3 * Copyright (C) 2010, University of Copenhagen DIKU and INRIA.
4 * Copyright (C) 2007, 2008 Ecole des Mines de Nantes
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License (GPL)
8 * version 2 as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * file license.txt for more details.
18 module TH
= Token_helpers
19 module TV
= Token_views_c
20 module LP
= Lexer_parser
22 module Stat
= Parsing_stat
28 (*****************************************************************************)
29 (* Some debugging functions *)
30 (*****************************************************************************)
32 let pr2, pr2_once
= Common.mk_pr2_wrappers
Flag_parsing_c.verbose_parsing
35 if !Flag_parsing_c.debug_cpp
36 then Common.pr2_once
("CPP-" ^ s
)
39 let msg_gen cond is_known printer s
=
42 if not
(!Flag_parsing_c.filter_msg
)
49 (* In the following, there are some harcoded names of types or macros
50 * but they are not used by our heuristics! They are just here to
51 * enable to detect false positive by printing only the typedef/macros
52 * that we don't know yet. If we print everything, then we can easily
53 * get lost with too much verbose tracing information. So those
54 * functions "filter" some messages. So our heuristics are still good,
55 * there is no more (or not that much) hardcoded linux stuff.
61 | "u_char" | "u_short" | "u_int" | "u_long"
62 | "u8" | "u16" | "u32" | "u64"
63 | "s8" | "s16" | "s32" | "s64"
64 | "__u8" | "__u16" | "__u32" | "__u64"
75 | s
when s
=~
".*_t$" -> true
80 (* note: cant use partial application with let msg_typedef =
81 * because it would compute msg_typedef at compile time when
82 * the flag debug_typedef is always false
85 incr
Stat.nTypedefInfer
;
86 msg_gen (!Flag_parsing_c.debug_typedef
)
89 pr2_cpp ("TYPEDEF: promoting: " ^ s
)
93 let msg_maybe_dangereous_typedef s
=
94 if not
(is_known_typdef s
)
97 ("PB MAYBE: dangerous typedef inference, maybe not a typedef: " ^ s
)
101 let msg_declare_macro s
=
102 incr
Stat.nMacroDecl
;
103 msg_gen (!Flag_parsing_c.debug_cpp
)
106 | "DECLARE_MUTEX" | "DECLARE_COMPLETION" | "DECLARE_RWSEM"
107 | "DECLARE_WAITQUEUE" | "DECLARE_WAIT_QUEUE_HEAD"
108 | "DEFINE_SPINLOCK" | "DEFINE_TIMER"
109 | "DEVICE_ATTR" | "CLASS_DEVICE_ATTR" | "DRIVER_ATTR"
110 | "SENSOR_DEVICE_ATTR"
112 | "DECLARE_WORK" | "DECLARE_TASKLET"
113 | "PORT_ATTR_RO" | "PORT_PMA_ATTR"
118 | s when s =~ "^DECLARE_.*" -> true
119 | s when s =~ ".*_ATTR$" -> true
120 | s when s =~ "^DEFINE_.*" -> true
126 (fun s
-> pr2_cpp ("MACRO: found declare-macro: " ^ s
))
131 incr
Stat.nIteratorHeuristic
;
132 pr2_cpp ("MACRO: found foreach: " ^ s
)
136 let msg_debug_macro s =
137 pr2_cpp ("MACRO: found debug-macro: " ^ s)
141 let msg_macro_noptvirg s
=
142 incr
Stat.nMacroStmt
;
143 pr2_cpp ("MACRO: found macro with param noptvirg: " ^ s
)
145 let msg_macro_toplevel_noptvirg s
=
146 incr
Stat.nMacroStmt
;
147 pr2_cpp ("MACRO: found toplevel macro noptvirg: " ^ s
)
149 let msg_macro_noptvirg_single s
=
150 incr
Stat.nMacroStmt
;
151 pr2_cpp ("MACRO: found single-macro noptvirg: " ^ s
)
156 let msg_macro_higher_order s
=
157 incr
Stat.nMacroHigherOrder
;
158 msg_gen (!Flag_parsing_c.debug_cpp
)
168 (fun s
-> pr2_cpp ("MACRO: found higher ordre macro : " ^ s
))
172 let msg_stringification s
=
173 incr
Stat.nMacroString
;
174 msg_gen (!Flag_parsing_c.debug_cpp
)
182 (* s when s =~ ".*STR.*" -> true *)
186 (fun s
-> pr2_cpp ("MACRO: found string-macro " ^ s
))
189 let msg_stringification_params s
=
190 incr
Stat.nMacroString
;
191 pr2_cpp ("MACRO: string-macro with params : " ^ s
)
195 let msg_apply_known_macro s
=
196 incr
Stat.nMacroExpand
;
197 pr2_cpp ("MACRO: found known macro = " ^ s
)
199 let msg_apply_known_macro_hint s
=
200 incr
Stat.nMacroHint
;
201 pr2_cpp ("MACRO: found known macro hint = " ^ s
)
206 let msg_ifdef_bool_passing is_ifdef_positif
=
207 incr
Stat.nIfdefZero
; (* of Version ? *)
209 then pr2_cpp "commenting parts of a #if 1 or #if LINUX_VERSION"
210 else pr2_cpp "commenting a #if 0 or #if LINUX_VERSION or __cplusplus"
213 let msg_ifdef_mid_something () =
214 incr
Stat.nIfdefExprPassing
;
215 pr2_cpp "found ifdef-mid-something"
217 let msg_ifdef_funheaders () =
218 incr
Stat.nIfdefFunheader
;
221 let msg_ifdef_cparen_else () =
222 incr
Stat.nIfdefPassing
;
223 pr2_cpp("found ifdef-cparen-else")
226 let msg_attribute s
=
227 incr
Stat.nMacroAttribute
;
232 (*****************************************************************************)
233 (* The regexp and basic view definitions *)
234 (*****************************************************************************)
236 (* opti: better to built then once and for all, especially regexp_foreach *)
238 let regexp_macro = Str.regexp
242 let regexp_annot = Str.regexp
246 let regexp_declare = Str.regexp
250 let regexp_foreach = Str.regexp_case_fold
251 ".*\\(for_?each\\|for_?all\\|iterate\\|loop\\|walk\\|scan\\|each\\|for\\)"
253 let regexp_typedef = Str.regexp
256 let false_typedef = [
261 let ok_typedef s
= not
(List.mem s
false_typedef)
264 not
(s
==~
regexp_annot)
269 (*****************************************************************************)
271 (*****************************************************************************)
273 (* ------------------------------------------------------------------------- *)
274 (* the pair is the status of '()' and '{}', ex: (-1,0)
275 * if too much ')' and good '{}'
276 * could do for [] too ?
277 * could do for ',' if encounter ',' at "toplevel", not inside () or {}
278 * then if have ifdef, then certainly can lead to a problem.
280 let (count_open_close_stuff_ifdef_clause
: TV.ifdef_grouped list
-> (int * int))=
282 let cnt_paren, cnt_brace
= ref 0, ref 0 in
283 xs
+> TV.iter_token_ifdef
(fun x
->
285 | x
when TH.is_opar x
-> incr
cnt_paren
286 | TOBrace _
-> incr cnt_brace
287 | x
when TH.is_cpar x
-> decr
cnt_paren
288 | TCBrace _
-> decr cnt_brace
292 !cnt_paren, !cnt_brace
295 (* ------------------------------------------------------------------------- *)
296 let forLOOKAHEAD = 30
299 (* look if there is a '{' just after the closing ')', and handling the
300 * possibility to have nested expressions inside nested parenthesis
302 * todo: use indentation instead of premier(statement) ?
304 let rec is_really_foreach xs
=
305 let rec is_foreach_aux = function
307 | TCPar _
::TOBrace _
::xs
-> true, xs
308 (* the following attempts to handle the cases where there is a
309 single statement in the body of the loop. undoubtedly more
311 todo: premier(statement) - suivant(funcall)
313 | TCPar _
::TIdent _
::xs
-> true, xs
314 | TCPar _
::Tif _
::xs
-> true, xs
315 | TCPar _
::Twhile _
::xs
-> true, xs
316 | TCPar _
::Tfor _
::xs
-> true, xs
317 | TCPar _
::Tswitch _
::xs
-> true, xs
318 | TCPar _
::Treturn _
::xs
-> true, xs
321 | TCPar _
::xs
-> false, xs
323 let (_
, xs'
) = is_foreach_aux xs
in
325 | x
::xs
-> is_foreach_aux xs
327 is_foreach_aux xs
+> fst
330 (* ------------------------------------------------------------------------- *)
331 let set_ifdef_token_parenthize_info cnt x
=
334 | TIfdefelse
(tag
, _
)
335 | TIfdefelif
(tag
, _
)
338 | TIfdefBool
(_
, tag
, _
)
339 | TIfdefMisc
(_
, tag
, _
)
340 | TIfdefVersion
(_
, tag
, _
)
344 | _
-> raise Impossible
348 let ifdef_paren_cnt = ref 0
351 let rec set_ifdef_parenthize_info xs
=
352 xs
+> List.iter
(function
353 | NotIfdefLine xs
-> ()
354 | Ifdefbool
(_
, xxs
, info_ifdef
)
355 | Ifdef
(xxs
, info_ifdef
) ->
357 incr
ifdef_paren_cnt;
358 let total_directives = List.length info_ifdef
in
360 info_ifdef
+> List.iter
(fun x
->
361 set_ifdef_token_parenthize_info (!ifdef_paren_cnt, total_directives)
363 xxs
+> List.iter
set_ifdef_parenthize_info
367 (*****************************************************************************)
368 (* The parsing hack for #define *)
369 (*****************************************************************************)
371 (* To parse macro definitions I need to do some tricks
372 * as some information can be get only at the lexing level. For instance
373 * the space after the name of the macro in '#define foo (x)' is meaningful
374 * but the grammar can not get this information. So define_ident below
375 * look at such space and generate a special TOpardefine. In a similar
376 * way macro definitions can contain some antislash and newlines
377 * and the grammar need to know where the macro ends (which is
378 * a line-level and so low token-level information). Hence the
379 * function 'define_line' below and the TDefEol.
381 * update: TDefEol is handled in a special way at different places,
382 * a little bit like EOF, especially for error recovery, so this
383 * is an important token that should not be retagged!
386 * ugly hack, a better solution perhaps would be to erase TDefEOL
387 * from the Ast and list of tokens in parse_c.
389 * note: I do a +1 somewhere, it's for the unparsing to correctly sync.
391 * note: can't replace mark_end_define by simply a fakeInfo(). The reason
392 * is where is the \n TCommentSpace. Normally there is always a last token
393 * to synchronize on, either EOF or the token of the next toplevel.
394 * In the case of the #define we got in list of token
395 * [TCommentSpace "\n"; TDefEOL] but if TDefEOL is a fakeinfo then we will
396 * not synchronize on it and so we will not print the "\n".
397 * A solution would be to put the TDefEOL before the "\n".
398 * (jll: tried to do this, see the comment "Put end of line..." below)
400 * todo?: could put a ExpandedTok for that ?
402 let mark_end_define ii
=
404 { Ast_c.pinfo
= Ast_c.OriginTok
{ (Ast_c.parse_info_of_info
ii) with
406 Common.charpos
= Ast_c.pos_of_info
ii + 1
408 cocci_tag
= ref Ast_c.emptyAnnot
;
409 comments_tag
= ref Ast_c.emptyComments
;
414 (* put the TDefEOL at the good place *)
415 let rec define_line_1 acc xs
=
419 let line = Ast_c.line_of_info
ii in
420 let acc = (TDefine
ii) :: acc in
421 define_line_2
acc line ii xs
422 | TCppEscapedNewline
ii::xs
->
423 pr2 ("SUSPICIOUS: a \\ character appears outside of a #define at");
424 pr2 (Ast_c.strloc_of_info
ii);
425 let acc = (TCommentSpace
ii) :: acc in
427 | x
::xs
-> define_line_1 (x
::acc) xs
429 and define_line_2
acc line lastinfo xs
=
432 (* should not happened, should meet EOF before *)
434 List.rev
(mark_end_define lastinfo
::acc)
436 let line'
= TH.line_of_tok x
in
437 let info = TH.info_of_tok x
in
441 let acc = (mark_end_define lastinfo
) :: acc in
442 let acc = (EOF
ii) :: acc in
444 | TCppEscapedNewline
ii ->
445 if (line'
<> line) then pr2 "PB: WEIRD: not same line number";
446 let acc = (TCommentSpace
ii) :: acc in
447 define_line_2
acc (line+1) info xs
450 then define_line_2
(x
::acc) line info xs
452 (* Put end of line token before the newline. A newline at least
453 must be there because the line changed and because we saw a
454 #define previously to get to this function at all *)
456 ((List.hd
acc)::(mark_end_define lastinfo
::(List.tl
acc)))
460 let rec define_ident acc xs
=
464 let acc = TDefine
ii :: acc in
466 | TCommentSpace i1
::TIdent
(s
,i2
)::TOPar
(i3
)::xs
->
467 (* Change also the kind of TIdent to avoid bad interaction
468 * with other parsing_hack tricks. For instant if keep TIdent then
469 * the stringication algo can believe the TIdent is a string-macro.
470 * So simpler to change the kind of the ident too.
472 (* if TOParDefine sticked to the ident, then
473 * it's a macro-function. Change token to avoid ambiguity
474 * between #define foo(x) and #define foo (x)
476 let acc = (TCommentSpace i1
) :: acc in
477 let acc = (TIdentDefine
(s
,i2
)) :: acc in
478 let acc = (TOParDefine i3
) :: acc in
481 | TCommentSpace i1
::TIdent
(s
,i2
)::xs
->
482 let acc = (TCommentSpace i1
) :: acc in
483 let acc = (TIdentDefine
(s
,i2
)) :: acc in
486 (* bugfix: ident of macro (as well as params, cf below) can be tricky
487 * note, do we need to subst in the body of the define ? no cos
488 * here the issue is the name of the macro, as in #define inline,
489 * so obviously the name of this macro will not be used in its
490 * body (it would be a recursive macro, which is forbidden).
493 | TCommentSpace i1
::t
::xs
->
495 let s = TH.str_of_tok t
in
496 let ii = TH.info_of_tok t
in
497 if s ==~
Common.regexp_alpha
499 pr2 (spf
"remapping: %s to an ident in macro name" s);
500 let acc = (TCommentSpace i1
) :: acc in
501 let acc = (TIdentDefine
(s,ii)) :: acc in
505 pr2 "WEIRD: weird #define body";
510 pr2 "WEIRD: weird #define body";
514 let acc = x
:: acc in
519 let fix_tokens_define2 xs
=
520 define_ident [] (define_line_1 [] xs
)
522 let fix_tokens_define a
=
523 Common.profile_code
"C parsing.fix_define" (fun () -> fix_tokens_define2 a
)
529 (* ------------------------------------------------------------------------- *)
530 (* Other parsing hacks related to cpp, Include/Define hacks *)
531 (* ------------------------------------------------------------------------- *)
533 (* Sometimes I prefer to generate a single token for a list of things in the
534 * lexer so that if I have to passed them, like for passing TInclude then
535 * it's easy. Also if I don't do a single token, then I need to
536 * parse the rest which may not need special stuff, like detecting
537 * end of line which the parser is not really ready for. So for instance
538 * could I parse a #include <a/b/c/xxx.h> as 2 or more tokens ? just
539 * lex #include ? so then need recognize <a/b/c/xxx.h> as one token ?
540 * but this kind of token is valid only after a #include and the
541 * lexing and parsing rules are different for such tokens so not that
542 * easy to parse such things in parser_c.mly. Hence the following hacks.
544 * less?: maybe could get rid of this like I get rid of some of fix_define.
549 (* used to generate new token from existing one *)
550 let new_info posadd str
ii =
552 Ast_c.OriginTok
{ (Ast_c.parse_info_of_info
ii) with
553 charpos
= Ast_c.pos_of_info
ii + posadd
;
555 column
= Ast_c.col_of_info
ii + posadd
;
557 (* must generate a new ref each time, otherwise share *)
558 cocci_tag
= ref Ast_c.emptyAnnot
;
559 comments_tag
= ref Ast_c.emptyComments
;
563 let rec comment_until_defeol xs
=
566 (* job not done in Cpp_token_c.define_parse ? *)
567 failwith
"cant find end of define token TDefEOL"
570 | Parser_c.TDefEOL i
->
571 Parser_c.TCommentCpp
(Token_c.CppDirective
, TH.info_of_tok x
)
575 (* bugfix: otherwise may lose a TComment token *)
576 if TH.is_real_comment
x
578 else Parser_c.TCommentCpp
(Token_c.CppPassingNormal
(*good?*), TH.info_of_tok
x)
580 x'
::comment_until_defeol xs
583 let drop_until_defeol xs
=
585 (Common.drop_until
(function Parser_c.TDefEOL _
-> true | _
-> false) xs
)
589 (* ------------------------------------------------------------------------- *)
590 (* returns a pair (replaced token, list of next tokens) *)
591 (* ------------------------------------------------------------------------- *)
593 let tokens_include (info, includes
, filename
, inifdef
) =
594 Parser_c.TIncludeStart
(Ast_c.rewrap_str includes
info, inifdef
),
595 [Parser_c.TIncludeFilename
596 (filename
, (new_info (String.length includes
) filename
info))
602 (*****************************************************************************)
603 (* CPP handling: macros, ifdefs, macros defs *)
604 (*****************************************************************************)
606 (* ------------------------------------------------------------------------- *)
607 (* special skip_start skip_end handling *)
608 (* ------------------------------------------------------------------------- *)
610 (* note: after this normally the token list should not contain any more the
611 * TCommentSkipTagStart and End tokens.
613 let rec commentize_skip_start_to_end xs
=
618 | {tok
= TCommentSkipTagStart
info} ->
620 let (before
, x2
, after
) =
621 xs
+> Common.split_when
(function
622 | {tok
= TCommentSkipTagEnd _
} -> true
626 let topass = x::x2
::before
in
627 topass +> List.iter
(fun tok
->
628 set_as_comment
Token_c.CppPassingExplicit tok
630 commentize_skip_start_to_end after
632 failwith
"could not find end of skip_start special comment"
634 | {tok
= TCommentSkipTagEnd
info} ->
635 failwith
"found skip_end comment but no skip_start"
637 commentize_skip_start_to_end xs
643 (* ------------------------------------------------------------------------- *)
644 (* ifdef keeping/passing *)
645 (* ------------------------------------------------------------------------- *)
647 (* #if 0, #if 1, #if LINUX_VERSION handling *)
648 let rec find_ifdef_bool xs
=
649 xs
+> List.iter
(function
650 | NotIfdefLine _
-> ()
651 | Ifdefbool
(is_ifdef_positif
, xxs
, info_ifdef_stmt
) ->
653 msg_ifdef_bool_passing is_ifdef_positif
;
656 | [] -> raise Impossible
657 | firstclause
::xxs
->
658 info_ifdef_stmt
+> List.iter
(set_as_comment
Token_c.CppDirective
);
661 then xxs
+> List.iter
662 (iter_token_ifdef
(set_as_comment
Token_c.CppPassingNormal
))
664 firstclause
+> iter_token_ifdef
(set_as_comment
Token_c.CppPassingNormal
);
665 (match List.rev xxs
with
669 (iter_token_ifdef
(set_as_comment
Token_c.CppPassingNormal
))
670 | [] -> (* not #else *) ()
675 | Ifdef
(xxs
, info_ifdef_stmt
) -> xxs
+> List.iter
find_ifdef_bool
680 let thresholdIfdefSizeMid = 6
682 (* infer ifdef involving not-closed expressions/statements *)
683 let rec find_ifdef_mid xs
=
684 xs
+> List.iter
(function
685 | NotIfdefLine _
-> ()
686 | Ifdef
(xxs
, info_ifdef_stmt
) ->
688 | [] -> raise Impossible
690 | first
::second
::rest
->
691 (* don't analyse big ifdef *)
692 if xxs
+> List.for_all
693 (fun xs
-> List.length xs
<= thresholdIfdefSizeMid) &&
694 (* don't want nested ifdef *)
695 xxs
+> List.for_all
(fun xs
->
697 (function NotIfdefLine _
-> true | _
-> false)
701 let counts = xxs
+> List.map count_open_close_stuff_ifdef_clause
in
702 let cnt1, cnt2
= List.hd
counts in
703 if cnt1 <> 0 || cnt2
<> 0 &&
704 counts +> List.for_all
(fun x -> x =*= (cnt1, cnt2
))
706 if counts +> List.exists (fun (cnt1, cnt2) ->
707 cnt1 <> 0 || cnt2 <> 0
711 msg_ifdef_mid_something();
713 (* keep only first, treat the rest as comment *)
714 info_ifdef_stmt
+> List.iter
(set_as_comment
Token_c.CppDirective
);
715 (second
::rest
) +> List.iter
716 (iter_token_ifdef
(set_as_comment
Token_c.CppPassingCosWouldGetError
));
720 List.iter
find_ifdef_mid xxs
722 (* no need complex analysis for ifdefbool *)
723 | Ifdefbool
(_
, xxs
, info_ifdef_stmt
) ->
724 List.iter
find_ifdef_mid xxs
730 let thresholdFunheaderLimit = 4
732 (* ifdef defining alternate function header, type *)
733 let rec find_ifdef_funheaders = function
735 | NotIfdefLine _
::xs
-> find_ifdef_funheaders xs
737 (* ifdef-funheader if ifdef with 2 lines and a '{' in next line *)
739 ([(NotIfdefLine
(({col
= 0} as _xline1
)::line1
))::ifdefblock1
;
740 (NotIfdefLine
(({col
= 0} as xline2
)::line2
))::ifdefblock2
743 ::NotIfdefLine
(({tok
= TOBrace i
; col
= 0})::line3
)
745 when List.length ifdefblock1
<= thresholdFunheaderLimit &&
746 List.length ifdefblock2
<= thresholdFunheaderLimit
748 find_ifdef_funheaders xs
;
750 msg_ifdef_funheaders ();
751 info_ifdef_stmt
+> List.iter
(set_as_comment
Token_c.CppDirective
);
752 let all_toks = [xline2
] @ line2
in
753 all_toks +> List.iter
(set_as_comment
Token_c.CppPassingCosWouldGetError
) ;
754 ifdefblock2
+> iter_token_ifdef
(set_as_comment
Token_c.CppPassingCosWouldGetError
);
756 (* ifdef with nested ifdef *)
758 ([[NotIfdefLine
(({col
= 0} as _xline1
)::line1
)];
760 ([[NotIfdefLine
(({col
= 0} as xline2
)::line2
)];
761 [NotIfdefLine
(({col
= 0} as xline3
)::line3
)];
767 ::NotIfdefLine
(({tok
= TOBrace i
; col
= 0})::line4
)
770 find_ifdef_funheaders xs
;
772 msg_ifdef_funheaders ();
773 info_ifdef_stmt
+> List.iter
(set_as_comment
Token_c.CppDirective
);
774 info_ifdef_stmt2
+> List.iter
(set_as_comment
Token_c.CppDirective
);
775 let all_toks = [xline2
;xline3
] @ line2
@ line3
in
776 all_toks +> List.iter
(set_as_comment
Token_c.CppPassingCosWouldGetError
);
778 (* ifdef with elseif *)
780 ([[NotIfdefLine
(({col
= 0} as _xline1
)::line1
)];
781 [NotIfdefLine
(({col
= 0} as xline2
)::line2
)];
782 [NotIfdefLine
(({col
= 0} as xline3
)::line3
)];
785 ::NotIfdefLine
(({tok
= TOBrace i
; col
= 0})::line4
)
788 find_ifdef_funheaders xs
;
790 msg_ifdef_funheaders ();
791 info_ifdef_stmt
+> List.iter
(set_as_comment
Token_c.CppDirective
);
792 let all_toks = [xline2
;xline3
] @ line2
@ line3
in
793 all_toks +> List.iter
(set_as_comment
Token_c.CppPassingCosWouldGetError
)
796 | Ifdef
(xxs
,info_ifdef_stmt
)::xs
797 | Ifdefbool
(_
, xxs
,info_ifdef_stmt
)::xs
->
798 List.iter
find_ifdef_funheaders xxs
;
799 find_ifdef_funheaders xs
804 let rec adjust_inifdef_include xs
=
805 xs
+> List.iter
(function
806 | NotIfdefLine _
-> ()
807 | Ifdef
(xxs
, info_ifdef_stmt
) | Ifdefbool
(_
, xxs
, info_ifdef_stmt
) ->
808 xxs
+> List.iter
(iter_token_ifdef
(fun tokext
->
809 match tokext
.tok
with
810 | Parser_c.TInclude
(s1
, s2
, inifdef_ref
, ii) ->
822 let rec find_ifdef_cparen_else xs
=
824 xs
+> List.iter
(function
825 | NotIfdefLine _
-> ()
826 | Ifdef
(xxs
, info_ifdef_stmt
) ->
828 | [] -> raise Impossible
830 | first
::second
::rest
->
832 (* found a closing ')' just after the #else *)
834 (* Too bad ocaml does not support better list pattern matching
835 * a la Prolog-III where can match the end of lists.
838 if List.length first
= 0 then false
840 let last_line = Common.last first
in
843 if List.length xs
= 0 then false
845 let last_tok = Common.last xs
in
846 TH.is_cpar
last_tok.tok
847 | Ifdef _
| Ifdefbool _
-> false
849 if condition then begin
850 msg_ifdef_cparen_else();
852 (* keep only first, treat the rest as comment *)
853 info_ifdef_stmt
+> List.iter
(set_as_comment
Token_c.CppDirective
);
854 (second
::rest
) +> List.iter
855 (iter_token_ifdef
(set_as_comment
Token_c.CppPassingCosWouldGetError
));
861 (* no need complex analysis for ifdefbool *)
862 | Ifdefbool
(_
, xxs
, info_ifdef_stmt
) ->
868 (* ------------------------------------------------------------------------- *)
869 (* cpp-builtin part2, macro, using standard.h or other defs *)
870 (* ------------------------------------------------------------------------- *)
872 (* now in cpp_token_c.ml *)
874 (* ------------------------------------------------------------------------- *)
875 (* stringification *)
876 (* ------------------------------------------------------------------------- *)
878 let rec find_string_macro_paren xs
=
881 | Parenthised
(xxs
, info_parens
)::xs
->
882 xxs
+> List.iter
(fun xs
->
884 (function PToken
({tok
= (TString _
| TMacroString _
)}) -> true | _
-> false) &&
886 (function PToken
({tok
= (TString _
| TMacroString _
)}) | PToken
({tok
= TIdent _
}) ->
889 xs
+> List.iter
(fun tok
->
891 | PToken
({tok
= TIdent
(s,_
)} as id
) ->
892 msg_stringification s;
893 id
.tok
<- TMacroString
(s, TH.info_of_tok id
.tok
);
897 find_string_macro_paren xs
899 find_string_macro_paren xs
901 find_string_macro_paren xs
904 (* ------------------------------------------------------------------------- *)
906 (* ------------------------------------------------------------------------- *)
908 (* don't forget to recurse in each case *)
909 let rec find_macro_paren xs
=
914 | PToken
({tok
= Tattribute _
} as id
)
915 ::Parenthised
(xxs
,info_parens
)
918 pr2_cpp ("MACRO: __attribute detected ");
919 [Parenthised
(xxs
, info_parens
)] +>
920 iter_token_paren
(set_as_comment
Token_c.CppAttr
);
921 set_as_comment
Token_c.CppAttr id
;
924 | PToken
({tok
= TattributeNoarg _
} as id
)
927 pr2_cpp ("MACRO: __attributenoarg detected ");
928 set_as_comment
Token_c.CppAttr id
;
932 (* attribute cpp, __xxx id *)
933 | PToken
({tok
= TIdent
(s,i1
)} as id
)
934 ::PToken
({tok
= TIdent
(s2
, i2
)} as id2
)
935 ::xs
when s ==~
regexp_annot
938 id
.tok
<- TMacroAttr
(s, i1
);
939 find_macro_paren ((PToken id2
)::xs
); (* recurse also on id2 ? *)
941 (* attribute cpp, id __xxx *)
942 | PToken
({tok
= TIdent
(s,i1
)} as _id
)
943 ::PToken
({tok
= TIdent
(s2
, i2
)} as id2
)
944 ::xs
when s2
==~
regexp_annot && (not
(s ==~
regexp_typedef))
947 id2
.tok
<- TMacroAttr
(s2
, i2
);
950 | PToken
({tok
= (Tstatic _
| Textern _
)} as tok1
)
951 ::PToken
({tok
= TIdent
(s,i1
)} as attr
)
952 ::xs
when s ==~
regexp_annot
954 pr2_cpp ("storage attribute: " ^
s);
955 attr
.tok
<- TMacroAttrStorage
(s,i1
);
956 (* recurse, may have other storage attributes *)
957 find_macro_paren (PToken
(tok1
)::xs
)
962 (* storage attribute *)
963 | PToken
({tok
= (Tstatic _
| Textern _
)} as tok1
)
964 ::PToken
({tok
= TMacroAttr
(s,i1
)} as attr
)::xs
966 pr2_cpp ("storage attribute: " ^
s);
967 attr
.tok
<- TMacroAttrStorage
(s,i1
);
968 (* recurse, may have other storage attributes *)
969 find_macro_paren (PToken
(tok1
)::xs
)
974 * the order of the matching clause is important
978 (* string macro with params, before case *)
979 | PToken
({tok
= (TString _
| TMacroString _
)})::PToken
({tok
= TIdent
(s,_
)} as id
)
980 ::Parenthised
(xxs
, info_parens
)
983 msg_stringification_params s;
984 id
.tok
<- TMacroString
(s, TH.info_of_tok id
.tok
);
985 [Parenthised
(xxs
, info_parens
)] +>
986 iter_token_paren
(set_as_comment
Token_c.CppMacro
);
990 | PToken
({tok
= TIdent
(s,_
)} as id
)
991 ::Parenthised
(xxs
, info_parens
)
992 ::PToken
({tok
= (TString _
| TMacroString _
)})
995 msg_stringification_params s;
996 id
.tok
<- TMacroString
(s, TH.info_of_tok id
.tok
);
997 [Parenthised
(xxs
, info_parens
)] +>
998 iter_token_paren
(set_as_comment
Token_c.CppMacro
);
1002 (* for the case where the string is not inside a funcall, but
1003 * for instance in an initializer.
1006 (* string macro variable, before case *)
1007 | PToken
({tok
= (TString _
| TMacroString _
)})::PToken
({tok
= TIdent
(s,_
)} as id
)
1010 msg_stringification s;
1011 id
.tok
<- TMacroString
(s, TH.info_of_tok id
.tok
);
1015 | PToken
({tok
= TIdent
(s,_
)} as id
)
1016 ::PToken
({tok
= (TString _
| TMacroString _
)})
1019 msg_stringification s;
1020 id
.tok
<- TMacroString
(s, TH.info_of_tok id
.tok
);
1028 | (PToken
x)::xs
-> find_macro_paren xs
1029 | (Parenthised
(xxs
, info_parens
))::xs
->
1030 xxs
+> List.iter
find_macro_paren;
1037 (* don't forget to recurse in each case *)
1038 let rec find_macro_lineparen xs
=
1042 (* linuxext: ex: static [const] DEVICE_ATTR(); *)
1045 [PToken
({tok
= Tstatic _
});
1046 PToken
({tok
= TIdent
(s,_
)} as macro
);
1047 Parenthised
(xxs
,info_parens
);
1048 PToken
({tok
= TPtVirg _
});
1052 when (s ==~
regexp_macro) ->
1054 msg_declare_macro s;
1055 let info = TH.info_of_tok macro
.tok
in
1056 macro
.tok
<- TMacroDecl
(Ast_c.str_of_info
info, info);
1058 find_macro_lineparen (xs
)
1060 (* the static const case *)
1063 [PToken
({tok
= Tstatic _
});
1064 PToken
({tok
= Tconst _
} as const
);
1065 PToken
({tok
= TIdent
(s,_
)} as macro
);
1066 Parenthised
(xxs
,info_parens
);
1067 PToken
({tok
= TPtVirg _
});
1073 when (s ==~
regexp_macro) ->
1075 msg_declare_macro s;
1076 let info = TH.info_of_tok macro
.tok
in
1077 macro
.tok
<- TMacroDecl
(Ast_c.str_of_info
info, info);
1079 (* need retag this const, otherwise ambiguity in grammar
1080 21: shift/reduce conflict (shift 121, reduce 137) on Tconst
1081 decl2 : Tstatic . TMacroDecl TOPar argument_list TCPar ...
1082 decl2 : Tstatic . Tconst TMacroDecl TOPar argument_list TCPar ...
1083 storage_class_spec : Tstatic . (137)
1085 const
.tok
<- TMacroDeclConst
(TH.info_of_tok const
.tok
);
1087 find_macro_lineparen (xs
)
1090 (* same but without trailing ';'
1092 * I do not put the final ';' because it can be on a multiline and
1093 * because of the way mk_line is coded, we will not have access to
1094 * this ';' on the next line, even if next to the ')' *)
1096 ([PToken
({tok
= Tstatic _
});
1097 PToken
({tok
= TIdent
(s,_
)} as macro
);
1098 Parenthised
(xxs
,info_parens
);
1102 when s ==~
regexp_macro ->
1104 msg_declare_macro s;
1105 let info = TH.info_of_tok macro
.tok
in
1106 macro
.tok
<- TMacroDecl
(Ast_c.str_of_info
info, info);
1108 find_macro_lineparen (xs
)
1113 (* on multiple lines *)
1116 (PToken
({tok
= Tstatic _
})::[]
1120 [PToken
({tok
= TIdent
(s,_
)} as macro
);
1121 Parenthised
(xxs
,info_parens
);
1122 PToken
({tok
= TPtVirg _
});
1127 when (s ==~
regexp_macro) ->
1129 msg_declare_macro s;
1130 let info = TH.info_of_tok macro
.tok
in
1131 macro
.tok
<- TMacroDecl
(Ast_c.str_of_info
info, info);
1133 find_macro_lineparen (xs
)
1136 (* linuxext: ex: DECLARE_BITMAP();
1138 * Here I use regexp_declare and not regexp_macro because
1139 * Sometimes it can be a FunCallMacro such as DEBUG(foo());
1140 * Here we don't have the preceding 'static' so only way to
1141 * not have positive is to restrict to .*DECLARE.* macros.
1143 * but there is a grammar rule for that, so don't need this case anymore
1144 * unless the parameter of the DECLARE_xxx are weird and can not be mapped
1145 * on a argument_list
1149 ([PToken
({tok
= TIdent
(s,_
)} as macro
);
1150 Parenthised
(xxs
,info_parens
);
1151 PToken
({tok
= TPtVirg _
});
1155 when (s ==~
regexp_declare) ->
1157 msg_declare_macro s;
1158 let info = TH.info_of_tok macro
.tok
in
1159 macro
.tok
<- TMacroDecl
(Ast_c.str_of_info
info, info);
1161 find_macro_lineparen (xs
)
1167 * Could also transform the TIdent in a TMacroTop but can have false
1168 * positive, so easier to just change the TCPar and so just solve
1169 * the end-of-stream pb of ocamlyacc
1172 ([PToken
({tok
= TIdent
(s,ii); col
= col1
; where
= ctx
} as _macro
);
1173 Parenthised
(xxs
,info_parens
);
1176 ::xs
when col1
=|= 0
1179 (* to reduce number of false positive *)
1181 | (Line
(PToken
({col
= col2
} as other
)::restline2
))::_
->
1182 TH.is_eof other
.tok
|| (col2
=|= 0 &&
1183 (match other
.tok
with
1184 | TOBrace _
-> false (* otherwise would match funcdecl *)
1185 | TCBrace _
when ctx
<> InFunction
-> false
1189 | tok
when TH.is_binary_operator tok
-> false
1200 msg_macro_toplevel_noptvirg s;
1201 (* just to avoid the end-of-stream pb of ocamlyacc *)
1202 let tcpar = Common.last info_parens
in
1203 tcpar.tok
<- TCParEOL
(TH.info_of_tok
tcpar.tok
);
1205 (*macro.tok <- TMacroTop (s, TH.info_of_tok macro.tok);*)
1209 find_macro_lineparen (xs
)
1213 (* macro with parameters
1218 ([PToken
({tok
= TIdent
(s,ii); col
= col1
; where
= ctx
} as macro
);
1219 Parenthised
(xxs
,info_parens
);
1223 (PToken
({col
= col2
} as other
)::restline2
1226 (* when s ==~ regexp_macro *)
1230 (match other
.tok
with
1231 | TOBrace _
-> false (* otherwise would match funcdecl *)
1232 | TCBrace _
when ctx
<> InFunction
-> false
1236 | tok
when TH.is_binary_operator tok
-> false
1243 (match other
.tok
, restline2
with
1244 | TCBrace _
, _
when ctx
=*= InFunction
-> true
1245 | Treturn _
, _
-> true
1247 | Telse _
, _
-> true
1249 (* case of label, usually put in first line *)
1250 | TIdent _
, (PToken
({tok
= TDotDot _
}))::_
->
1262 if col1
=|= 0 then ()
1264 msg_macro_noptvirg s;
1265 macro
.tok
<- TMacroStmt
(s, TH.info_of_tok macro
.tok
);
1266 [Parenthised
(xxs
, info_parens
)] +>
1267 iter_token_paren
(set_as_comment
Token_c.CppMacro
);
1270 find_macro_lineparen (line2
::xs
)
1272 (* linuxext:? single macro
1277 * todo: factorize code with previous rule ?
1280 ([PToken
({tok
= TIdent
(s,ii); col
= col1
; where
= ctx
} as macro
);
1284 (PToken
({col
= col2
} as other
)::restline2
1287 (* when s ==~ regexp_macro *)
1291 col1
<> 0 && (* otherwise can match typedef of fundecl*)
1292 (match other
.tok
with
1293 | TPtVirg _
-> false
1295 | TCBrace _
when ctx
<> InFunction
-> false
1296 | tok
when TH.is_binary_operator tok
-> false
1301 (match other
.tok
with
1302 | TCBrace _
when ctx
=*= InFunction
-> true
1312 msg_macro_noptvirg_single s;
1313 macro
.tok
<- TMacroStmt
(s, TH.info_of_tok macro
.tok
);
1315 find_macro_lineparen (line2
::xs
)
1318 find_macro_lineparen xs
1322 (* ------------------------------------------------------------------------- *)
1323 (* define tobrace init *)
1324 (* ------------------------------------------------------------------------- *)
1326 let rec find_define_init_brace_paren xs
=
1331 (* mainly for firefox *)
1332 | (PToken
{tok
= TDefine _
})
1333 ::(PToken
{tok
= TIdentDefine
(s,_
)})
1334 ::(PToken
({tok
= TOBrace i1
} as tokbrace
))
1339 match tok2
.tok
, tok3
.tok
with
1340 | TInt _
, TComma _
-> true
1341 | TString _
, TComma _
-> true
1342 | TIdent _
, TComma _
-> true
1348 pr2_cpp("found define initializer: " ^
s);
1349 tokbrace
.tok
<- TOBraceDefineInit i1
;
1354 (* mainly for linux, especially in sound/ *)
1355 | (PToken
{tok
= TDefine _
})
1356 ::(PToken
{tok
= TIdentDefine
(s,_
)})
1357 ::(Parenthised
(xxx
, info_parens
))
1358 ::(PToken
({tok
= TOBrace i1
} as tokbrace
))
1363 match tok2
.tok
, tok3
.tok
with
1364 | TInt _
, TComma _
-> true
1365 | TDot _
, TIdent _
-> true
1366 | TIdent _
, TComma _
-> true
1372 pr2_cpp("found define initializer with param: " ^
s);
1373 tokbrace
.tok
<- TOBraceDefineInit i1
;
1381 | (PToken
x)::xs
-> aux xs
1382 | (Parenthised
(xxs
, info_parens
))::xs
->
1383 (* not need for tobrace init:
1384 * xxs +> List.iter aux;
1391 (* ------------------------------------------------------------------------- *)
1393 (* ------------------------------------------------------------------------- *)
1395 (* obsolete now with macro expansion ? get some regression if comment.
1396 * todo: if do bad decision here, then it can influence other phases
1397 * and make it hard to parse. So maybe when have a parse error, should
1398 * undo some of the guess those heuristics have done, and restore
1399 * the original token value.
1402 let rec find_actions = function
1405 | PToken
({tok
= TIdent
(s,ii)})
1406 ::Parenthised
(xxs
,info_parens
)
1409 xxs
+> List.iter
find_actions;
1410 let modified = find_actions_params xxs
in
1412 then msg_macro_higher_order s
1417 and find_actions_params xxs
=
1418 xxs
+> List.fold_left
(fun acc xs
->
1419 let toks = tokens_of_paren xs
in
1420 if toks +> List.exists
(fun x -> TH.is_statement
x.tok
)
1421 (* undo: && List.length toks > 1
1422 * good for sparse, not good for linux
1425 xs
+> iter_token_paren
(fun x ->
1428 (* certainly because paren detection had a pb because of
1429 * some ifdef-exp. Do similar additional checking than
1430 * what is done in set_as_comment.
1432 pr2 "PB: weird, I try to tag an EOF token as an action"
1434 (* cf tests-bis/no_cpar_macro.c *)
1437 pr2 "PB: weird, I try to tag an EOM token as an action"
1439 x.tok
<- TAction
(TH.info_of_tok
x.tok
);
1448 (* ------------------------------------------------------------------------- *)
1449 (* main fix cpp function *)
1450 (* ------------------------------------------------------------------------- *)
1452 let filter_cpp_stuff xs
=
1456 | tok
when TH.is_comment tok
-> false
1457 (* don't want drop the define, or if drop, have to drop
1458 * also its body otherwise the line heuristics may be lost
1459 * by not finding the TDefine in column 0 but by finding
1460 * a TDefineIdent in a column > 0
1462 | Parser_c.TDefine _
-> true
1463 | tok
when TH.is_cpp_instruction tok
-> false
1468 let insert_virtual_positions l
=
1469 let strlen x = String.length
(Ast_c.str_of_info
x) in
1470 let rec loop prev offset
acc = function
1473 let ii = TH.info_of_tok
x in
1475 TH.visitor_info_of_tok
(function ii -> Ast_c.rewrap_pinfo pi
ii) x in
1476 match Ast_c.pinfo_of_info
ii with
1477 Ast_c.OriginTok pi
->
1478 let prev = Ast_c.parse_info_of_info
ii in
1479 loop prev (strlen ii) (x::acc) xs
1480 | Ast_c.ExpandedTok
(pi
,_
) ->
1481 let x'
= inject (Ast_c.ExpandedTok
(pi
,(prev,offset
))) in
1482 loop prev (offset
+ (strlen ii)) (x'
::acc) xs
1483 | Ast_c.FakeTok
(s,_
) ->
1484 let x'
= inject (Ast_c.FakeTok
(s,(prev,offset
))) in
1485 loop prev (offset
+ (strlen ii)) (x'
::acc) xs
1486 | Ast_c.AbstractLineTok _
-> failwith
"abstract not expected" in
1487 let rec skip_fake = function
1490 let ii = TH.info_of_tok
x in
1491 match Ast_c.pinfo_of_info
ii with
1492 | Ast_c.OriginTok pi
->
1493 let prev = Ast_c.parse_info_of_info
ii in
1494 let res = loop prev (strlen ii) [] xs
in
1496 | _
-> x::skip_fake xs
in
1500 (* ------------------------------------------------------------------------- *)
1501 let fix_tokens_cpp2 ~macro_defs tokens
=
1502 let tokens2 = ref (tokens
+> Common.acc_map
TV.mk_token_extended
) in
1505 (* the order is important, if you put the action heuristic first,
1506 * then because of ifdef, can have not closed paren
1507 * and so may believe that higher order macro
1508 * and it will eat too much tokens. So important to do
1511 * I recompute multiple times cleaner cos the mutable
1512 * can have be changed and so may have more comments
1513 * in the token original list.
1517 commentize_skip_start_to_end !tokens2;
1520 let cleaner = !tokens2 +> List.filter
(fun x ->
1521 (* is_comment will also filter the TCommentCpp created in
1522 * commentize_skip_start_to_end *)
1523 not
(TH.is_comment
x.tok
) (* could filter also #define/#include *)
1525 let ifdef_grouped = TV.mk_ifdef
cleaner in
1526 set_ifdef_parenthize_info ifdef_grouped;
1528 find_ifdef_funheaders ifdef_grouped;
1529 find_ifdef_bool ifdef_grouped;
1530 find_ifdef_mid ifdef_grouped;
1531 (* change order ? maybe cparen_else heuristic make some of the funheaders
1532 * heuristics irrelevant ?
1534 find_ifdef_cparen_else ifdef_grouped;
1535 adjust_inifdef_include ifdef_grouped;
1539 let cleaner = !tokens2 +> filter_cpp_stuff in
1541 let paren_grouped = TV.mk_parenthised
cleaner in
1542 Cpp_token_c.apply_macro_defs
1543 ~
msg_apply_known_macro
1544 ~
msg_apply_known_macro_hint
1545 macro_defs
paren_grouped;
1546 (* because the before field is used by apply_macro_defs *)
1547 tokens2 := TV.rebuild_tokens_extented
!tokens2;
1549 (* tagging contextual info (InFunc, InStruct, etc). Better to do
1550 * that after the "ifdef-simplification" phase.
1552 let cleaner = !tokens2 +> List.filter
(fun x ->
1553 not
(TH.is_comment
x.tok
) (* could filter also #define/#include *)
1556 let brace_grouped = TV.mk_braceised
cleaner in
1557 set_context_tag
brace_grouped;
1560 let cleaner = !tokens2 +> filter_cpp_stuff in
1562 let paren_grouped = TV.mk_parenthised
cleaner in
1563 let line_paren_grouped = TV.mk_line_parenthised
paren_grouped in
1564 find_define_init_brace_paren paren_grouped;
1565 find_string_macro_paren paren_grouped;
1566 find_macro_lineparen line_paren_grouped;
1567 find_macro_paren paren_grouped;
1570 (* obsolete: actions ? not yet *)
1571 let cleaner = !tokens2 +> filter_cpp_stuff in
1572 let paren_grouped = TV.mk_parenthised
cleaner in
1573 find_actions paren_grouped;
1577 insert_virtual_positions (!tokens2 +> Common.acc_map
(fun x -> x.tok
))
1580 let time_hack1 ~macro_defs a
=
1581 Common.profile_code_exclusif
"HACK" (fun () -> fix_tokens_cpp2 ~macro_defs a
)
1583 let fix_tokens_cpp ~macro_defs a
=
1584 Common.profile_code
"C parsing.fix_cpp" (fun () -> time_hack1 ~macro_defs a
)
1589 (*****************************************************************************)
1590 (* Lexing with lookahead *)
1591 (*****************************************************************************)
1593 (* Why using yet another parsing_hack technique ? The fix_xxx where do
1594 * some pre-processing on the full list of tokens is not enough ?
1595 * No cos sometimes we need more contextual info, and even if
1596 * set_context() tries to give some contextual info, it's not completely
1597 * accurate so the following code give yet another alternative, yet another
1598 * chance to transform some tokens.
1600 * todo?: maybe could try to get rid of this technique. Maybe a better
1601 * set_context() would make possible to move this code using a fix_xx
1604 * LALR(k) trick. We can do stuff by adding cases in lexer_c.mll, but
1605 * it is more general to do it via my LALR(k) tech. Because here we can
1606 * transform some token give some context information. So sometimes it
1607 * makes sense to transform a token in one context, sometimes not, and
1608 * lex can not provide us this context information. Note that the order
1609 * in the pattern matching in lookahead is important. Do not cut/paste.
1611 * Note that in next there is only "clean" tokens, there is no comment
1612 * or space tokens. This is done by the caller.
1616 open Lexer_parser
(* for the fields of lexer_hint type *)
1618 let not_struct_enum = function
1619 | (Parser_c.Tstruct _
| Parser_c.Tunion _
| Parser_c.Tenum _
)::_
-> false
1623 let lookahead2 ~pass next before
=
1625 match (next
, before
) with
1627 (*-------------------------------------------------------------*)
1628 (* typedef inference, parse_typedef_fix3 *)
1629 (*-------------------------------------------------------------*)
1631 | (TIdent
(s,i1
)::TIdent
(s2
,i2
)::_
, _
) when not_struct_enum before
&& s =$
= s2
1633 (* (take_safe 1 !passed_tok <> [TOPar]) -> *)
1635 (* parse_typedef_fix3:
1636 * acpi_object acpi_object;
1637 * etait mal parsé, car pas le temps d'appeler dt() dans le type_spec.
1638 * Le parser en interne a deja appelé le prochain token pour pouvoir
1639 * decider des choses.
1640 * => special case in lexer_heuristic, again
1642 if !Flag_parsing_c.debug_typedef
1643 then pr2 ("TYPEDEF: disable typedef cos special case: " ^
s);
1645 LP.disable_typedef
();
1647 msg_typedef s; LP.add_typedef_root
s;
1648 TypedefIdent
(s, i1
)
1651 | (TIdent
(s, i1
)::TIdent
(s2
, i2
)::_
, _
) when not_struct_enum before
1654 (* && not_annot s2 BUT lead to false positive*)
1656 msg_typedef s; LP.add_typedef_root
s;
1657 TypedefIdent
(s, i1
)
1661 | (TIdent
(s, i1
)::Tinline i2
::_
, _
) when not_struct_enum before
1664 msg_typedef s; LP.add_typedef_root
s;
1665 TypedefIdent
(s, i1
)
1668 (* [,(] xx [,)] AND param decl *)
1669 | (TIdent
(s, i1
)::(TComma _
|TCPar _
)::_
, (TComma _
|TOPar _
)::_
)
1670 when not_struct_enum before
&& (LP.current_context
() =*= LP.InParameter
)
1673 msg_typedef s; LP.add_typedef_root
s;
1674 TypedefIdent
(s, i1
)
1677 (* specialcase: [,(] xx* [,)] *)
1678 | (TIdent
(s, i1
)::TMul _
::(TComma _
|TCPar _
)::_
, (*(TComma _|TOPar _)::*)_
)
1679 when not_struct_enum before
1680 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1683 msg_typedef s; LP.add_typedef_root
s;
1684 TypedefIdent
(s, i1
)
1688 (* specialcase: [,(] xx** [,)] *)
1689 | (TIdent
(s, i1
)::TMul _
::TMul _
::(TComma _
|TCPar _
)::_
, (*(TComma _|TOPar _)::*)_
)
1690 when not_struct_enum before
1691 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1694 msg_typedef s; LP.add_typedef_root
s;
1695 TypedefIdent
(s, i1
)
1699 (* xx const * USELESS because of next rule ? *)
1700 | (TIdent
(s, i1
)::(Tconst _
|Tvolatile _
|Trestrict _
)::TMul _
::_
, _
)
1701 when not_struct_enum before
1702 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1706 msg_typedef s; LP.add_typedef_root
s;
1707 TypedefIdent
(s, i1
)
1710 | (TIdent
(s, i1
)::(Tconst _
|Tvolatile _
|Trestrict _
)::_
, _
)
1711 when not_struct_enum before
1713 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1716 msg_typedef s; LP.add_typedef_root
s;
1717 TypedefIdent
(s, i1
)
1721 | (TIdent
(s, i1
)::TMul _
::(Tconst _
| Tvolatile _
|Trestrict _
)::_
, _
)
1722 when not_struct_enum before
1725 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1727 msg_typedef s; LP.add_typedef_root
s;
1728 TypedefIdent
(s, i1
)
1732 | (TIdent
(s, i1
)::TCPar _
::_
, (Tconst _
| Tvolatile _
|Trestrict _
)::TOPar _
::_
) when
1734 msg_typedef s; LP.add_typedef_root
s;
1735 TypedefIdent
(s, i1
)
1739 (* ( xx ) [sizeof, ~] *)
1740 | (TIdent
(s, i1
)::TCPar _
::(Tsizeof _
|TTilde _
)::_
, TOPar _
::_
)
1741 when not_struct_enum before
1744 msg_typedef s; LP.add_typedef_root
s;
1745 TypedefIdent
(s, i1
)
1747 (* [(,] xx [ AND parameterdeclaration *)
1748 | (TIdent
(s, i1
)::TOCro _
::_
, (TComma _
|TOPar _
)::_
)
1749 when (LP.current_context
() =*= LP.InParameter
)
1752 msg_typedef s; LP.add_typedef_root
s;
1753 TypedefIdent
(s, i1
)
1755 (*------------------------------------------------------------*)
1756 (* if 'x*y' maybe an expr, maybe just a classic multiplication *)
1757 (* but if have a '=', or ',' I think not *)
1758 (*------------------------------------------------------------*)
1760 (* static xx * yy *)
1761 | (TIdent
(s, i1
)::TMul _
::TIdent
(s2
, i2
)::_
,
1762 (Tregister _
|Tstatic _
|Tvolatile _
|Tconst _
|Trestrict _
)::_
) when
1765 msg_typedef s; LP.add_typedef_root
s;
1766 TypedefIdent
(s, i1
)
1768 (* TODO xx * yy ; AND in start of compound element *)
1771 (* xx * yy, AND in paramdecl *)
1772 | (TIdent
(s, i1
)::TMul _
::TIdent
(s2
, i2
)::TComma _
::_
, _
)
1773 when not_struct_enum before
&& (LP.current_context
() =*= LP.InParameter
)
1777 msg_typedef s; LP.add_typedef_root
s;
1778 TypedefIdent
(s, i1
)
1781 (* xx * yy ; AND in Toplevel, except when have = before *)
1782 | (TIdent
(s, i1
)::TMul _
::TIdent
(s2
, i2
)::TPtVirg _
::_
, TEq _
::_
) ->
1784 | (TIdent
(s, i1
)::TMul _
::TIdent
(s2
, i2
)::TPtVirg _
::_
, _
)
1785 when not_struct_enum before
&& (LP.is_top_or_struct
(LP.current_context
()))
1787 msg_typedef s; LP.add_typedef_root
s;
1788 TypedefIdent
(s, i1
)
1790 (* xx * yy , AND in Toplevel *)
1791 | (TIdent
(s, i1
)::TMul _
::TIdent
(s2
, i2
)::TComma _
::_
, _
)
1792 when not_struct_enum before
&& (LP.current_context
() =*= LP.InTopLevel
)
1796 msg_typedef s; LP.add_typedef_root
s;
1797 TypedefIdent
(s, i1
)
1799 (* xx * yy ( AND in Toplevel *)
1800 | (TIdent
(s, i1
)::TMul _
::TIdent
(s2
, i2
)::TOPar _
::_
, _
)
1801 when not_struct_enum before
1802 && (LP.is_top_or_struct
(LP.current_context
()))
1805 msg_typedef s; LP.add_typedef_root
s;
1806 TypedefIdent
(s, i1
)
1809 (* todo? enough ? cos in struct def we can have some expression ! *)
1810 | (TIdent
(s, i1
)::TMul _
::TIdent
(s2
, i2
)::TOCro _
::_
, _
)
1811 when not_struct_enum before
&&
1812 (LP.is_top_or_struct
(LP.current_context
()))
1815 msg_typedef s; LP.add_typedef_root
s;
1816 TypedefIdent
(s, i1
)
1818 (* u16: 10; in struct *)
1819 | (TIdent
(s, i1
)::TDotDot _
::_
, (TOBrace _
| TPtVirg _
)::_
)
1820 when (LP.is_top_or_struct
(LP.current_context
()))
1823 msg_typedef s; LP.add_typedef_root
s;
1824 TypedefIdent
(s, i1
)
1827 (* why need TOPar condition as stated in preceding rule ? really needed ? *)
1828 (* YES cos at toplevel can have some expression !! for instance when *)
1829 (* enter in the dimension of an array *)
1831 | (TIdent s::TMul::TIdent s2::_ , _)
1832 when (take_safe 1 !passed_tok <> [Tstruct] &&
1833 (take_safe 1 !passed_tok <> [Tenum]))
1835 !LP._lexer_hint = Some LP.Toplevel ->
1836 msg_typedef s; LP.add_typedef_root s;
1841 | (TIdent
(s, i1
)::TMul _
::TIdent
(s2
, i2
)::TEq _
::_
, _
)
1842 when not_struct_enum before
1845 msg_typedef s; LP.add_typedef_root
s;
1846 TypedefIdent
(s, i1
)
1849 (* xx * yy) AND in paramdecl *)
1850 | (TIdent
(s, i1
)::TMul _
::TIdent
(s2
, i2
)::TCPar _
::_
, _
)
1851 when not_struct_enum before
&& (LP.current_context
() =*= LP.InParameter
)
1854 msg_typedef s; LP.add_typedef_root
s;
1855 TypedefIdent
(s, i1
)
1858 (* xx * yy; *) (* wrong ? *)
1859 | (TIdent
(s, i1
)::TMul _
::TIdent
(s2
, i2
)::TPtVirg _
::_
,
1860 (TOBrace _
| TPtVirg _
)::_
) when not_struct_enum before
1863 msg_typedef s; LP.add_typedef_root
s;
1864 msg_maybe_dangereous_typedef s;
1865 TypedefIdent
(s, i1
)
1868 (* xx * yy, and ';' before xx *) (* wrong ? *)
1869 | (TIdent
(s, i1
)::TMul _
::TIdent
(s2
, i2
)::TComma _
::_
,
1870 (TOBrace _
| TPtVirg _
)::_
) when
1873 msg_typedef s; LP.add_typedef_root
s;
1874 TypedefIdent
(s, i1
)
1878 | (TIdent
(s, i1
)::TMul _
::TIdent
(s2
, i2
)::_
, _
)
1879 when s ==~
regexp_typedef && not_struct_enum before
1880 (* struct user_info_t sometimes *)
1883 msg_typedef s; LP.add_typedef_root
s;
1884 TypedefIdent
(s, i1
)
1886 (* xx ** yy *) (* wrong ? *)
1887 | (TIdent
(s, i1
)::TMul _
::TMul _
::TIdent
(s2
, i2
)::_
, _
)
1888 when not_struct_enum before
1889 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1892 msg_typedef s; LP.add_typedef_root
s;
1893 TypedefIdent
(s, i1
)
1896 | (TIdent
(s, i1
)::TMul _
::TMul _
::TMul _
::TIdent
(s2
, i2
)::_
, _
)
1897 when not_struct_enum before
1899 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1901 msg_typedef s; LP.add_typedef_root
s;
1902 TypedefIdent
(s, i1
)
1905 | (TIdent
(s, i1
)::TMul _
::TMul _
::TCPar _
::_
, _
)
1906 when not_struct_enum before
1907 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1910 msg_typedef s; LP.add_typedef_root
s;
1911 TypedefIdent
(s, i1
)
1915 (* ----------------------------------- *)
1916 (* old: why not do like for other rules and start with TIdent ?
1917 * why do TOPar :: TIdent :: ..., _ and not TIdent :: ..., TOPAr::_ ?
1918 * new: prefer now start with TIdent because otherwise the add_typedef_root
1919 * may have no effect if in second pass or if have disable the add_typedef.
1923 | (TIdent
(s, i1
)::TCPar i2
::(TIdent
(_
,i3
)|TInt
(_
,i3
))::_
,
1925 when not
(TH.is_stuff_taking_parenthized
x) &&
1926 Ast_c.line_of_info i2
=|= Ast_c.line_of_info i3
1930 msg_typedef s; LP.add_typedef_root
s;
1932 TypedefIdent
(s, i1
)
1936 * but false positif: typedef int (xxx_t)(...), so do specialisation below.
1939 | (TIdent (s, i1)::TCPar _::TOPar _::_ , (TOPar info)::x::_)
1940 when not (TH.is_stuff_taking_parenthized x)
1943 msg_typedef s; LP.add_typedef_root s;
1945 TypedefIdent
(s, i1
)
1947 (* special case: = (xx) ( yy) *)
1948 | (TIdent
(s, i1
)::TCPar _
::TOPar _
::_
,
1949 (TOPar
info)::(TEq _
|TEqEq _
)::_
)
1952 msg_typedef s; LP.add_typedef_root
s;
1954 TypedefIdent
(s, i1
)
1958 | (TIdent
(s, i1
)::TMul _
::TCPar _
::TIdent
(s2
, i2
)::_
, (TOPar
info)::_
) when
1961 msg_typedef s; LP.add_typedef_root
s;
1966 (* (xx){ ... } constructor *)
1967 | (TIdent
(s, i1
)::TCPar _
::TOBrace _
::_
, TOPar _
::x::_
)
1968 when (*s ==~ regexp_typedef && *) not
(TH.is_stuff_taking_parenthized
x)
1971 msg_typedef s; LP.add_typedef_root
s;
1972 TypedefIdent
(s, i1
)
1975 (* can have sizeof on expression
1976 | (Tsizeof::TOPar::TIdent s::TCPar::_, _) ->
1977 msg_typedef s; LP.add_typedef_root s;
1982 (* ----------------------------------- *)
1983 (* x ( *y )(params), function pointer *)
1984 | (TIdent
(s, i1
)::TOPar _
::TMul _
::TIdent _
::TCPar _
::TOPar _
::_
, _
)
1985 when not_struct_enum before
1988 msg_typedef s; LP.add_typedef_root
s;
1989 TypedefIdent
(s, i1
)
1991 (* x* ( *y )(params), function pointer 2 *)
1992 | (TIdent
(s, i1
)::TMul _
::TOPar _
::TMul _
::TIdent _
::TCPar _
::TOPar _
::_
, _
)
1993 when not_struct_enum before
1996 msg_typedef s; LP.add_typedef_root
s;
1997 TypedefIdent
(s, i1
)
2000 (*-------------------------------------------------------------*)
2002 (*-------------------------------------------------------------*)
2003 | ((TIfdef
(_
,ii) |TIfdefelse
(_
,ii) |TIfdefelif
(_
,ii) |TEndif
(_
,ii) |
2004 TIfdefBool
(_
,_
,ii)|TIfdefMisc
(_
,_
,ii)|TIfdefVersion
(_
,_
,ii))
2009 if not !Flag_parsing_c.ifdef_to_if
2010 then TCommentCpp (Ast_c.CppDirective, ii)
2013 (* not !LP._lexer_hint.toplevel *)
2014 if !Flag_parsing_c.ifdef_directive_passing
2018 if (LP.current_context
() =*= LP.InInitializer
)
2020 pr2_cpp "In Initializer passing"; (* cheat: dont count in stat *)
2021 incr
Stat.nIfdefInitializer
;
2023 pr2_cpp("IFDEF: or related inside function. I treat it as comment");
2024 incr
Stat.nIfdefPassing
;
2026 TCommentCpp
(Token_c.CppDirective
, ii)
2030 | (TUndef
(id
, ii) as x)::_
, _
2034 pr2_cpp("UNDEF: I treat it as comment");
2035 TCommentCpp
(Token_c.CppDirective
, ii)
2039 | (TCppDirectiveOther
(ii) as x)::_
, _
2043 pr2_cpp ("OTHER directive: I treat it as comment");
2044 TCommentCpp
(Token_c.CppDirective
, ii)
2048 (* If ident contain a for_each, then certainly a macro. But to be
2049 * sure should look if there is a '{' after the ')', but it requires
2050 * to count the '('. Because this can be expensive, we do that only
2051 * when the token contains "for_each".
2053 | (TIdent
(s, i1
)::TOPar _
::rest
, _
)
2054 when not
(LP.current_context
() =*= LP.InTopLevel
)
2055 (* otherwise a function such as static void loopback_enable(int i) {
2056 * will be considered as a loop
2060 if s ==~
regexp_foreach &&
2061 is_really_foreach (Common.take_safe
forLOOKAHEAD rest
)
2065 TMacroIterator
(s, i1
)
2071 (*-------------------------------------------------------------*)
2073 | _
-> raise Impossible
2075 let lookahead ~pass a b
=
2076 Common.profile_code
"C parsing.lookahead" (fun () -> lookahead2 ~pass a b
)