record the types.
5299 (while (c-syntactic-re-search-forward c-symbol-start pos t)
5300 (goto-char (match-beginning 0))
5301 (unless (c-forward-type)
5302 (looking-at c-symbol-key) ; Always matches.
5303 (goto-char (match-end 0)))))
5304
5305 (goto-char pos)
5306 (c-forward-syntactic-ws)
5307 (setq safe-pos (point))))
5308
5309 ((and (c-keyword-member kwd-sym 'c-<>-sexp-kwds)
5310 (eq (char-after) ?<)
5311 (c-forward-<>-arglist (c-keyword-member kwd-sym 'c-<>-type-kwds)))
5312 (c-forward-syntactic-ws)
5313 (setq safe-pos (point)))
5314
5315 ((and (c-keyword-member kwd-sym 'c-nonsymbol-sexp-kwds)
5316 (not (looking-at c-symbol-start))
5317 (c-safe (c-forward-sexp) t))
5318 (c-forward-syntactic-ws)
5319 (setq safe-pos (point))))
5320
5321 (when (c-keyword-member kwd-sym 'c-colon-type-list-kwds)
5322 (if (eq (char-after) ?:)
5323 ;; If we are at the colon already, we move over the type
5324 ;; list after it.
5325 (progn
5326 (forward-char)
5327 (c-forward-syntactic-ws)
5328 (when (c-forward-keyword-prefixed-id type)
5329 (c-forward-id-comma-list type t)))
5330 ;; Not at the colon, so stop here. But the identifier
5331 ;; ranges in the type list later on should still be
5332 ;; recorded.
5333 (and c-record-type-identifiers
5334 (progn
5335 ;; If a keyword matched both one of the types above and
5336 ;; this one, we match `c-colon-type-list-re' after the
5337 ;; clause matched above.
5338 (goto-char safe-pos)
5339 (looking-at c-colon-type-list-re))
5340 (progn
5341 (goto-char (match-end 0))
5342 (c-forward-syntactic-ws)
5343 (c-forward-keyword-prefixed-id type))
5344 ;; There's a type after the `c-colon-type-list-re' match
5345 ;; after a keyword in `c-colon-type-list-kwds'.
5346 (c-forward-id-comma-list type nil))))
5347
5348 (goto-char safe-pos)
5349 t)))
5350
5351 (defun c-forward-<>-arglist (all-types)
5352 ;; The point is assumed to be at a "<". Try to treat it as the open
5353 ;; paren of an angle bracket arglist and move forward to the
5354 ;; corresponding ">". If successful, the point is left after the
5355 ;; ">" and t is returned, otherwise the point isn't moved and nil is
5356 ;; returned. If ALL-TYPES is t then all encountered arguments in
5357 ;; the arglist that might be types are treated as found types.
5358 ;;
5359 ;; The variable `c-parse-and-markup-<>-arglists' controls how this
5360 ;; function handles text properties on the angle brackets and argument
5361 ;; separating commas.
5362 ;;
5363 ;; `c-restricted-<>-arglists' controls how lenient the template
5364 ;; arglist recognition should be.
5365 ;;
5366 ;; This function records identifier ranges on
5367 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
5368 ;; `c-record-type-identifiers' is non-nil.
5369 ;;
5370 ;; This function might do hidden buffer changes.
5371
5372 (let ((start (point))
5373 ;; If `c-record-type-identifiers' is set then activate
5374 ;; recording of any found types that constitute an argument in
5375 ;; the arglist.
5376 (c-record-found-types (if c-record-type-identifiers t)))
5377 (if (catch 'angle-bracket-arglist-escape
5378 (setq c-record-found-types
5379 (c-forward-<>-arglist-recur all-types)))
5380 (progn
5381 (when (consp c-record-found-types)
5382 (setq c-record-type-identifiers
5383 ;; `nconc' doesn't mind that the tail of
5384 ;; `c-record-found-types' is t.
5385 (nconc c-record-found-types c-record-type-identifiers)))
5386 t)
5387
5388 (goto-char start)
5389 nil)))
5390
5391 (defun c-forward-<>-arglist-recur (all-types)
5392 ;; Recursive part of `c-forward-<>-arglist'.
5393 ;;
5394 ;; This function might do hidden buffer changes.
5395
5396 (let ((start (point)) res pos tmp
5397 ;; Cover this so that any recorded found type ranges are
5398 ;; automatically lost if it turns out to not be an angle
5399 ;; bracket arglist. It's propagated through the return value
5400 ;; on successful completion.
5401 (c-record-found-types c-record-found-types)
5402 ;; List that collects the positions after the argument
5403 ;; separating ',' in the arglist.
5404 arg-start-pos)
5405
5406 ;; If the '<' has paren open syntax then we've marked it as an angle
5407 ;; bracket arglist before, so skip to the end.
5408 (if (and (not c-parse-and-markup-<>-arglists)
5409 (c-get-char-property (point) 'syntax-table))
5410
5411 (progn
5412 (forward-char)
5413 (if (and (c-go-up-list-forward)
5414 (eq (char-before) ?>))
5415 t
5416
5417 ;; Got unmatched paren angle brackets. We don't clear the paren
5418 ;; syntax properties and retry, on the basis that it's very
5419 ;; unlikely that paren angle brackets become operators by code
5420 ;; manipulation. It's far more likely that it doesn't match due
5421 ;; to narrowing or some temporary change.
5422 (goto-char start)
5423 nil))
5424
5425 (forward-char)
5426 (unless (looking-at c-<-op-cont-regexp)
5427 (while (and
5428 (progn
5429
5430 (when c-record-type-identifiers
5431 (if all-types
5432
5433 ;; All encountered identifiers are types, so set the
5434 ;; promote flag and parse the type.
5435 (progn
5436 (c-forward-syntactic-ws)
5437 (when (looking-at c-identifier-start)
5438 (let ((c-promote-possible-types t))
5439 (c-forward-type))))
5440
5441 ;; Check if this arglist argument is a sole type. If
5442 ;; it's known then it's recorded in
5443 ;; `c-record-type-identifiers'. If it only is found
5444 ;; then it's recorded in `c-record-found-types' which we
5445 ;; might roll back if it turns out that this isn't an
5446 ;; angle bracket arglist afterall.
5447 (when (memq (char-before) '(?, ?<))
5448 (let ((orig-record-found-types c-record-found-types))
5449 (c-forward-syntactic-ws)
5450 (and (memq (c-forward-type) '(known found))
5451 (not (looking-at "[,>]"))
5452 ;; A found type was recorded but it's not the
5453 ;; only thing in the arglist argument, so reset
5454 ;; `c-record-found-types'.
5455 (setq c-record-found-types
5456 orig-record-found-types))))))
5457
5458 (setq pos (point))
5459 (or (when (eq (char-after) ?>)
5460 ;; Must check for '>' at the very start separately,
5461 ;; since the regexp below has to avoid ">>" without
5462 ;; using \\=.
5463 (forward-char)
5464 t)
5465
5466 ;; Note: These regexps exploit the match order in \| so
5467 ;; that "<>" is matched by "<" rather than "[^>:-]>".
5468 (c-syntactic-re-search-forward
5469 (if c-restricted-<>-arglists
5470 ;; Stop on ',', '|', '&', '+' and '-' to catch
5471 ;; common binary operators that could be between
5472 ;; two comparison expressions "a<b" and "c>d".
5473 "[<;{},|&+-]\\|\\([^>:-]>\\)"
5474 ;; Otherwise we still stop on ',' to find the
5475 ;; argument start positions.
5476 "[<;{},]\\|\\([^>:-]>\\)")
5477 nil 'move t t 1)
5478
5479 ;; If the arglist starter has lost its open paren
5480 ;; syntax but not the closer, we won't find the
5481 ;; closer above since we only search in the
5482 ;; balanced sexp. In that case we stop just short
5483 ;; of it so check if the following char is the closer.
5484 (when (eq (char-after) ?>)
5485 (forward-char)
5486 t)))
5487
5488 (cond
5489 ((eq (char-before) ?>)
5490 ;; Either an operator starting with '>' or the end of
5491 ;; the angle bracket arglist.
5492
5493 (if (looking-at c->-op-cont-regexp)
5494 (progn
5495 (goto-char (match-end 0))
5496 t) ; Continue the loop.
5497
5498 ;; The angle bracket arglist is finished.
5499 (when c-parse-and-markup-<>-arglists
5500 (while arg-start-pos
5501 (c-put-c-type-property (1- (car arg-start-pos))
5502 'c-<>-arg-sep)
5503 (setq arg-start-pos (cdr arg-start-pos)))
5504 (c-mark-<-as-paren start)
5505 (c-mark->-as-paren (1- (point))))
5506 (setq res t)
5507 nil)) ; Exit the loop.
5508
5509 ((eq (char-before) ?<)
5510 ;; Either an operator starting with '<' or a nested arglist.
5511
5512 (setq pos (point))
5513 (let (id-start id-end subres keyword-match)
5514 (if (if (looking-at c-<-op-cont-regexp)
5515 (setq tmp (match-end 0))
5516 (setq tmp pos)
5517 (backward-char)
5518 (not
5519 (and
5520
5521 (save-excursion
5522 ;; There's always an identifier before an angle
5523 ;; bracket arglist, or a keyword in
5524 ;; `c-<>-type-kwds' or `c-<>-arglist-kwds'.
5525 (c-backward-syntactic-ws)
5526 (setq id-end (point))
5527 (c-simple-skip-symbol-backward)
5528 (when (or (setq keyword-match
5529 (looking-at c-opt-<>-sexp-key))
5530 (not (looking-at c-keywords-regexp)))
5531 (setq id-start (point))))
5532
5533 (setq subres
5534 (let ((c-record-type-identifiers nil)
5535 (c-record-found-types nil))
5536 (c-forward-<>-arglist-recur
5537 (and keyword-match
5538 (c-keyword-member
5539 (c-keyword-sym (match-string 1))
5540 'c-<>-type-kwds)))))
5541 )))
5542
5543 ;; It was not an angle bracket arglist.
5544 (goto-char tmp)
5545
5546 ;; It was an angle bracket arglist.
5547 (setq c-record-found-types subres)
5548
5549 ;; Record the identifier before the template as a type
5550 ;; or reference depending on whether the arglist is last
5551 ;; in a qualified identifier.
5552 (when (and c-record-type-identifiers
5553 (not keyword-match))
5554 (if (and c-opt-identifier-concat-key
5555 (progn
5556 (c-forward-syntactic-ws)
5557 (looking-at c-opt-identifier-concat-key)))
5558 (c-record-ref-id (cons id-start id-end))
5559 (c-record-type-id (cons id-start id-end))))))
5560 t)
5561
5562 ((and (eq (char-before) ?,)
5563 (not c-restricted-<>-arglists))
5564 ;; Just another argument. Record the position. The
5565 ;; type check stuff that made us stop at it is at
5566 ;; the top of the loop.
5567 (setq arg-start-pos (cons (point) arg-start-pos)))
5568
5569 (t
5570 ;; Got a character that can't be in an angle bracket
5571 ;; arglist argument. Abort using `throw', since
5572 ;; it's useless to try to find a surrounding arglist
5573 ;; if we're nested.
5574 (throw 'angle-bracket-arglist-escape nil))))))
5575
5576 (if res
5577 (or c-record-found-types t)))))
5578
5579 (defun c-backward-<>-arglist (all-types &optional limit)
5580 ;; The point is assumed to be directly after a ">". Try to treat it
5581 ;; as the close paren of an angle bracket arglist and move back to
5582 ;; the corresponding "<". If successful, the point is left at
5583 ;; the "<" and t is returned, otherwise the point isn't moved and
5584 ;; nil is returned. ALL-TYPES is passed on to
5585 ;; `c-forward-<>-arglist'.
5586 ;;
5587 ;; If the optional LIMIT is given, it bounds the backward search.
5588 ;; It's then assumed to be at a syntactically relevant position.
5589 ;;
5590 ;; This is a wrapper around `c-forward-<>-arglist'. See that
5591 ;; function for more details.
5592
5593 (let ((start (point)))
5594 (backward-char)
5595 (if (and (not c-parse-and-markup-<>-arglists)
5596 (c-get-char-property (point) 'syntax-table))
5597
5598 (if (and (c-go-up-list-backward)
5599 (eq (char-after) ?<))
5600 t
5601 ;; See corresponding note in `c-forward-<>-arglist'.
5602 (goto-char start)
5603 nil)
5604
5605 (while (progn
5606 (c-syntactic-skip-backward "^<;{}" limit t)
5607
5608 (and
5609 (if (eq (char-before) ?<)
5610 t
5611 ;; Stopped at bob or a char that isn't allowed in an
5612 ;; arglist, so we've failed.
5613 (goto-char start)
5614 nil)
5615
5616 (if (> (point)
5617 (progn (c-beginning-of-current-token)
5618 (point)))
5619 ;; If we moved then the "<" was part of some
5620 ;; multicharacter token.
5621 t
5622
5623 (backward-char)
5624 (let ((beg-pos (point)))
5625 (if (c-forward-<>-arglist all-types)
5626 (cond ((= (point) start)
5627 ;; Matched the arglist. Break the while.
5628 (goto-char beg-pos)
5629 nil)
5630 ((> (point) start)
5631 ;; We started from a non-paren ">" inside an
5632 ;; arglist.
5633 (goto-char start)
5634 nil)
5635 (t
5636 ;; Matched a shorter arglist. Can be a nested
5637 ;; one so continue looking.
5638 (goto-char beg-pos)
5639 t))
5640 t))))))
5641
5642 (/= (point) start))))
5643
5644 (defun c-forward-name ()
5645 ;; Move forward over a complete name if at the beginning of one,
5646 ;; stopping at the next following token. If the point is not at
5647 ;; something that are recognized as name then it stays put. A name
5648 ;; could be something as simple as "foo" in C or something as
5649 ;; complex as "X<Y<class A<int>::B, BIT_MAX >> b>, ::operator<> ::
5650 ;; Z<(a>b)> :: operator const X<&foo>::T Q::G<unsigned short
5651 ;; int>::*volatile const" in C++ (this function is actually little
5652 ;; more than a `looking-at' call in all modes except those that,
5653 ;; like C++, have `c-recognize-<>-arglists' set). Return nil if no
5654 ;; name is found, 'template if it's an identifier ending with an
5655 ;; angle bracket arglist, 'operator of it's an operator identifier,
5656 ;; or t if it's some other kind of name.
5657 ;;
5658 ;; This function records identifier ranges on
5659 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
5660 ;; `c-record-type-identifiers' is non-nil.
5661 ;;
5662 ;; This function might do hidden buffer changes.
5663
5664 (let ((pos (point)) (start (point)) res id-start id-end
5665 ;; Turn off `c-promote-possible-types' here since we might
5666 ;; call `c-forward-<>-arglist' and we don't want it to promote
5667 ;; every suspect thing in the arglist to a type. We're
5668 ;; typically called from `c-forward-type' in this case, and
5669 ;; the caller only wants the top level type that it finds to
5670 ;; be promoted.
5671 c-promote-possible-types)
5672 (while
5673 (and
5674 (looking-at c-identifier-key)
5675
5676 (progn
5677 ;; Check for keyword. We go to the last symbol in
5678 ;; `c-identifier-key' first.
5679 (goto-char (setq id-end (match-end 0)))
5680 (c-simple-skip-symbol-backward)
5681 (setq id-start (point))
5682
5683 (if (looking-at c-keywords-regexp)
5684 (when (and (c-major-mode-is 'c++-mode)
5685 (looking-at
5686 (cc-eval-when-compile
5687 (concat "\\(operator\\|\\(template\\)\\)"
5688 "\\(" (c-lang-const c-nonsymbol-key c++)
5689 "\\|$\\)")))
5690 (if (match-beginning 2)
5691 ;; "template" is only valid inside an
5692 ;; identifier if preceded by "::".
5693 (save-excursion
5694 (c-backward-syntactic-ws)
5695 (and (c-safe (backward-char 2) t)
5696 (looking-at "::")))
5697 t))
5698
5699 ;; Handle a C++ operator or template identifier.
5700 (goto-char id-end)
5701 (c-forward-syntactic-ws)
5702 (cond ((eq (char-before id-end) ?e)
5703 ;; Got "... ::template".
5704 (let ((subres (c-forward-name)))
5705 (when subres
5706 (setq pos (point)
5707 res subres))))
5708
5709 ((looking-at c-identifier-start)
5710 ;; Got a cast operator.
5711 (when (c-forward-type)
5712 (setq pos (point)
5713 res 'operator)
5714 ;; Now we should match a sequence of either
5715 ;; '*', '&' or a name followed by ":: *",
5716 ;; where each can be followed by a sequence
5717 ;; of `c-opt-type-modifier-key'.
5718 (while (cond ((looking-at "[*&]")
5719 (goto-char (match-end 0))
5720 t)
5721 ((looking-at c-identifier-start)
5722 (and (c-forward-name)
5723 (looking-at "::")
5724 (progn
5725 (goto-char (match-end 0))
5726 (c-forward-syntactic-ws)
5727 (eq (char-after) ?*))
5728 (progn
5729 (forward-char)
5730 t))))
5731 (while (progn
5732 (c-forward-syntactic-ws)
5733 (setq pos (point))
5734 (looking-at c-opt-type-modifier-key))
5735 (goto-char (match-end 1))))))
5736
5737 ((looking-at c-overloadable-operators-regexp)
5738 ;; Got some other operator.
5739 (setq c-last-identifier-range
5740 (cons (point) (match-end 0)))
5741 (goto-char (match-end 0))
5742 (c-forward-syntactic-ws)
5743 (setq pos (point)
5744 res 'operator)))
5745
5746 nil)
5747
5748 ;; `id-start' is equal to `id-end' if we've jumped over
5749 ;; an identifier that doesn't end with a symbol token.
5750 ;; That can occur e.g. for Java import directives on the
5751 ;; form "foo.bar.*".
5752 (when (and id-start (/= id-start id-end))
5753 (setq c-last-identifier-range
5754 (cons id-start id-end)))
5755 (goto-char id-end)
5756 (c-forward-syntactic-ws)
5757 (setq pos (point)
5758 res t)))
5759
5760 (progn
5761 (goto-char pos)
5762 (when (or c-opt-identifier-concat-key
5763 c-recognize-<>-arglists)
5764
5765 (cond
5766 ((and c-opt-identifier-concat-key
5767 (looking-at c-opt-identifier-concat-key))
5768 ;; Got a concatenated identifier. This handles the
5769 ;; cases with tricky syntactic whitespace that aren't
5770 ;; covered in `c-identifier-key'.
5771 (goto-char (match-end 0))
5772 (c-forward-syntactic-ws)
5773 t)
5774
5775 ((and c-recognize-<>-arglists
5776 (eq (char-after) ?<))
5777 ;; Maybe an angle bracket arglist.
5778
5779 (when (let (c-record-type-identifiers
5780 c-record-found-types)
5781 (c-forward-<>-arglist nil))
5782
5783 (c-add-type start (1+ pos))
5784 (c-forward-syntactic-ws)
5785 (setq pos (point)
5786 c-last-identifier-range nil)
5787
5788 (if (and c-opt-identifier-concat-key
5789 (looking-at c-opt-identifier-concat-key))
5790
5791 ;; Continue if there's an identifier concatenation
5792 ;; operator after the template argument.
5793 (progn
5794 (when (and c-record-type-identifiers id-start)
5795 (c-record-ref-id (cons id-start id-end)))
5796 (forward-char 2)
5797 (c-forward-syntactic-ws)
5798 t)
5799
5800 (when (and c-record-type-identifiers id-start)
5801 (c-record-type-id (cons id-start id-end)))
5802 (setq res 'template)
5803 nil)))
5804 )))))
5805
5806 (goto-char pos)
5807 res))
5808
5809 (defun c-forward-type ()
5810 ;; Move forward over a type spec if at the beginning of one,
5811 ;; stopping at the next following token. Return t if it's a known
5812 ;; type that can't be a name or other expression, 'known if it's an
5813 ;; otherwise known type (according to `*-font-lock-extra-types'),
5814 ;; 'prefix if it's a known prefix of a type, 'found if it's a type
5815 ;; that matches one in `c-found-types', 'maybe if it's an identfier
5816 ;; that might be a type, or nil if it can't be a type (the point
5817 ;; isn't moved then). The point is assumed to be at the beginning
5818 ;; of a token.
5819 ;;
5820 ;; Note that this function doesn't skip past the brace definition
5821 ;; that might be considered part of the type, e.g.
5822 ;; "enum {a, b, c} foo".
5823 ;;
5824 ;; This function records identifier ranges on
5825 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
5826 ;; `c-record-type-identifiers' is non-nil.
5827 ;;
5828 ;; This function might do hidden buffer changes.
5829
5830 (let ((start (point)) pos res name-res id-start id-end id-range)
5831
5832 ;; Skip leading type modifiers. If any are found we know it's a
5833 ;; prefix of a type.
5834 (when c-opt-type-modifier-key
5835 (while (looking-at c-opt-type-modifier-key)
5836 (goto-char (match-end 1))
5837 (c-forward-syntactic-ws)
5838 (setq res 'prefix)))
5839
5840 (cond
5841 ((looking-at c-type-prefix-key)
5842 ;; Looking at a keyword that prefixes a type identifier,
5843 ;; e.g. "class".
5844 (goto-char (match-end 1))
5845 (c-forward-syntactic-ws)
5846 (setq pos (point))
5847 (if (memq (setq name-res (c-forward-name)) '(t template))
5848 (progn
5849 (when (eq name-res t)
5850 ;; In many languages the name can be used without the
5851 ;; prefix, so we add it to `c-found-types'.
5852 (c-add-type pos (point))
5853 (when (and c-record-type-identifiers
5854 c-last-identifier-range)
5855 (c-record-type-id c-last-identifier-range)))
5856 (setq res t))
5857 ;; Invalid syntax.
5858 (goto-char start)
5859 (setq res nil)))
5860
5861 ((progn
5862 (setq pos nil)
5863 (if (looking-at c-identifier-start)
5864 (save-excursion
5865 (setq id-start (point)
5866 name-res (c-forward-name))
5867 (when name-res
5868 (setq id-end (point)
5869 id-range c-last-identifier-range))))
5870 (and (cond ((looking-at c-primitive-type-key)
5871 (setq res t))
5872 ((c-with-syntax-table c-identifier-syntax-table
5873 (looking-at c-known-type-key))
5874 (setq res 'known)))
5875 (or (not id-end)
5876 (>= (save-excursion
5877 (save-match-data
5878 (goto-char (match-end 1))
5879 (c-forward-syntactic-ws)
5880 (setq pos (point))))
5881 id-end)
5882 (setq res nil))))
5883 ;; Looking at a primitive or known type identifier. We've
5884 ;; checked for a name first so that we don't go here if the
5885 ;; known type match only is a prefix of another name.
5886
5887 (setq id-end (match-end 1))
5888
5889 (when (and c-record-type-identifiers
5890 (or c-promote-possible-types (eq res t)))
5891 (c-record-type-id (cons (match-beginning 1) (match-end 1))))
5892
5893 (if (and c-opt-type-component-key
5894 (save-match-data
5895 (looking-at c-opt-type-component-key)))
5896 ;; There might be more keywords for the type.
5897 (let (safe-pos)
5898 (c-forward-keyword-clause 1)
5899 (while (progn
5900 (setq safe-pos (point))
5901 (looking-at c-opt-type-component-key))
5902 (when (and c-record-type-identifiers
5903 (looking-at c-primitive-type-key))
5904 (c-record-type-id (cons (match-beginning 1)
5905 (match-end 1))))
5906 (c-forward-keyword-clause 1))
5907 (if (looking-at c-primitive-type-key)
5908 (progn
5909 (when c-record-type-identifiers
5910 (c-record-type-id (cons (match-beginning 1)
5911 (match-end 1))))
5912 (c-forward-keyword-clause 1)
5913 (setq res t))
5914 (goto-char safe-pos)
5915 (setq res 'prefix)))
5916 (unless (save-match-data (c-forward-keyword-clause 1))
5917 (if pos
5918 (goto-char pos)
5919 (goto-char (match-end 1))
5920 (c-forward-syntactic-ws)))))
5921
5922 (name-res
5923 (cond ((eq name-res t)
5924 ;; A normal identifier.
5925 (goto-char id-end)
5926 (if (or res c-promote-possible-types)
5927 (progn
5928 (c-add-type id-start id-end)
5929 (when (and c-record-type-identifiers id-range)
5930 (c-record-type-id id-range))
5931 (unless res
5932 (setq res 'found)))
5933 (setq res (if (c-check-type id-start id-end)
5934 ;; It's an identifier that has been used as
5935 ;; a type somewhere else.
5936 'found
5937 ;; It's an identifier that might be a type.
5938 'maybe))))
5939 ((eq name-res 'template)
5940 ;; A template is a type.
5941 (goto-char id-end)
5942 (setq res t))
5943 (t
5944 ;; Otherwise it's an operator identifier, which is not a type.
5945 (goto-char start)
5946 (setq res nil)))))
5947
5948 (when res
5949 ;; Skip trailing type modifiers. If any are found we know it's
5950 ;; a type.
5951 (when c-opt-type-modifier-key
5952 (while (looking-at c-opt-type-modifier-key)
5953 (goto-char (match-end 1))
5954 (c-forward-syntactic-ws)
5955 (setq res t)))
5956
5957 ;; Step over any type suffix operator. Do not let the existence
5958 ;; of these alter the classification of the found type, since
5959 ;; these operators typically are allowed in normal expressions
5960 ;; too.
5961 (when c-opt-type-suffix-key
5962 (while (looking-at c-opt-type-suffix-key)
5963 (goto-char (match-end 1))
5964 (c-forward-syntactic-ws)))
5965
5966 (when c-opt-type-concat-key
5967 ;; Look for a trailing operator that concatenates the type
5968 ;; with a following one, and if so step past that one through
5969 ;; a recursive call. Note that we don't record concatenated
5970 ;; types in `c-found-types' - it's the component types that
5971 ;; are recorded when appropriate.
5972 (setq pos (point))
5973 (let* ((c-promote-possible-types (or (memq res '(t known))
5974 c-promote-possible-types))
5975 ;; If we can't promote then set `c-record-found-types' so that
5976 ;; we can merge in the types from the second part afterwards if
5977 ;; it turns out to be a known type there.
5978 (c-record-found-types (and c-record-type-identifiers
5979 (not c-promote-possible-types)))
5980 subres)
5981 (if (and (looking-at c-opt-type-concat-key)
5982
5983 (progn
5984 (goto-char (match-end 1))
5985 (c-forward-syntactic-ws)
5986 (setq subres (c-forward-type))))
5987
5988 (progn
5989 ;; If either operand certainly is a type then both are, but we
5990 ;; don't let the existence of the operator itself promote two
5991 ;; uncertain types to a certain one.
5992 (cond ((eq res t))
5993 ((eq subres t)
5994 (unless (eq name-res 'template)
5995 (c-add-type id-start id-end))
5996 (when (and c-record-type-identifiers id-range)
5997 (c-record-type-id id-range))
5998 (setq res t))
5999 ((eq res 'known))
6000 ((eq subres 'known)
6001 (setq res 'known))
6002 ((eq res 'found))
6003 ((eq subres 'found)
6004 (setq res 'found))
6005 (t
6006 (setq res 'maybe)))
6007
6008 (when (and (eq res t)
6009 (consp c-record-found-types))
6010 ;; Merge in the ranges of any types found by the second
6011 ;; `c-forward-type'.
6012 (setq c-record-type-identifiers
6013 ;; `nconc' doesn't mind that the tail of
6014 ;; `c-record-found-types' is t.
6015 (nconc c-record-found-types
6016 c-record-type-identifiers))))
6017
6018 (goto-char pos))))
6019
6020 (when (and c-record-found-types (memq res '(known found)) id-range)
6021 (setq c-record-found-types
6022 (cons id-range c-record-found-types))))
6023
6024 ;;(message "c-forward-type %s -> %s: %s" start (point) res)
6025
6026 res))
6027
6028 \f
6029 ;; Handling of large scale constructs like statements and declarations.
6030
6031 ;; Macro used inside `c-forward-decl-or-cast-1'. It ought to be a
6032 ;; defsubst or perhaps even a defun, but it contains lots of free
6033 ;; variables that refer to things inside `c-forward-decl-or-cast-1'.
6034 (defmacro c-fdoc-shift-type-backward (&optional short)
6035 ;; `c-forward-decl-or-cast-1' can consume an arbitrary length list
6036 ;; of types when parsing a declaration, which means that it
6037 ;; sometimes consumes the identifier in the declaration as a type.
6038 ;; This is used to "backtrack" and make the last type be treated as
6039 ;; an identifier instead.
6040 `(progn
6041 ,(unless short
6042 ;; These identifiers are bound only in the inner let.
6043 '(setq identifier-type at-type
6044 identifier-start type-start
6045 got-parens nil
6046 got-identifier t
6047 got-suffix t
6048 got-suffix-after-parens id-start
6049 paren-depth 0))
6050
6051 (if (setq at-type (if (eq backup-at-type 'prefix)
6052 t
6053 backup-at-type))
6054 (setq type-start backup-type-start
6055 id-start backup-id-start)
6056 (setq type-start start-pos
6057 id-start start-pos))
6058
6059 ;; When these flags already are set we've found specifiers that
6060 ;; unconditionally signal these attributes - backtracking doesn't
6061 ;; change that. So keep them set in that case.
6062 (or at-type-decl
6063 (setq at-type-decl backup-at-type-decl))
6064 (or maybe-typeless
6065 (setq maybe-typeless backup-maybe-typeless))
6066
6067 ,(unless short
6068 ;; This identifier is bound only in the inner let.
6069 '(setq start id-start))))
6070
6071 (defun c-forward-decl-or-cast-1 (preceding-token-end context last-cast-end)
6072 ;; Move forward over a declaration or a cast if at the start of one.
6073 ;; The point is assumed to be at the start of some token. Nil is
6074 ;; returned if no declaration or cast is recognized, and the point
6075 ;; is clobbered in that case.
6076 ;;
6077 ;; If a declaration is parsed:
6078 ;;
6079 ;; The point is left at the first token after the first complete
6080 ;; declarator, if there is one. The return value is a cons where
6081 ;; the car is the position of the first token in the declarator. (See
6082 ;; below for the cdr.)
6083 ;; Some examples:
6084 ;;
6085 ;; void foo (int a, char *b) stuff ...
6086 ;; car ^ ^ point
6087 ;; float (*a)[], b;
6088 ;; car ^ ^ point
6089 ;; unsigned int a = c_style_initializer, b;
6090 ;; car ^ ^ point
6091 ;; unsigned int a (cplusplus_style_initializer), b;
6092 ;; car ^ ^ point (might change)
6093 ;; class Foo : public Bar {}
6094 ;; car ^ ^ point
6095 ;; class PikeClass (int a, string b) stuff ...
6096 ;; car ^ ^ point
6097 ;; enum bool;
6098 ;; car ^ ^ point
6099 ;; enum bool flag;
6100 ;; car ^ ^ point
6101 ;; void cplusplus_function (int x) throw (Bad);
6102 ;; car ^ ^ point
6103 ;; Foo::Foo (int b) : Base (b) {}
6104 ;; car ^ ^ point
6105 ;;
6106 ;; The cdr of the return value is non-nil iff a `c-typedef-decl-kwds'
6107 ;; specifier (e.g. class, struct, enum, typedef) is found in the
6108 ;; declaration, i.e. the declared identifier(s) are types.
6109 ;;
6110 ;; If a cast is parsed:
6111 ;;
6112 ;; The point is left at the first token after the closing paren of
6113 ;; the cast. The return value is `cast'. Note that the start
6114 ;; position must be at the first token inside the cast parenthesis
6115 ;; to recognize it.
6116 ;;
6117 ;; PRECEDING-TOKEN-END is the first position after the preceding
6118 ;; token, i.e. on the other side of the syntactic ws from the point.
6119 ;; Use a value less than or equal to (point-min) if the point is at
6120 ;; the first token in (the visible part of) the buffer.
6121 ;;
6122 ;; CONTEXT is a symbol that describes the context at the point:
6123 ;; 'decl In a comma-separated declaration context (typically
6124 ;; inside a function declaration arglist).
6125 ;; '<> In an angle bracket arglist.
6126 ;; 'arglist Some other type of arglist.
6127 ;; nil Some other context or unknown context. Includes
6128 ;; within the parens of an if, for, ... construct.
6129 ;;
6130 ;; LAST-CAST-END is the first token after the closing paren of a
6131 ;; preceding cast, or nil if none is known. If
6132 ;; `c-forward-decl-or-cast-1' is used in succession, it should be
6133 ;; the position after the closest preceding call where a cast was
6134 ;; matched. In that case it's used to discover chains of casts like
6135 ;; "(a) (b) c".
6136 ;;
6137 ;; This function records identifier ranges on
6138 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
6139 ;; `c-record-type-identifiers' is non-nil.
6140 ;;
6141 ;; This function might do hidden buffer changes.
6142
6143 (let (;; `start-pos' is used below to point to the start of the
6144 ;; first type, i.e. after any leading specifiers. It might
6145 ;; also point at the beginning of the preceding syntactic
6146 ;; whitespace.
6147 (start-pos (point))
6148 ;; Set to the result of `c-forward-type'.
6149 at-type
6150 ;; The position of the first token in what we currently
6151 ;; believe is the type in the declaration or cast, after any
6152 ;; specifiers and their associated clauses.
6153 type-start
6154 ;; The position of the first token in what we currently
6155 ;; believe is the declarator for the first identifier. Set
6156 ;; when the type is found, and moved forward over any
6157 ;; `c-decl-hangon-kwds' and their associated clauses that
6158 ;; occurs after the type.
6159 id-start
6160 ;; These store `at-type', `type-start' and `id-start' of the
6161 ;; identifier before the one in those variables. The previous
6162 ;; identifier might turn out to be the real type in a
6163 ;; declaration if the last one has to be the declarator in it.
6164 ;; If `backup-at-type' is nil then the other variables have
6165 ;; undefined values.
6166 backup-at-type backup-type-start backup-id-start
6167 ;; Set if we've found a specifier that makes the defined
6168 ;; identifier(s) types.
6169 at-type-decl
6170 ;; Set if we've found a specifier that can start a declaration
6171 ;; where there's no type.
6172 maybe-typeless
6173 ;; If a specifier is found that also can be a type prefix,
6174 ;; these flags are set instead of those above. If we need to
6175 ;; back up an identifier, they are copied to the real flag
6176 ;; variables. Thus they only take effect if we fail to
6177 ;; interpret it as a type.
6178 backup-at-type-decl backup-maybe-typeless
6179 ;; Whether we've found a declaration or a cast. We might know
6180 ;; this before we've found the type in it. It's 'ids if we've
6181 ;; found two consecutive identifiers (usually a sure sign, but
6182 ;; we should allow that in labels too), and t if we've found a
6183 ;; specifier keyword (a 100% sure sign).
6184 at-decl-or-cast
6185 ;; Set when we need to back up to parse this as a declaration
6186 ;; but not as a cast.
6187 backup-if-not-cast
6188 ;; For casts, the return position.
6189 cast-end
6190 ;; Save `c-record-type-identifiers' and
6191 ;; `c-record-ref-identifiers' since ranges are recorded
6192 ;; speculatively and should be thrown away if it turns out
6193 ;; that it isn't a declaration or cast.
6194 (save-rec-type-ids c-record-type-identifiers)
6195 (save-rec-ref-ids c-record-ref-identifiers))
6196
6197 ;; Check for a type. Unknown symbols are treated as possible
6198 ;; types, but they could also be specifiers disguised through
6199 ;; macros like __INLINE__, so we recognize both types and known
6200 ;; specifiers after them too.
6201 (while
6202 (let* ((start (point)) kwd-sym kwd-clause-end found-type)
6203
6204 ;; Look for a specifier keyword clause.
6205 (when (looking-at c-prefix-spec-kwds-re)
6206 (setq kwd-sym (c-keyword-sym (match-string 1)))
6207 (save-excursion
6208 (c-forward-keyword-clause 1)
6209 (setq kwd-clause-end (point))))
6210
6211 (when (setq found-type (c-forward-type))
6212 ;; Found a known or possible type or a prefix of a known type.
6213
6214 (when at-type
6215 ;; Got two identifiers with nothing but whitespace
6216 ;; between them. That can only happen in declarations.
6217 (setq at-decl-or-cast 'ids)
6218
6219 (when (eq at-type 'found)
6220 ;; If the previous identifier is a found type we
6221 ;; record it as a real one; it might be some sort of
6222 ;; alias for a prefix like "unsigned".
6223 (save-excursion
6224 (goto-char type-start)
6225 (let ((c-promote-possible-types t))
6226 (c-forward-type)))))
6227
6228 (setq backup-at-type at-type
6229 backup-type-start type-start
6230 backup-id-start id-start
6231 at-type found-type
6232 type-start start
6233 id-start (point)
6234 ;; The previous ambiguous specifier/type turned out
6235 ;; to be a type since we've parsed another one after
6236 ;; it, so clear these backup flags.
6237 backup-at-type-decl nil
6238 backup-maybe-typeless nil))
6239
6240 (if kwd-sym
6241 (progn
6242 ;; Handle known specifier keywords and
6243 ;; `c-decl-hangon-kwds' which can occur after known
6244 ;; types.
6245
6246 (if (c-keyword-member kwd-sym 'c-decl-hangon-kwds)
6247 ;; It's a hang-on keyword that can occur anywhere.
6248 (progn
6249 (setq at-decl-or-cast t)
6250 (if at-type
6251 ;; Move the identifier start position if
6252 ;; we've passed a type.
6253 (setq id-start kwd-clause-end)
6254 ;; Otherwise treat this as a specifier and
6255 ;; move the fallback position.
6256 (setq start-pos kwd-clause-end))
6257 (goto-char kwd-clause-end))
6258
6259 ;; It's an ordinary specifier so we know that
6260 ;; anything before this can't be the type.
6261 (setq backup-at-type nil
6262 start-pos kwd-clause-end)
6263
6264 (if found-type
6265 ;; It's ambiguous whether this keyword is a
6266 ;; specifier or a type prefix, so set the backup
6267 ;; flags. (It's assumed that `c-forward-type'
6268 ;; moved further than `c-forward-keyword-clause'.)
6269 (progn
6270 (when (c-keyword-member kwd-sym 'c-typedef-decl-kwds)
6271 (setq backup-at-type-decl t))
6272 (when (c-keyword-member kwd-sym 'c-typeless-decl-kwds)
6273 (setq backup-maybe-typeless t)))
6274
6275 (when (c-keyword-member kwd-sym 'c-typedef-decl-kwds)
6276 (setq at-type-decl t))
6277 (when (c-keyword-member kwd-sym 'c-typeless-decl-kwds)
6278 (setq maybe-typeless t))
6279
6280 ;; Haven't matched a type so it's an umambiguous
6281 ;; specifier keyword and we know we're in a
6282 ;; declaration.
6283 (setq at-decl-or-cast t)
6284
6285 (goto-char kwd-clause-end))))
6286
6287 ;; If the type isn't known we continue so that we'll jump
6288 ;; over all specifiers and type identifiers. The reason
6289 ;; to do this for a known type prefix is to make things
6290 ;; like "unsigned INT16" work.
6291 (and found-type (not (eq found-type t))))))
6292
6293 (cond
6294 ((eq at-type t)
6295 ;; If a known type was found, we still need to skip over any
6296 ;; hangon keyword clauses after it. Otherwise it has already
6297 ;; been done in the loop above.
6298 (while (looking-at c-decl-hangon-key)
6299 (c-forward-keyword-clause 1))
6300 (setq id-start (point)))
6301
6302 ((eq at-type 'prefix)
6303 ;; A prefix type is itself a primitive type when it's not
6304 ;; followed by another type.
6305 (setq at-type t))
6306
6307 ((not at-type)
6308 ;; Got no type but set things up to continue anyway to handle
6309 ;; the various cases when a declaration doesn't start with a
6310 ;; type.
6311 (setq id-start start-pos))
6312
6313 ((and (eq at-type 'maybe)
6314 (c-major-mode-is 'c++-mode))
6315 ;; If it's C++ then check if the last "type" ends on the form
6316 ;; "foo::foo" or "foo::~foo", i.e. if it's the name of a
6317 ;; (con|de)structor.
6318 (save-excursion
6319 (let (name end-2 end-1)
6320 (goto-char id-start)
6321 (c-backward-syntactic-ws)
6322 (setq end-2 (point))
6323 (when (and
6324 (c-simple-skip-symbol-backward)
6325 (progn
6326 (setq name
6327 (buffer-substring-no-properties (point) end-2))
6328 ;; Cheating in the handling of syntactic ws below.
6329 (< (skip-chars-backward ":~ \t\n\r\v\f") 0))
6330 (progn
6331 (setq end-1 (point))
6332 (c-simple-skip-symbol-backward))
6333 (>= (point) type-start)
6334 (equal (buffer-substring-no-properties (point) end-1)
6335 name))
6336 ;; It is a (con|de)structor name. In that case the
6337 ;; declaration is typeless so zap out any preceding
6338 ;; identifier(s) that we might have taken as types.
6339 (goto-char type-start)
6340 (setq at-type nil
6341 backup-at-type nil
6342 id-start type-start))))))
6343
6344 ;; Check for and step over a type decl expression after the thing
6345 ;; that is or might be a type. This can't be skipped since we
6346 ;; need the correct end position of the declarator for
6347 ;; `max-type-decl-end-*'.
6348 (let ((start (point)) (paren-depth 0) pos
6349 ;; True if there's a non-open-paren match of
6350 ;; `c-type-decl-prefix-key'.
6351 got-prefix
6352 ;; True if the declarator is surrounded by a parenthesis pair.
6353 got-parens
6354 ;; True if there is an identifier in the declarator.
6355 got-identifier
6356 ;; True if there's a non-close-paren match of
6357 ;; `c-type-decl-suffix-key'.
6358 got-suffix
6359 ;; True if there's a prefix match outside the outermost
6360 ;; paren pair that surrounds the declarator.
6361 got-prefix-before-parens
6362 ;; True if there's a suffix match outside the outermost
6363 ;; paren pair that surrounds the declarator. The value is
6364 ;; the position of the first suffix match.
6365 got-suffix-after-parens
6366 ;; True if we've parsed the type decl to a token that is
6367 ;; known to end declarations in this context.
6368 at-decl-end
6369 ;; The earlier values of `at-type' and `type-start' if we've
6370 ;; shifted the type backwards.
6371 identifier-type identifier-start
6372 ;; If `c-parse-and-markup-<>-arglists' is set we need to
6373 ;; turn it off during the name skipping below to avoid
6374 ;; getting `c-type' properties that might be bogus. That
6375 ;; can happen since we don't know if
6376 ;; `c-restricted-<>-arglists' will be correct inside the
6377 ;; arglist paren that gets entered.
6378 c-parse-and-markup-<>-arglists)
6379
6380 (goto-char id-start)
6381
6382 ;; Skip over type decl prefix operators. (Note similar code in
6383 ;; `c-font-lock-declarators'.)
6384 (while (and (looking-at c-type-decl-prefix-key)
6385 (if (and (c-major-mode-is 'c++-mode)
6386 (match-beginning 2))
6387 ;; If the second submatch matches in C++ then
6388 ;; we're looking at an identifier that's a
6389 ;; prefix only if it specifies a member pointer.
6390 (when (setq got-identifier (c-forward-name))
6391 (if (looking-at "\\(::\\)")
6392 ;; We only check for a trailing "::" and
6393 ;; let the "*" that should follow be
6394 ;; matched in the next round.
6395 (progn (setq got-identifier nil) t)
6396 ;; It turned out to be the real identifier,
6397 ;; so stop.
6398 nil))
6399 t))
6400
6401 (if (eq (char-after) ?\()
6402 (progn
6403 (setq paren-depth (1+ paren-depth))
6404 (forward-char))
6405 (unless got-prefix-before-parens
6406 (setq got-prefix-before-parens (= paren-depth 0)))
6407 (setq got-prefix t)
6408 (goto-char (match-end 1)))
6409 (c-forward-syntactic-ws))
6410
6411 (setq got-parens (> paren-depth 0))
6412
6413 ;; Skip over an identifier.
6414 (or got-identifier
6415 (and (looking-at c-identifier-start)
6416 (setq got-identifier (c-forward-name))))
6417
6418 ;; Skip over type decl suffix operators.
6419 (while (if (looking-at c-type-decl-suffix-key)
6420
6421 (if (eq (char-after) ?\))
6422 (when (> paren-depth 0)
6423 (setq paren-depth (1- paren-depth))
6424 (forward-char)
6425 t)
6426 (when (if (save-match-data (looking-at "\\s\("))
6427 (c-safe (c-forward-sexp 1) t)
6428 (goto-char (match-end 1))
6429 t)
6430 (when (and (not got-suffix-after-parens)
6431 (= paren-depth 0))
6432 (setq got-suffix-after-parens (match-beginning 0)))
6433 (setq got-suffix t)))
6434
6435 ;; No suffix matched. We might have matched the
6436 ;; identifier as a type and the open paren of a
6437 ;; function arglist as a type decl prefix. In that
6438 ;; case we should "backtrack": Reinterpret the last
6439 ;; type as the identifier, move out of the arglist and
6440 ;; continue searching for suffix operators.
6441 ;;
6442 ;; Do this even if there's no preceding type, to cope
6443 ;; with old style function declarations in K&R C,
6444 ;; (con|de)structors in C++ and `c-typeless-decl-kwds'
6445 ;; style declarations. That isn't applicable in an
6446 ;; arglist context, though.
6447 (when (and (= paren-depth 1)
6448 (not got-prefix-before-parens)
6449 (not (eq at-type t))
6450 (or backup-at-type
6451 maybe-typeless
6452 backup-maybe-typeless
6453 (when c-recognize-typeless-decls
6454 (not context)))
6455 (setq pos (c-up-list-forward (point)))
6456 (eq (char-before pos) ?\)))
6457 (c-fdoc-shift-type-backward)
6458 (goto-char pos)
6459 t))
6460
6461 (c-forward-syntactic-ws))
6462
6463 (when (and (or maybe-typeless backup-maybe-typeless)
6464 (not got-identifier)
6465 (not got-prefix)
6466 at-type)
6467 ;; Have found no identifier but `c-typeless-decl-kwds' has
6468 ;; matched so we know we're inside a declaration. The
6469 ;; preceding type must be the identifier instead.
6470 (c-fdoc-shift-type-backward))
6471
6472 (setq
6473 at-decl-or-cast
6474 (catch 'at-decl-or-cast
6475
6476 ;; CASE 1
6477 (when (> paren-depth 0)
6478 ;; Encountered something inside parens that isn't matched by
6479 ;; the `c-type-decl-*' regexps, so it's not a type decl
6480 ;; expression. Try to skip out to the same paren depth to
6481 ;; not confuse the cast check below.
6482 (c-safe (goto-char (scan-lists (point) 1 paren-depth)))
6483 ;; If we've found a specifier keyword then it's a
6484 ;; declaration regardless.
6485 (throw 'at-decl-or-cast (eq at-decl-or-cast t)))
6486
6487 (setq at-decl-end
6488 (looking-at (cond ((eq context '<>) "[,>]")
6489 (context "[,\)]")
6490 (t "[,;]"))))
6491
6492 ;; Now we've collected info about various characteristics of
6493 ;; the construct we're looking at. Below follows a decision
6494 ;; tree based on that. It's ordered to check more certain
6495 ;; signs before less certain ones.
6496
6497 (if got-identifier
6498 (progn
6499
6500 ;; CASE 2
6501 (when (and (or at-type maybe-typeless)
6502 (not (or got-prefix got-parens)))
6503 ;; Got another identifier directly after the type, so it's a
6504 ;; declaration.
6505 (throw 'at-decl-or-cast t))
6506
6507 (when (and got-parens
6508 (not got-prefix)
6509 (not got-suffix-after-parens)
6510 (or backup-at-type
6511 maybe-typeless
6512 backup-maybe-typeless))
6513 ;; Got a declaration of the form "foo bar (gnu);" where we've
6514 ;; recognized "bar" as the type and "gnu" as the declarator.
6515 ;; In this case it's however more likely that "bar" is the
6516 ;; declarator and "gnu" a function argument or initializer (if
6517 ;; `c-recognize-paren-inits' is set), since the parens around
6518 ;; "gnu" would be superfluous if it's a declarator. Shift the
6519 ;; type one step backward.
6520 (c-fdoc-shift-type-backward)))
6521
6522 ;; Found no identifier.
6523
6524 (if backup-at-type
6525 (progn
6526
6527 ;; CASE 3
6528 (when (= (point) start)
6529 ;; Got a plain list of identifiers. If a colon follows it's
6530 ;; a valid label. Otherwise the last one probably is the
6531 ;; declared identifier and we should back up to the previous
6532 ;; type, providing it isn't a cast.
6533 (if (eq (char-after) ?:)
6534 ;; If we've found a specifier keyword then it's a
6535 ;; declaration regardless.
6536 (throw 'at-decl-or-cast (eq at-decl-or-cast t))
6537 (setq backup-if-not-cast t)
6538 (throw 'at-decl-or-cast t)))
6539
6540 ;; CASE 4
6541 (when (and got-suffix
6542 (not got-prefix)
6543 (not got-parens))
6544 ;; Got a plain list of identifiers followed by some suffix.
6545 ;; If this isn't a cast then the last identifier probably is
6546 ;; the declared one and we should back up to the previous
6547 ;; type.
6548 (setq backup-if-not-cast t)
6549 (throw 'at-decl-or-cast t)))
6550
6551 ;; CASE 5
6552 (when (eq at-type t)
6553 ;; If the type is known we know that there can't be any
6554 ;; identifier somewhere else, and it's only in declarations in
6555 ;; e.g. function prototypes and in casts that the identifier may
6556 ;; be left out.
6557 (throw 'at-decl-or-cast t))
6558
6559 (when (= (point) start)
6560 ;; Only got a single identifier (parsed as a type so far).
6561 ;; CASE 6
6562 (if (and
6563 ;; Check that the identifier isn't at the start of an
6564 ;; expression.
6565 at-decl-end
6566 (cond
6567 ((eq context 'decl)
6568 ;; Inside an arglist that contains declarations. If K&R
6569 ;; style declarations and parenthesis style initializers
6570 ;; aren't allowed then the single identifier must be a
6571 ;; type, else we require that it's known or found
6572 ;; (primitive types are handled above).
6573 (or (and (not c-recognize-knr-p)
6574 (not c-recognize-paren-inits))
6575 (memq at-type '(known found))))
6576 ((eq context '<>)
6577 ;; Inside a template arglist. Accept known and found
6578 ;; types; other identifiers could just as well be
6579 ;; constants in C++.
6580 (memq at-type '(known found)))))
6581 (throw 'at-decl-or-cast t)
6582 ;; CASE 7
6583 ;; Can't be a valid declaration or cast, but if we've found a
6584 ;; specifier it can't be anything else either, so treat it as
6585 ;; an invalid/unfinished declaration or cast.
6586 (throw 'at-decl-or-cast at-decl-or-cast))))
6587
6588 (if (and got-parens
6589 (not got-prefix)
6590 (not context)
6591 (not (eq at-type t))
6592 (or backup-at-type
6593 maybe-typeless
6594 backup-maybe-typeless
6595 (when c-recognize-typeless-decls
6596 (or (not got-suffix)
6597 (not (looking-at
6598 c-after-suffixed-type-maybe-decl-key))))))
6599 ;; Got an empty paren pair and a preceding type that probably
6600 ;; really is the identifier. Shift the type backwards to make
6601 ;; the last one the identifier. This is analogous to the
6602 ;; "backtracking" done inside the `c-type-decl-suffix-key' loop
6603 ;; above.
6604 ;;
6605 ;; Exception: In addition to the conditions in that
6606 ;; "backtracking" code, do not shift backward if we're not
6607 ;; looking at either `c-after-suffixed-type-decl-key' or "[;,]".
6608 ;; Since there's no preceding type, the shift would mean that
6609 ;; the declaration is typeless. But if the regexp doesn't match
6610 ;; then we will simply fall through in the tests below and not
6611 ;; recognize it at all, so it's better to try it as an abstract
6612 ;; declarator instead.
6613 (c-fdoc-shift-type-backward)
6614
6615 ;; Still no identifier.
6616 ;; CASE 8
6617 (when (and got-prefix (or got-parens got-suffix))
6618 ;; Require `got-prefix' together with either `got-parens' or
6619 ;; `got-suffix' to recognize it as an abstract declarator:
6620 ;; `got-parens' only is probably an empty function call.
6621 ;; `got-suffix' only can build an ordinary expression together
6622 ;; with the preceding identifier which we've taken as a type.
6623 ;; We could actually accept on `got-prefix' only, but that can
6624 ;; easily occur temporarily while writing an expression so we
6625 ;; avoid that case anyway. We could do a better job if we knew
6626 ;; the point when the fontification was invoked.
6627 (throw 'at-decl-or-cast t))
6628
6629 ;; CASE 9
6630 (when (and at-type
6631 (not got-prefix)
6632 (not got-parens)
6633 got-suffix-after-parens
6634 (eq (char-after got-suffix-after-parens) ?\())
6635 ;; Got a type, no declarator but a paren suffix. I.e. it's a
6636 ;; normal function call afterall (or perhaps a C++ style object
6637 ;; instantiation expression).
6638 (throw 'at-decl-or-cast nil))))
6639
6640 ;; CASE 10
6641 (when at-decl-or-cast
6642 ;; By now we've located the type in the declaration that we know
6643 ;; we're in.
6644 (throw 'at-decl-or-cast t))
6645
6646 ;; CASE 11
6647 (when (and got-identifier
6648 (not context)
6649 (looking-at c-after-suffixed-type-decl-key)
6650 (if (and got-parens
6651 (not got-prefix)
6652 (not got-suffix)
6653 (not (eq at-type t)))
6654 ;; Shift the type backward in the case that there's a
6655 ;; single identifier inside parens. That can only
6656 ;; occur in K&R style function declarations so it's
6657 ;; more likely that it really is a function call.
6658 ;; Therefore we only do this after
6659 ;; `c-after-suffixed-type-decl-key' has matched.
6660 (progn (c-fdoc-shift-type-backward) t)
6661 got-suffix-after-parens))
6662 ;; A declaration according to `c-after-suffixed-type-decl-key'.
6663 (throw 'at-decl-or-cast t))
6664
6665 ;; CASE 12
6666 (when (and (or got-prefix (not got-parens))
6667 (memq at-type '(t known)))
6668 ;; It's a declaration if a known type precedes it and it can't be a
6669 ;; function call.
6670 (throw 'at-decl-or-cast t))
6671
6672 ;; If we get here we can't tell if this is a type decl or a normal
6673 ;; expression by looking at it alone. (That's under the assumption
6674 ;; that normal expressions always can look like type decl expressions,
6675 ;; which isn't really true but the cases where it doesn't hold are so
6676 ;; uncommon (e.g. some placements of "const" in C++) it's not worth
6677 ;; the effort to look for them.)
6678
6679 (unless (or at-decl-end (looking-at "=[^=]"))
6680 ;; If this is a declaration it should end here or its initializer(*)
6681 ;; should start here, so check for allowed separation tokens. Note
6682 ;; that this rule doesn't work e.g. with a K&R arglist after a
6683 ;; function header.
6684 ;;
6685 ;; *) Don't check for C++ style initializers using parens
6686 ;; since those already have been matched as suffixes.
6687 ;;
6688 ;; If `at-decl-or-cast' is then we've found some other sign that
6689 ;; it's a declaration or cast, so then it's probably an
6690 ;; invalid/unfinished one.
6691 (throw 'at-decl-or-cast at-decl-or-cast))
6692
6693 ;; Below are tests that only should be applied when we're certain to
6694 ;; not have parsed halfway through an expression.
6695
6696 ;; CASE 14
6697 (when (memq at-type '(t known))
6698 ;; The expression starts with a known type so treat it as a
6699 ;; declaration.
6700 (throw 'at-decl-or-cast t))
6701
6702 ;; CASE 15
6703 (when (and (c-major-mode-is 'c++-mode)
6704 ;; In C++ we check if the identifier is a known type, since
6705 ;; (con|de)structors use the class name as identifier.
6706 ;; We've always shifted over the identifier as a type and
6707 ;; then backed up again in this case.
6708 identifier-type
6709 (or (memq identifier-type '(found known))
6710 (and (eq (char-after identifier-start) ?~)
6711 ;; `at-type' probably won't be 'found for
6712 ;; destructors since the "~" is then part of the
6713 ;; type name being checked against the list of
6714 ;; known types, so do a check without that
6715 ;; operator.
6716 (or (save-excursion
6717 (goto-char (1+ identifier-start))
6718 (c-forward-syntactic-ws)
6719 (c-with-syntax-table
6720 c-identifier-syntax-table
6721 (looking-at c-known-type-key)))
6722 (save-excursion
6723 (goto-char (1+ identifier-start))
6724 ;; We have already parsed the type earlier,
6725 ;; so it'd be possible to cache the end
6726 ;; position instead of redoing it here, but
6727 ;; then we'd need to keep track of another
6728 ;; position everywhere.
6729 (c-check-type (point)
6730 (progn (c-forward-type)
6731 (point))))))))
6732 (throw 'at-decl-or-cast t))
6733
6734 (if got-identifier
6735 (progn
6736 ;; CASE 16
6737 (when (and got-prefix-before-parens
6738 at-type
6739 (or at-decl-end (looking-at "=[^=]"))
6740 (not context)
6741 (not got-suffix))
6742 ;; Got something like "foo * bar;". Since we're not inside an
6743 ;; arglist it would be a meaningless expression because the
6744 ;; result isn't used. We therefore choose to recognize it as
6745 ;; a declaration. Do not allow a suffix since it could then
6746 ;; be a function call.
6747 (throw 'at-decl-or-cast t))
6748
6749 ;; CASE 17
6750 (when (and (or got-suffix-after-parens
6751 (looking-at "=[^=]"))
6752 (eq at-type 'found)
6753 (not (eq context 'arglist)))
6754 ;; Got something like "a (*b) (c);" or "a (b) = c;". It could
6755 ;; be an odd expression or it could be a declaration. Treat
6756 ;; it as a declaration if "a" has been used as a type
6757 ;; somewhere else (if it's a known type we won't get here).
6758 (throw 'at-decl-or-cast t)))
6759
6760 ;; CASE 18
6761 (when (and context
6762 (or got-prefix
6763 (and (eq context 'decl)
6764 (not c-recognize-paren-inits)
6765 (or got-parens got-suffix))))
6766 ;; Got a type followed by an abstract declarator. If `got-prefix'
6767 ;; is set it's something like "a *" without anything after it. If
6768 ;; `got-parens' or `got-suffix' is set it's "a()", "a[]", "a()[]",
6769 ;; or similar, which we accept only if the context rules out
6770 ;; expressions.
6771 (throw 'at-decl-or-cast t)))
6772
6773 ;; If we had a complete symbol table here (which rules out
6774 ;; `c-found-types') we should return t due to the disambiguation rule
6775 ;; (in at least C++) that anything that can be parsed as a declaration
6776 ;; is a declaration. Now we're being more defensive and prefer to
6777 ;; highlight things like "foo (bar);" as a declaration only if we're
6778 ;; inside an arglist that contains declarations.
6779 (eq context 'decl))))
6780
6781 ;; The point is now after the type decl expression.
6782
6783 (cond
6784 ;; Check for a cast.
6785 ((save-excursion
6786 (and
6787 c-cast-parens
6788
6789 ;; Should be the first type/identifier in a cast paren.
6790 (> preceding-token-end (point-min))
6791 (memq (char-before preceding-token-end) c-cast-parens)
6792
6793 ;; The closing paren should follow.
6794 (progn
6795 (c-forward-syntactic-ws)
6796 (looking-at "\\s\)"))
6797
6798 ;; There should be a primary expression after it.
6799 (let (pos)
6800 (forward-char)
6801 (c-forward-syntactic-ws)
6802 (setq cast-end (point))
6803 (and (looking-at c-primary-expr-regexp)
6804 (progn
6805 (setq pos (match-end 0))
6806 (or
6807 ;; Check if the expression begins with a prefix keyword.
6808 (match-beginning 2)
6809 (if (match-beginning 1)
6810 ;; Expression begins with an ambiguous operator. Treat
6811 ;; it as a cast if it's a type decl or if we've
6812 ;; recognized the type somewhere else.
6813 (or at-decl-or-cast
6814 (memq at-type '(t known found)))
6815 ;; Unless it's a keyword, it's the beginning of a primary
6816 ;; expression.
6817 (not (looking-at c-keywords-regexp)))))
6818 ;; If `c-primary-expr-regexp' matched a nonsymbol token, check
6819 ;; that it matched a whole one so that we don't e.g. confuse
6820 ;; the operator '-' with '->'. It's ok if it matches further,
6821 ;; though, since it e.g. can match the float '.5' while the
6822 ;; operator regexp only matches '.'.
6823 (or (not (looking-at c-nonsymbol-token-regexp))
6824 (<= (match-end 0) pos))))
6825
6826 ;; There should either be a cast before it or something that isn't an
6827 ;; identifier or close paren.
6828 (> preceding-token-end (point-min))
6829 (progn
6830 (goto-char (1- preceding-token-end))
6831 (or (eq (point) last-cast-end)
6832 (progn
6833 (c-backward-syntactic-ws)
6834 (if (< (skip-syntax-backward "w_") 0)
6835 ;; It's a symbol. Accept it only if it's one of the
6836 ;; keywords that can precede an expression (without
6837 ;; surrounding parens).
6838 (looking-at c-simple-stmt-key)
6839 (and
6840 ;; Check that it isn't a close paren (block close is ok,
6841 ;; though).
6842 (not (memq (char-before) '(?\) ?\])))
6843 ;; Check that it isn't a nonsymbol identifier.
6844 (not (c-on-identifier)))))))))
6845
6846 ;; Handle the cast.
6847 (when (and c-record-type-identifiers at-type (not (eq at-type t)))
6848 (let ((c-promote-possible-types t))
6849 (goto-char type-start)
6850 (c-forward-type)))
6851
6852 (goto-char cast-end)
6853 'cast)
6854
6855 (at-decl-or-cast
6856 ;; We're at a declaration. Highlight the type and the following
6857 ;; declarators.
6858
6859 (when backup-if-not-cast
6860 (c-fdoc-shift-type-backward t))
6861
6862 (when (and (eq context 'decl) (looking-at ","))
6863 ;; Make sure to propagate the `c-decl-arg-start' property to
6864 ;; the next argument if it's set in this one, to cope with
6865 ;; interactive refontification.
6866 (c-put-c-type-property (point) 'c-decl-arg-start))
6867
6868 (when (and c-record-type-identifiers at-type (not (eq at-type t)))
6869 (let ((c-promote-possible-types t))
6870 (save-excursion
6871 (goto-char type-start)
6872 (c-forward-type))))
6873
6874 (cons id-start at-type-decl))
6875
6876 (t
6877 ;; False alarm. Restore the recorded ranges.
6878 (setq c-record-type-identifiers save-rec-type-ids
6879 c-record-ref-identifiers save-rec-ref-ids)
6880 nil))))
6881
6882 (defun c-forward-label (&optional assume-markup preceding-token-end limit)
6883 ;; Assuming that point is at the beginning of a token, check if it starts a
6884 ;; label and if so move over it and return non-nil (t in default situations,
6885 ;; specific symbols (see below) for interesting situations), otherwise don't
6886 ;; move and return nil. "Label" here means "most things with a colon".
6887 ;;
6888 ;; More precisely, a "label" is regarded as one of:
6889 ;; (i) a goto target like "foo:" - returns the symbol `goto-target';
6890 ;; (ii) A case label - either the entire construct "case FOO:", or just the
6891 ;; bare "case", should the colon be missing. We return t;
6892 ;; (iii) a keyword which needs a colon, like "default:" or "private:"; We
6893 ;; return t;
6894 ;; (iv) One of QT's "extended" C++ variants of
6895 ;; "private:"/"protected:"/"public:"/"more:" looking like "public slots:".
6896 ;; Returns the symbol `qt-2kwds-colon'.
6897 ;; (v) QT's construct "signals:". Returns the symbol `qt-1kwd-colon'.
6898 ;; (vi) One of the keywords matched by `c-opt-extra-label-key' (without any
6899 ;; colon). Currently (2006-03), this applies only to Objective C's
6900 ;; keywords "@private", "@protected", and "@public". Returns t.
6901 ;;
6902 ;; One of the things which will NOT be recognised as a label is a bit-field
6903 ;; element of a struct, something like "int foo:5".
6904 ;;
6905 ;; The end of the label is taken to be just after the colon, or the end of
6906 ;; the first submatch in `c-opt-extra-label-key'. The point is directly
6907 ;; after the end on return. The terminating char gets marked with
6908 ;; `c-decl-end' to improve recognition of the following declaration or
6909 ;; statement.
6910 ;;
6911 ;; If ASSUME-MARKUP is non-nil, it's assumed that the preceding
6912 ;; label, if any, has already been marked up like that.
6913 ;;
6914 ;; If PRECEDING-TOKEN-END is given, it should be the first position
6915 ;; after the preceding token, i.e. on the other side of the
6916 ;; syntactic ws from the point. Use a value less than or equal to
6917 ;; (point-min) if the point is at the first token in (the visible
6918 ;; part of) the buffer.
6919 ;;
6920 ;; The optional LIMIT limits the forward scan for the colon.
6921 ;;
6922 ;; This function records the ranges of the label symbols on
6923 ;; `c-record-ref-identifiers' if `c-record-type-identifiers' (!) is
6924 ;; non-nil.
6925 ;;
6926 ;; This function might do hidden buffer changes.
6927
6928 (let ((start (point))
6929 label-end
6930 qt-symbol-idx
6931 macro-start ; if we're in one.
6932 label-type
6933 kwd)
6934 (cond
6935 ;; "case" or "default" (Doesn't apply to AWK).
6936 ((looking-at c-label-kwds-regexp)
6937 (let ((kwd-end (match-end 1)))
6938 ;; Record only the keyword itself for fontification, since in
6939 ;; case labels the following is a constant expression and not
6940 ;; a label.
6941 (when c-record-type-identifiers
6942 (c-record-ref-id (cons (match-beginning 1) kwd-end)))
6943
6944 ;; Find the label end.
6945 (goto-char kwd-end)
6946 (setq label-type
6947 (if (and (c-syntactic-re-search-forward
6948 ;; Stop on chars that aren't allowed in expressions,
6949 ;; and on operator chars that would be meaningless
6950 ;; there. FIXME: This doesn't cope with ?: operators.
6951 "[;{=,@]\\|\\(\\=\\|[^:]\\):\\([^:]\\|\\'\\)"
6952 limit t t nil 1)
6953 (match-beginning 2))
6954
6955 (progn ; there's a proper :
6956 (goto-char (match-beginning 2)) ; just after the :
6957 (c-put-c-type-property (1- (point)) 'c-decl-end)
6958 t)
6959
6960 ;; It's an unfinished label. We consider the keyword enough
6961 ;; to recognize it as a label, so that it gets fontified.
6962 ;; Leave the point at the end of it, but don't put any
6963 ;; `c-decl-end' marker.
6964 (goto-char kwd-end)
6965 t))))
6966
6967 ;; @private, @protected, @public, in Objective C, or similar.
6968 ((and c-opt-extra-label-key
6969 (looking-at c-opt-extra-label-key))
6970 ;; For a `c-opt-extra-label-key' match, we record the whole
6971 ;; thing for fontification. That's to get the leading '@' in
6972 ;; Objective-C protection labels fontified.
6973 (goto-char (match-end 1))
6974 (when c-record-type-identifiers
6975 (c-record-ref-id (cons (match-beginning 1) (point))))
6976 (c-put-c-type-property (1- (point)) 'c-decl-end)
6977 (setq label-type t))
6978
6979 ;; All other cases of labels.
6980 ((and c-recognize-colon-labels ; nil for AWK and IDL, otherwise t.
6981
6982 ;; A colon label must have something before the colon.
6983 (not (eq (char-after) ?:))
6984
6985 ;; Check that we're not after a token that can't precede a label.
6986 (or
6987 ;; Trivially succeeds when there's no preceding token.
6988 (if preceding-token-end
6989 (<= preceding-token-end (point-min))
6990 (save-excursion
6991 (c-backward-syntactic-ws)
6992 (setq preceding-token-end (point))
6993 (bobp)))
6994
6995 ;; Check if we're after a label, if we're after a closing
6996 ;; paren that belong to statement, and with
6997 ;; `c-label-prefix-re'. It's done in different order
6998 ;; depending on `assume-markup' since the checks have
6999 ;; different expensiveness.
7000 (if assume-markup
7001 (or
7002 (eq (c-get-char-property (1- preceding-token-end) 'c-type)
7003 'c-decl-end)
7004
7005 (save-excursion
7006 (goto-char (1- preceding-token-end))
7007 (c-beginning-of-current-token)
7008 (or (looking-at c-label-prefix-re)
7009 (looking-at c-block-stmt-1-key)))
7010
7011 (and (eq (char-before preceding-token-end) ?\))
7012 (c-after-conditional)))
7013
7014 (or
7015 (save-excursion
7016 (goto-char (1- preceding-token-end))
7017 (c-beginning-of-current-token)
7018 (or (looking-at c-label-prefix-re)
7019 (looking-at c-block-stmt-1-key)))
7020
7021 (cond
7022 ((eq (char-before preceding-token-end) ?\))
7023 (c-after-conditional))
7024
7025 ((eq (char-before preceding-token-end) ?:)
7026 ;; Might be after another label, so check it recursively.
7027 (save-restriction
7028 (save-excursion
7029 (goto-char (1- preceding-token-end))
7030 ;; Essentially the same as the
7031 ;; `c-syntactic-re-search-forward' regexp below.
7032 (setq macro-start
7033 (save-excursion (and (c-beginning-of-macro)
7034 (point))))
7035 (if macro-start (narrow-to-region macro-start (point-max)))
7036 (c-syntactic-skip-backward "^-]:?;}=*/%&|,<>!@+" nil t)
7037 ;; Note: the following should work instead of the
7038 ;; narrow-to-region above. Investigate why not,
7039 ;; sometime. ACM, 2006-03-31.
7040 ;; (c-syntactic-skip-backward "^-]:?;}=*/%&|,<>!@+"
7041 ;; macro-start t)
7042 (let ((pte (point))
7043 ;; If the caller turned on recording for us,
7044 ;; it shouldn't apply when we check the
7045 ;; preceding label.
7046 c-record-type-identifiers)
7047 ;; A label can't start at a cpp directive. Check for
7048 ;; this, since c-forward-syntactic-ws would foul up on it.
7049 (unless (and c-opt-cpp-prefix (looking-at c-opt-cpp-prefix))
7050 (c-forward-syntactic-ws)
7051 (c-forward-label nil pte start))))))))))
7052
7053 ;; Point is still at the beginning of the possible label construct.
7054 ;;
7055 ;; Check that the next nonsymbol token is ":", or that we're in one
7056 ;; of QT's "slots" declarations. Allow '(' for the sake of macro
7057 ;; arguments. FIXME: Should build this regexp from the language
7058 ;; constants.
7059 (cond
7060 ;; public: protected: private:
7061 ((and
7062 (c-major-mode-is 'c++-mode)
7063 (search-forward-regexp
7064 "\\=p\\(r\\(ivate\\|otected\\)\\|ublic\\)\\>[^_]" nil t)
7065 (progn (backward-char)
7066 (c-forward-syntactic-ws limit)
7067 (looking-at ":\\([^:]\\|\\'\\)"))) ; A single colon.
7068 (forward-char)
7069 (setq label-type t))
7070 ;; QT double keyword like "protected slots:" or goto target.
7071 ((progn (goto-char start) nil))
7072 ((when (c-syntactic-re-search-forward
7073 "[ \t\n[:?;{=*/%&|,<>!@+-]" limit t t) ; not at EOB
7074 (backward-char)
7075 (setq label-end (point))
7076 (setq qt-symbol-idx
7077 (and (c-major-mode-is 'c++-mode)
7078 (string-match
7079 "\\(p\\(r\\(ivate\\|otected\\)\\|ublic\\)\\|more\\)\\>"
7080 (buffer-substring start (point)))))
7081 (c-forward-syntactic-ws limit)
7082 (cond
7083 ((looking-at ":\\([^:]\\|\\'\\)") ; A single colon.
7084 (forward-char)
7085 (setq label-type
7086 (if (or (string= "signals" ; Special QT macro
7087 (setq kwd (buffer-substring-no-properties start label-end)))
7088 (string= "Q_SIGNALS" kwd))
7089 'qt-1kwd-colon
7090 'goto-target)))
7091 ((and qt-symbol-idx
7092 (search-forward-regexp "\\=\\(slots\\|Q_SLOTS\\)\\>" limit t)
7093 (progn (c-forward-syntactic-ws limit)
7094 (looking-at ":\\([^:]\\|\\'\\)"))) ; A single colon
7095 (forward-char)
7096 (setq label-type 'qt-2kwds-colon)))))))
7097
7098 (save-restriction
7099 (narrow-to-region start (point))
7100
7101 ;; Check that `c-nonlabel-token-key' doesn't match anywhere.
7102 (catch 'check-label
7103 (goto-char start)
7104 (while (progn
7105 (when (looking-at c-nonlabel-token-key)
7106 (goto-char start)
7107 (setq label-type nil)
7108 (throw 'check-label nil))
7109 (and (c-safe (c-forward-sexp)
7110 (c-forward-syntactic-ws)
7111 t)
7112 (not (eobp)))))
7113
7114 ;; Record the identifiers in the label for fontification, unless
7115 ;; it begins with `c-label-kwds' in which case the following
7116 ;; identifiers are part of a (constant) expression that
7117 ;; shouldn't be fontified.
7118 (when (and c-record-type-identifiers
7119 (progn (goto-char start)
7120 (not (looking-at c-label-kwds-regexp))))
7121 (while (c-syntactic-re-search-forward c-symbol-key nil t)
7122 (c-record-ref-id (cons (match-beginning 0)
7123 (match-end 0)))))
7124
7125 (c-put-c-type-property (1- (point-max)) 'c-decl-end)
7126 (goto-char (point-max)))))
7127
7128 (t
7129 ;; Not a label.
7130 (goto-char start)))
7131 label-type))
7132
7133 (defun c-forward-objc-directive ()
7134 ;; Assuming the point is at the beginning of a token, try to move
7135 ;; forward to the end of the Objective-C directive that starts
7136 ;; there. Return t if a directive was fully recognized, otherwise
7137 ;; the point is moved as far as one could be successfully parsed and
7138 ;; nil is returned.
7139 ;;
7140 ;; This function records identifier ranges on
7141 ;; `c-record-type-identifiers' and `c-record-ref-identifiers' if
7142 ;; `c-record-type-identifiers' is non-nil.
7143 ;;
7144 ;; This function might do hidden buffer changes.
7145
7146 (let ((start (point))
7147 start-char
7148 (c-promote-possible-types t)
7149 ;; Turn off recognition of angle bracket arglists while parsing
7150 ;; types here since the protocol reference list might then be
7151 ;; considered part of the preceding name or superclass-name.
7152 c-recognize-<>-arglists)
7153
7154 (if (or
7155 (when (looking-at
7156 (eval-when-compile
7157 (c-make-keywords-re t
7158 (append (c-lang-const c-protection-kwds objc)
7159 '("@end"))
7160 'objc-mode)))
7161 (goto-char (match-end 1))
7162 t)
7163
7164 (and
7165 (looking-at
7166 (eval-when-compile
7167 (c-make-keywords-re t
7168 '("@interface" "@implementation" "@protocol")
7169 'objc-mode)))
7170
7171 ;; Handle the name of the class itself.
7172 (progn
7173 ; (c-forward-token-2) ; 2006/1/13 This doesn't move if the token's
7174 ; at EOB.
7175 (goto-char (match-end 0))
7176 (c-skip-ws-forward)
7177 (c-forward-type))
7178
7179 (catch 'break
7180 ;; Look for ": superclass-name" or "( category-name )".
7181 (when (looking-at "[:\(]")
7182 (setq start-char (char-after))
7183 (forward-char)
7184 (c-forward-syntactic-ws)
7185 (unless (c-forward-type) (throw 'break nil))
7186 (when (eq start-char ?\()
7187 (unless (eq (char-after) ?\)) (throw 'break nil))
7188 (forward-char)
7189 (c-forward-syntactic-ws)))
7190
7191 ;; Look for a protocol reference list.
7192 (if (eq (char-after) ?<)
7193 (let ((c-recognize-<>-arglists t)
7194 (c-parse-and-markup-<>-arglists t)
7195 c-restricted-<>-arglists)
7196 (c-forward-<>-arglist t))
7197 t))))
7198
7199 (progn
7200 (c-backward-syntactic-ws)
7201 (c-clear-c-type-property start (1- (point)) 'c-decl-end)
7202 (c-put-c-type-property (1- (point)) 'c-decl-end)
7203 t)
7204
7205 (c-clear-c-type-property start (point) 'c-decl-end)
7206 nil)))
7207
7208 (defun c-beginning-of-inheritance-list (&optional lim)
7209 ;; Go to the first non-whitespace after the colon that starts a
7210 ;; multiple inheritance introduction. Optional LIM is the farthest
7211 ;; back we should search.
7212 ;;
7213 ;; This function might do hidden buffer changes.
7214 (c-with-syntax-table c++-template-syntax-table
7215 (c-backward-token-2 0 t lim)
7216 (while (and (or (looking-at c-symbol-start)
7217 (looking-at "[<,]\\|::"))
7218 (zerop (c-backward-token-2 1 t lim))))))
7219
7220 (defun c-in-method-def-p ()
7221 ;; Return nil if we aren't in a method definition, otherwise the
7222 ;; position of the initial [+-].
7223 ;;
7224 ;; This function might do hidden buffer changes.
7225 (save-excursion
7226 (beginning-of-line)
7227 (and c-opt-method-key
7228 (looking-at c-opt-method-key)
7229 (point))
7230 ))
7231
7232 ;; Contributed by Kevin Ryde <user42@zip.com.au>.
7233 (defun c-in-gcc-asm-p ()
7234 ;; Return non-nil if point is within a gcc \"asm\" block.
7235 ;;
7236 ;; This should be called with point inside an argument list.
7237 ;;
7238 ;; Only one level of enclosing parentheses is considered, so for
7239 ;; instance `nil' is returned when in a function call within an asm
7240 ;; operand.
7241 ;;
7242 ;; This function might do hidden buffer changes.
7243
7244 (and c-opt-asm-stmt-key
7245 (save-excursion
7246 (beginning-of-line)
7247 (backward-up-list 1)
7248 (c-beginning-of-statement-1 (point-min) nil t)
7249 (looking-at c-opt-asm-stmt-key))))
7250
7251 (defun c-at-toplevel-p ()
7252 "Return a determination as to whether point is \"at the top level\".
7253 Informally, \"at the top level\" is anywhere where you can write
7254 a function.
7255
7256 More precisely, being at the top-level means that point is either
7257 outside any enclosing block (such as a function definition), or
7258 directly inside a class, namespace or other block that contains
7259 another declaration level.
7260
7261 If point is not at the top-level (e.g. it is inside a method
7262 definition), then nil is returned. Otherwise, if point is at a
7263 top-level not enclosed within a class definition, t is returned.
7264 Otherwise, a 2-vector is returned where the zeroth element is the
7265 buffer position of the start of the class declaration, and the first
7266 element is the buffer position of the enclosing class's opening
7267 brace.
7268
7269 Note that this function might do hidden buffer changes. See the
7270 comment at the start of cc-engine.el for more info."
7271 (let ((paren-state (c-parse-state)))
7272 (or (not (c-most-enclosing-brace paren-state))
7273 (c-search-uplist-for-classkey paren-state))))
7274
7275 (defun c-just-after-func-arglist-p (&optional lim)
7276 ;; Return non-nil if the point is in the region after the argument
7277 ;; list of a function and its opening brace (or semicolon in case it
7278 ;; got no body). If there are K&R style argument declarations in
7279 ;; that region, the point has to be inside the first one for this
7280 ;; function to recognize it.
7281 ;;
7282 ;; If successful, the point is moved to the first token after the
7283 ;; function header (see `c-forward-decl-or-cast-1' for details) and
7284 ;; the position of the opening paren of the function arglist is
7285 ;; returned.
7286 ;;
7287 ;; The point is clobbered if not successful.
7288 ;;
7289 ;; LIM is used as bound for backward buffer searches.
7290 ;;
7291 ;; This function might do hidden buffer changes.
7292
7293 (let ((beg (point)) end id-start)
7294 (and
7295 (eq (c-beginning-of-statement-1 lim) 'same)
7296
7297 (not (or (c-major-mode-is 'objc-mode)
7298 (c-forward-objc-directive)))
7299
7300 (setq id-start
7301 (car-safe (c-forward-decl-or-cast-1 (c-point 'bosws) nil nil)))
7302 (< id-start beg)
7303
7304 ;; There should not be a '=' or ',' between beg and the
7305 ;; start of the declaration since that means we were in the
7306 ;; "expression part" of the declaration.
7307 (or (> (point) beg)
7308 (not (looking-at "[=,]")))
7309
7310 (save-excursion
7311 ;; Check that there's an arglist paren in the
7312 ;; declaration.
7313 (goto-char id-start)
7314 (cond ((eq (char-after) ?\()
7315 ;; The declarator is a paren expression, so skip past it
7316 ;; so that we don't get stuck on that instead of the
7317 ;; function arglist.
7318 (c-forward-sexp))
7319 ((and c-opt-op-identifier-prefix
7320 (looking-at c-opt-op-identifier-prefix))
7321 ;; Don't trip up on "operator ()".
7322 (c-forward-token-2 2 t)))
7323 (and (< (point) beg)
7324 (c-syntactic-re-search-forward "(" beg t t)
7325 (1- (point)))))))
7326
7327 (defun c-in-knr-argdecl (&optional lim)
7328 ;; Return the position of the first argument declaration if point is
7329 ;; inside a K&R style argument declaration list, nil otherwise.
7330 ;; `c-recognize-knr-p' is not checked. If LIM is non-nil, it's a
7331 ;; position that bounds the backward search for the argument list.
7332 ;;
7333 ;; Point must be within a possible K&R region, e.g. just before a top-level
7334 ;; "{". It must be outside of parens and brackets. The test can return
7335 ;; false positives otherwise.
7336 ;;
7337 ;; This function might do hidden buffer changes.
7338
7339 (save-excursion
7340 (save-restriction
7341 ;; If we're in a macro, our search range is restricted to it. Narrow to
7342 ;; the searchable range.
7343 (let* ((macro-start (c-query-macro-start))
7344 (lim (max (or lim (point-min)) (or macro-start (point-min))))
7345 before-lparen after-rparen
7346 (pp-count-out 20)) ; Max number of paren/brace constructs before we give up
7347 (narrow-to-region lim (c-point 'eol))
7348
7349 ;; Search backwards for the defun's argument list. We give up if we
7350 ;; encounter a "}" (end of a previous defun) or BOB.
7351 ;;
7352 ;; The criterion for a paren structure being the arg list is:
7353 ;; o - there is non-WS stuff after it but before any "{"; AND
7354 ;; o - the token after it isn't a ";" AND
7355 ;; o - it is preceded by either an identifier (the function name) or
7356 ;; a macro expansion like "DEFUN (...)"; AND
7357 ;; o - its content is a non-empty comma-separated list of identifiers
7358 ;; (an empty arg list won't have a knr region).
7359 ;;
7360 ;; The following snippet illustrates these rules:
7361 ;; int foo (bar, baz, yuk)
7362 ;; int bar [] ;
7363 ;; int (*baz) (my_type) ;
7364 ;; int (*) (void) (*yuk) (void) ;
7365 ;; {
7366
7367 (catch 'knr
7368 (while (> pp-count-out 0) ; go back one paren/bracket pair each time.
7369 (setq pp-count-out (1- pp-count-out))
7370 (c-syntactic-skip-backward "^)]}")
7371 (cond ((eq (char-before) ?\))
7372 (setq after-rparen (point)))
7373 ((eq (char-before) ?\])
7374 (setq after-rparen nil))
7375 (t ; either } (hit previous defun) or no more parens/brackets
7376 (throw 'knr nil)))
7377
7378 (if after-rparen
7379 ;; We're inside a paren. Could it be our argument list....?
7380 (if
7381 (and
7382 (progn
7383 (goto-char after-rparen)
7384 (unless (c-go-list-backward) (throw 'knr nil)) ;
7385 ;; FIXME!!! What about macros between the parens? 2007/01/20
7386 (setq before-lparen (point)))
7387
7388 ;; It can't be the arg list if next token is ; or {
7389 (progn (goto-char after-rparen)
7390 (c-forward-syntactic-ws)
7391 (not (memq (char-after) '(?\; ?\{))))
7392
7393 ;; Is the thing preceding the list an identifier (the
7394 ;; function name), or a macro expansion?
7395 (progn
7396 (goto-char before-lparen)
7397 (eq (c-backward-token-2) 0)
7398 (or (c-on-identifier)
7399 (and (eq (char-after) ?\))
7400 (c-go-up-list-backward)
7401 (eq (c-backward-token-2) 0)
7402 (c-on-identifier))))
7403
7404 ;; Have we got a non-empty list of comma-separated
7405 ;; identifiers?
7406 (progn
7407 (goto-char before-lparen)
7408 (c-forward-token-2) ; to first token inside parens
7409 (and
7410 (c-on-identifier)
7411 (c-forward-token-2)
7412 (catch 'id-list
7413 (while (eq (char-after) ?\,)
7414 (c-forward-token-2)
7415 (unless (c-on-identifier) (throw 'id-list nil))
7416 (c-forward-token-2))
7417 (eq (char-after) ?\))))))
7418
7419 ;; ...Yes. We've identified the function's argument list.
7420 (throw 'knr
7421 (progn (goto-char after-rparen)
7422 (c-forward-syntactic-ws)
7423 (point)))
7424
7425 ;; ...No. The current parens aren't the function's arg list.
7426 (goto-char before-lparen))
7427
7428 (or (c-go-list-backward) ; backwards over [ .... ]
7429 (throw 'knr nil)))))))))
7430
7431 (defun c-skip-conditional ()
7432 ;; skip forward over conditional at point, including any predicate
7433 ;; statements in parentheses. No error checking is performed.
7434 ;;
7435 ;; This function might do hidden buffer changes.
7436 (c-forward-sexp (cond
7437 ;; else if()
7438 ((looking-at (concat "\\<else"
7439 "\\([ \t\n]\\|\\\\\n\\)+"
7440 "if\\>\\([^_]\\|$\\)"))
7441 3)
7442 ;; do, else, try, finally
7443 ((looking-at (concat "\\<\\("
7444 "do\\|else\\|try\\|finally"
7445 "\\)\\>\\([^_]\\|$\\)"))
7446 1)
7447 ;; for, if, while, switch, catch, synchronized, foreach
7448 (t 2))))
7449
7450 (defun c-after-conditional (&optional lim)
7451 ;; If looking at the token after a conditional then return the
7452 ;; position of its start, otherwise return nil.
7453 ;;
7454 ;; This function might do hidden buffer changes.
7455 (save-excursion
7456 (and (zerop (c-backward-token-2 1 t lim))
7457 (or (looking-at c-block-stmt-1-key)
7458 (and (eq (char-after) ?\()
7459 (zerop (c-backward-token-2 1 t lim))
7460 (looking-at c-block-stmt-2-key)))
7461 (point))))
7462
7463 (defun c-after-special-operator-id (&optional lim)
7464 ;; If the point is after an operator identifier that isn't handled
7465 ;; like an ordinary symbol (i.e. like "operator =" in C++) then the
7466 ;; position of the start of that identifier is returned. nil is
7467 ;; returned otherwise. The point may be anywhere in the syntactic
7468 ;; whitespace after the last token of the operator identifier.
7469 ;;
7470 ;; This function might do hidden buffer changes.
7471 (save-excursion
7472 (and c-overloadable-operators-regexp
7473 (zerop (c-backward-token-2 1 nil lim))
7474 (looking-at c-overloadable-operators-regexp)
7475 (or (not c-opt-op-identifier-prefix)
7476 (and
7477 (zerop (c-backward-token-2 1 nil lim))
7478 (looking-at c-opt-op-identifier-prefix)))
7479 (point))))
7480
7481 (defsubst c-backward-to-block-anchor (&optional lim)
7482 ;; Assuming point is at a brace that opens a statement block of some
7483 ;; kind, move to the proper anchor point for that block. It might
7484 ;; need to be adjusted further by c-add-stmt-syntax, but the
7485 ;; position at return is suitable as start position for that
7486 ;; function.
7487 ;;
7488 ;; This function might do hidden buffer changes.
7489 (unless (= (point) (c-point 'boi))
7490 (let ((start (c-after-conditional lim)))
7491 (if start
7492 (goto-char start)))))
7493
7494 (defsubst c-backward-to-decl-anchor (&optional lim)
7495 ;; Assuming point is at a brace that opens the block of a top level
7496 ;; declaration of some kind, move to the proper anchor point for
7497 ;; that block.
7498 ;;
7499 ;; This function might do hidden buffer changes.
7500 (unless (= (point) (c-point 'boi))
7501 (c-beginning-of-statement-1 lim)))
7502
7503 (defun c-search-decl-header-end ()
7504 ;; Search forward for the end of the "header" of the current
7505 ;; declaration. That's the position where the definition body
7506 ;; starts, or the first variable initializer, or the ending
7507 ;; semicolon. I.e. search forward for the closest following
7508 ;; (syntactically relevant) '{', '=' or ';' token. Point is left
7509 ;; _after_ the first found token, or at point-max if none is found.
7510 ;;
7511 ;; This function might do hidden buffer changes.
7512
7513 (let ((base (point)))
7514 (if (c-major-mode-is 'c++-mode)
7515
7516 ;; In C++ we need to take special care to handle operator
7517 ;; tokens and those pesky template brackets.
7518 (while (and
7519 (c-syntactic-re-search-forward "[;{<=]" nil 'move t t)
7520 (or
7521 (c-end-of-current-token base)
7522 ;; Handle operator identifiers, i.e. ignore any
7523 ;; operator token preceded by "operator".
7524 (save-excursion
7525 (and (c-safe (c-backward-sexp) t)
7526 (looking-at c-opt-op-identifier-prefix)))
7527 (and (eq (char-before) ?<)
7528 (c-with-syntax-table c++-template-syntax-table
7529 (if (c-safe (goto-char (c-up-list-forward (point))))
7530 t
7531 (goto-char (point-max))
7532 nil)))))
7533 (setq base (point)))
7534
7535 (while (and
7536 (c-syntactic-re-search-forward "[;{=]" nil 'move t t)
7537 (c-end-of-current-token base))
7538 (setq base (point))))))
7539
7540 (defun c-beginning-of-decl-1 (&optional lim)
7541 ;; Go to the beginning of the current declaration, or the beginning
7542 ;; of the previous one if already at the start of it. Point won't
7543 ;; be moved out of any surrounding paren. Return a cons cell of the
7544 ;; form (MOVE . KNR-POS). MOVE is like the return value from
7545 ;; `c-beginning-of-statement-1'. If point skipped over some K&R
7546 ;; style argument declarations (and they are to be recognized) then
7547 ;; KNR-POS is set to the start of the first such argument
7548 ;; declaration, otherwise KNR-POS is nil. If LIM is non-nil, it's a
7549 ;; position that bounds the backward search.
7550 ;;
7551 ;; NB: Cases where the declaration continues after the block, as in
7552 ;; "struct foo { ... } bar;", are currently recognized as two
7553 ;; declarations, e.g. "struct foo { ... }" and "bar;" in this case.
7554 ;;
7555 ;; This function might do hidden buffer changes.
7556 (catch 'return
7557 (let* ((start (point))
7558 (last-stmt-start (point))
7559 (move (c-beginning-of-statement-1 lim nil t)))
7560
7561 ;; `c-beginning-of-statement-1' stops at a block start, but we
7562 ;; want to continue if the block doesn't begin a top level
7563 ;; construct, i.e. if it isn't preceded by ';', '}', ':', bob,
7564 ;; or an open paren.
7565 (let ((beg (point)) tentative-move)
7566 ;; Go back one "statement" each time round the loop until we're just
7567 ;; after a ;, }, or :, or at BOB or the start of a macro or start of
7568 ;; an ObjC method. This will move over a multiple declaration whose
7569 ;; components are comma separated.
7570 (while (and
7571 ;; Must check with c-opt-method-key in ObjC mode.
7572 (not (and c-opt-method-key
7573 (looking-at c-opt-method-key)))
7574 (/= last-stmt-start (point))
7575 (progn
7576 (c-backward-syntactic-ws lim)
7577 (not (memq (char-before) '(?\; ?} ?: nil))))
7578 (save-excursion
7579 (backward-char)
7580 (not (looking-at "\\s(")))
7581 ;; Check that we don't move from the first thing in a
7582 ;; macro to its header.
7583 (not (eq (setq tentative-move
7584 (c-beginning-of-statement-1 lim nil t))
7585 'macro)))
7586 (setq last-stmt-start beg
7587 beg (point)
7588 move tentative-move))
7589 (goto-char beg))
7590
7591 (when c-recognize-knr-p
7592 (let ((fallback-pos (point)) knr-argdecl-start)
7593 ;; Handle K&R argdecls. Back up after the "statement" jumped
7594 ;; over by `c-beginning-of-statement-1', unless it was the
7595 ;; function body, in which case we're sitting on the opening
7596 ;; brace now. Then test if we're in a K&R argdecl region and
7597 ;; that we started at the other side of the first argdecl in
7598 ;; it.
7599 (unless (eq (char-after) ?{)
7600 (goto-char last-stmt-start))
7601 (if (and (setq knr-argdecl-start (c-in-knr-argdecl lim))
7602 (< knr-argdecl-start start)
7603 (progn
7604 (goto-char knr-argdecl-start)
7605 (not (eq (c-beginning-of-statement-1 lim nil t) 'macro))))
7606 (throw 'return
7607 (cons (if (eq (char-after fallback-pos) ?{)
7608 'previous
7609 'same)
7610 knr-argdecl-start))
7611 (goto-char fallback-pos))))
7612
7613 ;; `c-beginning-of-statement-1' counts each brace block as a separate
7614 ;; statement, so the result will be 'previous if we've moved over any.
7615 ;; So change our result back to 'same if necessary.
7616 ;;
7617 ;; If they were brace list initializers we might not have moved over a
7618 ;; declaration boundary though, so change it to 'same if we've moved
7619 ;; past a '=' before '{', but not ';'. (This ought to be integrated
7620 ;; into `c-beginning-of-statement-1', so we avoid this extra pass which
7621 ;; potentially can search over a large amount of text.). Take special
7622 ;; pains not to get mislead by C++'s "operator=", and the like.
7623 (if (and (eq move 'previous)
7624 (c-with-syntax-table (if (c-major-mode-is 'c++-mode)
7625 c++-template-syntax-table
7626 (syntax-table))
7627 (save-excursion
7628 (and
7629 (progn
7630 (while ; keep going back to "[;={"s until we either find
7631 ; no more, or get to one which isn't an "operator ="
7632 (and (c-syntactic-re-search-forward "[;={]" start t t t)
7633 (eq (char-before) ?=)
7634 c-overloadable-operators-regexp
7635 c-opt-op-identifier-prefix
7636 (save-excursion
7637 (eq (c-backward-token-2) 0)
7638 (looking-at c-overloadable-operators-regexp)
7639 (eq (c-backward-token-2) 0)
7640 (looking-at c-opt-op-identifier-prefix))))
7641 (eq (char-before) ?=))
7642 (c-syntactic-re-search-forward "[;{]" start t t)
7643 (eq (char-before) ?{)
7644 (c-safe (goto-char (c-up-list-forward (point))) t)
7645 (not (c-syntactic-re-search-forward ";" start t t))))))
7646 (cons 'same nil)
7647 (cons move nil)))))
7648
7649 (defun c-end-of-decl-1 ()
7650 ;; Assuming point is at the start of a declaration (as detected by
7651 ;; e.g. `c-beginning-of-decl-1'), go to the end of it. Unlike
7652 ;; `c-beginning-of-decl-1', this function handles the case when a
7653 ;; block is followed by identifiers in e.g. struct declarations in C
7654 ;; or C++. If a proper end was found then t is returned, otherwise
7655 ;; point is moved as far as possible within the current sexp and nil
7656 ;; is returned. This function doesn't handle macros; use
7657 ;; `c-end-of-macro' instead in those cases.
7658 ;;
7659 ;; This function might do hidden buffer changes.
7660 (let ((start (point))
7661 (decl-syntax-table (if (c-major-mode-is 'c++-mode)
7662 c++-template-syntax-table
7663 (syntax-table))))
7664 (catch 'return
7665 (c-search-decl-header-end)
7666
7667 (when (and c-recognize-knr-p
7668 (eq (char-before) ?\;)
7669 (c-in-knr-argdecl start))
7670 ;; Stopped at the ';' in a K&R argdecl section which is
7671 ;; detected using the same criteria as in
7672 ;; `c-beginning-of-decl-1'. Move to the following block
7673 ;; start.
7674 (c-syntactic-re-search-forward "{" nil 'move t))
7675
7676 (when (eq (char-before) ?{)
7677 ;; Encountered a block in the declaration. Jump over it.
7678 (condition-case nil
7679 (goto-char (c-up-list-forward (point)))
7680 (error (goto-char (point-max))
7681 (throw 'return nil)))
7682 (if (or (not c-opt-block-decls-with-vars-key)
7683 (save-excursion
7684 (c-with-syntax-table decl-syntax-table
7685 (let ((lim (point)))
7686 (goto-char start)
7687 (not (and
7688 ;; Check for `c-opt-block-decls-with-vars-key'
7689 ;; before the first paren.
7690 (c-syntactic-re-search-forward
7691 (concat "[;=\(\[{]\\|\\("
7692 c-opt-block-decls-with-vars-key
7693 "\\)")
7694 lim t t t)
7695 (match-beginning 1)
7696 (not (eq (char-before) ?_))
7697 ;; Check that the first following paren is
7698 ;; the block.
7699 (c-syntactic-re-search-forward "[;=\(\[{]"
7700 lim t t t)
7701 (eq (char-before) ?{)))))))
7702 ;; The declaration doesn't have any of the
7703 ;; `c-opt-block-decls-with-vars' keywords in the
7704 ;; beginning, so it ends here at the end of the block.
7705 (throw 'return t)))
7706
7707 (c-with-syntax-table decl-syntax-table
7708 (while (progn
7709 (if (eq (char-before) ?\;)
7710 (throw 'return t))
7711 (c-syntactic-re-search-forward ";" nil 'move t))))
7712 nil)))
7713
7714 (defun c-looking-at-decl-block (containing-sexp goto-start &optional limit)
7715 ;; Assuming the point is at an open brace, check if it starts a
7716 ;; block that contains another declaration level, i.e. that isn't a
7717 ;; statement block or a brace list, and if so return non-nil.
7718 ;;
7719 ;; If the check is successful, the return value is the start of the
7720 ;; keyword that tells what kind of construct it is, i.e. typically
7721 ;; what `c-decl-block-key' matched. Also, if GOTO-START is set then
7722 ;; the point will be at the start of the construct, before any
7723 ;; leading specifiers, otherwise it's at the returned position.
7724 ;;
7725 ;; The point is clobbered if the check is unsuccessful.
7726 ;;
7727 ;; CONTAINING-SEXP is the position of the open of the surrounding
7728 ;; paren, or nil if none.
7729 ;;
7730 ;; The optional LIMIT limits the backward search for the start of
7731 ;; the construct. It's assumed to be at a syntactically relevant
7732 ;; position.
7733 ;;
7734 ;; If any template arglists are found in the searched region before
7735 ;; the open brace, they get marked with paren syntax.
7736 ;;
7737 ;; This function might do hidden buffer changes.
7738
7739 (let ((open-brace (point)) kwd-start first-specifier-pos)
7740 (c-syntactic-skip-backward c-block-prefix-charset limit t)
7741
7742 (when (and c-recognize-<>-arglists
7743 (eq (char-before) ?>))
7744 ;; Could be at the end of a template arglist.
7745 (let ((c-parse-and-markup-<>-arglists t)
7746 (c-disallow-comma-in-<>-arglists
7747 (and containing-sexp
7748 (not (eq (char-after containing-sexp) ?{)))))
7749 (while (and
7750 (c-backward-<>-arglist nil limit)
7751 (progn
7752 (c-syntactic-skip-backward c-block-prefix-charset limit t)
7753 (eq (char-before) ?>))))))
7754
7755 ;; Note: Can't get bogus hits inside template arglists below since they
7756 ;; have gotten paren syntax above.
7757 (when (and
7758 ;; If `goto-start' is set we begin by searching for the
7759 ;; first possible position of a leading specifier list.
7760 ;; The `c-decl-block-key' search continues from there since
7761 ;; we know it can't match earlier.
7762 (if goto-start
7763 (when (c-syntactic-re-search-forward c-symbol-start
7764 open-brace t t)
7765 (goto-char (setq first-specifier-pos (match-beginning 0)))
7766 t)
7767 t)
7768
7769 (cond
7770 ((c-syntactic-re-search-forward c-decl-block-key open-brace t t t)
7771 (goto-char (setq kwd-start (match-beginning 0)))
7772 (or
7773
7774 ;; Found a keyword that can't be a type?
7775 (match-beginning 1)
7776
7777 ;; Can be a type too, in which case it's the return type of a
7778 ;; function (under the assumption that no declaration level
7779 ;; block construct starts with a type).
7780 (not (c-forward-type))
7781
7782 ;; Jumped over a type, but it could be a declaration keyword
7783 ;; followed by the declared identifier that we've jumped over
7784 ;; instead (e.g. in "class Foo {"). If it indeed is a type
7785 ;; then we should be at the declarator now, so check for a
7786 ;; valid declarator start.
7787 ;;
7788 ;; Note: This doesn't cope with the case when a declared
7789 ;; identifier is followed by e.g. '(' in a language where '('
7790 ;; also might be part of a declarator expression. Currently
7791 ;; there's no such language.
7792 (not (or (looking-at c-symbol-start)
7793 (looking-at c-type-decl-prefix-key)))))
7794
7795 ;; In Pike a list of modifiers may be followed by a brace
7796 ;; to make them apply to many identifiers. Note that the
7797 ;; match data will be empty on return in this case.
7798 ((and (c-major-mode-is 'pike-mode)
7799 (progn
7800 (goto-char open-brace)
7801 (= (c-backward-token-2) 0))
7802 (looking-at c-specifier-key)
7803 ;; Use this variant to avoid yet another special regexp.
7804 (c-keyword-member (c-keyword-sym (match-string 1))
7805 'c-modifier-kwds))
7806 (setq kwd-start (point))
7807 t)))
7808
7809 ;; Got a match.
7810
7811 (if goto-start
7812 ;; Back up over any preceding specifiers and their clauses
7813 ;; by going forward from `first-specifier-pos', which is the
7814 ;; earliest possible position where the specifier list can
7815 ;; start.
7816 (progn
7817 (goto-char first-specifier-pos)
7818
7819 (while (< (point) kwd-start)
7820 (if (looking-at c-symbol-key)
7821 ;; Accept any plain symbol token on the ground that
7822 ;; it's a specifier masked through a macro (just
7823 ;; like `c-forward-decl-or-cast-1' skip forward over
7824 ;; such tokens).
7825 ;;
7826 ;; Could be more restrictive wrt invalid keywords,
7827 ;; but that'd only occur in invalid code so there's
7828 ;; no use spending effort on it.
7829 (let ((end (match-end 0)))
7830 (unless (c-forward-keyword-clause 0)
7831 (goto-char end)
7832 (c-forward-syntactic-ws)))
7833
7834 ;; Can't parse a declaration preamble and is still
7835 ;; before `kwd-start'. That means `first-specifier-pos'
7836 ;; was in some earlier construct. Search again.
7837 (if (c-syntactic-re-search-forward c-symbol-start
7838 kwd-start 'move t)
7839 (goto-char (setq first-specifier-pos (match-beginning 0)))
7840 ;; Got no preamble before the block declaration keyword.
7841 (setq first-specifier-pos kwd-start))))
7842
7843 (goto-char first-specifier-pos))
7844 (goto-char kwd-start))
7845
7846 kwd-start)))
7847
7848 (defun c-search-uplist-for-classkey (paren-state)
7849 ;; Check if the closest containing paren sexp is a declaration
7850 ;; block, returning a 2 element vector in that case. Aref 0
7851 ;; contains the bufpos at boi of the class key line, and aref 1
7852 ;; contains the bufpos of the open brace. This function is an
7853 ;; obsolete wrapper for `c-looking-at-decl-block'.
7854 ;;
7855 ;; This function might do hidden buffer changes.
7856 (let ((open-paren-pos (c-most-enclosing-brace paren-state)))
7857 (when open-paren-pos
7858 (save-excursion
7859 (goto-char open-paren-pos)
7860 (when (and (eq (char-after) ?{)
7861 (c-looking-at-decl-block
7862 (c-safe-position open-paren-pos paren-state)
7863 nil))
7864 (back-to-indentation)
7865 (vector (point) open-paren-pos))))))
7866
7867 (defun c-inside-bracelist-p (containing-sexp paren-state)
7868 ;; return the buffer position of the beginning of the brace list
7869 ;; statement if we're inside a brace list, otherwise return nil.
7870 ;; CONTAINING-SEXP is the buffer pos of the innermost containing
7871 ;; paren. PAREN-STATE is the remainder of the state of enclosing
7872 ;; braces
7873 ;;
7874 ;; N.B.: This algorithm can potentially get confused by cpp macros
7875 ;; placed in inconvenient locations. It's a trade-off we make for
7876 ;; speed.
7877 ;;
7878 ;; This function might do hidden buffer changes.
7879 (or
7880 ;; This will pick up brace list declarations.
7881 (c-safe
7882 (save-excursion
7883 (goto-char containing-sexp)
7884 (c-forward-sexp -1)
7885 (let (bracepos)
7886 (if (and (or (looking-at c-brace-list-key)
7887 (progn (c-forward-sexp -1)
7888 (looking-at c-brace-list-key)))
7889 (setq bracepos (c-down-list-forward (point)))
7890 (not (c-crosses-statement-barrier-p (point)
7891 (- bracepos 2))))
7892 (point)))))
7893 ;; this will pick up array/aggregate init lists, even if they are nested.
7894 (save-excursion
7895 (let ((class-key
7896 ;; Pike can have class definitions anywhere, so we must
7897 ;; check for the class key here.
7898 (and (c-major-mode-is 'pike-mode)
7899 c-decl-block-key))
7900 bufpos braceassignp lim next-containing)
7901 (while (and (not bufpos)
7902 containing-sexp)
7903 (when paren-state
7904 (if (consp (car paren-state))
7905 (setq lim (cdr (car paren-state))
7906 paren-state (cdr paren-state))
7907 (setq lim (car paren-state)))
7908 (when paren-state
7909 (setq next-containing (car paren-state)
7910 paren-state (cdr paren-state))))
7911 (goto-char containing-sexp)
7912 (if (c-looking-at-inexpr-block next-containing next-containing)
7913 ;; We're in an in-expression block of some kind. Do not
7914 ;; check nesting. We deliberately set the limit to the
7915 ;; containing sexp, so that c-looking-at-inexpr-block
7916 ;; doesn't check for an identifier before it.
7917 (setq containing-sexp nil)
7918 ;; see if the open brace is preceded by = or [...] in
7919 ;; this statement, but watch out for operator=
7920 (setq braceassignp 'dontknow)
7921 (c-backward-token-2 1 t lim)
7922 ;; Checks to do only on the first sexp before the brace.
7923 (when (and c-opt-inexpr-brace-list-key
7924 (eq (char-after) ?\[))
7925 ;; In Java, an initialization brace list may follow
7926 ;; directly after "new Foo[]", so check for a "new"
7927 ;; earlier.
7928 (while (eq braceassignp 'dontknow)
7929 (setq braceassignp
7930 (cond ((/= (c-backward-token-2 1 t lim) 0) nil)
7931 ((looking-at c-opt-inexpr-brace-list-key) t)
7932 ((looking-at "\\sw\\|\\s_\\|[.[]")
7933 ;; Carry on looking if this is an
7934 ;; identifier (may contain "." in Java)
7935 ;; or another "[]" sexp.
7936 'dontknow)
7937 (t nil)))))
7938 ;; Checks to do on all sexps before the brace, up to the
7939 ;; beginning of the statement.
7940 (while (eq braceassignp 'dontknow)
7941 (cond ((eq (char-after) ?\;)
7942 (setq braceassignp nil))
7943 ((and class-key
7944 (looking-at class-key))
7945 (setq braceassignp nil))
7946 ((eq (char-after) ?=)
7947 ;; We've seen a =, but must check earlier tokens so
7948 ;; that it isn't something that should be ignored.
7949 (setq braceassignp 'maybe)
7950 (while (and (eq braceassignp 'maybe)
7951 (zerop (c-backward-token-2 1 t lim)))
7952 (setq braceassignp
7953 (cond
7954 ;; Check for operator =
7955 ((and c-opt-op-identifier-prefix
7956 (looking-at c-opt-op-identifier-prefix))
7957 nil)
7958 ;; Check for `<opchar>= in Pike.
7959 ((and (c-major-mode-is 'pike-mode)
7960 (or (eq (char-after) ?`)
7961 ;; Special case for Pikes
7962 ;; `[]=, since '[' is not in
7963 ;; the punctuation class.
7964 (and (eq (char-after) ?\[)
7965 (eq (char-before) ?`))))
7966 nil)
7967 ((looking-at "\\s.") 'maybe)
7968 ;; make sure we're not in a C++ template
7969 ;; argument assignment
7970 ((and
7971 (c-major-mode-is 'c++-mode)
7972 (save-excursion
7973 (let ((here (point))
7974 (pos< (progn
7975 (skip-chars-backward "^<>")
7976 (point))))
7977 (and (eq (char-before) ?<)
7978 (not (c-crosses-statement-barrier-p
7979 pos< here))
7980 (not (c-in-literal))
7981 ))))
7982 nil)
7983 (t t))))))
7984 (if (and (eq braceassignp 'dontknow)
7985 (/= (c-backward-token-2 1 t lim) 0))
7986 (setq braceassignp nil)))
7987 (if (not braceassignp)
7988 (if (eq (char-after) ?\;)
7989 ;; Brace lists can't contain a semicolon, so we're done.
7990 (setq containing-sexp nil)
7991 ;; Go up one level.
7992 (setq containing-sexp next-containing
7993 lim nil
7994 next-containing nil))
7995 ;; we've hit the beginning of the aggregate list
7996 (c-beginning-of-statement-1
7997 (c-most-enclosing-brace paren-state))
7998 (setq bufpos (point))))
7999 )
8000 bufpos))
8001 ))
8002
8003 (defun c-looking-at-special-brace-list (&optional lim)
8004 ;; If we're looking at the start of a pike-style list, ie `({ })',
8005 ;; `([ ])', `(< >)' etc, a cons of a cons of its starting and ending
8006 ;; positions and its entry in c-special-brace-lists is returned, nil
8007 ;; otherwise. The ending position is nil if the list is still open.
8008 ;; LIM is the limit for forward search. The point may either be at
8009 ;; the `(' or at the following paren character. Tries to check the
8010 ;; matching closer, but assumes it's correct if no balanced paren is
8011 ;; found (i.e. the case `({ ... } ... )' is detected as _not_ being
8012 ;; a special brace list).
8013 ;;
8014 ;; This function might do hidden buffer changes.
8015 (if c-special-brace-lists
8016 (condition-case ()
8017 (save-excursion
8018 (let ((beg (point))
8019 inner-beg end type)
8020 (c-forward-syntactic-ws)
8021 (if (eq (char-after) ?\()
8022 (progn
8023 (forward-char 1)
8024 (c-forward-syntactic-ws)
8025 (setq inner-beg (point))
8026 (setq type (assq (char-after) c-special-brace-lists)))
8027 (if (setq type (assq (char-after) c-special-brace-lists))
8028 (progn
8029 (setq inner-beg (point))
8030 (c-backward-syntactic-ws)
8031 (forward-char -1)
8032 (setq beg (if (eq (char-after) ?\()
8033 (point)
8034 nil)))))
8035 (if (and beg type)
8036 (if (and (c-safe
8037 (goto-char beg)
8038 (c-forward-sexp 1)
8039 (setq end (point))
8040 (= (char-before) ?\)))
8041 (c-safe
8042 (goto-char inner-beg)
8043 (if (looking-at "\\s(")
8044 ;; Check balancing of the inner paren
8045 ;; below.
8046 (progn
8047 (c-forward-sexp 1)
8048 t)
8049 ;; If the inner char isn't a paren then
8050 ;; we can't check balancing, so just
8051 ;; check the char before the outer
8052 ;; closing paren.
8053 (goto-char end)
8054 (backward-char)
8055 (c-backward-syntactic-ws)
8056 (= (char-before) (cdr type)))))
8057 (if (or (/= (char-syntax (char-before)) ?\))
8058 (= (progn
8059 (c-forward-syntactic-ws)
8060 (point))
8061 (1- end)))
8062 (cons (cons beg end) type))
8063 (cons (list beg) type)))))
8064 (error nil))))
8065
8066 (defun c-looking-at-bos (&optional lim)
8067 ;; Return non-nil if between two statements or declarations, assuming
8068 ;; point is not inside a literal or comment.
8069 ;;
8070 ;; Obsolete - `c-at-statement-start-p' or `c-at-expression-start-p'
8071 ;; are recommended instead.
8072 ;;
8073 ;; This function might do hidden buffer changes.
8074 (c-at-statement-start-p))
8075 (make-obsolete 'c-looking-at-bos 'c-at-statement-start-p "22.1")
8076
8077 (defun c-looking-at-inexpr-block (lim containing-sexp &optional check-at-end)
8078 ;; Return non-nil if we're looking at the beginning of a block
8079 ;; inside an expression. The value returned is actually a cons of
8080 ;; either 'inlambda, 'inexpr-statement or 'inexpr-class and the
8081 ;; position of the beginning of the construct.
8082 ;;
8083 ;; LIM limits the backward search. CONTAINING-SEXP is the start
8084 ;; position of the closest containing list. If it's nil, the
8085 ;; containing paren isn't used to decide whether we're inside an
8086 ;; expression or not. If both LIM and CONTAINING-SEXP are used, LIM
8087 ;; needs to be farther back.
8088 ;;
8089 ;; If CHECK-AT-END is non-nil then extra checks at the end of the
8090 ;; brace block might be done. It should only be used when the
8091 ;; construct can be assumed to be complete, i.e. when the original
8092 ;; starting position was further down than that.
8093 ;;
8094 ;; This function might do hidden buffer changes.
8095
8096 (save-excursion
8097 (let ((res 'maybe) passed-paren
8098 (closest-lim (or containing-sexp lim (point-min)))
8099 ;; Look at the character after point only as a last resort
8100 ;; when we can't disambiguate.
8101 (block-follows (and (eq (char-after) ?{) (point))))
8102
8103 (while (and (eq res 'maybe)
8104 (progn (c-backward-syntactic-ws)
8105 (> (point) closest-lim))
8106 (not (bobp))
8107 (progn (backward-char)
8108 (looking-at "[\]\).]\\|\\w\\|\\s_"))
8109 (c-safe (forward-char)
8110 (goto-char (scan-sexps (point) -1))))
8111
8112 (setq res
8113 (if (looking-at c-keywords-regexp)
8114 (let ((kw-sym (c-keyword-sym (match-string 1))))
8115 (cond
8116 ((and block-follows
8117 (c-keyword-member kw-sym 'c-inexpr-class-kwds))
8118 (and (not (eq passed-paren ?\[))
8119 (or (not (looking-at c-class-key))
8120 ;; If the class definition is at the start of
8121 ;; a statement, we don't consider it an
8122 ;; in-expression class.
8123 (let ((prev (point)))
8124 (while (and
8125 (= (c-backward-token-2 1 nil closest-lim) 0)
8126 (eq (char-syntax (char-after)) ?w))
8127 (setq prev (point)))
8128 (goto-char prev)
8129 (not (c-at-statement-start-p)))
8130 ;; Also, in Pike we treat it as an
8131 ;; in-expression class if it's used in an
8132 ;; object clone expression.
8133 (save-excursion
8134 (and check-at-end
8135 (c-major-mode-is 'pike-mode)
8136 (progn (goto-char block-follows)
8137 (zerop (c-forward-token-2 1 t)))
8138 (eq (char-after) ?\())))
8139 (cons 'inexpr-class (point))))
8140 ((c-keyword-member kw-sym 'c-inexpr-block-kwds)
8141 (when (not passed-paren)
8142 (cons 'inexpr-statement (point))))
8143 ((c-keyword-member kw-sym 'c-lambda-kwds)
8144 (when (or (not passed-paren)
8145 (eq passed-paren ?\())
8146 (cons 'inlambda (point))))
8147 ((c-keyword-member kw-sym 'c-block-stmt-kwds)
8148 nil)
8149 (t
8150 'maybe)))
8151
8152 (if (looking-at "\\s(")
8153 (if passed-paren
8154 (if (and (eq passed-paren ?\[)
8155 (eq (char-after) ?\[))
8156 ;; Accept several square bracket sexps for
8157 ;; Java array initializations.
8158 'maybe)
8159 (setq passed-paren (char-after))
8160 'maybe)
8161 'maybe))))
8162
8163 (if (eq res 'maybe)
8164 (when (and c-recognize-paren-inexpr-blocks
8165 block-follows
8166 containing-sexp
8167 (eq (char-after containing-sexp) ?\())
8168 (goto-char containing-sexp)
8169 (if (or (save-excursion
8170 (c-backward-syntactic-ws lim)
8171 (and (> (point) (or lim (point-min)))
8172 (c-on-identifier)))
8173 (and c-special-brace-lists
8174 (c-looking-at-special-brace-list)))
8175 nil
8176 (cons 'inexpr-statement (point))))
8177
8178 res))))
8179
8180 (defun c-looking-at-inexpr-block-backward (paren-state)
8181 ;; Returns non-nil if we're looking at the end of an in-expression
8182 ;; block, otherwise the same as `c-looking-at-inexpr-block'.
8183 ;; PAREN-STATE is the paren state relevant at the current position.
8184 ;;
8185 ;; This function might do hidden buffer changes.
8186 (save-excursion
8187 ;; We currently only recognize a block.
8188 (let ((here (point))
8189 (elem (car-safe paren-state))
8190 containing-sexp)
8191 (when (and (consp elem)
8192 (progn (goto-char (cdr elem))
8193 (c-forward-syntactic-ws here)
8194 (= (point) here)))
8195 (goto-char (car elem))
8196 (if (setq paren-state (cdr paren-state))
8197 (setq containing-sexp (car-safe paren-state)))
8198 (c-looking-at-inexpr-block (c-safe-position containing-sexp
8199 paren-state)
8200 containing-sexp)))))
8201
8202 \f
8203 ;; `c-guess-basic-syntax' and the functions that precedes it below
8204 ;; implements the main decision tree for determining the syntactic
8205 ;; analysis of the current line of code.
8206
8207 ;; Dynamically bound to t when `c-guess-basic-syntax' is called during
8208 ;; auto newline analysis.
8209 (defvar c-auto-newline-analysis nil)
8210
8211 (defun c-brace-anchor-point (bracepos)
8212 ;; BRACEPOS is the position of a brace in a construct like "namespace
8213 ;; Bar {". Return the anchor point in this construct; this is the
8214 ;; earliest symbol on the brace's line which isn't earlier than
8215 ;; "namespace".
8216 ;;
8217 ;; Currently (2007-08-17), "like namespace" means "matches
8218 ;; c-other-block-decl-kwds". It doesn't work with "class" or "struct"
8219 ;; or anything like that.
8220 (save-excursion
8221 (let ((boi (c-point 'boi bracepos)))
8222 (goto-char bracepos)
8223 (while (and (> (point) boi)
8224 (not (looking-at c-other-decl-block-key)))
8225 (c-backward-token-2))
8226 (if (> (point) boi) (point) boi))))
8227
8228 (defsubst c-add-syntax (symbol &rest args)
8229 ;; A simple function to prepend a new syntax element to
8230 ;; `c-syntactic-context'. Using `setq' on it is unsafe since it
8231 ;; should always be dynamically bound but since we read it first
8232 ;; we'll fail properly anyway if this function is misused.
8233 (setq c-syntactic-context (cons (cons symbol args)
8234 c-syntactic-context)))
8235
8236 (defsubst c-append-syntax (symbol &rest args)
8237 ;; Like `c-add-syntax' but appends to the end of the syntax list.
8238 ;; (Normally not necessary.)
8239 (setq c-syntactic-context (nconc c-syntactic-context
8240 (list (cons symbol args)))))
8241
8242 (defun c-add-stmt-syntax (syntax-symbol
8243 syntax-extra-args
8244 stop-at-boi-only
8245 containing-sexp
8246 paren-state)
8247 ;; Add the indicated SYNTAX-SYMBOL to `c-syntactic-context', extending it as
8248 ;; needed with further syntax elements of the types `substatement',
8249 ;; `inexpr-statement', `arglist-cont-nonempty', `statement-block-intro', and
8250 ;; `defun-block-intro'.
8251 ;;
8252 ;; Do the generic processing to anchor the given syntax symbol on
8253 ;; the preceding statement: Skip over any labels and containing
8254 ;; statements on the same line, and then search backward until we
8255 ;; find a statement or block start that begins at boi without a
8256 ;; label or comment.
8257 ;;
8258 ;; Point is assumed to be at the prospective anchor point for the
8259 ;; given SYNTAX-SYMBOL. More syntax entries are added if we need to
8260 ;; skip past open parens and containing statements. Most of the added
8261 ;; syntax elements will get the same anchor point - the exception is
8262 ;; for an anchor in a construct like "namespace"[*] - this is as early
8263 ;; as possible in the construct but on the same line as the {.
8264 ;;
8265 ;; [*] i.e. with a keyword matching c-other-block-decl-kwds.
8266 ;;
8267 ;; SYNTAX-EXTRA-ARGS are a list of the extra arguments for the
8268 ;; syntax symbol. They are appended after the anchor point.
8269 ;;
8270 ;; If STOP-AT-BOI-ONLY is nil, we can stop in the middle of the line
8271 ;; if the current statement starts there.
8272 ;;
8273 ;; Note: It's not a problem if PAREN-STATE "overshoots"
8274 ;; CONTAINING-SEXP, i.e. contains info about parens further down.
8275 ;;
8276 ;; This function might do hidden buffer changes.
8277
8278 (if (= (point) (c-point 'boi))
8279 ;; This is by far the most common case, so let's give it special
8280 ;; treatment.
8281 (apply 'c-add-syntax syntax-symbol (point) syntax-extra-args)
8282
8283 (let ((syntax-last c-syntactic-context)
8284 (boi (c-point 'boi))
8285 ;; Set when we're on a label, so that we don't stop there.
8286 ;; FIXME: To be complete we should check if we're on a label
8287 ;; now at the start.
8288 on-label)
8289
8290 ;; Use point as the anchor point for "namespace", "extern", etc.
8291 (apply 'c-add-syntax syntax-symbol
8292 (if (rassq syntax-symbol c-other-decl-block-key-in-symbols-alist)
8293 (point) nil)
8294 syntax-extra-args)
8295
8296 ;; Loop while we have to back out of containing blocks.
8297 (while
8298 (and
8299 (catch 'back-up-block
8300
8301 ;; Loop while we have to back up statements.
8302 (while (or (/= (point) boi)
8303 on-label
8304 (looking-at c-comment-start-regexp))
8305
8306 ;; Skip past any comments that stands between the
8307 ;; statement start and boi.
8308 (let ((savepos (point)))
8309 (while (and (/= savepos boi)
8310 (c-backward-single-comment))
8311 (setq savepos (point)
8312 boi (c-point 'boi)))
8313 (goto-char savepos))
8314
8315 ;; Skip to the beginning of this statement or backward
8316 ;; another one.
8317 (let ((old-pos (point))
8318 (old-boi boi)
8319 (step-type (c-beginning-of-statement-1 containing-sexp)))
8320 (setq boi (c-point 'boi)
8321 on-label (eq step-type 'label))
8322
8323 (cond ((= (point) old-pos)
8324 ;; If we didn't move we're at the start of a block and
8325 ;; have to continue outside it.
8326 (throw 'back-up-block t))
8327
8328 ((and (eq step-type 'up)
8329 (>= (point) old-boi)
8330 (looking-at "else\\>[^_]")
8331 (save-excursion
8332 (goto-char old-pos)
8333 (looking-at "if\\>[^_]")))
8334 ;; Special case to avoid deeper and deeper indentation
8335 ;; of "else if" clauses.
8336 )
8337
8338 ((and (not stop-at-boi-only)
8339 (/= old-pos old-boi)
8340 (memq step-type '(up previous)))
8341 ;; If stop-at-boi-only is nil, we shouldn't back up
8342 ;; over previous or containing statements to try to
8343 ;; reach boi, so go back to the last position and
8344 ;; exit.
8345 (goto-char old-pos)
8346 (throw 'back-up-block nil))
8347
8348 (t
8349 (if (and (not stop-at-boi-only)
8350 (memq step-type '(up previous beginning)))
8351 ;; If we've moved into another statement then we
8352 ;; should no longer try to stop in the middle of a
8353 ;; line.
8354 (setq stop-at-boi-only t))
8355
8356 ;; Record this as a substatement if we skipped up one
8357 ;; level.
8358 (when (eq step-type 'up)
8359 (c-add-syntax 'substatement nil))))
8360 )))
8361
8362 containing-sexp)
8363
8364 ;; Now we have to go out of this block.
8365 (goto-char containing-sexp)
8366
8367 ;; Don't stop in the middle of a special brace list opener
8368 ;; like "({".
8369 (when c-special-brace-lists
8370 (let ((special-list (c-looking-at-special-brace-list)))
8371 (when (and special-list
8372 (< (car (car special-list)) (point)))
8373 (setq containing-sexp (car (car special-list)))
8374 (goto-char containing-sexp))))
8375
8376 (setq paren-state (c-whack-state-after containing-sexp paren-state)
8377 containing-sexp (c-most-enclosing-brace paren-state)
8378 boi (c-point 'boi))
8379
8380 ;; Analyze the construct in front of the block we've stepped out
8381 ;; from and add the right syntactic element for it.
8382 (let ((paren-pos (point))
8383 (paren-char (char-after))
8384 step-type)
8385
8386 (if (eq paren-char ?\()
8387 ;; Stepped out of a parenthesis block, so we're in an
8388 ;; expression now.
8389 (progn
8390 (when (/= paren-pos boi)
8391 (if (and c-recognize-paren-inexpr-blocks
8392 (progn
8393 (c-backward-syntactic-ws containing-sexp)
8394 (or (not (looking-at "\\>"))
8395 (not (c-on-identifier))))
8396 (save-excursion
8397 (goto-char (1+ paren-pos))
8398 (c-forward-syntactic-ws)
8399 (eq (char-after) ?{)))
8400 ;; Stepped out of an in-expression statement. This
8401 ;; syntactic element won't get an anchor pos.
8402 (c-add-syntax 'inexpr-statement)
8403
8404 ;; A parenthesis normally belongs to an arglist.
8405 (c-add-syntax 'arglist-cont-nonempty nil paren-pos)))
8406
8407 (goto-char (max boi
8408 (if containing-sexp
8409 (1+ containing-sexp)
8410 (point-min))))
8411 (setq step-type 'same
8412 on-label nil))
8413
8414 ;; Stepped out of a brace block.
8415 (setq step-type (c-beginning-of-statement-1 containing-sexp)
8416 on-label (eq step-type 'label))
8417
8418 (if (and (eq step-type 'same)
8419 (/= paren-pos (point)))
8420 (let (inexpr)
8421 (cond
8422 ((save-excursion
8423 (goto-char paren-pos)
8424 (setq inexpr (c-looking-at-inexpr-block
8425 (c-safe-position containing-sexp paren-state)
8426 containing-sexp)))
8427 (c-add-syntax (if (eq (car inexpr) 'inlambda)
8428 'defun-block-intro
8429 'statement-block-intro)
8430 nil))
8431 ((looking-at c-other-decl-block-key)
8432 (c-add-syntax
8433 (cdr (assoc (match-string 1)
8434 c-other-decl-block-key-in-symbols-alist))
8435 (max (c-point 'boi paren-pos) (point))))
8436 (t (c-add-syntax 'defun-block-intro nil))))
8437
8438 (c-add-syntax 'statement-block-intro nil)))
8439
8440 (if (= paren-pos boi)
8441 ;; Always done if the open brace was at boi. The
8442 ;; c-beginning-of-statement-1 call above is necessary
8443 ;; anyway, to decide the type of block-intro to add.
8444 (goto-char paren-pos)
8445 (setq boi (c-point 'boi)))
8446 ))
8447
8448 ;; Fill in the current point as the anchor for all the symbols
8449 ;; added above.
8450 (let ((p c-syntactic-context) q)
8451 (while (not (eq p syntax-last))
8452 (setq q (cdr (car p))) ; e.g. (nil 28) [from (arglist-cont-nonempty nil 28)]
8453 (while q
8454 (unless (car q)
8455 (setcar q (point)))
8456 (setq q (cdr q)))
8457 (setq p (cdr p))))
8458 )))
8459
8460 (defun c-add-class-syntax (symbol
8461 containing-decl-open
8462 containing-decl-start
8463 containing-decl-kwd
8464 paren-state)
8465 ;; The inclass and class-close syntactic symbols are added in
8466 ;; several places and some work is needed to fix everything.
8467 ;; Therefore it's collected here.
8468 ;;
8469 ;; This function might do hidden buffer changes.
8470 (goto-char containing-decl-open)
8471 (if (and (eq symbol 'inclass) (= (point) (c-point 'boi)))
8472 (progn
8473 (c-add-syntax symbol containing-decl-open)
8474 containing-decl-open)
8475 (goto-char containing-decl-start)
8476 ;; Ought to use `c-add-stmt-syntax' instead of backing up to boi
8477 ;; here, but we have to do like this for compatibility.
8478 (back-to-indentation)
8479 (c-add-syntax symbol (point))
8480 (if (and (c-keyword-member containing-decl-kwd
8481 'c-inexpr-class-kwds)
8482 (/= containing-decl-start (c-point 'boi containing-decl-start)))
8483 (c-add-syntax 'inexpr-class))
8484 (point)))
8485
8486 (defun c-guess-continued-construct (indent-point
8487 char-after-ip
8488 beg-of-same-or-containing-stmt
8489 containing-sexp
8490 paren-state)
8491 ;; This function contains the decision tree reached through both
8492 ;; cases 18 and 10. It's a continued statement or top level
8493 ;; construct of some kind.
8494 ;;
8495 ;; This function might do hidden buffer changes.
8496
8497 (let (special-brace-list)
8498 (goto-char indent-point)
8499 (skip-chars-forward " \t")
8500
8501 (cond
8502 ;; (CASE A removed.)
8503 ;; CASE B: open braces for class or brace-lists
8504 ((setq special-brace-list
8505 (or (and c-special-brace-lists
8506 (c-looking-at-special-brace-list))
8507 (eq char-after-ip ?{)))
8508
8509 (cond
8510 ;; CASE B.1: class-open
8511 ((save-excursion
8512 (and (eq (char-after) ?{)
8513 (c-looking-at-decl-block containing-sexp t)
8514 (setq beg-of-same-or-containing-stmt (point))))
8515 (c-add-syntax 'class-open beg-of-same-or-containing-stmt))
8516
8517 ;; CASE B.2: brace-list-open
8518 ((or (consp special-brace-list)
8519 (save-excursion
8520 (goto-char beg-of-same-or-containing-stmt)
8521 (c-syntactic-re-search-forward "=\\([^=]\\|$\\)"
8522 indent-point t t t)))
8523 ;; The most semantically accurate symbol here is
8524 ;; brace-list-open, but we normally report it simply as a
8525 ;; statement-cont. The reason is that one normally adjusts
8526 ;; brace-list-open for brace lists as top-level constructs,
8527 ;; and brace lists inside statements is a completely different
8528 ;; context. C.f. case 5A.3.
8529 (c-beginning-of-statement-1 containing-sexp)
8530 (c-add-stmt-syntax (if c-auto-newline-analysis
8531 ;; Turn off the dwim above when we're
8532 ;; analyzing the nature of the brace
8533 ;; for the auto newline feature.
8534 'brace-list-open
8535 'statement-cont)
8536 nil nil
8537 containing-sexp paren-state))
8538
8539 ;; CASE B.3: The body of a function declared inside a normal
8540 ;; block. Can occur e.g. in Pike and when using gcc
8541 ;; extensions, but watch out for macros followed by blocks.
8542 ;; C.f. cases E, 16F and 17G.
8543 ((and (not (c-at-statement-start-p))
8544 (eq (c-beginning-of-statement-1 containing-sexp nil nil t)
8545 'same)
8546 (save-excursion
8547 (let ((c-recognize-typeless-decls nil))
8548 ;; Turn off recognition of constructs that lacks a
8549 ;; type in this case, since that's more likely to be
8550 ;; a macro followed by a block.
8551 (c-forward-decl-or-cast-1 (c-point 'bosws) nil nil))))
8552 (c-add-stmt-syntax 'defun-open nil t
8553 containing-sexp paren-state))
8554
8555 ;; CASE B.4: Continued statement with block open. The most
8556 ;; accurate analysis is perhaps `statement-cont' together with
8557 ;; `block-open' but we play DWIM and use `substatement-open'
8558 ;; instead. The rationaly is that this typically is a macro
8559 ;; followed by a block which makes it very similar to a
8560 ;; statement with a substatement block.
8561 (t
8562 (c-add-stmt-syntax 'substatement-open nil nil
8563 containing-sexp paren-state))
8564 ))
8565
8566 ;; CASE C: iostream insertion or extraction operator
8567 ((and (looking-at "\\(<<\\|>>\\)\\([^=]\\|$\\)")
8568 (save-excursion
8569 (goto-char beg-of-same-or-containing-stmt)
8570 ;; If there is no preceding streamop in the statement
8571 ;; then indent this line as a normal statement-cont.
8572 (when (c-syntactic-re-search-forward
8573 "\\(<<\\|>>\\)\\([^=]\\|$\\)" indent-point 'move t t)
8574 (c-add-syntax 'stream-op (c-point 'boi))
8575 t))))
8576
8577 ;; CASE E: In the "K&R region" of a function declared inside a
8578 ;; normal block. C.f. case B.3.
8579 ((and (save-excursion
8580 ;; Check that the next token is a '{'. This works as
8581 ;; long as no language that allows nested function
8582 ;; definitions allows stuff like member init lists, K&R
8583 ;; declarations or throws clauses there.
8584 ;;
8585 ;; Note that we do a forward search for something ahead
8586 ;; of the indentation line here. That's not good since
8587 ;; the user might not have typed it yet. Unfortunately
8588 ;; it's exceedingly tricky to recognize a function
8589 ;; prototype in a code block without resorting to this.
8590 (c-forward-syntactic-ws)
8591 (eq (char-after) ?{))
8592 (not (c-at-statement-start-p))
8593 (eq (c-beginning-of-statement-1 containing-sexp nil nil t)
8594 'same)
8595 (save-excursion
8596 (let ((c-recognize-typeless-decls nil))
8597 ;; Turn off recognition of constructs that lacks a
8598 ;; type in this case, since that's more likely to be
8599 ;; a macro followed by a block.
8600 (c-forward-decl-or-cast-1 (c-point 'bosws) nil nil))))
8601 (c-add-stmt-syntax 'func-decl-cont nil t
8602 containing-sexp paren-state))
8603
8604 ;; CASE D: continued statement.
8605 (t
8606 (c-beginning-of-statement-1 containing-sexp)
8607 (c-add-stmt-syntax 'statement-cont nil nil
8608 containing-sexp paren-state))
8609 )))
8610
8611 ;; The next autoload was added by RMS on 2005/8/9 - don't know why (ACM,
8612 ;; 2005/11/29).
8613 ;;;###autoload
8614 (defun c-guess-basic-syntax ()
8615 "Return the syntactic context of the current line."
8616 (save-excursion
8617 (beginning-of-line)
8618 (c-save-buffer-state
8619 ((indent-point (point))
8620 (case-fold-search nil)
8621 ;; A whole ugly bunch of various temporary variables. Have
8622 ;; to declare them here since it's not possible to declare
8623 ;; a variable with only the scope of a cond test and the
8624 ;; following result clauses, and most of this function is a
8625 ;; single gigantic cond. :P
8626 literal char-before-ip before-ws-ip char-after-ip macro-start
8627 in-macro-expr c-syntactic-context placeholder c-in-literal-cache
8628 step-type tmpsymbol keyword injava-inher special-brace-list tmp-pos
8629 containing-<
8630 ;; The following record some positions for the containing
8631 ;; declaration block if we're directly within one:
8632 ;; `containing-decl-open' is the position of the open
8633 ;; brace. `containing-decl-start' is the start of the
8634 ;; declaration. `containing-decl-kwd' is the keyword
8635 ;; symbol of the keyword that tells what kind of block it
8636 ;; is.
8637 containing-decl-open
8638 containing-decl-start
8639 containing-decl-kwd
8640 ;; The open paren of the closest surrounding sexp or nil if
8641 ;; there is none.
8642 containing-sexp
8643 ;; The position after the closest preceding brace sexp
8644 ;; (nested sexps are ignored), or the position after
8645 ;; `containing-sexp' if there is none, or (point-min) if
8646 ;; `containing-sexp' is nil.
8647 lim
8648 ;; The paren state outside `containing-sexp', or at
8649 ;; `indent-point' if `containing-sexp' is nil.
8650 (paren-state (c-parse-state))
8651 ;; There's always at most one syntactic element which got
8652 ;; an anchor pos. It's stored in syntactic-relpos.
8653 syntactic-relpos
8654 (c-stmt-delim-chars c-stmt-delim-chars))
8655
8656 ;; Check if we're directly inside an enclosing declaration
8657 ;; level block.
8658 (when (and (setq containing-sexp
8659 (c-most-enclosing-brace paren-state))
8660 (progn
8661 (goto-char containing-sexp)
8662 (eq (char-after) ?{))
8663 (setq placeholder
8664 (c-looking-at-decl-block
8665 (c-most-enclosing-brace paren-state
8666 containing-sexp)
8667 t)))
8668 (setq containing-decl-open containing-sexp
8669 containing-decl-start (point)
8670 containing-sexp nil)
8671 (goto-char placeholder)
8672 (setq containing-decl-kwd (and (looking-at c-keywords-regexp)
8673 (c-keyword-sym (match-string 1)))))
8674
8675 ;; Init some position variables.
8676 (if c-state-cache
8677 (progn
8678 (setq containing-sexp (car paren-state)
8679 paren-state (cdr paren-state))
8680 (if (consp containing-sexp)
8681 (progn
8682 (setq lim (cdr containing-sexp))
8683 (if (cdr c-state-cache)
8684 ;; Ignore balanced paren. The next entry
8685 ;; can't be another one.
8686 (setq containing-sexp (car (cdr c-state-cache))
8687 paren-state (cdr paren-state))
8688 ;; If there is no surrounding open paren then
8689 ;; put the last balanced pair back on paren-state.
8690 (setq paren-state (cons containing-sexp paren-state)
8691 containing-sexp nil)))
8692 (setq lim (1+ containing-sexp))))
8693 (setq lim (point-min)))
8694
8695 ;; If we're in a parenthesis list then ',' delimits the
8696 ;; "statements" rather than being an operator (with the
8697 ;; exception of the "for" clause). This difference is
8698 ;; typically only noticeable when statements are used in macro
8699 ;; arglists.
8700 (when (and containing-sexp
8701 (eq (char-after containing-sexp) ?\())
8702 (setq c-stmt-delim-chars c-stmt-delim-chars-with-comma))
8703
8704 ;; cache char before and after indent point, and move point to
8705 ;; the most likely position to perform the majority of tests
8706 (goto-char indent-point)
8707 (c-backward-syntactic-ws lim)
8708 (setq before-ws-ip (point)
8709 char-before-ip (char-before))
8710 (goto-char indent-point)
8711 (skip-chars-forward " \t")
8712 (setq char-after-ip (char-after))
8713
8714 ;; are we in a literal?
8715 (setq literal (c-in-literal lim))
8716
8717 ;; now figure out syntactic qualities of the current line
8718 (cond
8719
8720 ;; CASE 1: in a string.
8721 ((eq literal 'string)
8722 (c-add-syntax 'string (c-point 'bopl)))
8723
8724 ;; CASE 2: in a C or C++ style comment.
8725 ((and (memq literal '(c c++))
8726 ;; This is a kludge for XEmacs where we use
8727 ;; `buffer-syntactic-context', which doesn't correctly
8728 ;; recognize "\*/" to end a block comment.
8729 ;; `parse-partial-sexp' which is used by
8730 ;; `c-literal-limits' will however do that in most
8731 ;; versions, which results in that we get nil from
8732 ;; `c-literal-limits' even when `c-in-literal' claims
8733 ;; we're inside a comment.
8734 (setq placeholder (c-literal-limits lim)))
8735 (c-add-syntax literal (car placeholder)))
8736
8737 ;; CASE 3: in a cpp preprocessor macro continuation.
8738 ((and (save-excursion
8739 (when (c-beginning-of-macro)
8740 (setq macro-start (point))))
8741 (/= macro-start (c-point 'boi))
8742 (progn
8743 (setq tmpsymbol 'cpp-macro-cont)
8744 (or (not c-syntactic-indentation-in-macros)
8745 (save-excursion
8746 (goto-char macro-start)
8747 ;; If at the beginning of the body of a #define
8748 ;; directive then analyze as cpp-define-intro
8749 ;; only. Go on with the syntactic analysis
8750 ;; otherwise. in-macro-expr is set if we're in a
8751 ;; cpp expression, i.e. before the #define body
8752 ;; or anywhere in a non-#define directive.
8753 (if (c-forward-to-cpp-define-body)
8754 (let ((indent-boi (c-point 'boi indent-point)))
8755 (setq in-macro-expr (> (point) indent-boi)
8756 tmpsymbol 'cpp-define-intro)
8757 (= (point) indent-boi))
8758 (setq in-macro-expr t)
8759 nil)))))
8760 (c-add-syntax tmpsymbol macro-start)
8761 (setq macro-start nil))
8762
8763 ;; CASE 11: an else clause?
8764 ((looking-at "else\\>[^_]")
8765 (c-beginning-of-statement-1 containing-sexp)
8766 (c-add-stmt-syntax 'else-clause nil t
8767 containing-sexp paren-state))
8768
8769 ;; CASE 12: while closure of a do/while construct?
8770 ((and (looking-at "while\\>[^_]")
8771 (save-excursion
8772 (prog1 (eq (c-beginning-of-statement-1 containing-sexp)
8773 'beginning)
8774 (setq placeholder (point)))))
8775 (goto-char placeholder)
8776 (c-add-stmt-syntax 'do-while-closure nil t
8777 containing-sexp paren-state))
8778
8779 ;; CASE 13: A catch or finally clause? This case is simpler
8780 ;; than if-else and do-while, because a block is required
8781 ;; after every try, catch and finally.
8782 ((save-excursion
8783 (and (cond ((c-major-mode-is 'c++-mode)
8784 (looking-at "catch\\>[^_]"))
8785 ((c-major-mode-is 'java-mode)
8786 (looking-at "\\(catch\\|finally\\)\\>[^_]")))
8787 (and (c-safe (c-backward-syntactic-ws)
8788 (c-backward-sexp)
8789 t)
8790 (eq (char-after) ?{)
8791 (c-safe (c-backward-syntactic-ws)
8792 (c-backward-sexp)
8793 t)
8794 (if (eq (char-after) ?\()
8795 (c-safe (c-backward-sexp) t)
8796 t))
8797 (looking-at "\\(try\\|catch\\)\\>[^_]")
8798 (setq placeholder (point))))
8799 (goto-char placeholder)
8800 (c-add-stmt-syntax 'catch-clause nil t
8801 containing-sexp paren-state))
8802
8803 ;; CASE 18: A substatement we can recognize by keyword.
8804 ((save-excursion
8805 (and c-opt-block-stmt-key
8806 (not (eq char-before-ip ?\;))
8807 (not (c-at-vsemi-p before-ws-ip))
8808 (not (memq char-after-ip '(?\) ?\] ?,)))
8809 (or (not (eq char-before-ip ?}))
8810 (c-looking-at-inexpr-block-backward c-state-cache))
8811 (> (point)
8812 (progn
8813 ;; Ought to cache the result from the
8814 ;; c-beginning-of-statement-1 calls here.
8815 (setq placeholder (point))
8816 (while (eq (setq step-type
8817 (c-beginning-of-statement-1 lim))
8818 'label))
8819 (if (eq step-type 'previous)
8820 (goto-char placeholder)
8821 (setq placeholder (point))
8822 (if (and (eq step-type 'same)
8823 (not (looking-at c-opt-block-stmt-key)))
8824 ;; Step up to the containing statement if we
8825 ;; stayed in the same one.
8826 (let (step)
8827 (while (eq
8828 (setq step
8829 (c-beginning-of-statement-1 lim))
8830 'label))
8831 (if (eq step 'up)
8832 (setq placeholder (point))
8833 ;; There was no containing statement afterall.
8834 (goto-char placeholder)))))
8835 placeholder))
8836 (if (looking-at c-block-stmt-2-key)
8837 ;; Require a parenthesis after these keywords.
8838 ;; Necessary to catch e.g. synchronized in Java,
8839 ;; which can be used both as statement and
8840 ;; modifier.
8841 (and (zerop (c-forward-token-2 1 nil))
8842 (eq (char-after) ?\())
8843 (looking-at c-opt-block-stmt-key))))
8844
8845 (if (eq step-type 'up)
8846 ;; CASE 18A: Simple substatement.
8847 (progn
8848 (goto-char placeholder)
8849 (cond
8850 ((eq char-after-ip ?{)
8851 (c-add-stmt-syntax 'substatement-open nil nil
8852 containing-sexp paren-state))
8853 ((save-excursion
8854 (goto-char indent-point)
8855 (back-to-indentation)
8856 (c-forward-label))
8857 (c-add-stmt-syntax 'substatement-label nil nil
8858 containing-sexp paren-state))
8859 (t
8860 (c-add-stmt-syntax 'substatement nil nil
8861 containing-sexp paren-state))))
8862
8863 ;; CASE 18B: Some other substatement. This is shared
8864 ;; with case 10.
8865 (c-guess-continued-construct indent-point
8866 char-after-ip
8867 placeholder
8868 lim
8869 paren-state)))
8870
8871 ;; CASE 14: A case or default label
8872 ((looking-at c-label-kwds-regexp)
8873 (if containing-sexp
8874 (progn
8875 (goto-char containing-sexp)
8876 (setq lim (c-most-enclosing-brace c-state-cache
8877 containing-sexp))
8878 (c-backward-to-block-anchor lim)
8879 (c-add-stmt-syntax 'case-label nil t lim paren-state))
8880 ;; Got a bogus label at the top level. In lack of better
8881 ;; alternatives, anchor it on (point-min).
8882 (c-add-syntax 'case-label (point-min))))
8883
8884 ;; CASE 15: any other label
8885 ((save-excursion
8886 (back-to-indentation)
8887 (and (not (looking-at c-syntactic-ws-start))
8888 (c-forward-label)))
8889 (cond (containing-decl-open
8890 (setq placeholder (c-add-class-syntax 'inclass
8891 containing-decl-open
8892 containing-decl-start
8893 containing-decl-kwd
8894 paren-state))
8895 ;; Append access-label with the same anchor point as
8896 ;; inclass gets.
8897 (c-append-syntax 'access-label placeholder))
8898
8899 (containing-sexp
8900 (goto-char containing-sexp)
8901 (setq lim (c-most-enclosing-brace c-state-cache
8902 containing-sexp))
8903 (save-excursion
8904 (setq tmpsymbol
8905 (if (and (eq (c-beginning-of-statement-1 lim) 'up)
8906 (looking-at "switch\\>[^_]"))
8907 ;; If the surrounding statement is a switch then
8908 ;; let's analyze all labels as switch labels, so
8909 ;; that they get lined up consistently.
8910 'case-label
8911 'label)))
8912 (c-backward-to-block-anchor lim)
8913 (c-add-stmt-syntax tmpsymbol nil t lim paren-state))
8914
8915 (t
8916 ;; A label on the top level. Treat it as a class
8917 ;; context. (point-min) is the closest we get to the
8918 ;; class open brace.
8919 (c-add-syntax 'access-label (point-min)))))
8920
8921 ;; CASE 4: In-expression statement. C.f. cases 7B, 16A and
8922 ;; 17E.
8923 ((setq placeholder (c-looking-at-inexpr-block
8924 (c-safe-position containing-sexp paren-state)
8925 containing-sexp
8926 ;; Have to turn on the heuristics after
8927 ;; the point even though it doesn't work
8928 ;; very well. C.f. test case class-16.pike.
8929 t))
8930 (setq tmpsymbol (assq (car placeholder)
8931 '((inexpr-class . class-open)
8932 (inexpr-statement . block-open))))
8933 (if tmpsymbol
8934 ;; It's a statement block or an anonymous class.
8935 (setq tmpsymbol (cdr tmpsymbol))
8936 ;; It's a Pike lambda. Check whether we are between the
8937 ;; lambda keyword and the argument list or at the defun
8938 ;; opener.
8939 (setq tmpsymbol (if (eq char-after-ip ?{)
8940 'inline-open
8941 'lambda-intro-cont)))
8942 (goto-char (cdr placeholder))
8943 (back-to-indentation)
8944 (c-add-stmt-syntax tmpsymbol nil t
8945 (c-most-enclosing-brace c-state-cache (point))
8946 paren-state)
8947 (unless (eq (point) (cdr placeholder))
8948 (c-add-syntax (car placeholder))))
8949
8950 ;; CASE 5: Line is inside a declaration level block or at top level.
8951 ((or containing-decl-open (null containing-sexp))
8952 (cond
8953
8954 ;; CASE 5A: we are looking at a defun, brace list, class,
8955 ;; or inline-inclass method opening brace
8956 ((setq special-brace-list
8957 (or (and c-special-brace-lists
8958 (c-looking-at-special-brace-list))
8959 (eq char-after-ip ?{)))
8960 (cond
8961
8962 ;; CASE 5A.1: Non-class declaration block open.
8963 ((save-excursion
8964 (let (tmp)
8965 (and (eq char-after-ip ?{)
8966 (setq tmp (c-looking-at-decl-block containing-sexp t))
8967 (progn
8968 (setq placeholder (point))
8969 (goto-char tmp)
8970 (looking-at c-symbol-key))
8971 (c-keyword-member
8972 (c-keyword-sym (setq keyword (match-string 0)))
8973 'c-other-block-decl-kwds))))
8974 (goto-char placeholder)
8975 (c-add-stmt-syntax
8976 (if (string-equal keyword "extern")
8977 ;; Special case for extern-lang-open.
8978 'extern-lang-open
8979 (intern (concat keyword "-open")))
8980 nil t containing-sexp paren-state))
8981
8982 ;; CASE 5A.2: we are looking at a class opening brace
8983 ((save-excursion
8984 (goto-char indent-point)
8985 (skip-chars-forward " \t")
8986 (and (eq (char-after) ?{)
8987 (c-looking-at-decl-block containing-sexp t)
8988 (setq placeholder (point))))
8989 (c-add-syntax 'class-open placeholder))
8990
8991 ;; CASE 5A.3: brace list open
8992 ((save-excursion
8993 (c-beginning-of-decl-1 lim)
8994 (while (looking-at c-specifier-key)
8995 (goto-char (match-end 1))
8996 (c-forward-syntactic-ws indent-point))
8997 (setq placeholder (c-point 'boi))
8998 (or (consp special-brace-list)
8999 (and (or (save-excursion
9000 (goto-char indent-point)
9001 (setq tmpsymbol nil)
9002 (while (and (> (point) placeholder)
9003 (zerop (c-backward-token-2 1 t))
9004 (/= (char-after) ?=))
9005 (and c-opt-inexpr-brace-list-key
9006 (not tmpsymbol)
9007 (looking-at c-opt-inexpr-brace-list-key)
9008 (setq tmpsymbol 'topmost-intro-cont)))
9009 (eq (char-after) ?=))
9010 (looking-at c-brace-list-key))
9011 (save-excursion
9012 (while (and (< (point) indent-point)
9013 (zerop (c-forward-token-2 1 t))
9014 (not (memq (char-after) '(?\; ?\()))))
9015 (not (memq (char-after) '(?\; ?\()))
9016 ))))
9017 (if (and (not c-auto-newline-analysis)
9018 (c-major-mode-is 'java-mode)
9019 (eq tmpsymbol 'topmost-intro-cont))
9020 ;; We're in Java and have found that the open brace
9021 ;; belongs to a "new Foo[]" initialization list,
9022 ;; which means the brace list is part of an
9023 ;; expression and not a top level definition. We
9024 ;; therefore treat it as any topmost continuation
9025 ;; even though the semantically correct symbol still
9026 ;; is brace-list-open, on the same grounds as in
9027 ;; case B.2.
9028 (progn
9029 (c-beginning-of-statement-1 lim)
9030 (c-add-syntax 'topmost-intro-cont (c-point 'boi)))
9031 (c-add-syntax 'brace-list-open placeholder)))
9032
9033 ;; CASE 5A.4: inline defun open
9034 ((and containing-decl-open
9035 (not (c-keyword-member containing-decl-kwd
9036 'c-other-block-decl-kwds)))
9037 (c-add-syntax 'inline-open)
9038 (c-add-class-syntax 'inclass
9039 containing-decl-open
9040 containing-decl-start
9041 containing-decl-kwd
9042 paren-state))
9043
9044 ;; CASE 5A.5: ordinary defun open
9045 (t
9046 (save-excursion
9047 (c-beginning-of-decl-1 lim)
9048 (while (looking-at c-specifier-key)
9049 (goto-char (match-end 1))
9050 (c-forward-syntactic-ws indent-point))
9051 (c-add-syntax 'defun-open (c-point 'boi))
9052 ;; Bogus to use bol here, but it's the legacy. (Resolved,
9053 ;; 2007-11-09)
9054 ))))
9055
9056 ;; CASE 5B: After a function header but before the body (or
9057 ;; the ending semicolon if there's no body).
9058 ((save-excursion
9059 (when (setq placeholder (c-just-after-func-arglist-p lim))
9060 (setq tmp-pos (point))))
9061 (cond
9062
9063 ;; CASE 5B.1: Member init list.
9064 ((eq (char-after tmp-pos) ?:)
9065 (if (or (> tmp-pos indent-point)
9066 (= (c-point 'bosws) (1+ tmp-pos)))
9067 (progn
9068 ;; There is no preceding member init clause.
9069 ;; Indent relative to the beginning of indentation
9070 ;; for the topmost-intro line that contains the
9071 ;; prototype's open paren.
9072 (goto-char placeholder)
9073 (c-add-syntax 'member-init-intro (c-point 'boi)))
9074 ;; Indent relative to the first member init clause.
9075 (goto-char (1+ tmp-pos))
9076 (c-forward-syntactic-ws)
9077 (c-add-syntax 'member-init-cont (point))))
9078
9079 ;; CASE 5B.2: K&R arg decl intro
9080 ((and c-recognize-knr-p
9081 (c-in-knr-argdecl lim))
9082 (c-beginning-of-statement-1 lim)
9083 (c-add-syntax 'knr-argdecl-intro (c-point 'boi))
9084 (if containing-decl-open
9085 (c-add-class-syntax 'inclass
9086 containing-decl-open
9087 containing-decl-start
9088 containing-decl-kwd
9089 paren-state)))
9090
9091 ;; CASE 5B.4: Nether region after a C++ or Java func
9092 ;; decl, which could include a `throws' declaration.
9093 (t
9094 (c-beginning-of-statement-1 lim)
9095 (c-add-syntax 'func-decl-cont (c-point 'boi))
9096 )))
9097
9098 ;; CASE 5C: inheritance line. could be first inheritance
9099 ;; line, or continuation of a multiple inheritance
9100 ((or (and (c-major-mode-is 'c++-mode)
9101 (progn
9102 (when (eq char-after-ip ?,)
9103 (skip-chars-forward " \t")
9104 (forward-char))
9105 (looking-at c-opt-postfix-decl-spec-key)))
9106 (and (or (eq char-before-ip ?:)
9107 ;; watch out for scope operator
9108 (save-excursion
9109 (and (eq char-after-ip ?:)
9110 (c-safe (forward-char 1) t)
9111 (not (eq (char-after) ?:))
9112 )))
9113 (save-excursion
9114 (c-backward-syntactic-ws lim)
9115 (if (eq char-before-ip ?:)
9116 (progn
9117 (forward-char -1)
9118 (c-backward-syntactic-ws lim)))
9119 (back-to-indentation)
9120 (looking-at c-class-key)))
9121 ;; for Java
9122 (and (c-major-mode-is 'java-mode)
9123 (let ((fence (save-excursion
9124 (c-beginning-of-statement-1 lim)
9125 (point)))
9126 cont done)
9127 (save-excursion
9128 (while (not done)
9129 (cond ((looking-at c-opt-postfix-decl-spec-key)
9130 (setq injava-inher (cons cont (point))
9131 done t))
9132 ((or (not (c-safe (c-forward-sexp -1) t))
9133 (<= (point) fence))
9134 (setq done t))
9135 )
9136 (setq cont t)))
9137 injava-inher)
9138 (not (c-crosses-statement-barrier-p (cdr injava-inher)
9139 (point)))
9140 ))
9141 (cond
9142
9143 ;; CASE 5C.1: non-hanging colon on an inher intro
9144 ((eq char-after-ip ?:)
9145 (c-beginning-of-statement-1 lim)
9146 (c-add-syntax 'inher-intro (c-point 'boi))
9147 ;; don't add inclass symbol since relative point already
9148 ;; contains any class offset
9149 )
9150
9151 ;; CASE 5C.2: hanging colon on an inher intro
9152 ((eq char-before-ip ?:)
9153 (c-beginning-of-statement-1 lim)
9154 (c-add-syntax 'inher-intro (c-point 'boi))
9155 (if containing-decl-open
9156 (c-add-class-syntax 'inclass
9157 containing-decl-open
9158 containing-decl-start
9159 containing-decl-kwd
9160 paren-state)))
9161
9162 ;; CASE 5C.3: in a Java implements/extends
9163 (injava-inher
9164 (let ((where (cdr injava-inher))
9165 (cont (car injava-inher)))
9166 (goto-char where)
9167 (cond ((looking-at "throws\\>[^_]")
9168 (c-add-syntax 'func-decl-cont
9169 (progn (c-beginning-of-statement-1 lim)
9170 (c-point 'boi))))
9171 (cont (c-add-syntax 'inher-cont where))
9172 (t (c-add-syntax 'inher-intro
9173 (progn (goto-char (cdr injava-inher))
9174 (c-beginning-of-statement-1 lim)
9175 (point))))
9176 )))
9177
9178 ;; CASE 5C.4: a continued inheritance line
9179 (t
9180 (c-beginning-of-inheritance-list lim)
9181 (c-add-syntax 'inher-cont (point))
9182 ;; don't add inclass symbol since relative point already
9183 ;; contains any class offset
9184 )))
9185
9186 ;; CASE 5D: this could be a top-level initialization, a
9187 ;; member init list continuation, or a template argument
9188 ;; list continuation.
9189 ((save-excursion
9190 ;; Note: We use the fact that lim is always after any
9191 ;; preceding brace sexp.
9192 (if c-recognize-<>-arglists
9193 (while (and
9194 (progn
9195 (c-syntactic-skip-backward "^;,=<>" lim t)
9196 (> (point) lim))
9197 (or
9198 (when c-overloadable-operators-regexp
9199 (when (setq placeholder (c-after-special-operator-id lim))
9200 (goto-char placeholder)
9201 t))
9202 (cond
9203 ((eq (char-before) ?>)
9204 (or (c-backward-<>-arglist nil lim)
9205 (backward-char))
9206 t)
9207 ((eq (char-before) ?<)
9208 (backward-char)
9209 (if (save-excursion
9210 (c-forward-<>-arglist nil))
9211 (progn (forward-char)
9212 nil)
9213 t))
9214 (t nil)))))
9215 ;; NB: No c-after-special-operator-id stuff in this
9216 ;; clause - we assume only C++ needs it.
9217 (c-syntactic-skip-backward "^;,=" lim t))
9218 (memq (char-before) '(?, ?= ?<)))
9219 (cond
9220
9221 ;; CASE 5D.3: perhaps a template list continuation?
9222 ((and (c-major-mode-is 'c++-mode)
9223 (save-excursion
9224 (save-restriction
9225 (c-with-syntax-table c++-template-syntax-table
9226 (goto-char indent-point)
9227 (setq placeholder (c-up-list-backward))
9228 (and placeholder
9229 (eq (char-after placeholder) ?<))))))
9230 (c-with-syntax-table c++-template-syntax-table
9231 (goto-char placeholder)
9232 (c-beginning-of-statement-1 lim t)
9233 (if (save-excursion
9234 (c-backward-syntactic-ws lim)
9235 (eq (char-before) ?<))
9236 ;; In a nested template arglist.
9237 (progn
9238 (goto-char placeholder)
9239 (c-syntactic-skip-backward "^,;" lim t)
9240 (c-forward-syntactic-ws))
9241 (back-to-indentation)))
9242 ;; FIXME: Should use c-add-stmt-syntax, but it's not yet
9243 ;; template aware.
9244 (c-add-syntax 'template-args-cont (point) placeholder))
9245
9246 ;; CASE 5D.4: perhaps a multiple inheritance line?
9247 ((and (c-major-mode-is 'c++-mode)
9248 (save-excursion
9249 (c-beginning-of-statement-1 lim)
9250 (setq placeholder (point))
9251 (if (looking-at "static\\>[^_]")
9252 (c-forward-token-2 1 nil indent-point))
9253 (and (looking-at c-class-key)
9254 (zerop (c-forward-token-2 2 nil indent-point))
9255 (if (eq (char-after) ?<)
9256 (c-with-syntax-table c++-template-syntax-table
9257 (zerop (c-forward-token-2 1 t indent-point)))
9258 t)
9259 (eq (char-after) ?:))))
9260 (goto-char placeholder)
9261 (c-add-syntax 'inher-cont (c-point 'boi)))
9262
9263 ;; CASE 5D.5: Continuation of the "expression part" of a
9264 ;; top level construct. Or, perhaps, an unrecognised construct.
9265 (t
9266 (while (and (setq placeholder (point))
9267 (eq (car (c-beginning-of-decl-1 containing-sexp))
9268 'same)
9269 (save-excursion
9270 (c-backward-syntactic-ws)
9271 (eq (char-before) ?}))
9272 (< (point) placeholder)))
9273 (c-add-stmt-syntax
9274 (cond
9275 ((eq (point) placeholder) 'statement) ; unrecognised construct
9276 ;; A preceding comma at the top level means that a
9277 ;; new variable declaration starts here. Use
9278 ;; topmost-intro-cont for it, for consistency with
9279 ;; the first variable declaration. C.f. case 5N.
9280 ((eq char-before-ip ?,) 'topmost-intro-cont)
9281 (t 'statement-cont))
9282 nil nil containing-sexp paren-state))
9283 ))
9284
9285 ;; CASE 5F: Close of a non-class declaration level block.
9286 ((and (eq char-after-ip ?})
9287 (c-keyword-member containing-decl-kwd
9288 'c-other-block-decl-kwds))
9289 ;; This is inconsistent: Should use `containing-decl-open'
9290 ;; here if it's at boi, like in case 5J.
9291 (goto-char containing-decl-start)
9292 (c-add-stmt-syntax
9293 (if (string-equal (symbol-name containing-decl-kwd) "extern")
9294 ;; Special case for compatibility with the
9295 ;; extern-lang syntactic symbols.
9296 'extern-lang-close
9297 (intern (concat (symbol-name containing-decl-kwd)
9298 "-close")))
9299 nil t
9300 (c-most-enclosing-brace paren-state (point))
9301 paren-state))
9302
9303 ;; CASE 5G: we are looking at the brace which closes the
9304 ;; enclosing nested class decl
9305 ((and containing-sexp
9306 (eq char-after-ip ?})
9307 (eq containing-decl-open containing-sexp))
9308 (c-add-class-syntax 'class-close
9309 containing-decl-open
9310 containing-decl-start
9311 containing-decl-kwd
9312 paren-state))
9313
9314 ;; CASE 5H: we could be looking at subsequent knr-argdecls
9315 ((and c-recognize-knr-p
9316 (not containing-sexp) ; can't be knr inside braces.
9317 (not (eq char-before-ip ?}))
9318 (save-excursion
9319 (setq placeholder (cdr (c-beginning-of-decl-1 lim)))
9320 (and placeholder
9321 ;; Do an extra check to avoid tripping up on
9322 ;; statements that occur in invalid contexts
9323 ;; (e.g. in macro bodies where we don't really
9324 ;; know the context of what we're looking at).
9325 (not (and c-opt-block-stmt-key
9326 (looking-at c-opt-block-stmt-key)))))
9327 (< placeholder indent-point))
9328 (goto-char placeholder)
9329 (c-add-syntax 'knr-argdecl (point)))
9330
9331 ;; CASE 5I: ObjC method definition.
9332 ((and c-opt-method-key
9333 (looking-at c-opt-method-key))
9334 (c-beginning-of-statement-1 nil t)
9335 (if (= (point) indent-point)
9336 ;; Handle the case when it's the first (non-comment)
9337 ;; thing in the buffer. Can't look for a 'same return
9338 ;; value from cbos1 since ObjC directives currently
9339 ;; aren't recognized fully, so that we get 'same
9340 ;; instead of 'previous if it moved over a preceding
9341 ;; directive.
9342 (goto-char (point-min)))
9343 (c-add-syntax 'objc-method-intro (c-point 'boi)))
9344
9345 ;; CASE 5P: AWK pattern or function or continuation
9346 ;; thereof.
9347 ((c-major-mode-is 'awk-mode)
9348 (setq placeholder (point))
9349 (c-add-stmt-syntax
9350 (if (and (eq (c-beginning-of-statement-1) 'same)
9351 (/= (point) placeholder))
9352 'topmost-intro-cont
9353 'topmost-intro)
9354 nil nil
9355 containing-sexp paren-state))
9356
9357 ;; CASE 5N: At a variable declaration that follows a class
9358 ;; definition or some other block declaration that doesn't
9359 ;; end at the closing '}'. C.f. case 5D.5.
9360 ((progn
9361 (c-backward-syntactic-ws lim)
9362 (and (eq (char-before) ?})
9363 (save-excursion
9364 (let ((start (point)))
9365 (if (and c-state-cache
9366 (consp (car c-state-cache))
9367 (eq (cdar c-state-cache) (point)))
9368 ;; Speed up the backward search a bit.
9369 (goto-char (caar c-state-cache)))
9370 (c-beginning-of-decl-1 containing-sexp)
9371 (setq placeholder (point))
9372 (if (= start (point))
9373 ;; The '}' is unbalanced.
9374 nil
9375 (c-end-of-decl-1)
9376 (>= (point) indent-point))))))
9377 (goto-char placeholder)
9378 (c-add-stmt-syntax 'topmost-intro-cont nil nil
9379 containing-sexp paren-state))
9380
9381 ;; NOTE: The point is at the end of the previous token here.
9382
9383 ;; CASE 5J: we are at the topmost level, make
9384 ;; sure we skip back past any access specifiers
9385 ((and
9386 ;; A macro continuation line is never at top level.
9387 (not (and macro-start
9388 (> indent-point macro-start)))
9389 (save-excursion
9390 (setq placeholder (point))
9391 (or (memq char-before-ip '(?\; ?{ ?} nil))
9392 (c-at-vsemi-p before-ws-ip)
9393 (when (and (eq char-before-ip ?:)
9394 (eq (c-beginning-of-statement-1 lim)
9395 'label))
9396 (c-backward-syntactic-ws lim)
9397 (setq placeholder (point)))
9398 (and (c-major-mode-is 'objc-mode)
9399 (catch 'not-in-directive
9400 (c-beginning-of-statement-1 lim)
9401 (setq placeholder (point))
9402 (while (and (c-forward-objc-directive)
9403 (< (point) indent-point))
9404 (c-forward-syntactic-ws)
9405 (if (>= (point) indent-point)
9406 (throw 'not-in-directive t))
9407 (setq placeholder (point)))
9408 nil)))))
9409 ;; For historic reasons we anchor at bol of the last
9410 ;; line of the previous declaration. That's clearly
9411 ;; highly bogus and useless, and it makes our lives hard
9412 ;; to remain compatible. :P
9413 (goto-char placeholder)
9414 (c-add-syntax 'topmost-intro (c-point 'bol))
9415 (if containing-decl-open
9416 (if (c-keyword-member containing-decl-kwd
9417 'c-other-block-decl-kwds)
9418 (progn
9419 (goto-char (c-brace-anchor-point containing-decl-open))
9420 (c-add-stmt-syntax
9421 (if (string-equal (symbol-name containing-decl-kwd)
9422 "extern")
9423 ;; Special case for compatibility with the
9424 ;; extern-lang syntactic symbols.
9425 'inextern-lang
9426 (intern (concat "in"
9427 (symbol-name containing-decl-kwd))))
9428 nil t
9429 (c-most-enclosing-brace paren-state (point))
9430 paren-state))
9431 (c-add-class-syntax 'inclass
9432 containing-decl-open
9433 containing-decl-start
9434 containing-decl-kwd
9435 paren-state)))
9436 (when (and c-syntactic-indentation-in-macros
9437 macro-start
9438 (/= macro-start (c-point 'boi indent-point)))
9439 (c-add-syntax 'cpp-define-intro)
9440 (setq macro-start nil)))
9441
9442 ;; CASE 5K: we are at an ObjC method definition
9443 ;; continuation line.
9444 ((and c-opt-method-key
9445 (save-excursion
9446 (c-beginning-of-statement-1 lim)
9447 (beginning-of-line)
9448 (when (looking-at c-opt-method-key)
9449 (setq placeholder (point)))))
9450 (c-add-syntax 'objc-method-args-cont placeholder))
9451
9452 ;; CASE 5L: we are at the first argument of a template
9453 ;; arglist that begins on the previous line.
9454 ((and c-recognize-<>-arglists
9455 (eq (char-before) ?<)
9456 (setq placeholder (1- (point)))
9457 (not (and c-overloadable-operators-regexp
9458 (c-after-special-operator-id lim))))
9459 (c-beginning-of-statement-1 (c-safe-position (point) paren-state))
9460 (c-add-syntax 'template-args-cont (c-point 'boi) placeholder))
9461
9462 ;; CASE 5Q: we are at a statement within a macro.
9463 (macro-start
9464 (c-beginning-of-statement-1 containing-sexp)
9465 (c-add-stmt-syntax 'statement nil t containing-sexp paren-state))
9466
9467 ;; CASE 5M: we are at a topmost continuation line
9468 (t
9469 (c-beginning-of-statement-1 (c-safe-position (point) paren-state))
9470 (when (c-major-mode-is 'objc-mode)
9471 (setq placeholder (point))
9472 (while (and (c-forward-objc-directive)
9473 (< (point) indent-point))
9474 (c-forward-syntactic-ws)
9475 (setq placeholder (point)))
9476 (goto-char placeholder))
9477 (c-add-syntax 'topmost-intro-cont (c-point 'boi)))
9478 ))
9479
9480 ;; (CASE 6 has been removed.)
9481
9482 ;; CASE 19: line is an expression, not a statement, and is directly
9483 ;; contained by a template delimiter. Most likely, we are in a
9484 ;; template arglist within a statement. This case is based on CASE
9485 ;; 7. At some point in the future, we may wish to create more
9486 ;; syntactic symbols such as `template-intro',
9487 ;; `template-cont-nonempty', etc., and distinguish between them as we
9488 ;; do for `arglist-intro' etc. (2009-12-07).
9489 ((and c-recognize-<>-arglists
9490 (setq containing-< (c-up-list-backward indent-point containing-sexp))
9491 (eq (char-after containing-<) ?\<))
9492 (setq placeholder (c-point 'boi containing-<))
9493 (goto-char containing-sexp) ; Most nested Lbrace/Lparen (but not
9494 ; '<') before indent-point.
9495 (if (>= (point) placeholder)
9496 (progn
9497 (forward-char)
9498 (skip-chars-forward " \t"))
9499 (goto-char placeholder))
9500 (c-add-stmt-syntax 'template-args-cont (list containing-<) t
9501 (c-most-enclosing-brace c-state-cache (point))
9502 paren-state))
9503
9504
9505 ;; CASE 7: line is an expression, not a statement. Most
9506 ;; likely we are either in a function prototype or a function
9507 ;; call argument list, or a template argument list.
9508 ((not (or (and c-special-brace-lists
9509 (save-excursion
9510 (goto-char containing-sexp)
9511 (c-looking-at-special-brace-list)))
9512 (eq (char-after containing-sexp) ?{)
9513 (eq (char-after containing-sexp) ?<)))
9514 (cond
9515
9516 ;; CASE 7A: we are looking at the arglist closing paren.
9517 ;; C.f. case 7F.
9518 ((memq char-after-ip '(?\) ?\]))
9519 (goto-char containing-sexp)
9520 (setq placeholder (c-point 'boi))
9521 (if (and (c-safe (backward-up-list 1) t)
9522 (>= (point) placeholder))
9523 (progn
9524 (forward-char)
9525 (skip-chars-forward " \t"))
9526 (goto-char placeholder))
9527 (c-add-stmt-syntax 'arglist-close (list containing-sexp) t
9528 (c-most-enclosing-brace paren-state (point))
9529 paren-state))
9530
9531 ;; CASE 7B: Looking at the opening brace of an
9532 ;; in-expression block or brace list. C.f. cases 4, 16A
9533 ;; and 17E.
9534 ((and (eq char-after-ip ?{)
9535 (progn
9536 (setq placeholder (c-inside-bracelist-p (point)
9537 paren-state))
9538 (if placeholder
9539 (setq tmpsymbol '(brace-list-open . inexpr-class))
9540 (setq tmpsymbol '(block-open . inexpr-statement)
9541 placeholder
9542 (cdr-safe (c-looking-at-inexpr-block
9543 (c-safe-position containing-sexp
9544 paren-state)
9545 containing-sexp)))
9546 ;; placeholder is nil if it's a block directly in
9547 ;; a function arglist. That makes us skip out of
9548 ;; this case.
9549 )))
9550 (goto-char placeholder)
9551 (back-to-indentation)
9552 (c-add-stmt-syntax (car tmpsymbol) nil t
9553 (c-most-enclosing-brace paren-state (point))
9554 paren-state)
9555 (if (/= (point) placeholder)
9556 (c-add-syntax (cdr tmpsymbol))))
9557
9558 ;; CASE 7C: we are looking at the first argument in an empty
9559 ;; argument list. Use arglist-close if we're actually
9560 ;; looking at a close paren or bracket.
9561 ((memq char-before-ip '(?\( ?\[))
9562 (goto-char containing-sexp)
9563 (setq placeholder (c-point 'boi))
9564 (if (and (c-safe (backward-up-list 1) t)
9565 (>= (point) placeholder))
9566 (progn
9567 (forward-char)
9568 (skip-chars-forward " \t"))
9569 (goto-char placeholder))
9570 (c-add-stmt-syntax 'arglist-intro (list containing-sexp) t
9571 (c-most-enclosing-brace paren-state (point))
9572 paren-state))
9573
9574 ;; CASE 7D: we are inside a conditional test clause. treat
9575 ;; these things as statements
9576 ((progn
9577 (goto-char containing-sexp)
9578 (and (c-safe (c-forward-sexp -1) t)
9579 (looking-at "\\<for\\>[^_]")))
9580 (goto-char (1+ containing-sexp))
9581 (c-forward-syntactic-ws indent-point)
9582 (if (eq char-before-ip ?\;)
9583 (c-add-syntax 'statement (point))
9584 (c-add-syntax 'statement-cont (point))
9585 ))
9586
9587 ;; CASE 7E: maybe a continued ObjC method call. This is the
9588 ;; case when we are inside a [] bracketed exp, and what
9589 ;; precede the opening bracket is not an identifier.
9590 ((and c-opt-method-key
9591 (eq (char-after containing-sexp) ?\[)
9592 (progn
9593 (goto-char (1- containing-sexp))
9594 (c-backward-syntactic-ws (c-point 'bod))
9595 (if (not (looking-at c-symbol-key))
9596 (c-add-syntax 'objc-method-call-cont containing-sexp))
9597 )))
9598
9599 ;; CASE 7F: we are looking at an arglist continuation line,
9600 ;; but the preceding argument is on the same line as the
9601 ;; opening paren. This case includes multi-line
9602 ;; mathematical paren groupings, but we could be on a
9603 ;; for-list continuation line. C.f. case 7A.
9604 ((progn
9605 (goto-char (1+ containing-sexp))
9606 (< (save-excursion
9607 (c-forward-syntactic-ws)
9608 (point))
9609 (c-point 'bonl)))
9610 (goto-char containing-sexp) ; paren opening the arglist
9611 (setq placeholder (c-point 'boi))
9612 (if (and (c-safe (backward-up-list 1) t)
9613 (>= (point) placeholder))
9614 (progn
9615 (forward-char)
9616 (skip-chars-forward " \t"))
9617 (goto-char placeholder))
9618 (c-add-stmt-syntax 'arglist-cont-nonempty (list containing-sexp) t
9619 (c-most-enclosing-brace c-state-cache (point))
9620 paren-state))
9621
9622 ;; CASE 7G: we are looking at just a normal arglist
9623 ;; continuation line
9624 (t (c-forward-syntactic-ws indent-point)
9625 (c-add-syntax 'arglist-cont (c-point 'boi)))
9626 ))
9627
9628 ;; CASE 8: func-local multi-inheritance line
9629 ((and (c-major-mode-is 'c++-mode)
9630 (save-excursion
9631 (goto-char indent-point)
9632 (skip-chars-forward " \t")
9633 (looking-at c-opt-postfix-decl-spec-key)))
9634 (goto-char indent-point)
9635 (skip-chars-forward " \t")
9636 (cond
9637
9638 ;; CASE 8A: non-hanging colon on an inher intro
9639 ((eq char-after-ip ?:)
9640 (c-backward-syntactic-ws lim)
9641 (c-add-syntax 'inher-intro (c-point 'boi)))
9642
9643 ;; CASE 8B: hanging colon on an inher intro
9644 ((eq char-before-ip ?:)
9645 (c-add-syntax 'inher-intro (c-point 'boi)))
9646
9647 ;; CASE 8C: a continued inheritance line
9648 (t
9649 (c-beginning-of-inheritance-list lim)
9650 (c-add-syntax 'inher-cont (point))
9651 )))
9652
9653 ;; CASE 9: we are inside a brace-list
9654 ((and (not (c-major-mode-is 'awk-mode)) ; Maybe this isn't needed (ACM, 2002/3/29)
9655 (setq special-brace-list
9656 (or (and c-special-brace-lists ;;;; ALWAYS NIL FOR AWK!!
9657 (save-excursion
9658 (goto-char containing-sexp)
9659 (c-looking-at-special-brace-list)))
9660 (c-inside-bracelist-p containing-sexp paren-state))))
9661 (cond
9662
9663 ;; CASE 9A: In the middle of a special brace list opener.
9664 ((and (consp special-brace-list)
9665 (save-excursion
9666 (goto-char containing-sexp)
9667 (eq (char-after) ?\())
9668 (eq char-after-ip (car (cdr special-brace-list))))
9669 (goto-char (car (car special-brace-list)))
9670 (skip-chars-backward " \t")
9671 (if (and (bolp)
9672 (assoc 'statement-cont
9673 (setq placeholder (c-guess-basic-syntax))))
9674 (setq c-syntactic-context placeholder)
9675 (c-beginning-of-statement-1
9676 (c-safe-position (1- containing-sexp) paren-state))
9677 (c-forward-token-2 0)
9678 (while (looking-at c-specifier-key)
9679 (goto-char (match-end 1))
9680 (c-forward-syntactic-ws))
9681 (c-add-syntax 'brace-list-open (c-point 'boi))))
9682
9683 ;; CASE 9B: brace-list-close brace
9684 ((if (consp special-brace-list)
9685 ;; Check special brace list closer.
9686 (progn
9687 (goto-char (car (car special-brace-list)))
9688 (save-excursion
9689 (goto-char indent-point)
9690 (back-to-indentation)
9691 (or
9692 ;; We were between the special close char and the `)'.
9693 (and (eq (char-after) ?\))
9694 (eq (1+ (point)) (cdr (car special-brace-list))))
9695 ;; We were before the special close char.
9696 (and (eq (char-after) (cdr (cdr special-brace-list)))
9697 (zerop (c-forward-token-2))
9698 (eq (1+ (point)) (cdr (car special-brace-list)))))))
9699 ;; Normal brace list check.
9700 (and (eq char-after-ip ?})
9701 (c-safe (goto-char (c-up-list-backward (point))) t)
9702 (= (point) containing-sexp)))
9703 (if (eq (point) (c-point 'boi))
9704 (c-add-syntax 'brace-list-close (point))
9705 (setq lim (c-most-enclosing-brace c-state-cache (point)))
9706 (c-beginning-of-statement-1 lim)
9707 (c-add-stmt-syntax 'brace-list-close nil t lim paren-state)))
9708
9709 (t
9710 ;; Prepare for the rest of the cases below by going to the
9711 ;; token following the opening brace
9712 (if (consp special-brace-list)
9713 (progn
9714 (goto-char (car (car special-brace-list)))
9715 (c-forward-token-2 1 nil indent-point))
9716 (goto-char containing-sexp))
9717 (forward-char)
9718 (let ((start (point)))
9719 (c-forward-syntactic-ws indent-point)
9720 (goto-char (max start (c-point 'bol))))
9721 (c-skip-ws-forward indent-point)
9722 (cond
9723
9724 ;; CASE 9C: we're looking at the first line in a brace-list
9725 ((= (point) indent-point)
9726 (if (consp special-brace-list)
9727 (goto-char (car (car special-brace-list)))
9728 (goto-char containing-sexp))
9729 (if (eq (point) (c-point 'boi))
9730 (c-add-syntax 'brace-list-intro (point))
9731 (setq lim (c-most-enclosing-brace c-state-cache (point)))
9732 (c-beginning-of-statement-1 lim)
9733 (c-add-stmt-syntax 'brace-list-intro nil t lim paren-state)))
9734
9735 ;; CASE 9D: this is just a later brace-list-entry or
9736 ;; brace-entry-open
9737 (t (if (or (eq char-after-ip ?{)
9738 (and c-special-brace-lists
9739 (save-excursion
9740 (goto-char indent-point)
9741 (c-forward-syntactic-ws (c-point 'eol))
9742 (c-looking-at-special-brace-list (point)))))
9743 (c-add-syntax 'brace-entry-open (point))
9744 (c-add-syntax 'brace-list-entry (point))
9745 ))
9746 ))))
9747
9748 ;; CASE 10: A continued statement or top level construct.
9749 ((and (not (memq char-before-ip '(?\; ?:)))
9750 (not (c-at-vsemi-p before-ws-ip))
9751 (or (not (eq char-before-ip ?}))
9752 (c-looking-at-inexpr-block-backward c-state-cache))
9753 (> (point)
9754 (save-excursion
9755 (c-beginning-of-statement-1 containing-sexp)
9756 (setq placeholder (point))))
9757 (/= placeholder containing-sexp))
9758 ;; This is shared with case 18.
9759 (c-guess-continued-construct indent-point
9760 char-after-ip
9761 placeholder
9762 containing-sexp
9763 paren-state))
9764
9765 ;; CASE 16: block close brace, possibly closing the defun or
9766 ;; the class
9767 ((eq char-after-ip ?})
9768 ;; From here on we have the next containing sexp in lim.
9769 (setq lim (c-most-enclosing-brace paren-state))
9770 (goto-char containing-sexp)
9771 (cond
9772
9773 ;; CASE 16E: Closing a statement block? This catches
9774 ;; cases where it's preceded by a statement keyword,
9775 ;; which works even when used in an "invalid" context,
9776 ;; e.g. a macro argument.
9777 ((c-after-conditional)
9778 (c-backward-to-block-anchor lim)
9779 (c-add-stmt-syntax 'block-close nil t lim paren-state))
9780
9781 ;; CASE 16A: closing a lambda defun or an in-expression
9782 ;; block? C.f. cases 4, 7B and 17E.
9783 ((setq placeholder (c-looking-at-inexpr-block
9784 (c-safe-position containing-sexp paren-state)
9785 nil))
9786 (setq tmpsymbol (if (eq (car placeholder) 'inlambda)
9787 'inline-close
9788 'block-close))
9789 (goto-char containing-sexp)
9790 (back-to-indentation)
9791 (if (= containing-sexp (point))
9792 (c-add-syntax tmpsymbol (point))
9793 (goto-char (cdr placeholder))
9794 (back-to-indentation)
9795 (c-add-stmt-syntax tmpsymbol nil t
9796 (c-most-enclosing-brace paren-state (point))
9797 paren-state)
9798 (if (/= (point) (cdr placeholder))
9799 (c-add-syntax (car placeholder)))))
9800
9801 ;; CASE 16B: does this close an inline or a function in
9802 ;; a non-class declaration level block?
9803 ((save-excursion
9804 (and lim
9805 (progn
9806 (goto-char lim)
9807 (c-looking-at-decl-block
9808 (c-most-enclosing-brace paren-state lim)
9809 nil))
9810 (setq placeholder (point))))
9811 (c-backward-to-decl-anchor lim)
9812 (back-to-indentation)
9813 (if (save-excursion
9814 (goto-char placeholder)
9815 (looking-at c-other-decl-block-key))
9816 (c-add-syntax 'defun-close (point))
9817 (c-add-syntax 'inline-close (point))))
9818
9819 ;; CASE 16F: Can be a defun-close of a function declared
9820 ;; in a statement block, e.g. in Pike or when using gcc
9821 ;; extensions, but watch out for macros followed by
9822 ;; blocks. Let it through to be handled below.
9823 ;; C.f. cases B.3 and 17G.
9824 ((save-excursion
9825 (and (not (c-at-statement-start-p))
9826 (eq (c-beginning-of-statement-1 lim nil nil t) 'same)
9827 (setq placeholder (point))
9828 (let ((c-recognize-typeless-decls nil))
9829 ;; Turn off recognition of constructs that
9830 ;; lacks a type in this case, since that's more
9831 ;; likely to be a macro followed by a block.
9832 (c-forward-decl-or-cast-1 (c-point 'bosws) nil nil))))
9833 (back-to-indentation)
9834 (if (/= (point) containing-sexp)
9835 (goto-char placeholder))
9836 (c-add-stmt-syntax 'defun-close nil t lim paren-state))
9837
9838 ;; CASE 16C: If there is an enclosing brace then this is
9839 ;; a block close since defun closes inside declaration
9840 ;; level blocks have been handled above.
9841 (lim
9842 ;; If the block is preceded by a case/switch label on
9843 ;; the same line, we anchor at the first preceding label
9844 ;; at boi. The default handling in c-add-stmt-syntax
9845 ;; really fixes it better, but we do like this to keep
9846 ;; the indentation compatible with version 5.28 and
9847 ;; earlier. C.f. case 17H.
9848 (while (and (/= (setq placeholder (point)) (c-point 'boi))
9849 (eq (c-beginning-of-statement-1 lim) 'label)))
9850 (goto-char placeholder)
9851 (if (looking-at c-label-kwds-regexp)
9852 (c-add-syntax 'block-close (point))
9853 (goto-char containing-sexp)
9854 ;; c-backward-to-block-anchor not necessary here; those
9855 ;; situations are handled in case 16E above.
9856 (c-add-stmt-syntax 'block-close nil t lim paren-state)))
9857
9858 ;; CASE 16D: Only top level defun close left.
9859 (t
9860 (goto-char containing-sexp)
9861 (c-backward-to-decl-anchor lim)
9862 (c-add-stmt-syntax 'defun-close nil nil
9863 (c-most-enclosing-brace paren-state)
9864 paren-state))
9865 ))
9866
9867 ;; CASE 17: Statement or defun catchall.
9868 (t
9869 (goto-char indent-point)
9870 ;; Back up statements until we find one that starts at boi.
9871 (while (let* ((prev-point (point))
9872 (last-step-type (c-beginning-of-statement-1
9873 containing-sexp)))
9874 (if (= (point) prev-point)
9875 (progn
9876 (setq step-type (or step-type last-step-type))
9877 nil)
9878 (setq step-type last-step-type)
9879 (/= (point) (c-point 'boi)))))
9880 (cond
9881
9882 ;; CASE 17B: continued statement
9883 ((and (eq step-type 'same)
9884 (/= (point) indent-point))
9885 (c-add-stmt-syntax 'statement-cont nil nil
9886 containing-sexp paren-state))
9887
9888 ;; CASE 17A: After a case/default label?
9889 ((progn
9890 (while (and (eq step-type 'label)
9891 (not (looking-at c-label-kwds-regexp)))
9892 (setq step-type
9893 (c-beginning-of-statement-1 containing-sexp)))
9894 (eq step-type 'label))
9895 (c-add-stmt-syntax (if (eq char-after-ip ?{)
9896 'statement-case-open
9897 'statement-case-intro)
9898 nil t containing-sexp paren-state))
9899
9900 ;; CASE 17D: any old statement
9901 ((progn
9902 (while (eq step-type 'label)
9903 (setq step-type
9904 (c-beginning-of-statement-1 containing-sexp)))
9905 (eq step-type 'previous))
9906 (c-add-stmt-syntax 'statement nil t
9907 containing-sexp paren-state)
9908 (if (eq char-after-ip ?{)
9909 (c-add-syntax 'block-open)))
9910
9911 ;; CASE 17I: Inside a substatement block.
9912 ((progn
9913 ;; The following tests are all based on containing-sexp.
9914 (goto-char containing-sexp)
9915 ;; From here on we have the next containing sexp in lim.
9916 (setq lim (c-most-enclosing-brace paren-state containing-sexp))
9917 (c-after-conditional))
9918 (c-backward-to-block-anchor lim)
9919 (c-add-stmt-syntax 'statement-block-intro nil t
9920 lim paren-state)
9921 (if (eq char-after-ip ?{)
9922 (c-add-syntax 'block-open)))
9923
9924 ;; CASE 17E: first statement in an in-expression block.
9925 ;; C.f. cases 4, 7B and 16A.
9926 ((setq placeholder (c-looking-at-inexpr-block
9927 (c-safe-position containing-sexp paren-state)
9928 nil))
9929 (setq tmpsymbol (if (eq (car placeholder) 'inlambda)
9930 'defun-block-intro
9931 'statement-block-intro))
9932 (back-to-indentation)
9933 (if (= containing-sexp (point))
9934 (c-add-syntax tmpsymbol (point))
9935 (goto-char (cdr placeholder))
9936 (back-to-indentation)
9937 (c-add-stmt-syntax tmpsymbol nil t
9938 (c-most-enclosing-brace c-state-cache (point))
9939 paren-state)
9940 (if (/= (point) (cdr placeholder))
9941 (c-add-syntax (car placeholder))))
9942 (if (eq char-after-ip ?{)
9943 (c-add-syntax 'block-open)))
9944
9945 ;; CASE 17F: first statement in an inline, or first
9946 ;; statement in a top-level defun. we can tell this is it
9947 ;; if there are no enclosing braces that haven't been
9948 ;; narrowed out by a class (i.e. don't use bod here).
9949 ((save-excursion
9950 (or (not (setq placeholder (c-most-enclosing-brace
9951 paren-state)))
9952 (and (progn
9953 (goto-char placeholder)
9954 (eq (char-after) ?{))
9955 (c-looking-at-decl-block (c-most-enclosing-brace
9956 paren-state (point))
9957 nil))))
9958 (c-backward-to-decl-anchor lim)
9959 (back-to-indentation)
9960 (c-add-syntax 'defun-block-intro (point)))
9961
9962 ;; CASE 17G: First statement in a function declared inside
9963 ;; a normal block. This can occur in Pike and with
9964 ;; e.g. the gcc extensions, but watch out for macros
9965 ;; followed by blocks. C.f. cases B.3 and 16F.
9966 ((save-excursion
9967 (and (not (c-at-statement-start-p))
9968 (eq (c-beginning-of-statement-1 lim nil nil t) 'same)
9969 (setq placeholder (point))
9970 (let ((c-recognize-typeless-decls nil))
9971 ;; Turn off recognition of constructs that lacks
9972 ;; a type in this case, since that's more likely
9973 ;; to be a macro followed by a block.
9974 (c-forward-decl-or-cast-1 (c-point 'bosws) nil nil))))
9975 (back-to-indentation)
9976 (if (/= (point) containing-sexp)
9977 (goto-char placeholder))
9978 (c-add-stmt-syntax 'defun-block-intro nil t
9979 lim paren-state))
9980
9981 ;; CASE 17H: First statement in a block.
9982 (t
9983 ;; If the block is preceded by a case/switch label on the
9984 ;; same line, we anchor at the first preceding label at
9985 ;; boi. The default handling in c-add-stmt-syntax is
9986 ;; really fixes it better, but we do like this to keep the
9987 ;; indentation compatible with version 5.28 and earlier.
9988 ;; C.f. case 16C.
9989 (while (and (/= (setq placeholder (point)) (c-point 'boi))
9990 (eq (c-beginning-of-statement-1 lim) 'label)))
9991 (goto-char placeholder)
9992 (if (looking-at c-label-kwds-regexp)
9993 (c-add-syntax 'statement-block-intro (point))
9994 (goto-char containing-sexp)
9995 ;; c-backward-to-block-anchor not necessary here; those
9996 ;; situations are handled in case 17I above.
9997 (c-add-stmt-syntax 'statement-block-intro nil t
9998 lim paren-state))
9999 (if (eq char-after-ip ?{)
10000 (c-add-syntax 'block-open)))
10001 ))
10002 )
10003
10004 ;; now we need to look at any modifiers
10005 (goto-char indent-point)
10006 (skip-chars-forward " \t")
10007
10008 ;; are we looking at a comment only line?
10009 (when (and (looking-at c-comment-start-regexp)
10010 (/= (c-forward-token-2 0 nil (c-point 'eol)) 0))
10011 (c-append-syntax 'comment-intro))
10012
10013 ;; we might want to give additional offset to friends (in C++).
10014 (when (and c-opt-friend-key
10015 (looking-at c-opt-friend-key))
10016 (c-append-syntax 'friend))
10017
10018 ;; Set syntactic-relpos.
10019 (let ((p c-syntactic-context))
10020 (while (and p
10021 (if (integerp (c-langelem-pos (car p)))
10022 (progn
10023 (setq syntactic-relpos (c-langelem-pos (car p)))
10024 nil)
10025 t))
10026 (setq p (cdr p))))
10027
10028 ;; Start of or a continuation of a preprocessor directive?
10029 (if (and macro-start
10030 (eq macro-start (c-point 'boi))
10031 (not (and (c-major-mode-is 'pike-mode)
10032 (eq (char-after (1+ macro-start)) ?\"))))
10033 (c-append-syntax 'cpp-macro)
10034 (when (and c-syntactic-indentation-in-macros macro-start)
10035 (if in-macro-expr
10036 (when (or
10037 (< syntactic-relpos macro-start)
10038 (not (or
10039 (assq 'arglist-intro c-syntactic-context)
10040 (assq 'arglist-cont c-syntactic-context)
10041 (assq 'arglist-cont-nonempty c-syntactic-context)
10042 (assq 'arglist-close c-syntactic-context))))
10043 ;; If inside a cpp expression, i.e. anywhere in a
10044 ;; cpp directive except a #define body, we only let
10045 ;; through the syntactic analysis that is internal
10046 ;; in the expression. That means the arglist
10047 ;; elements, if they are anchored inside the cpp
10048 ;; expression.
10049 (setq c-syntactic-context nil)
10050 (c-add-syntax 'cpp-macro-cont macro-start))
10051 (when (and (eq macro-start syntactic-relpos)
10052 (not (assq 'cpp-define-intro c-syntactic-context))
10053 (save-excursion
10054 (goto-char macro-start)
10055 (or (not (c-forward-to-cpp-define-body))
10056 (<= (point) (c-point 'boi indent-point)))))
10057 ;; Inside a #define body and the syntactic analysis is
10058 ;; anchored on the start of the #define. In this case
10059 ;; we add cpp-define-intro to get the extra
10060 ;; indentation of the #define body.
10061 (c-add-syntax 'cpp-define-intro)))))
10062
10063 ;; return the syntax
10064 c-syntactic-context)))
10065
10066 \f
10067 ;; Indentation calculation.
10068
10069 (defun c-evaluate-offset (offset langelem symbol)
10070 ;; offset can be a number, a function, a variable, a list, or one of
10071 ;; the symbols + or -
10072 ;;
10073 ;; This function might do hidden buffer changes.
10074 (let ((res
10075 (cond
10076 ((numberp offset) offset)
10077 ((vectorp offset) offset)
10078 ((null offset) nil)
10079
10080 ((eq offset '+) c-basic-offset)
10081 ((eq offset '-) (- c-basic-offset))
10082 ((eq offset '++) (* 2 c-basic-offset))
10083 ((eq offset '--) (* 2 (- c-basic-offset)))
10084 ((eq offset '*) (/ c-basic-offset 2))
10085 ((eq offset '/) (/ (- c-basic-offset) 2))
10086
10087 ((functionp offset)
10088 (c-evaluate-offset
10089 (funcall offset
10090 (cons (c-langelem-sym langelem)
10091 (c-langelem-pos langelem)))
10092 langelem symbol))
10093
10094 ((listp offset)
10095 (cond
10096 ((eq (car offset) 'quote)
10097 (c-benign-error "The offset %S for %s was mistakenly quoted"
10098 offset symbol)
10099 nil)
10100
10101 ((memq (car offset) '(min max))
10102 (let (res val (method (car offset)))
10103 (setq offset (cdr offset))
10104 (while offset
10105 (setq val (c-evaluate-offset (car offset) langelem symbol))
10106 (cond
10107 ((not val))
10108 ((not res)
10109 (setq res val))
10110 ((integerp val)
10111 (if (vectorp res)
10112 (c-benign-error "\
10113 Error evaluating offset %S for %s: \
10114 Cannot combine absolute offset %S with relative %S in `%s' method"
10115 (car offset) symbol res val method)
10116 (setq res (funcall method res val))))
10117 (t
10118 (if (integerp res)
10119 (c-benign-error "\
10120 Error evaluating offset %S for %s: \
10121 Cannot combine relative offset %S with absolute %S in `%s' method"
10122 (car offset) symbol res val method)
10123 (setq res (vector (funcall method (aref res 0)
10124 (aref val 0)))))))
10125 (setq offset (cdr offset)))
10126 res))
10127
10128 ((eq (car offset) 'add)
10129 (let (res val)
10130 (setq offset (cdr offset))
10131 (while offset
10132 (setq val (c-evaluate-offset (car offset) langelem symbol))
10133 (cond
10134 ((not val))
10135 ((not res)
10136 (setq res val))
10137 ((integerp val)
10138 (if (vectorp res)
10139 (setq res (vector (+ (aref res 0) val)))
10140 (setq res (+ res val))))
10141 (t
10142 (if (vectorp res)
10143 (c-benign-error "\
10144 Error evaluating offset %S for %s: \
10145 Cannot combine absolute offsets %S and %S in `add' method"
10146 (car offset) symbol res val)
10147 (setq res val)))) ; Override.
10148 (setq offset (cdr offset)))
10149 res))
10150
10151 (t
10152 (let (res)
10153 (when (eq (car offset) 'first)
10154 (setq offset (cdr offset)))
10155 (while (and (not res) offset)
10156 (setq res (c-evaluate-offset (car offset) langelem symbol)
10157 offset (cdr offset)))
10158 res))))
10159
10160 ((and (symbolp offset) (boundp offset))
10161 (symbol-value offset))
10162
10163 (t
10164 (c-benign-error "Unknown offset format %S for %s" offset symbol)
10165 nil))))
10166
10167 (if (or (null res) (integerp res)
10168 (and (vectorp res) (= (length res) 1) (integerp (aref res 0))))
10169 res
10170 (c-benign-error "Error evaluating offset %S for %s: Got invalid value %S"
10171 offset symbol res)
10172 nil)))
10173
10174 (defun c-calc-offset (langelem)
10175 ;; Get offset from LANGELEM which is a list beginning with the
10176 ;; syntactic symbol and followed by any analysis data it provides.
10177 ;; That data may be zero or more elements, but if at least one is
10178 ;; given then the first is the anchor position (or nil). The symbol
10179 ;; is matched against `c-offsets-alist' and the offset calculated
10180 ;; from that is returned.
10181 ;;
10182 ;; This function might do hidden buffer changes.
10183 (let* ((symbol (c-langelem-sym langelem))
10184 (match (assq symbol c-offsets-alist))
10185 (offset (cdr-safe match)))
10186 (if match
10187 (setq offset (c-evaluate-offset offset langelem symbol))
10188 (if c-strict-syntax-p
10189 (c-benign-error "No offset found for syntactic symbol %s" symbol))
10190 (setq offset 0))
10191 (if (vectorp offset)
10192 offset
10193 (or (and (numberp offset) offset)
10194 (and (symbolp offset) (symbol-value offset))
10195 0))
10196 ))
10197
10198 (defun c-get-offset (langelem)
10199 ;; This is a compatibility wrapper for `c-calc-offset' in case
10200 ;; someone is calling it directly. It takes an old style syntactic
10201 ;; element on the form (SYMBOL . ANCHOR-POS) and converts it to the
10202 ;; new list form.
10203 ;;
10204 ;; This function might do hidden buffer changes.
10205 (if (c-langelem-pos langelem)
10206 (c-calc-offset (list (c-langelem-sym langelem)
10207 (c-langelem-pos langelem)))
10208 (c-calc-offset langelem)))
10209
10210 (defun c-get-syntactic-indentation (langelems)
10211 ;; Calculate the syntactic indentation from a syntactic description
10212 ;; as returned by `c-guess-syntax'.
10213 ;;
10214 ;; Note that topmost-intro always has an anchor position at bol, for
10215 ;; historical reasons. It's often used together with other symbols
10216 ;; that has more sane positions. Since we always use the first
10217 ;; found anchor position, we rely on that these other symbols always
10218 ;; precede topmost-intro in the LANGELEMS list.
10219 ;;
10220 ;; This function might do hidden buffer changes.
10221 (let ((indent 0) anchor)
10222
10223 (while langelems
10224 (let* ((c-syntactic-element (car langelems))
10225 (res (c-calc-offset c-syntactic-element)))
10226
10227 (if (vectorp res)
10228 ;; Got an absolute column that overrides any indentation
10229 ;; we've collected so far, but not the relative
10230 ;; indentation we might get for the nested structures
10231 ;; further down the langelems list.
10232 (setq indent (elt res 0)
10233 anchor (point-min)) ; A position at column 0.
10234
10235 ;; Got a relative change of the current calculated
10236 ;; indentation.
10237 (setq indent (+ indent res))
10238
10239 ;; Use the anchor position from the first syntactic
10240 ;; element with one.
10241 (unless anchor
10242 (setq anchor (c-langelem-pos (car langelems)))))
10243
10244 (setq langelems (cdr langelems))))
10245
10246 (if anchor
10247 (+ indent (save-excursion
10248 (goto-char anchor)
10249 (current-column)))
10250 indent)))
10251
10252 \f
10253 (cc-provide 'cc-engine)
10254
10255 ;; arch-tag: 149add18-4673-4da5-ac47-6805e4eae089
10256 ;;; cc-engine.el ends here