3 * Copyright (C) 2010, University of Copenhagen DIKU and INRIA.
4 * Copyright (C) 2002, 2006, 2007, 2008, 2009 Yoann Padioleau
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.
17 (*****************************************************************************)
18 (* The AST C related types *)
19 (*****************************************************************************)
21 * Some stuff are tagged semantic: which means that they are computed
24 * This means that some elements in this AST are present only if
25 * some annotation/transformation has been done on the original AST returned
26 * by the parser. Cf type_annotater, comment_annotater, cpp_ast_c, etc.
30 (* ------------------------------------------------------------------------- *)
32 (* ------------------------------------------------------------------------- *)
34 (* To allow some transformations over the AST, we keep as much information
35 * as possible in the AST such as the tokens content and their locations.
36 * Those info are called 'info' (how original) and can be tagged.
37 * For instance one tag may say that the unparser should remove this token.
39 * Update: Now I use a ref! in those 'info' so take care.
40 * That means that modifications of the info of tokens can have
41 * an effect on the info stored in the ast (which is sometimes
42 * convenient, cf unparse_c.ml or comment_annotater_c.ml)
44 * convention: I often use 'ii' for the name of a list of info.
46 * Sometimes we want to add someting at the beginning or at the end
47 * of a construct. For 'function' and 'decl' we want to add something
48 * to their left and for 'if' 'while' et 'for' and so on at their right.
49 * We want some kinds of "virtual placeholders" that represent the start or
50 * end of a construct. We use fakeInfo for that purpose.
51 * To identify those cases I have added a fakestart/fakeend comment.
53 * cocci: Each token will be decorated in the future by the mcodekind
54 * of cocci. It is the job of the pretty printer to look at this
55 * information and decide to print or not the token (and also the
56 * pending '+' associated sometimes with the token).
58 * The first time that we parse the original C file, the mcodekind is
59 * empty, or more precisely all is tagged as a CONTEXT with NOTHING
60 * associated. This is what I call a "clean" expr/statement/....
62 * Each token will also be decorated in the future with an environment,
63 * because the pending '+' may contain metavariables that refer to some
70 type posl
= int * int (* line-col, for MetaPosValList, for position variables *)
73 (* the virtual position is set in Parsing_hacks.insert_virtual_positions *)
74 type virtual_position
= Common.parse_info
* int (* character offset *)
78 (* Present both in ast and list of tokens *)
79 | OriginTok
of Common.parse_info
80 (* Present only in ast and generated after parsing. Used mainly
81 * by Julia, to add stuff at virtual places, beginning of func or decl *)
82 | FakeTok
of string * virtual_position
83 (* Present both in ast and list of tokens. *)
84 | ExpandedTok
of Common.parse_info
* virtual_position
86 (* Present neither in ast nor in list of tokens
87 * but only in the '+' of the mcode of some tokens. Those kind of tokens
88 * are used to be able to use '=' to compare big ast portions.
90 | AbstractLineTok
of Common.parse_info
(* local to the abstracted thing *)
96 (* this cocci_tag can be changed, which is how we can express some program
97 * transformations by tagging the tokens involved in this transformation.
99 cocci_tag
: (Ast_cocci.mcodekind
* metavars_binding list
) option ref;
100 (* set in comment_annotater_c.ml *)
101 comments_tag
: comments_around
ref;
103 (* todo? token_info : sometimes useful to know what token it was *)
107 (* wrap2 is like wrap, except that I use it often for separator such
108 * as ','. In that case the info is associated to the argument that
109 * follows, so in 'a,b' I will have in the list [(a,[]); (b,[','])].
111 * wrap3 is like wrap, except that I use it in case sometimes it
112 * will be empty because the info will be included in a nested
113 * entity (e.g. for Ident in expr because it's inlined in the name)
114 * so user should never assume List.length wrap3 > 0.
116 and 'a wrap
= 'a
* il
117 and 'a wrap2
= 'a
* il
118 and 'a wrap3
= 'a
* il
(* * evotype*)
120 (* ------------------------------------------------------------------------- *)
122 (* ------------------------------------------------------------------------- *)
124 (* was called 'ident' before, but 'name' is I think better
125 * as concatenated strings can be used not only for identifiers and for
126 * declarators, but also for fields, for labels, etc.
128 * Note: because now the info is embeded in the name, the info for
129 * expression like Ident, or types like Typename, are not anymore
130 * stored in the expression or type. Hence if you assume this,
131 * which was true before, you are now wrong. So never write code like
132 * let (unwrape,_), ii = e and use 'ii' believing it contains
133 * the local ii to e. If you want to do that, use the appropiate
134 * wrapper get_local_ii_of_expr_inlining_ii_of_name.
137 | RegularName
of string wrap
138 | CppConcatenatedName
of (string wrap
) wrap2
(* the ## separators *) list
139 (* normally only used inside list of things, as in parameters or arguments
140 * in which case, cf cpp-manual, it has a special meaning *)
141 | CppVariadicName
of string wrap
(* ## s *)
142 | CppIdentBuilder
of string wrap
(* s ( ) *) *
143 ((string wrap
) wrap2 list
) (* arguments *)
146 (* ------------------------------------------------------------------------- *)
148 (* ------------------------------------------------------------------------- *)
149 (* Could have more precise type in fullType, in expression, etc, but
150 * it requires to do too much things in parsing such as checking no
151 * conflicting structname, computing value, etc. Better to separate
152 * concern. So I put '=>' to mean what we would really like. In fact
153 * what we really like is defining another fullType, expression, etc
154 * from scratch, because many stuff are just sugar.
156 * invariant: Array and FunctionType have also typeQualifier but they
157 * dont have sense. I put this to factorise some code. If you look in
158 * the grammar, you see that we can never specify const for the array
159 * himself (but we can do it for pointer) or function, we always
160 * have in the action rule of the grammar a { (nQ, FunctionType ...) }.
163 * Because of ExprStatement, we can have more 'new scope' events, but
164 * rare I think. For instance with 'array of constExpression' there can
165 * have an exprStatement and a new (local) struct defined. Same for
171 and fullType
= typeQualifier
* typeC
172 and typeC
= typeCbis wrap
(* todo reput wrap3 *)
175 NoType
(* for c++ only *)
176 | BaseType
of baseType
178 | Pointer
of fullType
179 | Array
of constExpression
option * fullType
180 | FunctionType
of functionType
182 | Enum
of string option * enumType
183 | StructUnion
of structUnion
* string option * structType
(* new scope *)
186 | StructUnionName
of structUnion
* string
188 | TypeName
of name
* fullType
option (* semantic: filled later *)
190 | ParenType
of fullType
(* for unparser: *)
192 (* gccext: TypeOfType below may seems useless; Why declare a
193 * __typeof__(int) x; ?
194 * When used with macros, it allows to fix a problem of C which
195 * is that type declaration can be spread around the ident. Indeed it
196 * may be difficult to have a macro such as
197 * '#define macro(type, ident) type ident;'
198 * because when you want to do a
199 * macro(char[256], x),
200 * then it will generate invalid code, but with a
201 * '#define macro(type, ident) __typeof(type) ident;'
204 | TypeOfExpr
of expression
205 | TypeOfType
of fullType
207 (* cppext: IfdefType TODO *)
209 (* -------------------------------------- *)
212 | FloatType
of floatType
217 (* stdC: type section
219 * note: char and signed char are semantically different!!
221 and intType
= CChar
(* obsolete? | CWchar *)
224 and signed
= sign
* base
225 and base
= CChar2
| CShort
| CInt
| CLong
| CLongLong
(* gccext: *)
226 and sign
= Signed
| UnSigned
228 and floatType
= CFloat
| CDouble
| CLongDouble
231 (* -------------------------------------- *)
232 and structUnion
= Struct
| Union
233 and structType
= field list
235 | DeclarationField
of field_declaration
240 | MacroDeclField
of (string * argument wrap2 list
)
241 wrap
(* optional ';'*)
244 | CppDirectiveStruct
of cpp_directive
245 | IfdefStruct
of ifdef_directive
(* * field list list *)
248 (* before unparser, I didn't have a FieldDeclList but just a Field. *)
249 and field_declaration
=
250 | FieldDeclList
of fieldkind wrap2 list
(* , *) wrap
(* ; *)
252 (* At first I thought that a bitfield could be only Signed/Unsigned.
253 * But it seems that gcc allow char i:4. C rule must say that you
254 * can cast into int so enum too, ...
257 | Simple
of name
option * fullType
258 | BitField
of name
option * fullType
*
259 info
(* : *) * constExpression
260 (* fullType => BitFieldInt | BitFieldUnsigned *)
263 (* -------------------------------------- *)
264 and enumType
= oneEnumType wrap2
(* , *) list
265 (* => string * int list *)
267 and oneEnumType
= name
* (info
(* = *) * constExpression
) option
269 (* -------------------------------------- *)
270 (* return * (params * has "...") *)
271 and functionType
= fullType
* (parameterType wrap2 list
* bool wrap
)
273 { p_namei
: name
option;
274 p_register
: bool wrap
;
277 (* => (bool (register) * fullType) list * bool *)
280 and typeQualifier
= typeQualifierbis wrap
281 and typeQualifierbis
= {const
: bool; volatile
: bool}
283 (* gccext: cppext: *)
284 and attribute
= attributebis wrap
286 | Attribute
of string
288 (* ------------------------------------------------------------------------- *)
290 (* ------------------------------------------------------------------------- *)
291 and expression
= (expressionbis
* exp_info
ref (* semantic: *)) wrap3
292 and exp_info
= exp_type
option * test
293 and exp_type
= fullType
(* Type_c.completed_and_simplified *) * local
294 and local
= LocalVar
of parse_info
| StaticLocalVar
of parse_info
295 | NotLocalVar
(* cocci: *)
296 and test
= Test
| NotTest
(* cocci: *)
300 (* Ident can be a enumeration constant, a simple variable, a name of a func.
301 * With cppext, Ident can also be the name of a macro. Sparse says
302 * "an identifier with a meaning is a symbol" *)
303 | Ident
of name
(* todo? more semantic info such as LocalFunc *)
305 | Constant
of constant
306 | FunCall
of expression
* argument wrap2
(* , *) list
307 (* gccext: x ? /* empty */ : y <=> x ? x : y; hence the 'option' below *)
308 | CondExpr
of expression
* expression
option * expression
310 (* should be considered as statements, bad C langage *)
311 | Sequence
of expression
* expression
312 | Assignment
of expression
* assignOp
* expression
315 | Postfix
of expression
* fixOp
316 | Infix
of expression
* fixOp
318 | Unary
of expression
* unaryOp
319 | Binary
of expression
* binaryOp
* expression
321 | ArrayAccess
of expression
* expression
323 (* field ident access *)
324 | RecordAccess
of expression
* name
325 | RecordPtAccess
of expression
* name
326 (* redundant normally, could replace it by DeRef RecordAcces *)
328 | SizeOfExpr
of expression
329 | SizeOfType
of fullType
330 | Cast
of fullType
* expression
333 | StatementExpr
of compound wrap
(* ( ) new scope *)
334 | Constructor
of fullType
* initialiser
337 | ParenExpr
of expression
341 | Delete
of expression
343 (* cppext: IfdefExpr TODO *)
345 (* cppext: normally just expression *)
346 and argument
= (expression
, weird_argument
) Common.either
348 | ArgType
of parameterType
349 | ArgAction
of action_macro
351 (* todo: ArgStatement of statement, possibly have ghost token *)
355 (* I put string for Int and Float because int would not be enough because
356 * OCaml int are 31 bits. So simpler to do string. Same reason to have
357 * string instead of int list for the String case.
359 * note: -2 is not a constant, it is the unary operator '-'
360 * applied to constant 2. So the string must represent a positive
364 | String
of (string * isWchar
)
365 | MultiString
of string list
(* can contain MacroString, todo: more info *)
366 | Char
of (string * isWchar
) (* normally it is equivalent to Int *)
367 | Int
of (string * intType
)
368 | Float
of (string * floatType
)
370 and isWchar
= IsWchar
| IsChar
373 and unaryOp
= GetRef
| DeRef
| UnPlus
| UnMinus
| Tilde
| Not
374 | GetRefLabel
(* gccext: GetRefLabel, via &&label notation *)
375 and assignOp
= SimpleAssign
| OpAssign
of arithOp
376 and fixOp
= Dec
| Inc
378 and binaryOp
= Arith
of arithOp
| Logical
of logicalOp
381 | Plus
| Minus
| Mul
| Div
| Mod
386 | Inf
| Sup
| InfEq
| SupEq
390 and constExpression
= expression
(* => int *)
392 (* ------------------------------------------------------------------------- *)
394 (* ------------------------------------------------------------------------- *)
395 (* note: that assignement is not a statement but an expression;
396 * wonderful C langage.
398 * note: I use 'and' for type definition cos gccext allow statement as
399 * expression, so need mutual recursive type definition.
403 and statement
= statementbis wrap3
406 | Compound
of compound
(* new scope *)
407 | ExprStatement
of exprStatement
408 | Selection
of selection
(* have fakeend *)
409 | Iteration
of iteration
(* have fakeend *)
412 (* simplify cocci: only at the beginning of a compound normally *)
413 | Decl
of declaration
417 | NestedFunc
of definition
424 and labeled
= Label
of name
* statement
425 | Case
of expression
* statement
426 | CaseRange
of expression
* expression
* statement
(* gccext: *)
427 | Default
of statement
430 * old: compound = (declaration list * statement list)
431 * old: (declaration, statement) either list
432 * Simplify cocci to just have statement list, by integrating Decl in stmt.
434 * update: now introduce also the _sequencable to allow ifdef in the middle.
435 * Indeed, I now allow ifdefs in the ast but they must be only between
436 * "sequencable" elements. They can be put in a type only if this type
437 * is used in a list, like at the toplevel, used in a 'toplevel list',
438 * or inside a compound, used in a 'statement list'. I must not allow
439 * ifdef anywhere. For instance I can not make ifdef a statement
440 * cos some instruction like If accept only one statement and the
441 * ifdef directive must not take the place of a legitimate instruction.
442 * We had a similar phenomena in SmPL where we have the notion
443 * of statement and sequencable statement too. Once you have
444 * such a type of sequencable thing, then s/xx list/xx_sequencable list/
445 * and introduce the ifdef.
447 * update: those ifdefs are either passed, or present in the AST but in
448 * a flat form. To structure those flat ifdefs you have to run
449 * a transformation that will put in a tree the statements inside
450 * ifdefs branches. Cf cpp_ast_c.ml. This is for instance the difference
451 * between a IfdefStmt (flat) and IfdefStmt2 (tree structured).
454 and compound
= statement_sequencable list
456 (* cppext: easier to put at statement_list level than statement level *)
457 and statement_sequencable
=
458 | StmtElem
of statement
461 | CppDirectiveStmt
of cpp_directive
462 | IfdefStmt
of ifdef_directive
464 (* this will be build in cpp_ast_c from the previous flat IfdefStmt *)
465 | IfdefStmt2
of ifdef_directive list
* (statement_sequencable list
) list
467 and exprStatement
= expression
option
469 (* for Switch, need check that all elements in the compound start
470 * with a case:, otherwise unreachable code.
473 | If
of expression
* statement
* statement
474 | Switch
of expression
* statement
478 | While
of expression
* statement
479 | DoWhile
of statement
* expression
480 | For
of exprStatement wrap
* exprStatement wrap
* exprStatement wrap
*
483 | MacroIteration
of string * argument wrap2 list
* statement
485 and jump
= Goto
of name
487 | Return
| ReturnExpr
of expression
488 | GotoComputed
of expression
(* gccext: goto *exp ';' *)
492 and asmbody
= il
(* string list *) * colon wrap
(* : *) list
493 and colon
= Colon
of colon_option wrap2 list
494 and colon_option
= colon_option_bis wrap
495 and colon_option_bis
= ColonMisc
| ColonExpr
of expression
498 (* ------------------------------------------------------------------------- *)
500 (* ------------------------------------------------------------------------- *)
501 (* (string * ...) option cos can have empty declaration or struct tag
504 * Before I had a Typedef constructor, but why make this special case and not
505 * have StructDef, EnumDef, ... so that 'struct t {...} v' will generate 2
506 * declarations ? So I try to generalise and not have Typedef either. This
507 * requires more work in parsing. Better to separate concern.
509 * Before the need for unparser, I didn't have a DeclList but just a Decl.
511 * I am not sure what it means to declare a prototype inline, but gcc
516 | DeclList
of onedecl wrap2
(* , *) list wrap
(* ; fakestart sto *)
518 (* bool is true if there is a ; at the end *)
519 | MacroDecl
of (string * argument wrap2 list
* bool) wrap
(* fakestart *)
522 { v_namei
: (name
* v_init
) option;
524 (* semantic: set in type annotated and used in cocci_vs_c
525 * when we transform some initialisation into affectation
527 v_type_bis
: fullType
(* Type_c.completed_and_simplified *) option ref;
529 v_local
: local_decl
; (* cocci: *)
530 v_attr
: attribute list
; (* gccext: *)
533 NoInit
| ValInit
of info
* initialiser
534 | ConstrInit
of argument wrap2
(* , *) list wrap
535 and storage
= storagebis
* bool (* gccext: inline or not *)
536 and storagebis
= NoSto
| StoTypedef
| Sto
of storageClass
537 and storageClass
= Auto
| Static
| Register
| Extern
539 and local_decl
= LocalDecl
| NotLocalDecl
541 (* fullType is the type used if the type should be converted to
542 an assignment. It can be adjusted in the type annotation
543 phase when typedef information is availalble *)
544 and initialiser
= initialiserbis wrap
546 | InitExpr
of expression
547 | InitList
of initialiser wrap2
(* , *) list
549 | InitDesignators
of designator list
* initialiser
550 | InitFieldOld
of string * initialiser
551 | InitIndexOld
of expression
* initialiser
553 (* ex: [2].y = x, or .y[2] or .y.x. They can be nested *)
554 and designator
= designatorbis wrap
556 | DesignatorField
of string
557 | DesignatorIndex
of expression
558 | DesignatorRange
of expression
* expression
560 (* ------------------------------------------------------------------------- *)
561 (* Function definition *)
562 (* ------------------------------------------------------------------------- *)
563 (* Normally we should define another type functionType2 because there
564 * are more restrictions on what can define a function than a pointer
565 * function. For instance a function declaration can omit the name of the
566 * parameter whereas a function definition can not. But, in some cases such
567 * as 'f(void) {', there is no name too, so I simplified and reused the
568 * same functionType type for both declaration and function definition.
570 * Also old style C does not have type in the parameter, so again simpler
571 * to abuse the functionType and allow missing type.
573 and definition
= definitionbis wrap
(* ( ) { } fakestart sto *)
576 f_type
: functionType
; (* less? a functionType2 ? *)
579 f_attr
: attribute list
; (* gccext: *)
580 f_old_c_style
: declaration list
option;
582 (* cppext: IfdefFunHeader TODO *)
584 (* ------------------------------------------------------------------------- *)
585 (* cppext: cpp directives, #ifdef, #define and #include body *)
586 (* ------------------------------------------------------------------------- *)
591 (*| Ifdef ? no, ifdefs are handled differently, cf ifdef_directive below *)
593 and define
= string wrap
(* #define s eol *) * (define_kind
* define_val
)
596 | DefineFunc
of ((string wrap
) wrap2 list
) wrap
(* () *)
599 (* most common case; e.g. to define int constant *)
600 | DefineExpr
of expression
602 | DefineStmt
of statement
603 | DefineType
of fullType
604 | DefineDoWhileZero
of (statement
* expression
) wrap
(* do { } while(0) *)
606 | DefineFunction
of definition
607 | DefineInit
of initialiser
(* in practice only { } with possible ',' *)
609 (* TODO DefineMulti of define_val list *)
611 | DefineText
of string wrap
619 { i_include
: inc_file wrap
; (* #include s *)
620 (* cocci: computed in ? *)
621 i_rel_pos
: include_rel_pos
option ref;
622 (* cocci: cf -test incl *)
624 (* cf cpp_ast_c.ml. set to None at parsing time. *)
625 i_content
: (Common.filename
(* full path *) * program
) option;
628 | Local
of inc_elem list
629 | NonLocal
of inc_elem list
630 | Weird
of string (* ex: #include SYSTEM_H *)
631 and inc_elem
= string
633 (* cocci: to tag the first of #include <xx/> and last of #include <yy/>
635 * The first_of and last_of store the list of prefixes that was
636 * introduced by the include. On #include <a/b/x>, if the include was
637 * the first in the file, it would give in first_of the following
638 * prefixes a/b/c; a/b/; a/ ; <empty>
640 * This is set after parsing, in cocci.ml, in update_rel_pos.
642 and include_rel_pos
= {
643 first_of
: string list list
;
644 last_of
: string list list
;
649 (* todo? to specialize if someone need more info *)
650 and ifdef_directive
= (* or and 'a ifdefed = 'a list wrap *)
651 | IfdefDirective
of (ifdefkind
* matching_tag
) wrap
653 | Ifdef
(* todo? of string ? of formula_cpp ? *)
654 | IfdefElseif
(* same *)
655 | IfdefElse
(* same *)
657 (* set in Parsing_hacks.set_ifdef_parenthize_info. It internally use
658 * a global so it means if you parse the same file twice you may get
659 * different id. I try now to avoid this pb by resetting it each
660 * time I parse a file.
663 IfdefTag
of (int (* tag *) * int (* total with this tag *))
669 (* ------------------------------------------------------------------------- *)
670 (* The toplevels elements *)
671 (* ------------------------------------------------------------------------- *)
673 | Declaration
of declaration
674 | Definition
of definition
677 | CppTop
of cpp_directive
678 | IfdefTop
of ifdef_directive
(* * toplevel list *)
681 | MacroTop
of string * argument wrap2 list
* il
683 | EmptyDef
of il
(* gccext: allow redundant ';' *)
684 | NotParsedCorrectly
of il
686 | FinalDef
of info
(* EOF *)
688 (* ------------------------------------------------------------------------- *)
689 and program
= toplevel list
691 (*****************************************************************************)
693 (*****************************************************************************)
694 (* Was previously in pattern.ml, but because of the transformer,
695 * we need to decorate each token with some cocci code AND the environment
696 * for this cocci code.
698 and metavars_binding
= (Ast_cocci.meta_name
, metavar_binding_kind
) assoc
699 and metavar_binding_kind
=
700 | MetaIdVal
of string *
701 Ast_cocci.meta_name list
(* negative constraints *)
702 | MetaFuncVal
of string
703 | MetaLocalFuncVal
of string
705 | MetaExprVal
of expression
(* a "clean expr" *) *
706 (*subterm constraints, currently exprs*)
707 Ast_cocci.meta_name list
708 | MetaExprListVal
of argument wrap2 list
709 | MetaParamVal
of parameterType
710 | MetaParamListVal
of parameterType wrap2 list
712 | MetaTypeVal
of fullType
713 | MetaInitVal
of initialiser
714 | MetaInitListVal
of initialiser wrap2 list
715 | MetaDeclVal
of declaration
716 | MetaFieldVal
of field
717 | MetaFieldListVal
of field list
718 | MetaStmtVal
of statement
720 (* Could also be in Lib_engine.metavars_binding2 with the ParenVal,
721 * because don't need to have the value for a position in the env of
722 * a '+'. But ParenVal or LabelVal are used only by CTL, they are not
723 * variables accessible via SmPL whereas the position can be one day
724 * so I think it's better to put MetaPosVal here *)
725 | MetaPosVal
of (Ast_cocci.fixpos
* Ast_cocci.fixpos
) (* max, min *)
727 (Common.filename
* string (*element*) * posl
* posl
) list
(* min, max *)
728 | MetaListlenVal
of int
731 (*****************************************************************************)
733 (*****************************************************************************)
735 (* convention: I often use "m" for comments as I can not use "c"
736 * (already use for c stuff) and "com" is too long.
739 (* this type will be associated to each token.
741 and comments_around
= {
742 mbefore
: Token_c.comment_like_token list
;
743 mafter
: Token_c.comment_like_token list
;
745 (* less: could remove ? do something simpler than CComment for
746 * coccinelle, cf above. *)
747 mbefore2
: comment_and_relative_pos list
;
748 mafter2
: comment_and_relative_pos list
;
750 and comment_and_relative_pos
= {
752 minfo
: Common.parse_info
;
753 (* the int represent the number of lines of difference between the
754 * current token and the comment. When on same line, this number is 0.
755 * When previous line, -1. In some way the after/before in previous
756 * record is useless because the sign of the integer can helps
757 * do the difference too, but I keep it that way.
761 * cppbetween: bool; touse? if false positive
762 * is_alone_in_line: bool; (*for labels, to avoid false positive*)
766 and comment
= Common.parse_info
767 and com
= comment list
ref
772 (*****************************************************************************)
773 (* Some constructors *)
774 (*****************************************************************************)
775 let nullQualif = ({const
=false; volatile
= false}, [])
778 let defaultInt = (BaseType
(IntType
(Si
(Signed
, CInt
))))
780 let noType () = ref (None
,NotTest
)
781 let noInstr = (ExprStatement
(None
), [])
782 let noTypedefDef () = None
784 let emptyMetavarsBinding =
785 ([]: metavars_binding
)
787 let emptyAnnotCocci =
788 (Ast_cocci.CONTEXT
(Ast_cocci.NoPos
,Ast_cocci.NOTHING
),
789 ([] : metavars_binding list
))
792 (None
: (Ast_cocci.mcodekind
* metavars_binding list
) option)
794 (* compatibility mode *)
795 let mcode_and_env_of_cocciref aref
=
798 | None
-> emptyAnnotCocci
809 (* for include, some meta information needed by cocci *)
811 ref (None
: include_rel_pos
option)
816 (* When want add some info in ast that does not correspond to
817 * an existing C element.
818 * old: or when don't want 'synchronize' on it in unparse_c.ml
819 * (now have other mark for tha matter).
821 let no_virt_pos = ({str
="";charpos
=0;line
=0;column
=0;file
=""},-1)
824 { pinfo
= FakeTok
("",no_virt_pos);
825 cocci_tag
= ref emptyAnnot;
826 comments_tag
= ref emptyComments;
831 let noi_content = (None
: ((Common.filename
* program
) option))
833 (*****************************************************************************)
835 (*****************************************************************************)
840 let unwrap_expr ((unwrap_e
, typ
), iie
) = unwrap_e
841 let rewrap_expr ((_old_unwrap_e
, typ
), iie
) newe
= ((newe
, typ
), iie
)
843 let unwrap_typeC (qu
, (typeC
, ii
)) = typeC
844 let rewrap_typeC (qu
, (typeC
, ii
)) newtypeC
= (qu
, (newtypeC
, ii
))
846 let unwrap_typeCbis (typeC
, ii
) = typeC
848 let unwrap_st (unwrap_st, ii
) = unwrap_st
850 (* ------------------------------------------------------------------------- *)
851 let mk_e unwrap_e ii
= (unwrap_e
, noType()), ii
852 let mk_e_bis unwrap_e ty ii
= (unwrap_e
, ty
), ii
854 let mk_ty typeC ii
= nQ, (typeC
, ii
)
855 let mk_tybis typeC ii
= (typeC
, ii
)
857 let mk_st unwrap_st ii
= (unwrap_st, ii
)
859 (* ------------------------------------------------------------------------- *)
860 let get_ii_typeC_take_care (typeC
, ii
) = ii
861 let get_ii_st_take_care (st
, ii
) = ii
862 let get_ii_expr_take_care (e
, ii
) = ii
864 let get_st_and_ii (st
, ii
) = st
, ii
865 let get_ty_and_ii (qu
, (typeC
, ii
)) = qu
, (typeC
, ii
)
866 let get_e_and_ii (e
, ii
) = e
, ii
869 (* ------------------------------------------------------------------------- *)
870 let get_type_expr ((unwrap_e
, typ
), iie
) = !typ
871 let set_type_expr ((unwrap_e
, oldtyp
), iie
) newtyp
=
873 (* old: (unwrap_e, newtyp), iie *)
875 let get_onlytype_expr ((unwrap_e
, typ
), iie
) =
877 | Some
(ft
,_local
), _test
-> Some ft
880 let get_onlylocal_expr ((unwrap_e
, typ
), iie
) =
882 | Some
(ft
,local
), _test
-> Some local
885 (* ------------------------------------------------------------------------- *)
886 let rewrap_str s ii
=
889 OriginTok pi
-> OriginTok
{ pi
with Common.str
= s
;}
890 | ExpandedTok
(pi
,vpi
) -> ExpandedTok
({ pi
with Common.str
= s
;},vpi
)
891 | FakeTok
(_
,vpi
) -> FakeTok
(s
,vpi
)
892 | AbstractLineTok pi
-> OriginTok
{ pi
with Common.str
= s
;})}
894 let rewrap_pinfo pi ii
=
899 (* info about the current location *)
900 let get_pi = function
902 | ExpandedTok
(_
,(pi
,_
)) -> pi
903 | FakeTok
(_
,(pi
,_
)) -> pi
904 | AbstractLineTok pi
-> pi
907 let get_opi = function
909 | ExpandedTok
(pi
,_
) -> pi
(* diff with get_pi *)
910 | FakeTok
(_
,_
) -> failwith
"no position information"
911 | AbstractLineTok pi
-> pi
915 OriginTok pi
-> pi
.Common.str
916 | ExpandedTok
(pi
,_
) -> pi
.Common.str
918 | AbstractLineTok pi
-> pi
.Common.str
923 | ExpandedTok
(_
,(pi
,_
)) -> f pi
924 | FakeTok
(_
,(pi
,_
)) -> f pi
925 | AbstractLineTok pi
-> f pi
927 let get_orig_info f ii
=
930 | ExpandedTok
(pi
,_
) -> f pi
(* diff with get_info *)
931 | FakeTok
(_
,(pi
,_
)) -> f pi
932 | AbstractLineTok pi
-> f pi
934 let make_expanded ii
=
935 {ii
with pinfo
= ExpandedTok
(get_opi ii
.pinfo
,no_virt_pos)}
937 let pos_of_info ii
= get_info (function x
-> x
.Common.charpos
) ii
938 let opos_of_info ii
= get_orig_info (function x
-> x
.Common.charpos
) ii
939 let line_of_info ii
= get_orig_info (function x
-> x
.Common.line
) ii
940 let col_of_info ii
= get_orig_info (function x
-> x
.Common.column
) ii
941 let file_of_info ii
= get_orig_info (function x
-> x
.Common.file
) ii
942 let mcode_of_info ii
= fst
(mcode_and_env_of_cocciref ii
.cocci_tag
)
943 let pinfo_of_info ii
= ii
.pinfo
944 let parse_info_of_info ii
= get_pi ii
.pinfo
946 let strloc_of_info ii
=
947 spf
"%s:%d" (file_of_info ii
) (line_of_info ii
)
951 FakeTok
(_
,_
) -> true
954 let is_origintok ii
=
956 | OriginTok pi
-> true
959 (* ------------------------------------------------------------------------- *)
960 type posrv
= Real
of Common.parse_info
| Virt
of virtual_position
962 let compare_pos ii1 ii2
=
963 let get_pos = function
964 OriginTok pi
-> Real pi
965 | FakeTok
(s
,vpi
) -> Virt vpi
966 | ExpandedTok
(pi
,vpi
) -> Virt vpi
967 | AbstractLineTok pi
-> Real pi
in (* used for printing *)
968 let pos1 = get_pos (pinfo_of_info ii1
) in
969 let pos2 = get_pos (pinfo_of_info ii2
) in
970 match (pos1,pos2) with
971 (Real p1
, Real p2
) ->
972 compare p1
.Common.charpos p2
.Common.charpos
973 | (Virt
(p1
,_
), Real p2
) ->
974 if (compare p1
.Common.charpos p2
.Common.charpos
) =|= (-1) then (-1) else 1
975 | (Real p1
, Virt
(p2
,_
)) ->
976 if (compare p1
.Common.charpos p2
.Common.charpos
) =|= 1 then 1 else (-1)
977 | (Virt
(p1
,o1
), Virt
(p2
,o2
)) ->
978 let poi1 = p1
.Common.charpos
in
979 let poi2 = p2
.Common.charpos
in
980 match compare
poi1 poi2 with
985 let equal_posl (l1
,c1
) (l2
,c2
) =
986 (l1
=|= l2
) && (c1
=|= c2
)
988 let info_to_fixpos ii
=
989 match pinfo_of_info ii
with
990 OriginTok pi
-> Ast_cocci.Real pi
.Common.charpos
991 | ExpandedTok
(_
,(pi
,offset
)) ->
992 Ast_cocci.Virt
(pi
.Common.charpos
,offset
)
993 | FakeTok
(_
,(pi
,offset
)) ->
994 Ast_cocci.Virt
(pi
.Common.charpos
,offset
)
995 | AbstractLineTok pi
-> failwith
"unexpected abstract"
998 let is_test (e
: expression
) =
999 let (_
,info
), _
= e
in
1000 let (_
,test
) = !info
in
1003 (*****************************************************************************)
1005 (*****************************************************************************)
1007 (* When we have extended the C Ast to add some info to the tokens,
1008 * such as its line number in the file, we can not use anymore the
1009 * ocaml '=' to compare Ast elements. To overcome this problem, to be
1010 * able to use again '=', we just have to get rid of all those extra
1011 * information, to "abstract those line" (al) information.
1013 * Julia then modifies it a little to have a tokenindex, so the original
1014 * true al_info is in fact real_al_info.
1017 let al_info tokenindex x
=
1020 {charpos
= tokenindex
;
1022 column
= tokenindex
;
1024 str
= str_of_info x
});
1025 cocci_tag
= ref emptyAnnot;
1026 comments_tag
= ref emptyComments;
1029 let semi_al_info x
=
1031 cocci_tag
= ref emptyAnnot;
1032 comments_tag
= ref emptyComments;
1035 let magic_real_number = -10
1037 let real_al_info x
=
1040 {charpos
= magic_real_number;
1041 line
= magic_real_number;
1042 column
= magic_real_number;
1044 str
= str_of_info x
});
1045 cocci_tag
= ref emptyAnnot;
1046 comments_tag
= ref emptyComments;
1051 List.filter
(function (Token_c.TCommentCpp _
,_
) -> true | _
-> false) l
in
1053 (x
,{i
with Common.charpos
= magic_real_number;
1054 Common.line
= magic_real_number;
1055 Common.column
= magic_real_number}) in
1056 {mbefore
= []; (* duplicates mafter of the previous token *)
1057 mafter
= List.map
al_com (keep_cpp x
.mafter
);
1062 let al_info_cpp tokenindex x
=
1065 {charpos
= tokenindex
;
1067 column
= tokenindex
;
1069 str
= str_of_info x
});
1070 cocci_tag
= ref emptyAnnot;
1071 comments_tag
= ref (al_comments !(x
.comments_tag
));
1074 let semi_al_info_cpp x
=
1076 cocci_tag
= ref emptyAnnot;
1077 comments_tag
= ref (al_comments !(x
.comments_tag
));
1080 let real_al_info_cpp x
=
1083 {charpos
= magic_real_number;
1084 line
= magic_real_number;
1085 column
= magic_real_number;
1087 str
= str_of_info x
});
1088 cocci_tag
= ref emptyAnnot;
1089 comments_tag
= ref (al_comments !(x
.comments_tag
));
1093 (*****************************************************************************)
1095 (*****************************************************************************)
1097 (* Transform a list of arguments (or parameters) where the commas are
1098 * represented via the wrap2 and associated with an element, with
1099 * a list where the comma are on their own. f(1,2,2) was
1100 * [(1,[]); (2,[,]); (2,[,])] and become [1;',';2;',';2].
1102 * Used in cocci_vs_c.ml, to have a more direct correspondance between
1103 * the ast_cocci of julia and ast_c.
1105 let rec (split_comma
: 'a wrap2 list
-> ('a
, il
) either list
) =
1110 then (Left e
)::split_comma xs
1111 else Right ii
::Left e
::split_comma xs
1113 let rec (unsplit_comma
: ('a
, il
) either list
-> 'a wrap2 list
) =
1116 | Right ii
::Left e
::xs
->
1117 (e
, ii
)::unsplit_comma xs
1119 let empty_ii = [] in
1120 (e
, empty_ii)::unsplit_comma xs
1127 (*****************************************************************************)
1128 (* Helpers, could also be put in lib_parsing_c.ml instead *)
1129 (*****************************************************************************)
1131 (* should maybe be in pretty_print_c ? *)
1133 let s_of_inc_file inc_file
=
1135 | Local xs
-> xs
+> Common.join
"/"
1136 | NonLocal xs
-> xs
+> Common.join
"/"
1139 let s_of_inc_file_bis inc_file
=
1141 | Local xs
-> "\"" ^ xs
+> Common.join
"/" ^
"\""
1142 | NonLocal xs
-> "<" ^ xs
+> Common.join
"/" ^
">"
1145 let fieldname_of_fieldkind fieldkind
=
1146 match fieldkind
with
1147 | Simple
(sopt
, ft
) -> sopt
1148 | BitField
(sopt
, ft
, info
, expr
) -> sopt
1151 let s_of_attr attr
=
1153 +> List.map
(fun (Attribute s
, ii
) -> s
)
1157 (* ------------------------------------------------------------------------- *)
1158 let str_of_name ident
=
1160 | RegularName
(s
,ii
) -> s
1161 | CppConcatenatedName xs
->
1162 xs
+> List.map
(fun (x
,iiop
) -> unwrap x
) +> Common.join
"##"
1163 | CppVariadicName
(s
, ii
) -> "##" ^ s
1164 | CppIdentBuilder
((s
,iis
), xs
) ->
1166 (xs
+> List.map
(fun ((x
,iix
), iicomma
) -> x
) +> Common.join
",") ^
1169 let get_s_and_ii_of_name name
=
1171 | RegularName
(s
, iis
) -> s
, iis
1172 | CppIdentBuilder
((s
, iis
), xs
) -> s
, iis
1173 | CppVariadicName
(s
,iis
) ->
1174 let (iop
, iis
) = Common.tuple_of_list2 iis
in
1176 | CppConcatenatedName xs
->
1178 | [] -> raise Impossible
1179 | ((s
,iis
),noiiop
)::xs
->
1183 let get_s_and_info_of_name name
=
1184 let (s
,ii
) = get_s_and_ii_of_name name
in
1187 let info_of_name name
=
1188 let (s
,ii
) = get_s_and_ii_of_name name
in
1191 let ii_of_name name
=
1192 let (s
,ii
) = get_s_and_ii_of_name name
in
1195 let get_local_ii_of_expr_inlining_ii_of_name e
=
1196 let (ebis
,_
),ii
= e
in
1198 | Ident name
, noii ->
1201 | RecordAccess
(e
, name
), ii
->
1202 ii
@ ii_of_name name
1203 | RecordPtAccess
(e
, name
), ii
->
1204 ii
@ ii_of_name name
1208 let get_local_ii_of_tybis_inlining_ii_of_name ty
=
1210 | TypeName
(name
, _typ
), [] -> ii_of_name name
1213 (* the following is used to obtain the argument to LocalVar *)
1214 let info_of_type ft
=
1215 let (qu
, ty
) = ft
in
1216 (* bugfix: because of string->name, the ii can be deeper *)
1217 let ii = get_local_ii_of_tybis_inlining_ii_of_name ty
in
1220 | [] -> failwith
"type has no text; need to think again"
1222 (* only Label and Goto have name *)
1223 let get_local_ii_of_st_inlining_ii_of_name st
=
1225 | Labeled
(Label
(name
, st
)), ii -> ii_of_name name
@ ii
1226 | Jump
(Goto name
), ii ->
1227 let (i1
, i3
) = Common.tuple_of_list2
ii in
1228 [i1
] @ ii_of_name name
@ [i3
]
1233 (* ------------------------------------------------------------------------- *)
1234 let name_of_parameter param
=
1235 param
.p_namei
+> Common.map_option
(str_of_name)