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$" -> true
80 (* note: cant use partial application with let msg_typedef =
81 * because it would compute msg_typedef at compile time when
82 * the flag debug_typedef is always false
85 incr
Stat.nTypedefInfer
;
86 msg_gen (!Flag_parsing_c.debug_typedef
)
89 (*pr2_cpp (Printf.sprintf "TYPEDEF: promoting(%d): %s" n s)*)
90 pr2_cpp (Printf.sprintf
"TYPEDEF: promoting: %s" s
)
94 let msg_maybe_dangereous_typedef s
=
95 if not
(is_known_typdef s
)
98 ("PB MAYBE: dangerous typedef inference, maybe not a typedef: " ^ s
)
102 let msg_declare_macro s
=
103 incr
Stat.nMacroDecl
;
104 msg_gen (!Flag_parsing_c.debug_cpp
)
107 | "DECLARE_MUTEX" | "DECLARE_COMPLETION" | "DECLARE_RWSEM"
108 | "DECLARE_WAITQUEUE" | "DECLARE_WAIT_QUEUE_HEAD"
109 | "DEFINE_SPINLOCK" | "DEFINE_TIMER"
110 | "DEVICE_ATTR" | "CLASS_DEVICE_ATTR" | "DRIVER_ATTR"
111 | "SENSOR_DEVICE_ATTR"
113 | "DECLARE_WORK" | "DECLARE_TASKLET"
114 | "PORT_ATTR_RO" | "PORT_PMA_ATTR"
119 | s when s =~ "^DECLARE_.*" -> true
120 | s when s =~ ".*_ATTR$" -> true
121 | s when s =~ "^DEFINE_.*" -> true
127 (fun s
-> pr2_cpp ("MACRO: found declare-macro: " ^ s
))
132 incr
Stat.nIteratorHeuristic
;
133 pr2_cpp ("MACRO: found foreach: " ^ s
)
137 let msg_debug_macro s =
138 pr2_cpp ("MACRO: found debug-macro: " ^ s)
142 let msg_macro_noptvirg s
=
143 incr
Stat.nMacroStmt
;
144 pr2_cpp ("MACRO: found macro with param noptvirg: " ^ s
)
146 let msg_macro_toplevel_noptvirg s
=
147 incr
Stat.nMacroStmt
;
148 pr2_cpp ("MACRO: found toplevel macro noptvirg: " ^ s
)
150 let msg_macro_noptvirg_single s
=
151 incr
Stat.nMacroStmt
;
152 pr2_cpp ("MACRO: found single-macro noptvirg: " ^ s
)
157 let msg_macro_higher_order s
=
158 incr
Stat.nMacroHigherOrder
;
159 msg_gen (!Flag_parsing_c.debug_cpp
)
169 (fun s
-> pr2_cpp ("MACRO: found higher ordre macro : " ^ s
))
173 let msg_stringification s
=
174 incr
Stat.nMacroString
;
175 msg_gen (!Flag_parsing_c.debug_cpp
)
183 (* s when s =~ ".*STR.*" -> true *)
187 (fun s
-> pr2_cpp ("MACRO: found string-macro " ^ s
))
190 let msg_stringification_params s
=
191 incr
Stat.nMacroString
;
192 pr2_cpp ("MACRO: string-macro with params : " ^ s
)
196 let msg_apply_known_macro s
=
197 incr
Stat.nMacroExpand
;
198 pr2_cpp ("MACRO: found known macro = " ^ s
)
200 let msg_apply_known_macro_hint s
=
201 incr
Stat.nMacroHint
;
202 pr2_cpp ("MACRO: found known macro hint = " ^ s
)
207 let msg_ifdef_bool_passing is_ifdef_positif
=
208 incr
Stat.nIfdefZero
; (* of Version ? *)
210 then pr2_cpp "commenting parts of a #if 1 or #if LINUX_VERSION"
211 else pr2_cpp "commenting a #if 0 or #if LINUX_VERSION or __cplusplus"
214 let msg_ifdef_mid_something () =
215 incr
Stat.nIfdefExprPassing
;
216 pr2_cpp "found ifdef-mid-something"
218 let msg_ifdef_funheaders () =
219 incr
Stat.nIfdefFunheader
;
222 let msg_ifdef_cparen_else () =
223 incr
Stat.nIfdefPassing
;
224 pr2_cpp("found ifdef-cparen-else")
227 let msg_attribute s
=
228 incr
Stat.nMacroAttribute
;
233 (*****************************************************************************)
234 (* The regexp and basic view definitions *)
235 (*****************************************************************************)
237 (* opti: better to built then once and for all, especially regexp_foreach *)
239 let regexp_macro = Str.regexp
243 let regexp_annot = Str.regexp
247 let regexp_declare = Str.regexp
251 let regexp_foreach = Str.regexp_case_fold
252 ".*\\(for_?each\\|for_?all\\|iterate\\|loop\\|walk\\|scan\\|each\\|for\\)"
254 let regexp_typedef = Str.regexp
257 let false_typedef = [
262 let ok_typedef s
= not
(List.mem s
false_typedef)
265 not
(s
==~
regexp_annot)
270 (*****************************************************************************)
272 (*****************************************************************************)
274 (* ------------------------------------------------------------------------- *)
275 (* the pair is the status of '()' and '{}', ex: (-1,0)
276 * if too much ')' and good '{}'
277 * could do for [] too ?
278 * could do for ',' if encounter ',' at "toplevel", not inside () or {}
279 * then if have ifdef, then certainly can lead to a problem.
281 let (count_open_close_stuff_ifdef_clause
: TV.ifdef_grouped list
-> (int * int))=
283 let cnt_paren, cnt_brace
= ref 0, ref 0 in
284 xs
+> TV.iter_token_ifdef
(fun x
->
286 | x
when TH.is_opar x
-> incr
cnt_paren
287 | TOBrace _
-> incr cnt_brace
288 | x
when TH.is_cpar x
-> decr
cnt_paren
289 | TCBrace _
-> decr cnt_brace
293 !cnt_paren, !cnt_brace
296 (* ------------------------------------------------------------------------- *)
297 let forLOOKAHEAD = 30
300 (* look if there is a '{' just after the closing ')', and handling the
301 * possibility to have nested expressions inside nested parenthesis
303 * todo: use indentation instead of premier(statement) ?
305 let rec is_really_foreach xs
=
306 let rec is_foreach_aux = function
308 | TCPar _
::TOBrace _
::xs
-> true, xs
309 (* the following attempts to handle the cases where there is a
310 single statement in the body of the loop. undoubtedly more
312 todo: premier(statement) - suivant(funcall)
314 | TCPar _
::TIdent _
::xs
-> true, xs
315 | TCPar _
::Tif _
::xs
-> true, xs
316 | TCPar _
::Twhile _
::xs
-> true, xs
317 | TCPar _
::Tfor _
::xs
-> true, xs
318 | TCPar _
::Tswitch _
::xs
-> true, xs
319 | TCPar _
::Treturn _
::xs
-> true, xs
322 | TCPar _
::xs
-> false, xs
324 let (_
, xs'
) = is_foreach_aux xs
in
326 | x
::xs
-> is_foreach_aux xs
328 is_foreach_aux xs
+> fst
331 (* ------------------------------------------------------------------------- *)
332 let set_ifdef_token_parenthize_info cnt x
=
335 | TIfdefelse
(tag
, _
)
336 | TIfdefelif
(tag
, _
)
339 | TIfdefBool
(_
, tag
, _
)
340 | TIfdefMisc
(_
, tag
, _
)
341 | TIfdefVersion
(_
, tag
, _
)
345 | _
-> raise Impossible
349 let ifdef_paren_cnt = ref 0
352 let rec set_ifdef_parenthize_info xs
=
353 xs
+> List.iter
(function
354 | NotIfdefLine xs
-> ()
355 | Ifdefbool
(_
, xxs
, info_ifdef
)
356 | Ifdef
(xxs
, info_ifdef
) ->
358 incr
ifdef_paren_cnt;
359 let total_directives = List.length info_ifdef
in
361 info_ifdef
+> List.iter
(fun x
->
362 set_ifdef_token_parenthize_info (!ifdef_paren_cnt, total_directives)
364 xxs
+> List.iter
set_ifdef_parenthize_info
368 (*****************************************************************************)
369 (* The parsing hack for #define *)
370 (*****************************************************************************)
372 (* To parse macro definitions I need to do some tricks
373 * as some information can be get only at the lexing level. For instance
374 * the space after the name of the macro in '#define foo (x)' is meaningful
375 * but the grammar can not get this information. So define_ident below
376 * look at such space and generate a special TOpardefine. In a similar
377 * way macro definitions can contain some antislash and newlines
378 * and the grammar need to know where the macro ends (which is
379 * a line-level and so low token-level information). Hence the
380 * function 'define_line' below and the TDefEol.
382 * update: TDefEol is handled in a special way at different places,
383 * a little bit like EOF, especially for error recovery, so this
384 * is an important token that should not be retagged!
387 * ugly hack, a better solution perhaps would be to erase TDefEOL
388 * from the Ast and list of tokens in parse_c.
390 * note: I do a +1 somewhere, it's for the unparsing to correctly sync.
392 * note: can't replace mark_end_define by simply a fakeInfo(). The reason
393 * is where is the \n TCommentSpace. Normally there is always a last token
394 * to synchronize on, either EOF or the token of the next toplevel.
395 * In the case of the #define we got in list of token
396 * [TCommentSpace "\n"; TDefEOL] but if TDefEOL is a fakeinfo then we will
397 * not synchronize on it and so we will not print the "\n".
398 * A solution would be to put the TDefEOL before the "\n".
399 * (jll: tried to do this, see the comment "Put end of line..." below)
401 * todo?: could put a ExpandedTok for that ?
403 let mark_end_define ii
=
405 { Ast_c.pinfo
= Ast_c.OriginTok
{ (Ast_c.parse_info_of_info
ii) with
407 Common.charpos
= Ast_c.pos_of_info
ii + 1
409 cocci_tag
= ref Ast_c.emptyAnnot
;
410 comments_tag
= ref Ast_c.emptyComments
;
415 (* put the TDefEOL at the good place *)
416 let rec define_line_1 acc xs
=
420 let line = Ast_c.line_of_info
ii in
421 let acc = (TDefine
ii) :: acc in
422 define_line_2
acc line ii xs
424 let line = Ast_c.line_of_info
ii in
425 let acc = (TUndef
ii) :: acc in
426 define_line_2
acc line ii xs
427 | TCppEscapedNewline
ii::xs
->
428 pr2 ("SUSPICIOUS: a \\ character appears outside of a #define at");
429 pr2 (Ast_c.strloc_of_info
ii);
430 let acc = (TCommentSpace
ii) :: acc in
432 | x
::xs
-> define_line_1 (x
::acc) xs
434 and define_line_2
acc line lastinfo xs
=
437 (* should not happened, should meet EOF before *)
439 List.rev
(mark_end_define lastinfo
::acc)
441 let line'
= TH.line_of_tok x
in
442 let info = TH.info_of_tok x
in
446 let acc = (mark_end_define lastinfo
) :: acc in
447 let acc = (EOF
ii) :: acc in
449 | TCppEscapedNewline
ii ->
450 if (line'
<> line) then pr2 "PB: WEIRD: not same line number";
451 let acc = (TCommentSpace
ii) :: acc in
452 define_line_2
acc (line+1) info xs
455 then define_line_2
(x
::acc) line info xs
457 (* Put end of line token before the newline. A newline at least
458 must be there because the line changed and because we saw a
459 #define previously to get to this function at all *)
461 ((List.hd
acc)::(mark_end_define lastinfo
::(List.tl
acc)))
465 let rec define_ident acc xs
=
469 let acc = TUndef
ii :: acc in
471 TCommentSpace i1
::TIdent
(s
,i2
)::xs
->
472 let acc = (TCommentSpace i1
) :: acc in
473 let acc = (TIdentDefine
(s
,i2
)) :: acc in
476 pr2 "WEIRD: weird #define body";
480 let acc = TDefine
ii :: acc in
482 | TCommentSpace i1
::TIdent
(s
,i2
)::TOPar
(i3
)::xs
->
483 (* Change also the kind of TIdent to avoid bad interaction
484 * with other parsing_hack tricks. For instant if keep TIdent then
485 * the stringication algo can believe the TIdent is a string-macro.
486 * So simpler to change the kind of the ident too.
488 (* if TOParDefine sticked to the ident, then
489 * it's a macro-function. Change token to avoid ambiguity
490 * between #define foo(x) and #define foo (x)
492 let acc = (TCommentSpace i1
) :: acc in
493 let acc = (TIdentDefine
(s
,i2
)) :: acc in
494 let acc = (TOParDefine i3
) :: acc in
497 | TCommentSpace i1
::TIdent
(s
,i2
)::xs
->
498 let acc = (TCommentSpace i1
) :: acc in
499 let acc = (TIdentDefine
(s
,i2
)) :: acc in
502 (* bugfix: ident of macro (as well as params, cf below) can be tricky
503 * note, do we need to subst in the body of the define ? no cos
504 * here the issue is the name of the macro, as in #define inline,
505 * so obviously the name of this macro will not be used in its
506 * body (it would be a recursive macro, which is forbidden).
509 | TCommentSpace i1
::t
::xs
->
511 let s = TH.str_of_tok t
in
512 let ii = TH.info_of_tok t
in
513 if s ==~
Common.regexp_alpha
515 pr2 (spf
"remapping: %s to an ident in macro name" s);
516 let acc = (TCommentSpace i1
) :: acc in
517 let acc = (TIdentDefine
(s,ii)) :: acc in
521 pr2 "WEIRD: weird #define body";
526 pr2 "WEIRD: weird #define body";
530 let acc = x
:: acc in
535 let fix_tokens_define2 xs
=
536 define_ident [] (define_line_1 [] xs
)
538 let fix_tokens_define a
=
539 Common.profile_code
"C parsing.fix_define" (fun () -> fix_tokens_define2 a
)
545 (* ------------------------------------------------------------------------- *)
546 (* Other parsing hacks related to cpp, Include/Define hacks *)
547 (* ------------------------------------------------------------------------- *)
549 (* Sometimes I prefer to generate a single token for a list of things in the
550 * lexer so that if I have to passed them, like for passing TInclude then
551 * it's easy. Also if I don't do a single token, then I need to
552 * parse the rest which may not need special stuff, like detecting
553 * end of line which the parser is not really ready for. So for instance
554 * could I parse a #include <a/b/c/xxx.h> as 2 or more tokens ? just
555 * lex #include ? so then need recognize <a/b/c/xxx.h> as one token ?
556 * but this kind of token is valid only after a #include and the
557 * lexing and parsing rules are different for such tokens so not that
558 * easy to parse such things in parser_c.mly. Hence the following hacks.
560 * less?: maybe could get rid of this like I get rid of some of fix_define.
565 (* used to generate new token from existing one *)
566 let new_info posadd str
ii =
568 Ast_c.OriginTok
{ (Ast_c.parse_info_of_info
ii) with
569 charpos
= Ast_c.pos_of_info
ii + posadd
;
571 column
= Ast_c.col_of_info
ii + posadd
;
573 (* must generate a new ref each time, otherwise share *)
574 cocci_tag
= ref Ast_c.emptyAnnot
;
575 comments_tag
= ref Ast_c.emptyComments
;
579 let rec comment_until_defeol xs
=
582 (* job not done in Cpp_token_c.define_parse ? *)
583 failwith
"cant find end of define token TDefEOL"
586 | Parser_c.TDefEOL i
->
587 Parser_c.TCommentCpp
(Token_c.CppDirective
, TH.info_of_tok x
)
591 (* bugfix: otherwise may lose a TComment token *)
592 if TH.is_real_comment
x
594 else Parser_c.TCommentCpp
(Token_c.CppPassingNormal
(*good?*), TH.info_of_tok
x)
596 x'
::comment_until_defeol xs
599 let drop_until_defeol xs
=
601 (Common.drop_until
(function Parser_c.TDefEOL _
-> true | _
-> false) xs
)
605 (* ------------------------------------------------------------------------- *)
606 (* returns a pair (replaced token, list of next tokens) *)
607 (* ------------------------------------------------------------------------- *)
609 let tokens_include (info, includes
, filename
, inifdef
) =
610 Parser_c.TIncludeStart
(Ast_c.rewrap_str includes
info, inifdef
),
611 [Parser_c.TIncludeFilename
612 (filename
, (new_info (String.length includes
) filename
info))
618 (*****************************************************************************)
619 (* CPP handling: macros, ifdefs, macros defs *)
620 (*****************************************************************************)
622 (* ------------------------------------------------------------------------- *)
623 (* special skip_start skip_end handling *)
624 (* ------------------------------------------------------------------------- *)
626 (* note: after this normally the token list should not contain any more the
627 * TCommentSkipTagStart and End tokens.
629 let rec commentize_skip_start_to_end xs
=
634 | {tok
= TCommentSkipTagStart
info} ->
636 let (before
, x2
, after
) =
637 xs
+> Common.split_when
(function
638 | {tok
= TCommentSkipTagEnd _
} -> true
642 let topass = x::x2
::before
in
643 topass +> List.iter
(fun tok
->
644 set_as_comment
Token_c.CppPassingExplicit tok
646 commentize_skip_start_to_end after
648 failwith
"could not find end of skip_start special comment"
650 | {tok
= TCommentSkipTagEnd
info} ->
651 failwith
"found skip_end comment but no skip_start"
653 commentize_skip_start_to_end xs
659 (* ------------------------------------------------------------------------- *)
660 (* ifdef keeping/passing *)
661 (* ------------------------------------------------------------------------- *)
663 (* #if 0, #if 1, #if LINUX_VERSION handling *)
664 let rec find_ifdef_bool xs
=
665 xs
+> List.iter
(function
666 | NotIfdefLine _
-> ()
667 | Ifdefbool
(is_ifdef_positif
, xxs
, info_ifdef_stmt
) ->
669 msg_ifdef_bool_passing is_ifdef_positif
;
672 | [] -> raise Impossible
673 | firstclause
::xxs
->
674 info_ifdef_stmt
+> List.iter
(set_as_comment
Token_c.CppDirective
);
677 then xxs
+> List.iter
678 (iter_token_ifdef
(set_as_comment
Token_c.CppPassingNormal
))
680 firstclause
+> iter_token_ifdef
(set_as_comment
Token_c.CppPassingNormal
);
681 (match List.rev xxs
with
685 (iter_token_ifdef
(set_as_comment
Token_c.CppPassingNormal
))
686 | [] -> (* not #else *) ()
691 | Ifdef
(xxs
, info_ifdef_stmt
) -> xxs
+> List.iter
find_ifdef_bool
696 let thresholdIfdefSizeMid = 6
698 (* infer ifdef involving not-closed expressions/statements *)
699 let rec find_ifdef_mid xs
=
700 xs
+> List.iter
(function
701 | NotIfdefLine _
-> ()
702 | Ifdef
(xxs
, info_ifdef_stmt
) ->
704 | [] -> raise Impossible
706 | first
::second
::rest
->
707 (* don't analyse big ifdef *)
708 if xxs
+> List.for_all
709 (fun xs
-> List.length xs
<= thresholdIfdefSizeMid) &&
710 (* don't want nested ifdef *)
711 xxs
+> List.for_all
(fun xs
->
713 (function NotIfdefLine _
-> true | _
-> false)
717 let counts = xxs
+> List.map count_open_close_stuff_ifdef_clause
in
718 let cnt1, cnt2
= List.hd
counts in
719 if cnt1 <> 0 || cnt2
<> 0 &&
720 counts +> List.for_all
(fun x -> x =*= (cnt1, cnt2
))
722 if counts +> List.exists (fun (cnt1, cnt2) ->
723 cnt1 <> 0 || cnt2 <> 0
727 msg_ifdef_mid_something();
729 (* keep only first, treat the rest as comment *)
730 info_ifdef_stmt
+> List.iter
(set_as_comment
Token_c.CppDirective
);
731 (second
::rest
) +> List.iter
732 (iter_token_ifdef
(set_as_comment
Token_c.CppPassingCosWouldGetError
));
736 List.iter
find_ifdef_mid xxs
738 (* no need complex analysis for ifdefbool *)
739 | Ifdefbool
(_
, xxs
, info_ifdef_stmt
) ->
740 List.iter
find_ifdef_mid xxs
746 let thresholdFunheaderLimit = 4
748 (* ifdef defining alternate function header, type *)
749 let rec find_ifdef_funheaders = function
751 | NotIfdefLine _
::xs
-> find_ifdef_funheaders xs
753 (* ifdef-funheader if ifdef with 2 lines and a '{' in next line *)
755 ([(NotIfdefLine
(({col
= 0} as _xline1
)::line1
))::ifdefblock1
;
756 (NotIfdefLine
(({col
= 0} as xline2
)::line2
))::ifdefblock2
759 ::NotIfdefLine
(({tok
= TOBrace i
; col
= 0})::line3
)
761 when List.length ifdefblock1
<= thresholdFunheaderLimit &&
762 List.length ifdefblock2
<= thresholdFunheaderLimit
764 find_ifdef_funheaders xs
;
766 msg_ifdef_funheaders ();
767 info_ifdef_stmt
+> List.iter
(set_as_comment
Token_c.CppDirective
);
768 let all_toks = [xline2
] @ line2
in
769 all_toks +> List.iter
(set_as_comment
Token_c.CppPassingCosWouldGetError
) ;
770 ifdefblock2
+> iter_token_ifdef
(set_as_comment
Token_c.CppPassingCosWouldGetError
);
772 (* ifdef with nested ifdef *)
774 ([[NotIfdefLine
(({col
= 0} as _xline1
)::line1
)];
776 ([[NotIfdefLine
(({col
= 0} as xline2
)::line2
)];
777 [NotIfdefLine
(({col
= 0} as xline3
)::line3
)];
783 ::NotIfdefLine
(({tok
= TOBrace i
; col
= 0})::line4
)
786 find_ifdef_funheaders xs
;
788 msg_ifdef_funheaders ();
789 info_ifdef_stmt
+> List.iter
(set_as_comment
Token_c.CppDirective
);
790 info_ifdef_stmt2
+> List.iter
(set_as_comment
Token_c.CppDirective
);
791 let all_toks = [xline2
;xline3
] @ line2
@ line3
in
792 all_toks +> List.iter
(set_as_comment
Token_c.CppPassingCosWouldGetError
);
794 (* ifdef with elseif *)
796 ([[NotIfdefLine
(({col
= 0} as _xline1
)::line1
)];
797 [NotIfdefLine
(({col
= 0} as xline2
)::line2
)];
798 [NotIfdefLine
(({col
= 0} as xline3
)::line3
)];
801 ::NotIfdefLine
(({tok
= TOBrace i
; col
= 0})::line4
)
804 find_ifdef_funheaders xs
;
806 msg_ifdef_funheaders ();
807 info_ifdef_stmt
+> List.iter
(set_as_comment
Token_c.CppDirective
);
808 let all_toks = [xline2
;xline3
] @ line2
@ line3
in
809 all_toks +> List.iter
(set_as_comment
Token_c.CppPassingCosWouldGetError
)
812 | Ifdef
(xxs
,info_ifdef_stmt
)::xs
813 | Ifdefbool
(_
, xxs
,info_ifdef_stmt
)::xs
->
814 List.iter
find_ifdef_funheaders xxs
;
815 find_ifdef_funheaders xs
820 let rec adjust_inifdef_include xs
=
821 xs
+> List.iter
(function
822 | NotIfdefLine _
-> ()
823 | Ifdef
(xxs
, info_ifdef_stmt
) | Ifdefbool
(_
, xxs
, info_ifdef_stmt
) ->
824 xxs
+> List.iter
(iter_token_ifdef
(fun tokext
->
825 match tokext
.tok
with
826 | Parser_c.TInclude
(s1
, s2
, inifdef_ref
, ii) ->
838 let rec find_ifdef_cparen_else xs
=
840 xs
+> List.iter
(function
841 | NotIfdefLine _
-> ()
842 | Ifdef
(xxs
, info_ifdef_stmt
) ->
844 | [] -> raise Impossible
846 | first
::second
::rest
->
848 (* found a closing ')' just after the #else *)
850 (* Too bad ocaml does not support better list pattern matching
851 * a la Prolog-III where can match the end of lists.
854 if List.length first
= 0 then false
856 let last_line = Common.last first
in
859 if List.length xs
= 0 then false
861 let last_tok = Common.last xs
in
862 TH.is_cpar
last_tok.tok
863 | Ifdef _
| Ifdefbool _
-> false
865 if condition then begin
866 msg_ifdef_cparen_else();
868 (* keep only first, treat the rest as comment *)
869 info_ifdef_stmt
+> List.iter
(set_as_comment
Token_c.CppDirective
);
870 (second
::rest
) +> List.iter
871 (iter_token_ifdef
(set_as_comment
Token_c.CppPassingCosWouldGetError
));
877 (* no need complex analysis for ifdefbool *)
878 | Ifdefbool
(_
, xxs
, info_ifdef_stmt
) ->
884 (* ------------------------------------------------------------------------- *)
885 (* cpp-builtin part2, macro, using standard.h or other defs *)
886 (* ------------------------------------------------------------------------- *)
888 (* now in cpp_token_c.ml *)
890 (* ------------------------------------------------------------------------- *)
891 (* stringification *)
892 (* ------------------------------------------------------------------------- *)
894 let rec find_string_macro_paren xs
=
897 | Parenthised
(xxs
, info_parens
)::xs
->
898 xxs
+> List.iter
(fun xs
->
900 (function PToken
({tok
= (TString _
| TMacroString _
)}) -> true | _
-> false) &&
902 (function PToken
({tok
= (TString _
| TMacroString _
)}) | PToken
({tok
= TIdent _
}) ->
905 xs
+> List.iter
(fun tok
->
907 | PToken
({tok
= TIdent
(s,_
)} as id
) ->
908 msg_stringification s;
909 id
.tok
<- TMacroString
(s, TH.info_of_tok id
.tok
);
913 find_string_macro_paren xs
915 find_string_macro_paren xs
917 find_string_macro_paren xs
920 (* ------------------------------------------------------------------------- *)
922 (* ------------------------------------------------------------------------- *)
924 (* don't forget to recurse in each case *)
925 let rec find_macro_paren xs
=
930 | PToken
({tok
= Tattribute _
} as id
)
931 ::Parenthised
(xxs
,info_parens
)
934 pr2_cpp ("MACRO: __attribute detected ");
935 [Parenthised
(xxs
, info_parens
)] +>
936 iter_token_paren
(set_as_comment
Token_c.CppAttr
);
937 set_as_comment
Token_c.CppAttr id
;
940 | PToken
({tok
= TattributeNoarg _
} as id
)
943 pr2_cpp ("MACRO: __attributenoarg detected ");
944 set_as_comment
Token_c.CppAttr id
;
948 (* attribute cpp, __xxx id *)
949 | PToken
({tok
= TIdent
(s,i1
)} as id
)
950 ::PToken
({tok
= TIdent
(s2
, i2
)} as id2
)
951 ::xs
when s ==~
regexp_annot
954 id
.tok
<- TMacroAttr
(s, i1
);
955 find_macro_paren ((PToken id2
)::xs
); (* recurse also on id2 ? *)
957 (* attribute cpp, id __xxx *)
958 | PToken
({tok
= TIdent
(s,i1
)} as _id
)
959 ::PToken
({tok
= TIdent
(s2
, i2
)} as id2
)
960 ::xs
when s2
==~
regexp_annot && (not
(s ==~
regexp_typedef))
963 id2
.tok
<- TMacroAttr
(s2
, i2
);
966 | PToken
({tok
= (Tstatic _
| Textern _
)} as tok1
)
967 ::PToken
({tok
= TIdent
(s,i1
)} as attr
)
968 ::xs
when s ==~
regexp_annot
970 pr2_cpp ("storage attribute: " ^
s);
971 attr
.tok
<- TMacroAttrStorage
(s,i1
);
972 (* recurse, may have other storage attributes *)
973 find_macro_paren (PToken
(tok1
)::xs
)
978 (* storage attribute *)
979 | PToken
({tok
= (Tstatic _
| Textern _
)} as tok1
)
980 ::PToken
({tok
= TMacroAttr
(s,i1
)} as attr
)::xs
982 pr2_cpp ("storage attribute: " ^
s);
983 attr
.tok
<- TMacroAttrStorage
(s,i1
);
984 (* recurse, may have other storage attributes *)
985 find_macro_paren (PToken
(tok1
)::xs
)
990 * the order of the matching clause is important
994 (* string macro with params, before case *)
995 | PToken
({tok
= (TString _
| TMacroString _
)})::PToken
({tok
= TIdent
(s,_
)} as id
)
996 ::Parenthised
(xxs
, info_parens
)
999 msg_stringification_params s;
1000 id
.tok
<- TMacroString
(s, TH.info_of_tok id
.tok
);
1001 [Parenthised
(xxs
, info_parens
)] +>
1002 iter_token_paren
(set_as_comment
Token_c.CppMacro
);
1006 | PToken
({tok
= TIdent
(s,_
)} as id
)
1007 ::Parenthised
(xxs
, info_parens
)
1008 ::PToken
({tok
= (TString _
| TMacroString _
)})
1011 msg_stringification_params s;
1012 id
.tok
<- TMacroString
(s, TH.info_of_tok id
.tok
);
1013 [Parenthised
(xxs
, info_parens
)] +>
1014 iter_token_paren
(set_as_comment
Token_c.CppMacro
);
1018 (* for the case where the string is not inside a funcall, but
1019 * for instance in an initializer.
1022 (* string macro variable, before case *)
1023 | PToken
({tok
= (TString _
| TMacroString _
)})::PToken
({tok
= TIdent
(s,_
)} as id
)
1026 msg_stringification s;
1027 id
.tok
<- TMacroString
(s, TH.info_of_tok id
.tok
);
1031 | PToken
({tok
= TIdent
(s,_
)} as id
)
1032 ::PToken
({tok
= (TString _
| TMacroString _
)})
1035 msg_stringification s;
1036 id
.tok
<- TMacroString
(s, TH.info_of_tok id
.tok
);
1044 | (PToken
x)::xs
-> find_macro_paren xs
1045 | (Parenthised
(xxs
, info_parens
))::xs
->
1046 xxs
+> List.iter
find_macro_paren;
1053 (* don't forget to recurse in each case *)
1054 let rec find_macro_lineparen xs
=
1058 (* linuxext: ex: static [const] DEVICE_ATTR(); *)
1061 [PToken
({tok
= Tstatic _
});
1062 PToken
({tok
= TIdent
(s,_
)} as macro
);
1063 Parenthised
(xxs
,info_parens
);
1064 PToken
({tok
= TPtVirg _
});
1068 when (s ==~
regexp_macro) ->
1070 msg_declare_macro s;
1071 let info = TH.info_of_tok macro
.tok
in
1072 macro
.tok
<- TMacroDecl
(Ast_c.str_of_info
info, info);
1074 find_macro_lineparen (xs
)
1076 (* the static const case *)
1079 [PToken
({tok
= Tstatic _
});
1080 PToken
({tok
= Tconst _
} as const
);
1081 PToken
({tok
= TIdent
(s,_
)} as macro
);
1082 Parenthised
(xxs
,info_parens
);
1083 PToken
({tok
= TPtVirg _
});
1089 when (s ==~
regexp_macro) ->
1091 msg_declare_macro s;
1092 let info = TH.info_of_tok macro
.tok
in
1093 macro
.tok
<- TMacroDecl
(Ast_c.str_of_info
info, info);
1095 (* need retag this const, otherwise ambiguity in grammar
1096 21: shift/reduce conflict (shift 121, reduce 137) on Tconst
1097 decl2 : Tstatic . TMacroDecl TOPar argument_list TCPar ...
1098 decl2 : Tstatic . Tconst TMacroDecl TOPar argument_list TCPar ...
1099 storage_class_spec : Tstatic . (137)
1101 const
.tok
<- TMacroDeclConst
(TH.info_of_tok const
.tok
);
1103 find_macro_lineparen (xs
)
1106 (* same but without trailing ';'
1108 * I do not put the final ';' because it can be on a multiline and
1109 * because of the way mk_line is coded, we will not have access to
1110 * this ';' on the next line, even if next to the ')' *)
1112 ([PToken
({tok
= Tstatic _
});
1113 PToken
({tok
= TIdent
(s,_
)} as macro
);
1114 Parenthised
(xxs
,info_parens
);
1118 when s ==~
regexp_macro ->
1120 msg_declare_macro s;
1121 let info = TH.info_of_tok macro
.tok
in
1122 macro
.tok
<- TMacroDecl
(Ast_c.str_of_info
info, info);
1124 find_macro_lineparen (xs
)
1129 (* on multiple lines *)
1132 (PToken
({tok
= Tstatic _
})::[]
1136 [PToken
({tok
= TIdent
(s,_
)} as macro
);
1137 Parenthised
(xxs
,info_parens
);
1138 PToken
({tok
= TPtVirg _
});
1143 when (s ==~
regexp_macro) ->
1145 msg_declare_macro s;
1146 let info = TH.info_of_tok macro
.tok
in
1147 macro
.tok
<- TMacroDecl
(Ast_c.str_of_info
info, info);
1149 find_macro_lineparen (xs
)
1152 (* linuxext: ex: DECLARE_BITMAP();
1154 * Here I use regexp_declare and not regexp_macro because
1155 * Sometimes it can be a FunCallMacro such as DEBUG(foo());
1156 * Here we don't have the preceding 'static' so only way to
1157 * not have positive is to restrict to .*DECLARE.* macros.
1159 * but there is a grammar rule for that, so don't need this case anymore
1160 * unless the parameter of the DECLARE_xxx are weird and can not be mapped
1161 * on a argument_list
1165 ([PToken
({tok
= TIdent
(s,_
)} as macro
);
1166 Parenthised
(xxs
,info_parens
);
1167 PToken
({tok
= TPtVirg _
});
1171 when (s ==~
regexp_declare) ->
1173 msg_declare_macro s;
1174 let info = TH.info_of_tok macro
.tok
in
1175 macro
.tok
<- TMacroDecl
(Ast_c.str_of_info
info, info);
1177 find_macro_lineparen (xs
)
1183 * Could also transform the TIdent in a TMacroTop but can have false
1184 * positive, so easier to just change the TCPar and so just solve
1185 * the end-of-stream pb of ocamlyacc
1188 ([PToken
({tok
= TIdent
(s,ii); col
= col1
; where
= ctx
} as _macro
);
1189 Parenthised
(xxs
,info_parens
);
1192 ::xs
when col1
=|= 0
1195 (* to reduce number of false positive *)
1197 | (Line
(PToken
({col
= col2
} as other
)::restline2
))::_
->
1198 TH.is_eof other
.tok
|| (col2
=|= 0 &&
1199 (match other
.tok
with
1200 | TOBrace _
-> false (* otherwise would match funcdecl *)
1201 | TCBrace _
when ctx
<> InFunction
-> false
1205 | tok
when TH.is_binary_operator tok
-> false
1216 msg_macro_toplevel_noptvirg s;
1217 (* just to avoid the end-of-stream pb of ocamlyacc *)
1218 let tcpar = Common.last info_parens
in
1219 tcpar.tok
<- TCParEOL
(TH.info_of_tok
tcpar.tok
);
1221 (*macro.tok <- TMacroTop (s, TH.info_of_tok macro.tok);*)
1225 find_macro_lineparen (xs
)
1229 (* macro with parameters
1234 ([PToken
({tok
= TIdent
(s,ii); col
= col1
; where
= ctx
} as macro
);
1235 Parenthised
(xxs
,info_parens
);
1239 (PToken
({col
= col2
} as other
)::restline2
1242 (* when s ==~ regexp_macro *)
1246 (match other
.tok
with
1247 | TOBrace _
-> false (* otherwise would match funcdecl *)
1248 | TCBrace _
when ctx
<> InFunction
-> false
1252 | tok
when TH.is_binary_operator tok
-> false
1259 (match other
.tok
, restline2
with
1260 | TCBrace _
, _
when ctx
=*= InFunction
-> true
1261 | Treturn _
, _
-> true
1263 | Telse _
, _
-> true
1265 (* case of label, usually put in first line *)
1266 | TIdent _
, (PToken
({tok
= TDotDot _
}))::_
->
1278 if col1
=|= 0 then ()
1280 msg_macro_noptvirg s;
1281 macro
.tok
<- TMacroStmt
(s, TH.info_of_tok macro
.tok
);
1282 [Parenthised
(xxs
, info_parens
)] +>
1283 iter_token_paren
(set_as_comment
Token_c.CppMacro
);
1286 find_macro_lineparen (line2
::xs
)
1288 (* linuxext:? single macro
1293 * todo: factorize code with previous rule ?
1296 ([PToken
({tok
= TIdent
(s,ii); col
= col1
; where
= ctx
} as macro
);
1300 (PToken
({col
= col2
} as other
)::restline2
1303 (* when s ==~ regexp_macro *)
1307 col1
<> 0 && (* otherwise can match typedef of fundecl*)
1308 (match other
.tok
with
1309 | TPtVirg _
-> false
1311 | TCBrace _
when ctx
<> InFunction
-> false
1312 | tok
when TH.is_binary_operator tok
-> false
1317 (match other
.tok
with
1318 | TCBrace _
when ctx
=*= InFunction
-> true
1328 msg_macro_noptvirg_single s;
1329 macro
.tok
<- TMacroStmt
(s, TH.info_of_tok macro
.tok
);
1331 find_macro_lineparen (line2
::xs
)
1334 find_macro_lineparen xs
1338 (* ------------------------------------------------------------------------- *)
1339 (* define tobrace init *)
1340 (* ------------------------------------------------------------------------- *)
1342 let rec find_define_init_brace_paren xs
=
1347 (* mainly for firefox *)
1348 | (PToken
{tok
= TDefine _
})
1349 ::(PToken
{tok
= TIdentDefine
(s,_
)})
1350 ::(PToken
({tok
= TOBrace i1
} as tokbrace
))
1355 match tok2
.tok
, tok3
.tok
with
1356 | TInt _
, TComma _
-> true
1357 | TString _
, TComma _
-> true
1358 | TIdent _
, TComma _
-> true
1364 pr2_cpp("found define initializer: " ^
s);
1365 tokbrace
.tok
<- TOBraceDefineInit i1
;
1370 (* mainly for linux, especially in sound/ *)
1371 | (PToken
{tok
= TDefine _
})
1372 ::(PToken
{tok
= TIdentDefine
(s,_
)})
1373 ::(Parenthised
(xxx
, info_parens
))
1374 ::(PToken
({tok
= TOBrace i1
} as tokbrace
))
1379 match tok2
.tok
, tok3
.tok
with
1380 | TInt _
, TComma _
-> true
1381 | TDot _
, TIdent _
-> true
1382 | TIdent _
, TComma _
-> true
1388 pr2_cpp("found define initializer with param: " ^
s);
1389 tokbrace
.tok
<- TOBraceDefineInit i1
;
1397 | (PToken
x)::xs
-> aux xs
1398 | (Parenthised
(xxs
, info_parens
))::xs
->
1399 (* not need for tobrace init:
1400 * xxs +> List.iter aux;
1407 (* ------------------------------------------------------------------------- *)
1409 (* ------------------------------------------------------------------------- *)
1411 (* obsolete now with macro expansion ? get some regression if comment.
1412 * todo: if do bad decision here, then it can influence other phases
1413 * and make it hard to parse. So maybe when have a parse error, should
1414 * undo some of the guess those heuristics have done, and restore
1415 * the original token value.
1418 let rec find_actions = function
1421 | PToken
({tok
= TIdent
(s,ii)})
1422 ::Parenthised
(xxs
,info_parens
)
1425 xxs
+> List.iter
find_actions;
1426 let modified = find_actions_params xxs
in
1428 then msg_macro_higher_order s
1433 and find_actions_params xxs
=
1434 xxs
+> List.fold_left
(fun acc xs
->
1435 let toks = tokens_of_paren xs
in
1436 if toks +> List.exists
(fun x -> TH.is_statement
x.tok
)
1437 (* undo: && List.length toks > 1
1438 * good for sparse, not good for linux
1441 xs
+> iter_token_paren
(fun x ->
1444 (* certainly because paren detection had a pb because of
1445 * some ifdef-exp. Do similar additional checking than
1446 * what is done in set_as_comment.
1448 pr2 "PB: weird, I try to tag an EOF token as an action"
1450 (* cf tests-bis/no_cpar_macro.c *)
1453 pr2 "PB: weird, I try to tag an EOM token as an action"
1455 x.tok
<- TAction
(TH.info_of_tok
x.tok
);
1464 (* ------------------------------------------------------------------------- *)
1465 (* main fix cpp function *)
1466 (* ------------------------------------------------------------------------- *)
1468 let filter_cpp_stuff xs
=
1472 | tok
when TH.is_comment tok
-> false
1473 (* don't want drop the define, or if drop, have to drop
1474 * also its body otherwise the line heuristics may be lost
1475 * by not finding the TDefine in column 0 but by finding
1476 * a TDefineIdent in a column > 0
1478 | Parser_c.TDefine _
-> true
1479 | tok
when TH.is_cpp_instruction tok
-> false
1484 let insert_virtual_positions l
=
1485 let strlen x = String.length
(Ast_c.str_of_info
x) in
1486 let rec loop prev offset
acc = function
1489 let ii = TH.info_of_tok
x in
1491 TH.visitor_info_of_tok
(function ii -> Ast_c.rewrap_pinfo pi
ii) x in
1492 match Ast_c.pinfo_of_info
ii with
1493 Ast_c.OriginTok pi
->
1494 let prev = Ast_c.parse_info_of_info
ii in
1495 loop prev (strlen ii) (x::acc) xs
1496 | Ast_c.ExpandedTok
(pi
,_
) ->
1497 let x'
= inject (Ast_c.ExpandedTok
(pi
,(prev,offset
))) in
1498 loop prev (offset
+ (strlen ii)) (x'
::acc) xs
1499 | Ast_c.FakeTok
(s,_
) ->
1500 let x'
= inject (Ast_c.FakeTok
(s,(prev,offset
))) in
1501 loop prev (offset
+ (strlen ii)) (x'
::acc) xs
1502 | Ast_c.AbstractLineTok _
-> failwith
"abstract not expected" in
1503 let rec skip_fake = function
1506 let ii = TH.info_of_tok
x in
1507 match Ast_c.pinfo_of_info
ii with
1508 | Ast_c.OriginTok pi
->
1509 let prev = Ast_c.parse_info_of_info
ii in
1510 let res = loop prev (strlen ii) [] xs
in
1512 | _
-> x::skip_fake xs
in
1515 (* ------------------------------------------------------------------------- *)
1517 let fix_tokens_cpp2 ~macro_defs tokens
=
1518 let tokens2 = ref (tokens
+> Common.acc_map
TV.mk_token_extended
) in
1521 (* the order is important, if you put the action heuristic first,
1522 * then because of ifdef, can have not closed paren
1523 * and so may believe that higher order macro
1524 * and it will eat too much tokens. So important to do
1527 * I recompute multiple times cleaner cos the mutable
1528 * can have be changed and so may have more comments
1529 * in the token original list.
1533 commentize_skip_start_to_end !tokens2;
1536 let cleaner = !tokens2 +> List.filter
(fun x ->
1537 (* is_comment will also filter the TCommentCpp created in
1538 * commentize_skip_start_to_end *)
1539 not
(TH.is_comment
x.tok
) (* could filter also #define/#include *)
1541 let ifdef_grouped = TV.mk_ifdef
cleaner in
1542 set_ifdef_parenthize_info ifdef_grouped;
1544 find_ifdef_funheaders ifdef_grouped;
1545 find_ifdef_bool ifdef_grouped;
1546 find_ifdef_mid ifdef_grouped;
1547 (* change order ? maybe cparen_else heuristic make some of the funheaders
1548 * heuristics irrelevant ?
1550 find_ifdef_cparen_else ifdef_grouped;
1551 adjust_inifdef_include ifdef_grouped;
1555 let cleaner = !tokens2 +> filter_cpp_stuff in
1557 let paren_grouped = TV.mk_parenthised
cleaner in
1558 Cpp_token_c.apply_macro_defs
1559 ~
msg_apply_known_macro
1560 ~
msg_apply_known_macro_hint
1561 macro_defs
paren_grouped;
1562 (* because the before field is used by apply_macro_defs *)
1563 tokens2 := TV.rebuild_tokens_extented
!tokens2;
1565 (* tagging contextual info (InFunc, InStruct, etc). Better to do
1566 * that after the "ifdef-simplification" phase.
1568 let cleaner = !tokens2 +> List.filter
(fun x ->
1569 not
(TH.is_comment
x.tok
) (* could filter also #define/#include *)
1572 let brace_grouped = TV.mk_braceised
cleaner in
1573 set_context_tag
brace_grouped;
1576 let cleaner = !tokens2 +> filter_cpp_stuff in
1578 let paren_grouped = TV.mk_parenthised
cleaner in
1579 let line_paren_grouped = TV.mk_line_parenthised
paren_grouped in
1580 find_define_init_brace_paren paren_grouped;
1581 find_string_macro_paren paren_grouped;
1582 find_macro_lineparen line_paren_grouped;
1583 find_macro_paren paren_grouped;
1586 (* obsolete: actions ? not yet *)
1587 let cleaner = !tokens2 +> filter_cpp_stuff in
1588 let paren_grouped = TV.mk_parenthised
cleaner in
1589 find_actions paren_grouped;
1593 insert_virtual_positions (!tokens2 +> Common.acc_map
(fun x -> x.tok
))
1596 let time_hack1 ~macro_defs a
=
1597 Common.profile_code_exclusif
"HACK" (fun () -> fix_tokens_cpp2 ~macro_defs a
)
1599 let fix_tokens_cpp ~macro_defs a
=
1600 Common.profile_code
"C parsing.fix_cpp" (fun () -> time_hack1 ~macro_defs a
)
1605 (*****************************************************************************)
1606 (* Lexing with lookahead *)
1607 (*****************************************************************************)
1609 (* Why using yet another parsing_hack technique ? The fix_xxx where do
1610 * some pre-processing on the full list of tokens is not enough ?
1611 * No cos sometimes we need more contextual info, and even if
1612 * set_context() tries to give some contextual info, it's not completely
1613 * accurate so the following code give yet another alternative, yet another
1614 * chance to transform some tokens.
1616 * todo?: maybe could try to get rid of this technique. Maybe a better
1617 * set_context() would make possible to move this code using a fix_xx
1620 * LALR(k) trick. We can do stuff by adding cases in lexer_c.mll, but
1621 * it is more general to do it via my LALR(k) tech. Because here we can
1622 * transform some token give some context information. So sometimes it
1623 * makes sense to transform a token in one context, sometimes not, and
1624 * lex can not provide us this context information. Note that the order
1625 * in the pattern matching in lookahead is important. Do not cut/paste.
1627 * Note that in next there is only "clean" tokens, there is no comment
1628 * or space tokens. This is done by the caller.
1632 open Lexer_parser
(* for the fields of lexer_hint type *)
1634 let not_struct_enum = function
1635 | (Parser_c.Tstruct _
| Parser_c.Tunion _
| Parser_c.Tenum _
)::_
-> false
1638 let pointer = function
1640 | TAnd _
when !Flag.c_plus_plus
-> true
1643 let lookahead2 ~pass next before
=
1645 match (next
, before
) with
1648 (* yy xx( and in function *)
1649 | TOPar i1
::_
, TIdent
(s,i2
)::TypedefIdent _
::_
1650 when !Flag.c_plus_plus
&& (LP.current_context
() = (LP.InFunction
)) ->
1651 pr2_cpp("constructed_object: " ^
s);
1652 TOParCplusplusInit i1
1653 | TypedefIdent
(s,i
)::TOPar i1
::_
,_
1654 when !Flag.c_plus_plus
&& (LP.current_context
() = (LP.InFunction
)) ->
1657 (*-------------------------------------------------------------*)
1658 (* typedef inference, parse_typedef_fix3 *)
1659 (*-------------------------------------------------------------*)
1661 | (TIdent
(s,i1
)::TIdent
(s2
,i2
)::_
, _
) when not_struct_enum before
&& s =$
= s2
1663 (* (take_safe 1 !passed_tok <> [TOPar]) -> *)
1665 (* parse_typedef_fix3:
1666 * acpi_object acpi_object;
1667 * etait mal parsé, car pas le temps d'appeler dt() dans le type_spec.
1668 * Le parser en interne a deja appelé le prochain token pour pouvoir
1669 * decider des choses.
1670 * => special case in lexer_heuristic, again
1672 if !Flag_parsing_c.debug_typedef
1673 then pr2 ("TYPEDEF: disable typedef cos special case: " ^
s);
1675 LP.disable_typedef
();
1677 msg_typedef s 1; LP.add_typedef_root
s;
1678 TypedefIdent
(s, i1
)
1681 | (TIdent
(s, i1
)::TIdent
(s2
, i2
)::_
, _
) when not_struct_enum before
1684 (* && not_annot s2 BUT lead to false positive*)
1686 msg_typedef s 2; LP.add_typedef_root
s;
1687 TypedefIdent
(s, i1
)
1691 | (TIdent
(s, i1
)::Tinline i2
::_
, _
) when not_struct_enum before
1694 msg_typedef s 3; LP.add_typedef_root
s;
1695 TypedefIdent
(s, i1
)
1698 (* [,(] xx [,)] AND param decl *)
1699 | (TIdent
(s, i1
)::(TComma _
|TCPar _
)::_
, (TComma _
|TOPar _
)::_
)
1700 when not_struct_enum before
&& (LP.current_context
() =*= LP.InParameter
)
1703 msg_typedef s 4; LP.add_typedef_root
s;
1704 TypedefIdent
(s, i1
)
1707 (* specialcase: [,(] xx* [,)] *)
1708 | (TIdent
(s, i1
)::ptr
::(TComma _
|TCPar _
)::_
, (*(TComma _|TOPar _)::*)_
)
1709 when pointer ptr
&& not_struct_enum before
1710 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1713 msg_typedef s 5; LP.add_typedef_root
s;
1714 TypedefIdent
(s, i1
)
1718 (* specialcase: [,(] xx** [,)] *)
1719 | (TIdent
(s, i1
)::TMul _
::TMul _
::(TComma _
|TCPar _
)::_
, (*(TComma _|TOPar _)::*)_
)
1720 when not_struct_enum before
1721 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1724 msg_typedef s 6; LP.add_typedef_root
s;
1725 TypedefIdent
(s, i1
)
1729 (* xx const * USELESS because of next rule ? *)
1730 | (TIdent
(s, i1
)::(Tconst _
|Tvolatile _
|Trestrict _
)::TMul _
::_
, _
)
1731 when not_struct_enum before
1732 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1736 msg_typedef s 7; LP.add_typedef_root
s;
1737 TypedefIdent
(s, i1
)
1740 | (TIdent
(s, i1
)::(Tconst _
|Tvolatile _
|Trestrict _
)::_
, _
)
1741 when not_struct_enum before
1743 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1746 msg_typedef s 8; LP.add_typedef_root
s;
1747 TypedefIdent
(s, i1
)
1751 | (TIdent
(s, i1
)::ptr
::(Tconst _
| Tvolatile _
|Trestrict _
)::_
, _
)
1752 when pointer ptr
&& not_struct_enum before
1755 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1757 msg_typedef s 9; LP.add_typedef_root
s;
1758 TypedefIdent
(s, i1
)
1762 | (TIdent
(s, i1
)::TCPar _
::_
, (Tconst _
| Tvolatile _
|Trestrict _
)::TOPar _
::_
) when
1764 msg_typedef s 10; LP.add_typedef_root
s;
1765 TypedefIdent
(s, i1
)
1769 (* ( xx ) [sizeof, ~] *)
1770 | (TIdent
(s, i1
)::TCPar _
::(Tsizeof _
|TTilde _
)::_
, TOPar _
::_
)
1771 when not_struct_enum before
1774 msg_typedef s 11; LP.add_typedef_root
s;
1775 TypedefIdent
(s, i1
)
1777 (* [(,] xx [ AND parameterdeclaration *)
1778 | (TIdent
(s, i1
)::TOCro _
::_
, (TComma _
|TOPar _
)::_
)
1779 when (LP.current_context
() =*= LP.InParameter
)
1782 msg_typedef s 12; LP.add_typedef_root
s;
1783 TypedefIdent
(s, i1
)
1785 (*------------------------------------------------------------*)
1786 (* if 'x*y' maybe an expr, maybe just a classic multiplication *)
1787 (* but if have a '=', or ',' I think not *)
1788 (*------------------------------------------------------------*)
1790 (* static xx * yy *)
1791 | (TIdent
(s, i1
)::ptr
::TIdent
(s2
, i2
)::_
,
1792 (Tregister _
|Tstatic _
|Tvolatile _
|Tconst _
|Trestrict _
)::_
) when
1793 pointer ptr
&& ok_typedef s
1795 msg_typedef s 13; LP.add_typedef_root
s;
1796 TypedefIdent
(s, i1
)
1798 (* TODO xx * yy ; AND in start of compound element *)
1801 (* xx * yy, AND in paramdecl *)
1802 | (TIdent
(s, i1
)::ptr
::TIdent
(s2
, i2
)::TComma _
::_
, _
)
1803 when not_struct_enum before
&& (LP.current_context
() =*= LP.InParameter
)
1804 && pointer ptr
&& ok_typedef s
1807 msg_typedef s 14; LP.add_typedef_root
s;
1808 TypedefIdent
(s, i1
)
1811 (* xx * yy ; AND in Toplevel, except when have = before *)
1812 | (TIdent
(s, i1
)::TMul _
::TIdent
(s2
, i2
)::TPtVirg _
::_
, TEq _
::_
) ->
1814 | (TIdent
(s, i1
)::ptr
::TIdent
(s2
, i2
)::TPtVirg _
::_
, _
)
1815 when not_struct_enum before
&& pointer ptr
&&
1816 (LP.is_top_or_struct
(LP.current_context
()))
1818 msg_typedef s 15; LP.add_typedef_root
s;
1819 TypedefIdent
(s, i1
)
1821 (* xx * yy , AND in Toplevel *)
1822 | (TIdent
(s, i1
)::ptr
::TIdent
(s2
, i2
)::TComma _
::_
, _
)
1823 when not_struct_enum before
&& (LP.current_context
() =*= LP.InTopLevel
)
1824 && ok_typedef s && pointer ptr
1827 msg_typedef s 16; LP.add_typedef_root
s;
1828 TypedefIdent
(s, i1
)
1830 (* xx * yy ( AND in Toplevel *)
1831 | (TIdent
(s, i1
)::ptr
::TIdent
(s2
, i2
)::TOPar _
::_
, _
)
1832 when not_struct_enum before
1833 && (LP.is_top_or_struct
(LP.current_context
()))
1834 && ok_typedef s && pointer ptr
1836 msg_typedef s 17; LP.add_typedef_root
s;
1837 TypedefIdent
(s, i1
)
1840 (* todo? enough ? cos in struct def we can have some expression ! *)
1841 | (TIdent
(s, i1
)::ptr
::TIdent
(s2
, i2
)::TOCro _
::_
, _
)
1842 when not_struct_enum before
&&
1843 (LP.is_top_or_struct
(LP.current_context
()))
1844 && ok_typedef s && pointer ptr
1846 msg_typedef s 18; LP.add_typedef_root
s;
1847 TypedefIdent
(s, i1
)
1849 (* u16: 10; in struct *)
1850 | (TIdent
(s, i1
)::TDotDot _
::_
, (TOBrace _
| TPtVirg _
)::_
)
1851 when (LP.is_top_or_struct
(LP.current_context
()))
1854 msg_typedef s 19; LP.add_typedef_root
s;
1855 TypedefIdent
(s, i1
)
1858 (* why need TOPar condition as stated in preceding rule ? really needed ? *)
1859 (* YES cos at toplevel can have some expression !! for instance when *)
1860 (* enter in the dimension of an array *)
1862 | (TIdent s::TMul::TIdent s2::_ , _)
1863 when (take_safe 1 !passed_tok <> [Tstruct] &&
1864 (take_safe 1 !passed_tok <> [Tenum]))
1866 !LP._lexer_hint = Some LP.Toplevel ->
1867 msg_typedef s 20; LP.add_typedef_root s;
1872 | (TIdent
(s, i1
)::ptr
::TIdent
(s2
, i2
)::TEq _
::_
, _
)
1873 when not_struct_enum before
1874 && ok_typedef s && pointer ptr
1876 msg_typedef s 21; LP.add_typedef_root
s;
1877 TypedefIdent
(s, i1
)
1880 (* xx * yy) AND in paramdecl *)
1881 | (TIdent
(s, i1
)::ptr
::TIdent
(s2
, i2
)::TCPar _
::_
, _
)
1882 when not_struct_enum before
&& (LP.current_context
() =*= LP.InParameter
)
1883 && ok_typedef s && pointer ptr
1885 msg_typedef s 22; LP.add_typedef_root
s;
1886 TypedefIdent
(s, i1
)
1889 (* xx * yy; *) (* wrong ? *)
1890 | (TIdent
(s, i1
)::ptr
::TIdent
(s2
, i2
)::TPtVirg _
::_
,
1891 (TOBrace _
| TPtVirg _
)::_
) when not_struct_enum before
1892 && ok_typedef s & pointer ptr
1894 msg_typedef s 23; LP.add_typedef_root
s;
1895 msg_maybe_dangereous_typedef s;
1896 TypedefIdent
(s, i1
)
1899 (* xx * yy, and ';' before xx *) (* wrong ? *)
1900 | (TIdent
(s, i1
)::ptr
::TIdent
(s2
, i2
)::TComma _
::_
,
1901 (TOBrace _
| TPtVirg _
)::_
) when
1902 ok_typedef s && pointer ptr
1904 msg_typedef s 24; LP.add_typedef_root
s;
1905 TypedefIdent
(s, i1
)
1909 | (TIdent
(s, i1
)::ptr
::TIdent
(s2
, i2
)::_
, _
)
1910 when s ==~
regexp_typedef && not_struct_enum before
1911 (* struct user_info_t sometimes *)
1912 && ok_typedef s && pointer ptr
1914 msg_typedef s 25; LP.add_typedef_root
s;
1915 TypedefIdent
(s, i1
)
1917 (* xx ** yy *) (* wrong ? *)
1918 | (TIdent
(s, i1
)::TMul _
::TMul _
::TIdent
(s2
, i2
)::_
, _
)
1919 when not_struct_enum before
1920 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1923 msg_typedef s 26; LP.add_typedef_root
s;
1924 TypedefIdent
(s, i1
)
1927 | (TIdent
(s, i1
)::TMul _
::TMul _
::TMul _
::TIdent
(s2
, i2
)::_
, _
)
1928 when not_struct_enum before
1930 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1932 msg_typedef s 27; LP.add_typedef_root
s;
1933 TypedefIdent
(s, i1
)
1936 | (TIdent
(s, i1
)::TMul _
::TMul _
::TCPar _
::_
, _
)
1937 when not_struct_enum before
1938 (* && !LP._lexer_hint = Some LP.ParameterDeclaration *)
1941 msg_typedef s 28; LP.add_typedef_root
s;
1942 TypedefIdent
(s, i1
)
1946 (* ----------------------------------- *)
1947 (* old: why not do like for other rules and start with TIdent ?
1948 * why do TOPar :: TIdent :: ..., _ and not TIdent :: ..., TOPAr::_ ?
1949 * new: prefer now start with TIdent because otherwise the add_typedef_root
1950 * may have no effect if in second pass or if have disable the add_typedef.
1954 | (TIdent
(s, i1
)::TCPar i2
::(TIdent
(_
,i3
)|TInt
(_
,i3
))::_
,
1956 when not
(TH.is_stuff_taking_parenthized
x) &&
1957 Ast_c.line_of_info i2
=|= Ast_c.line_of_info i3
1961 msg_typedef s 29; LP.add_typedef_root
s;
1963 TypedefIdent
(s, i1
)
1967 * but false positif: typedef int (xxx_t)(...), so do specialisation below.
1970 | (TIdent (s, i1)::TCPar _::TOPar _::_ , (TOPar info)::x::_)
1971 when not (TH.is_stuff_taking_parenthized x)
1974 msg_typedef s 30; LP.add_typedef_root s;
1976 TypedefIdent
(s, i1
)
1978 (* special case: = (xx) ( yy) *)
1979 | (TIdent
(s, i1
)::TCPar _
::TOPar _
::_
,
1980 (TOPar
info)::(TEq _
|TEqEq _
)::_
)
1983 msg_typedef s 31; LP.add_typedef_root
s;
1985 TypedefIdent
(s, i1
)
1989 | (TIdent
(s, i1
)::ptr
::TCPar _
::TIdent
(s2
, i2
)::_
, (TOPar
info)::_
)
1990 when ok_typedef s && pointer ptr
1992 msg_typedef s 32; LP.add_typedef_root
s;
1997 (* (xx){ ... } constructor *)
1998 | (TIdent
(s, i1
)::TCPar _
::TOBrace _
::_
, TOPar _
::x::_
)
1999 when (*s ==~ regexp_typedef && *) not
(TH.is_stuff_taking_parenthized
x)
2002 msg_typedef s 33; LP.add_typedef_root
s;
2003 TypedefIdent
(s, i1
)
2006 (* can have sizeof on expression
2007 | (Tsizeof::TOPar::TIdent s::TCPar::_, _) ->
2008 msg_typedef s; LP.add_typedef_root s;
2013 (* ----------------------------------- *)
2014 (* x ( *y )(params), function pointer *)
2015 | (TIdent
(s, i1
)::TOPar _
::TMul _
::TIdent _
::TCPar _
::TOPar _
::_
, _
)
2016 when not_struct_enum before
2019 msg_typedef s 34; LP.add_typedef_root
s;
2020 TypedefIdent
(s, i1
)
2022 (* x* ( *y )(params), function pointer 2 *)
2023 | (TIdent
(s, i1
)::TMul _
::TOPar _
::TMul _
::TIdent _
::TCPar _
::TOPar _
::_
, _
)
2024 when not_struct_enum before
2027 msg_typedef s 35; LP.add_typedef_root
s;
2028 TypedefIdent
(s, i1
)
2031 (*-------------------------------------------------------------*)
2033 (*-------------------------------------------------------------*)
2034 | ((TIfdef
(_
,ii) |TIfdefelse
(_
,ii) |TIfdefelif
(_
,ii) |TEndif
(_
,ii) |
2035 TIfdefBool
(_
,_
,ii)|TIfdefMisc
(_
,_
,ii)|TIfdefVersion
(_
,_
,ii))
2040 if not !Flag_parsing_c.ifdef_to_if
2041 then TCommentCpp (Ast_c.CppDirective, ii)
2044 (* not !LP._lexer_hint.toplevel *)
2045 if !Flag_parsing_c.ifdef_directive_passing
2049 if (LP.current_context
() =*= LP.InInitializer
)
2051 pr2_cpp "In Initializer passing"; (* cheat: dont count in stat *)
2052 incr
Stat.nIfdefInitializer
;
2054 pr2_cpp("IFDEF: or related inside function. I treat it as comment");
2055 incr
Stat.nIfdefPassing
;
2057 TCommentCpp
(Token_c.CppDirective
, ii)
2061 | (TUndef
(ii) as x)::_
, _
2065 pr2_cpp("UNDEF: I treat it as comment");
2066 TCommentCpp
(Token_c.CppDirective
, ii)
2070 | (TCppDirectiveOther
(ii) as x)::_
, _
2074 pr2_cpp ("OTHER directive: I treat it as comment");
2075 TCommentCpp
(Token_c.CppDirective
, ii)
2079 (* If ident contain a for_each, then certainly a macro. But to be
2080 * sure should look if there is a '{' after the ')', but it requires
2081 * to count the '('. Because this can be expensive, we do that only
2082 * when the token contains "for_each".
2084 | (TIdent
(s, i1
)::TOPar _
::rest
, _
)
2085 when not
(LP.current_context
() =*= LP.InTopLevel
)
2086 (* otherwise a function such as static void loopback_enable(int i) {
2087 * will be considered as a loop
2091 if s ==~
regexp_foreach &&
2092 is_really_foreach (Common.take_safe
forLOOKAHEAD rest
)
2096 TMacroIterator
(s, i1
)
2102 (*-------------------------------------------------------------*)
2104 | _
-> raise Impossible
2106 let lookahead ~pass a b
=
2107 Common.profile_code
"C parsing.lookahead" (fun () -> lookahead2 ~pass a b
)