Commit | Line | Data |
---|---|---|
15637ed4 RG |
1 | /*- |
2 | * This code is derived from software copyrighted by the Free Software | |
3 | * Foundation. | |
4 | * | |
5 | * Modified 1991 by Donn Seeley at UUNET Technologies, Inc. | |
6 | */ | |
7 | ||
8 | #ifndef lint | |
9 | static char sccsid[] = "@(#)cpp.c 6.3 (Berkeley) 5/8/91"; | |
10 | #endif /* not lint */ | |
11 | ||
12 | /* C Compatible Compiler Preprocessor (CCCP) | |
13 | Copyright (C) 1986, 1987, 1989 Free Software Foundation, Inc. | |
14 | Written by Paul Rubin, June 1986 | |
15 | Adapted to ANSI C, Richard Stallman, Jan 1987 | |
16 | ||
17 | This program is free software; you can redistribute it and/or modify it | |
18 | under the terms of the GNU General Public License as published by the | |
19 | Free Software Foundation; either version 1, or (at your option) any | |
20 | later version. | |
21 | ||
22 | This program is distributed in the hope that it will be useful, | |
23 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
24 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
25 | GNU General Public License for more details. | |
26 | ||
27 | You should have received a copy of the GNU General Public License | |
28 | along with this program; if not, write to the Free Software | |
29 | Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
30 | ||
31 | In other words, you are welcome to use, share and improve this program. | |
32 | You are forbidden to forbid anyone else to use, share and improve | |
33 | what you give them. Help stamp out software-hoarding! */ | |
34 | \f | |
35 | typedef unsigned char U_CHAR; | |
36 | ||
37 | #ifdef EMACS | |
38 | #define NO_SHORTNAMES | |
39 | #include "../src/config.h" | |
40 | #ifdef open | |
41 | #undef open | |
42 | #undef read | |
43 | #undef write | |
44 | #endif /* open */ | |
45 | #endif /* EMACS */ | |
46 | ||
47 | #ifndef EMACS | |
48 | #include "config.h" | |
49 | #endif /* not EMACS */ | |
50 | ||
51 | #ifndef STDC_VALUE | |
52 | #define STDC_VALUE 1 | |
53 | #endif | |
54 | ||
55 | /* In case config.h defines these. */ | |
56 | #undef bcopy | |
57 | #undef bzero | |
58 | #undef bcmp | |
59 | ||
60 | #include <sys/types.h> | |
61 | #include <sys/stat.h> | |
62 | #include <ctype.h> | |
63 | #include <stdio.h> | |
64 | #include <signal.h> | |
65 | ||
66 | #ifndef VMS | |
67 | #include <sys/file.h> | |
68 | #ifndef USG | |
69 | #include <sys/time.h> /* for __DATE__ and __TIME__ */ | |
70 | #include <sys/resource.h> | |
71 | #else | |
72 | #define index strchr | |
73 | #define rindex strrchr | |
74 | #include <time.h> | |
75 | #include <fcntl.h> | |
76 | #endif /* USG */ | |
77 | #endif /* not VMS */ | |
78 | ||
79 | /* VMS-specific definitions */ | |
80 | #ifdef VMS | |
81 | #include <time.h> | |
82 | #include <errno.h> /* This defines "errno" properly */ | |
83 | #include <perror.h> /* This defines sys_errlist/sys_nerr properly */ | |
84 | #define O_RDONLY 0 /* Open arg for Read/Only */ | |
85 | #define O_WRONLY 1 /* Open arg for Write/Only */ | |
86 | #define read(fd,buf,size) VAX11_C_read(fd,buf,size) | |
87 | #define write(fd,buf,size) VAX11_C_write(fd,buf,size) | |
88 | #ifdef __GNUC__ | |
89 | #define BSTRING /* VMS/GCC supplies the bstring routines */ | |
90 | #endif /* __GNUC__ */ | |
91 | #endif /* VMS */ | |
92 | ||
93 | #ifndef O_RDONLY | |
94 | #define O_RDONLY 0 | |
95 | #endif | |
96 | ||
97 | #define max(a,b) ((a) > (b) ? (a) : (b)) | |
98 | ||
99 | #ifndef S_ISREG | |
100 | #define S_ISREG(m) (((m) & S_IFMT) == S_IFREG) | |
101 | #endif | |
102 | ||
103 | /* External declarations. */ | |
104 | ||
105 | void bcopy (), bzero (); | |
106 | int bcmp (); | |
107 | extern char *getenv (); | |
108 | extern char *version_string; | |
109 | ||
110 | /* Forward declarations. */ | |
111 | ||
112 | struct directive; | |
113 | struct file_buf; | |
114 | struct arglist; | |
115 | struct argdata; | |
116 | ||
117 | int do_define (), do_line (), do_include (), do_undef (), do_error (), | |
118 | do_pragma (), do_if (), do_xifdef (), do_else (), | |
119 | do_elif (), do_endif (), do_sccs (), do_once (); | |
120 | ||
121 | struct hashnode *install (); | |
122 | struct hashnode *lookup (); | |
123 | ||
124 | char *xmalloc (), *xrealloc (), *xcalloc (), *savestring (); | |
125 | void fatal (), fancy_abort (), pfatal_with_name (), perror_with_name (); | |
126 | ||
127 | void macroexpand (); | |
128 | void dump_all_macros (); | |
129 | void conditional_skip (); | |
130 | void skip_if_group (); | |
131 | void output_line_command (); | |
132 | /* Last arg to output_line_command. */ | |
133 | enum file_change_code {same_file, enter_file, leave_file}; | |
134 | ||
135 | int grow_outbuf (); | |
136 | int handle_directive (); | |
137 | void memory_full (); | |
138 | ||
139 | U_CHAR *macarg1 (); | |
140 | char *macarg (); | |
141 | ||
142 | U_CHAR *skip_to_end_of_comment (); | |
143 | U_CHAR *skip_quoted_string (); | |
144 | \f | |
145 | #ifndef FATAL_EXIT_CODE | |
146 | #define FATAL_EXIT_CODE 33 /* gnu cc command understands this */ | |
147 | #endif | |
148 | ||
149 | #ifndef SUCCESS_EXIT_CODE | |
150 | #define SUCCESS_EXIT_CODE 0 /* 0 means success on Unix. */ | |
151 | #endif | |
152 | ||
153 | /* Name under which this program was invoked. */ | |
154 | ||
155 | char *progname; | |
156 | ||
157 | /* Nonzero means handle C++ comment syntax and use | |
158 | extra default include directories for C++. */ | |
159 | ||
160 | int cplusplus; | |
161 | ||
162 | /* Current maximum length of directory names in the search path | |
163 | for include files. (Altered as we get more of them.) */ | |
164 | ||
165 | int max_include_len; | |
166 | ||
167 | /* Nonzero means copy comments into the output file. */ | |
168 | ||
169 | int put_out_comments = 0; | |
170 | ||
171 | /* Nonzero means don't process the ANSI trigraph sequences. */ | |
172 | ||
173 | int no_trigraphs = 0; | |
174 | ||
175 | /* Nonzero means print the names of included files rather than | |
176 | the preprocessed output. 1 means just the #include "...", | |
177 | 2 means #include <...> as well. */ | |
178 | ||
179 | int print_deps = 0; | |
180 | ||
181 | /* Nonzero means don't output line number information. */ | |
182 | ||
183 | int no_line_commands; | |
184 | ||
185 | /* Nonzero means inhibit output of the preprocessed text | |
186 | and instead output the definitions of all user-defined macros | |
187 | in a form suitable for use as input to cccp. */ | |
188 | ||
189 | int dump_macros; | |
190 | ||
191 | /* Nonzero means give all the error messages the ANSI standard requires. */ | |
192 | ||
193 | int pedantic; | |
194 | ||
195 | /* Nonzero means don't print warning messages. -w. */ | |
196 | ||
197 | int inhibit_warnings = 0; | |
198 | ||
199 | /* Nonzero means warn if slash-star appears in a comment. */ | |
200 | ||
201 | int warn_comments; | |
202 | ||
203 | /* Nonzero means warn if there are any trigraphs. */ | |
204 | ||
205 | int warn_trigraphs; | |
206 | ||
207 | /* Nonzero means try to imitate old fashioned non-ANSI preprocessor. */ | |
208 | ||
209 | int traditional; | |
210 | ||
211 | /* Nonzero causes output not to be done, | |
212 | but directives such as #define that have side effects | |
213 | are still obeyed. */ | |
214 | ||
215 | int no_output; | |
216 | \f | |
217 | /* I/O buffer structure. | |
218 | The `fname' field is nonzero for source files and #include files | |
219 | and for the dummy text used for -D and -U. | |
220 | It is zero for rescanning results of macro expansion | |
221 | and for expanding macro arguments. */ | |
222 | #define INPUT_STACK_MAX 200 | |
223 | struct file_buf { | |
224 | char *fname; | |
225 | int lineno; | |
226 | int length; | |
227 | U_CHAR *buf; | |
228 | U_CHAR *bufp; | |
229 | /* Macro that this level is the expansion of. | |
230 | Included so that we can reenable the macro | |
231 | at the end of this level. */ | |
232 | struct hashnode *macro; | |
233 | /* Value of if_stack at start of this file. | |
234 | Used to prohibit unmatched #endif (etc) in an include file. */ | |
235 | struct if_stack *if_stack; | |
236 | /* Object to be freed at end of input at this level. */ | |
237 | U_CHAR *free_ptr; | |
238 | } instack[INPUT_STACK_MAX]; | |
239 | ||
240 | /* Current nesting level of input sources. | |
241 | `instack[indepth]' is the level currently being read. */ | |
242 | int indepth = -1; | |
243 | #define CHECK_DEPTH(code) \ | |
244 | if (indepth >= (INPUT_STACK_MAX - 1)) \ | |
245 | { \ | |
246 | error_with_line (line_for_error (instack[indepth].lineno), \ | |
247 | "macro or #include recursion too deep"); \ | |
248 | code; \ | |
249 | } | |
250 | ||
251 | /* Current depth in #include directives that use <...>. */ | |
252 | int system_include_depth = 0; | |
253 | ||
254 | typedef struct file_buf FILE_BUF; | |
255 | ||
256 | /* The output buffer. Its LENGTH field is the amount of room allocated | |
257 | for the buffer, not the number of chars actually present. To get | |
258 | that, subtract outbuf.buf from outbuf.bufp. */ | |
259 | ||
260 | #define OUTBUF_SIZE 10 /* initial size of output buffer */ | |
261 | FILE_BUF outbuf; | |
262 | ||
263 | /* Grow output buffer OBUF points at | |
264 | so it can hold at least NEEDED more chars. */ | |
265 | ||
266 | #define check_expand(OBUF, NEEDED) \ | |
267 | (((OBUF)->length - ((OBUF)->bufp - (OBUF)->buf) <= (NEEDED)) \ | |
268 | ? grow_outbuf ((OBUF), (NEEDED)) : 0) | |
269 | ||
270 | struct file_name_list | |
271 | { | |
272 | struct file_name_list *next; | |
273 | char *fname; | |
274 | }; | |
275 | ||
276 | /* #include "file" looks in source file dir, then stack. */ | |
277 | /* #include <file> just looks in the stack. */ | |
278 | /* -I directories are added to the end, then the defaults are added. */ | |
279 | struct file_name_list include_defaults[] = | |
280 | { | |
281 | #ifndef VMS | |
282 | { &include_defaults[1], GCC_INCLUDE_DIR }, | |
283 | { &include_defaults[2], "/usr/include" }, | |
284 | { 0, "/usr/local/include" } | |
285 | #else | |
286 | { &include_defaults[1], "GNU_CC_INCLUDE:" }, /* GNU includes */ | |
287 | { &include_defaults[2], "SYS$SYSROOT:[SYSLIB.]" }, /* VAX-11 "C" includes */ | |
288 | { 0, "" }, /* This makes normal VMS filespecs work OK */ | |
289 | #endif /* VMS */ | |
290 | }; | |
291 | ||
292 | /* These are used instead of the above, for C++. */ | |
293 | struct file_name_list cplusplus_include_defaults[] = | |
294 | { | |
295 | #ifndef VMS | |
296 | /* Pick up GNU C++ specific include files. */ | |
297 | { &cplusplus_include_defaults[1], GPLUSPLUS_INCLUDE_DIR }, | |
298 | /* Use GNU CC specific header files. */ | |
299 | { &cplusplus_include_defaults[2], GCC_INCLUDE_DIR }, | |
300 | { 0, "/usr/include" } | |
301 | #else | |
302 | { &cplusplus_include_defaults[1], "GNU_GXX_INCLUDE:" }, | |
303 | { &cplusplus_include_defaults[2], "GNU_CC_INCLUDE:" }, | |
304 | /* VAX-11 C includes */ | |
305 | { &cplusplus_include_defaults[3], "SYS$SYSROOT:[SYSLIB.]" }, | |
306 | { 0, "" }, /* This makes normal VMS filespecs work OK */ | |
307 | #endif /* VMS */ | |
308 | }; | |
309 | ||
310 | struct file_name_list *include = 0; /* First dir to search */ | |
311 | /* First dir to search for <file> */ | |
312 | struct file_name_list *first_bracket_include = 0; | |
313 | struct file_name_list *last_include = 0; /* Last in chain */ | |
314 | ||
315 | /* List of included files that contained #once. */ | |
316 | struct file_name_list *dont_repeat_files = 0; | |
317 | ||
318 | /* List of other included files. */ | |
319 | struct file_name_list *all_include_files = 0; | |
320 | \f | |
321 | /* Structure allocated for every #define. For a simple replacement | |
322 | such as | |
323 | #define foo bar , | |
324 | nargs = -1, the `pattern' list is null, and the expansion is just | |
325 | the replacement text. Nargs = 0 means a functionlike macro with no args, | |
326 | e.g., | |
327 | #define getchar() getc (stdin) . | |
328 | When there are args, the expansion is the replacement text with the | |
329 | args squashed out, and the reflist is a list describing how to | |
330 | build the output from the input: e.g., "3 chars, then the 1st arg, | |
331 | then 9 chars, then the 3rd arg, then 0 chars, then the 2nd arg". | |
332 | The chars here come from the expansion. Whatever is left of the | |
333 | expansion after the last arg-occurrence is copied after that arg. | |
334 | Note that the reflist can be arbitrarily long--- | |
335 | its length depends on the number of times the arguments appear in | |
336 | the replacement text, not how many args there are. Example: | |
337 | #define f(x) x+x+x+x+x+x+x would have replacement text "++++++" and | |
338 | pattern list | |
339 | { (0, 1), (1, 1), (1, 1), ..., (1, 1), NULL } | |
340 | where (x, y) means (nchars, argno). */ | |
341 | ||
342 | typedef struct definition DEFINITION; | |
343 | struct definition { | |
344 | int nargs; | |
345 | int length; /* length of expansion string */ | |
346 | U_CHAR *expansion; | |
347 | struct reflist { | |
348 | struct reflist *next; | |
349 | char stringify; /* nonzero if this arg was preceded by a | |
350 | # operator. */ | |
351 | char raw_before; /* Nonzero if a ## operator before arg. */ | |
352 | char raw_after; /* Nonzero if a ## operator after arg. */ | |
353 | int nchars; /* Number of literal chars to copy before | |
354 | this arg occurrence. */ | |
355 | int argno; /* Number of arg to substitute (origin-0) */ | |
356 | } *pattern; | |
357 | /* Names of macro args, concatenated in reverse order | |
358 | with comma-space between them. | |
359 | The only use of this is that we warn on redefinition | |
360 | if this differs between the old and new definitions. */ | |
361 | U_CHAR *argnames; | |
362 | }; | |
363 | ||
364 | /* different kinds of things that can appear in the value field | |
365 | of a hash node. Actually, this may be useless now. */ | |
366 | union hashval { | |
367 | int ival; | |
368 | char *cpval; | |
369 | DEFINITION *defn; | |
370 | }; | |
371 | ||
372 | ||
373 | /* The structure of a node in the hash table. The hash table | |
374 | has entries for all tokens defined by #define commands (type T_MACRO), | |
375 | plus some special tokens like __LINE__ (these each have their own | |
376 | type, and the appropriate code is run when that type of node is seen. | |
377 | It does not contain control words like "#define", which are recognized | |
378 | by a separate piece of code. */ | |
379 | ||
380 | /* different flavors of hash nodes --- also used in keyword table */ | |
381 | enum node_type { | |
382 | T_DEFINE = 1, /* the `#define' keyword */ | |
383 | T_INCLUDE, /* the `#include' keyword */ | |
384 | T_IFDEF, /* the `#ifdef' keyword */ | |
385 | T_IFNDEF, /* the `#ifndef' keyword */ | |
386 | T_IF, /* the `#if' keyword */ | |
387 | T_ELSE, /* `#else' */ | |
388 | T_PRAGMA, /* `#pragma' */ | |
389 | T_ELIF, /* `#else' */ | |
390 | T_UNDEF, /* `#undef' */ | |
391 | T_LINE, /* `#line' */ | |
392 | T_ERROR, /* `#error' */ | |
393 | T_ENDIF, /* `#endif' */ | |
394 | T_SCCS, /* `#sccs', used on system V. */ | |
395 | T_IDENT, /* `#ident', used on system V. */ | |
396 | T_SPECLINE, /* special symbol `__LINE__' */ | |
397 | T_DATE, /* `__DATE__' */ | |
398 | T_FILE, /* `__FILE__' */ | |
399 | T_BASE_FILE, /* `__BASE_FILE__' */ | |
400 | T_INCLUDE_LEVEL, /* `__INCLUDE_LEVEL__' */ | |
401 | T_VERSION, /* `__VERSION__' */ | |
402 | T_TIME, /* `__TIME__' */ | |
403 | T_CONST, /* Constant value, used by `__STDC__' */ | |
404 | T_MACRO, /* macro defined by `#define' */ | |
405 | T_DISABLED, /* macro temporarily turned off for rescan */ | |
406 | T_SPEC_DEFINED, /* special `defined' macro for use in #if statements */ | |
407 | T_UNUSED /* Used for something not defined. */ | |
408 | }; | |
409 | ||
410 | struct hashnode { | |
411 | struct hashnode *next; /* double links for easy deletion */ | |
412 | struct hashnode *prev; | |
413 | struct hashnode **bucket_hdr; /* also, a back pointer to this node's hash | |
414 | chain is kept, in case the node is the head | |
415 | of the chain and gets deleted. */ | |
416 | enum node_type type; /* type of special token */ | |
417 | int length; /* length of token, for quick comparison */ | |
418 | U_CHAR *name; /* the actual name */ | |
419 | union hashval value; /* pointer to expansion, or whatever */ | |
420 | }; | |
421 | ||
422 | typedef struct hashnode HASHNODE; | |
423 | ||
424 | /* Some definitions for the hash table. The hash function MUST be | |
425 | computed as shown in hashf () below. That is because the rescan | |
426 | loop computes the hash value `on the fly' for most tokens, | |
427 | in order to avoid the overhead of a lot of procedure calls to | |
428 | the hashf () function. Hashf () only exists for the sake of | |
429 | politeness, for use when speed isn't so important. */ | |
430 | ||
431 | #define HASHSIZE 1403 | |
432 | HASHNODE *hashtab[HASHSIZE]; | |
433 | #define HASHSTEP(old, c) ((old << 2) + c) | |
434 | #define MAKE_POS(v) (v & ~0x80000000) /* make number positive */ | |
435 | ||
436 | /* Symbols to predefine. */ | |
437 | ||
438 | #ifdef CPP_PREDEFINES | |
439 | char *predefs = CPP_PREDEFINES; | |
440 | #else | |
441 | char *predefs = ""; | |
442 | #endif | |
443 | ||
444 | /* `struct directive' defines one #-directive, including how to handle it. */ | |
445 | ||
446 | struct directive { | |
447 | int length; /* Length of name */ | |
448 | int (*func)(); /* Function to handle directive */ | |
449 | char *name; /* Name of directive */ | |
450 | enum node_type type; /* Code which describes which directive. */ | |
451 | char angle_brackets; /* Nonzero => <...> is special. */ | |
452 | char traditional_comments; /* Nonzero: keep comments if -traditional. */ | |
453 | char pass_thru; /* Copy preprocessed directive to output file. */ | |
454 | }; | |
455 | ||
456 | /* Here is the actual list of #-directives, most-often-used first. */ | |
457 | ||
458 | struct directive directive_table[] = { | |
459 | { 6, do_define, "define", T_DEFINE, 0, 1}, | |
460 | { 2, do_if, "if", T_IF}, | |
461 | { 5, do_xifdef, "ifdef", T_IFDEF}, | |
462 | { 6, do_xifdef, "ifndef", T_IFNDEF}, | |
463 | { 5, do_endif, "endif", T_ENDIF}, | |
464 | { 4, do_else, "else", T_ELSE}, | |
465 | { 4, do_elif, "elif", T_ELIF}, | |
466 | { 4, do_line, "line", T_LINE}, | |
467 | { 7, do_include, "include", T_INCLUDE, 1}, | |
468 | { 5, do_undef, "undef", T_UNDEF}, | |
469 | { 5, do_error, "error", T_ERROR}, | |
470 | #ifdef SCCS_DIRECTIVE | |
471 | { 4, do_sccs, "sccs", T_SCCS}, | |
472 | #endif | |
473 | { 6, do_pragma, "pragma", T_PRAGMA, 0, 0, 1}, | |
474 | { -1, 0, "", T_UNUSED}, | |
475 | }; | |
476 | ||
477 | /* table to tell if char can be part of a C identifier. */ | |
478 | U_CHAR is_idchar[256]; | |
479 | /* table to tell if char can be first char of a c identifier. */ | |
480 | U_CHAR is_idstart[256]; | |
481 | /* table to tell if c is horizontal space. */ | |
482 | U_CHAR is_hor_space[256]; | |
483 | /* table to tell if c is horizontal or vertical space. */ | |
484 | U_CHAR is_space[256]; | |
485 | ||
486 | #define SKIP_WHITE_SPACE(p) do { while (is_hor_space[*p]) p++; } while (0) | |
487 | #define SKIP_ALL_WHITE_SPACE(p) do { while (is_space[*p]) p++; } while (0) | |
488 | ||
489 | int errors = 0; /* Error counter for exit code */ | |
490 | ||
491 | /* Zero means dollar signs are punctuation. | |
492 | -$ stores 0; -traditional, stores 1. Default is 1 for VMS, 0 otherwise. | |
493 | This must be 0 for correct processing of this ANSI C program: | |
494 | #define foo(a) #a | |
495 | #define lose(b) foo(b) | |
496 | #define test$ | |
497 | lose(test) */ | |
498 | #ifndef DOLLARS_IN_IDENTIFIERS | |
499 | #define DOLLARS_IN_IDENTIFIERS 0 | |
500 | #endif | |
501 | int dollars_in_ident = DOLLARS_IN_IDENTIFIERS; | |
502 | ||
503 | FILE_BUF expand_to_temp_buffer (); | |
504 | ||
505 | DEFINITION *collect_expansion (); | |
506 | ||
507 | /* Stack of conditionals currently in progress | |
508 | (including both successful and failing conditionals). */ | |
509 | ||
510 | struct if_stack { | |
511 | struct if_stack *next; /* for chaining to the next stack frame */ | |
512 | char *fname; /* copied from input when frame is made */ | |
513 | int lineno; /* similarly */ | |
514 | int if_succeeded; /* true if a leg of this if-group | |
515 | has been passed through rescan */ | |
516 | enum node_type type; /* type of last directive seen in this group */ | |
517 | }; | |
518 | typedef struct if_stack IF_STACK_FRAME; | |
519 | IF_STACK_FRAME *if_stack = NULL; | |
520 | ||
521 | /* Buffer of -M output. */ | |
522 | ||
523 | char *deps_buffer; | |
524 | ||
525 | /* Number of bytes allocated in above. */ | |
526 | int deps_allocated_size; | |
527 | ||
528 | /* Number of bytes used. */ | |
529 | int deps_size; | |
530 | ||
531 | /* Number of bytes since the last newline. */ | |
532 | int deps_column; | |
533 | ||
534 | /* Nonzero means -I- has been seen, | |
535 | so don't look for #include "foo" the source-file directory. */ | |
536 | int ignore_srcdir; | |
537 | \f | |
538 | /* Handler for SIGPIPE. */ | |
539 | ||
540 | static void | |
541 | pipe_closed () | |
542 | { | |
543 | fatal ("output pipe has been closed"); | |
544 | } | |
545 | ||
546 | int | |
547 | main (argc, argv) | |
548 | int argc; | |
549 | char **argv; | |
550 | { | |
551 | int st_mode; | |
552 | long st_size; | |
553 | char *in_fname, *out_fname; | |
554 | int f, i; | |
555 | FILE_BUF *fp; | |
556 | char **pend_files = (char **) xmalloc (argc * sizeof (char *)); | |
557 | char **pend_defs = (char **) xmalloc (argc * sizeof (char *)); | |
558 | char **pend_undefs = (char **) xmalloc (argc * sizeof (char *)); | |
559 | int inhibit_predefs = 0; | |
560 | int no_standard_includes = 0; | |
561 | ||
562 | /* Non-0 means don't output the preprocessed program. */ | |
563 | int inhibit_output = 0; | |
564 | ||
565 | /* Stream on which to print the dependency information. */ | |
566 | FILE *deps_stream = 0; | |
567 | /* Target-name to write with the dependency information. */ | |
568 | char *deps_target = 0; | |
569 | ||
570 | #ifdef RLIMIT_STACK | |
571 | /* Get rid of any avoidable limit on stack size. */ | |
572 | { | |
573 | struct rlimit rlim; | |
574 | ||
575 | /* Set the stack limit huge so that alloca (particularly stringtab | |
576 | * in dbxread.c) does not fail. */ | |
577 | getrlimit (RLIMIT_STACK, &rlim); | |
578 | rlim.rlim_cur = rlim.rlim_max; | |
579 | setrlimit (RLIMIT_STACK, &rlim); | |
580 | } | |
581 | #endif /* RLIMIT_STACK defined */ | |
582 | ||
583 | progname = argv[0]; | |
584 | #ifdef VMS | |
585 | { | |
586 | /* Remove directories from PROGNAME. */ | |
587 | char *s; | |
588 | extern char *rindex (); | |
589 | ||
590 | progname = savestring (argv[0]); | |
591 | ||
592 | if (!(s = rindex (progname, ']'))) | |
593 | s = rindex (progname, ':'); | |
594 | if (s) | |
595 | strcpy (progname, s+1); | |
596 | if (s = rindex (progname, '.')) | |
597 | *s = '\0'; | |
598 | } | |
599 | #endif | |
600 | ||
601 | in_fname = NULL; | |
602 | out_fname = NULL; | |
603 | ||
604 | /* Initialize is_idchar to allow $. */ | |
605 | dollars_in_ident = 1; | |
606 | initialize_char_syntax (); | |
607 | dollars_in_ident = DOLLARS_IN_IDENTIFIERS; | |
608 | ||
609 | no_line_commands = 0; | |
610 | no_trigraphs = 1; | |
611 | dump_macros = 0; | |
612 | no_output = 0; | |
613 | cplusplus = 0; | |
614 | #ifdef CPLUSPLUS | |
615 | cplusplus = 1; | |
616 | #endif | |
617 | ||
618 | signal (SIGPIPE, pipe_closed); | |
619 | ||
620 | #ifndef VMS | |
621 | max_include_len | |
622 | = max (max (sizeof (GCC_INCLUDE_DIR), | |
623 | sizeof (GPLUSPLUS_INCLUDE_DIR)), | |
624 | sizeof ("/usr/include/CC")); | |
625 | #else /* VMS */ | |
626 | max_include_len | |
627 | = sizeof("SYS$SYSROOT:[SYSLIB.]"); | |
628 | #endif /* VMS */ | |
629 | ||
630 | bzero (pend_files, argc * sizeof (char *)); | |
631 | bzero (pend_defs, argc * sizeof (char *)); | |
632 | bzero (pend_undefs, argc * sizeof (char *)); | |
633 | ||
634 | /* Process switches and find input file name. */ | |
635 | ||
636 | for (i = 1; i < argc; i++) { | |
637 | if (argv[i][0] != '-') { | |
638 | if (out_fname != NULL) | |
639 | fatal ("Usage: %s [switches] input output", argv[0]); | |
640 | else if (in_fname != NULL) | |
641 | out_fname = argv[i]; | |
642 | else | |
643 | in_fname = argv[i]; | |
644 | } else { | |
645 | switch (argv[i][1]) { | |
646 | ||
647 | case 'i': | |
648 | if (argv[i][2] != 0) | |
649 | pend_files[i] = argv[i] + 2; | |
650 | else if (i + 1 == argc) | |
651 | fatal ("Filename missing after -i option"); | |
652 | else | |
653 | pend_files[i] = argv[i+1], i++; | |
654 | break; | |
655 | ||
656 | case 'o': | |
657 | if (out_fname != NULL) | |
658 | fatal ("Output filename specified twice"); | |
659 | if (i + 1 == argc) | |
660 | fatal ("Filename missing after -o option"); | |
661 | out_fname = argv[++i]; | |
662 | if (!strcmp (out_fname, "-")) | |
663 | out_fname = ""; | |
664 | break; | |
665 | ||
666 | case 'p': | |
667 | pedantic = 1; | |
668 | break; | |
669 | ||
670 | case 't': | |
671 | if (!strcmp (argv[i], "-traditional")) { | |
672 | traditional = 1; | |
673 | dollars_in_ident = 1; | |
674 | } else if (!strcmp (argv[i], "-trigraphs")) { | |
675 | no_trigraphs = 0; | |
676 | } | |
677 | break; | |
678 | ||
679 | case '+': | |
680 | cplusplus = 1; | |
681 | break; | |
682 | ||
683 | case 'w': | |
684 | inhibit_warnings = 1; | |
685 | break; | |
686 | ||
687 | case 'W': | |
688 | if (!strcmp (argv[i], "-Wtrigraphs")) { | |
689 | warn_trigraphs = 1; | |
690 | } | |
691 | if (!strcmp (argv[i], "-Wcomments")) | |
692 | warn_comments = 1; | |
693 | if (!strcmp (argv[i], "-Wcomment")) | |
694 | warn_comments = 1; | |
695 | if (!strcmp (argv[i], "-Wall")) { | |
696 | warn_trigraphs = 1; | |
697 | warn_comments = 1; | |
698 | } | |
699 | break; | |
700 | ||
701 | case 'M': | |
702 | if (!strcmp (argv[i], "-M")) | |
703 | print_deps = 2; | |
704 | else if (!strcmp (argv[i], "-MM")) | |
705 | print_deps = 1; | |
706 | inhibit_output = 1; | |
707 | break; | |
708 | ||
709 | case 'd': | |
710 | dump_macros = 1; | |
711 | no_output = 1; | |
712 | break; | |
713 | ||
714 | case 'v': | |
715 | fprintf (stderr, "GNU CPP version %s\n", version_string); | |
716 | break; | |
717 | ||
718 | case 'D': | |
719 | { | |
720 | char *p, *p1; | |
721 | ||
722 | if (argv[i][2] != 0) | |
723 | p = argv[i] + 2; | |
724 | else if (i + 1 == argc) | |
725 | fatal ("Macro name missing after -D option"); | |
726 | else | |
727 | p = argv[++i]; | |
728 | ||
729 | if ((p1 = (char *) index (p, '=')) != NULL) | |
730 | *p1 = ' '; | |
731 | pend_defs[i] = p; | |
732 | } | |
733 | break; | |
734 | ||
735 | case 'U': /* JF #undef something */ | |
736 | if (argv[i][2] != 0) | |
737 | pend_undefs[i] = argv[i] + 2; | |
738 | else if (i + 1 == argc) | |
739 | fatal ("Macro name missing after -U option"); | |
740 | else | |
741 | pend_undefs[i] = argv[i+1], i++; | |
742 | break; | |
743 | ||
744 | case 'C': | |
745 | put_out_comments = 1; | |
746 | break; | |
747 | ||
748 | case 'E': /* -E comes from cc -E; ignore it. */ | |
749 | break; | |
750 | ||
751 | case 'P': | |
752 | no_line_commands = 1; | |
753 | break; | |
754 | ||
755 | case '$': /* Don't include $ in identifiers. */ | |
756 | dollars_in_ident = 0; | |
757 | break; | |
758 | ||
759 | case 'I': /* Add directory to path for includes. */ | |
760 | { | |
761 | struct file_name_list *dirtmp; | |
762 | ||
763 | if (! ignore_srcdir && !strcmp (argv[i] + 2, "-")) | |
764 | ignore_srcdir = 1; | |
765 | else { | |
766 | dirtmp = (struct file_name_list *) | |
767 | xmalloc (sizeof (struct file_name_list)); | |
768 | dirtmp->next = 0; /* New one goes on the end */ | |
769 | if (include == 0) | |
770 | include = dirtmp; | |
771 | else | |
772 | last_include->next = dirtmp; | |
773 | last_include = dirtmp; /* Tail follows the last one */ | |
774 | if (argv[i][2] != 0) | |
775 | dirtmp->fname = argv[i] + 2; | |
776 | else if (i + 1 == argc) | |
777 | fatal ("Directory name missing after -I option"); | |
778 | else | |
779 | dirtmp->fname = argv[++i]; | |
780 | if (strlen (dirtmp->fname) > max_include_len) | |
781 | max_include_len = strlen (dirtmp->fname); | |
782 | if (dirtmp->fname[0] == '.' && dirtmp->fname[1] == '\0') | |
783 | dirtmp->fname = 0; /* Current directory */ | |
784 | if (ignore_srcdir && first_bracket_include == 0) | |
785 | first_bracket_include = dirtmp; | |
786 | } | |
787 | } | |
788 | break; | |
789 | ||
790 | case 'n': | |
791 | /* -nostdinc causes no default include directories. | |
792 | You must specify all include-file directories with -I. */ | |
793 | no_standard_includes = 1; | |
794 | break; | |
795 | ||
796 | case 'u': | |
797 | /* Sun compiler passes undocumented switch "-undef". | |
798 | Let's assume it means to inhibit the predefined symbols. */ | |
799 | inhibit_predefs = 1; | |
800 | break; | |
801 | ||
802 | case '\0': /* JF handle '-' as file name meaning stdin or stdout */ | |
803 | if (in_fname == NULL) { | |
804 | in_fname = ""; | |
805 | break; | |
806 | } else if (out_fname == NULL) { | |
807 | out_fname = ""; | |
808 | break; | |
809 | } /* else fall through into error */ | |
810 | ||
811 | default: | |
812 | fatal ("Invalid option `%s'", argv[i]); | |
813 | } | |
814 | } | |
815 | } | |
816 | ||
817 | /* Now that dollars_in_ident is known, initialize is_idchar. */ | |
818 | initialize_char_syntax (); | |
819 | ||
820 | /* Install __LINE__, etc. Must follow initialize_char_syntax | |
821 | and option processing. */ | |
822 | initialize_builtins (); | |
823 | ||
824 | /* Do standard #defines that identify processor type. */ | |
825 | ||
826 | if (!inhibit_predefs) { | |
827 | char *p = (char *) alloca (strlen (predefs) + 1); | |
828 | strcpy (p, predefs); | |
829 | while (*p) { | |
830 | char *q; | |
831 | if (p[0] != '-' || p[1] != 'D') | |
832 | abort (); | |
833 | q = &p[2]; | |
834 | while (*p && *p != ' ') p++; | |
835 | if (*p != 0) | |
836 | *p++= 0; | |
837 | make_definition (q); | |
838 | } | |
839 | } | |
840 | ||
841 | /* Do defines specified with -D. */ | |
842 | for (i = 1; i < argc; i++) | |
843 | if (pend_defs[i]) | |
844 | make_definition (pend_defs[i]); | |
845 | ||
846 | /* Do undefines specified with -U. */ | |
847 | for (i = 1; i < argc; i++) | |
848 | if (pend_undefs[i]) | |
849 | make_undef (pend_undefs[i]); | |
850 | ||
851 | /* Unless -fnostdinc, | |
852 | tack on the standard include file dirs to the specified list */ | |
853 | if (!no_standard_includes) { | |
854 | if (include == 0) | |
855 | include = (cplusplus ? cplusplus_include_defaults : include_defaults); | |
856 | else | |
857 | last_include->next | |
858 | = (cplusplus ? cplusplus_include_defaults : include_defaults); | |
859 | /* Make sure the list for #include <...> also has the standard dirs. */ | |
860 | if (ignore_srcdir && first_bracket_include == 0) | |
861 | first_bracket_include | |
862 | = (cplusplus ? cplusplus_include_defaults : include_defaults); | |
863 | } | |
864 | ||
865 | /* Initialize output buffer */ | |
866 | ||
867 | outbuf.buf = (U_CHAR *) xmalloc (OUTBUF_SIZE); | |
868 | outbuf.bufp = outbuf.buf; | |
869 | outbuf.length = OUTBUF_SIZE; | |
870 | ||
871 | /* Scan the -i files before the main input. | |
872 | Much like #including them, but with no_output set | |
873 | so that only their macro definitions matter. */ | |
874 | ||
875 | no_output++; | |
876 | for (i = 1; i < argc; i++) | |
877 | if (pend_files[i]) { | |
878 | int fd = open (pend_files[i], O_RDONLY, 0666); | |
879 | if (fd < 0) { | |
880 | perror_with_name (pend_files[i]); | |
881 | return FATAL_EXIT_CODE; | |
882 | } | |
883 | finclude (fd, pend_files[i], &outbuf); | |
884 | } | |
885 | no_output--; | |
886 | ||
887 | /* Create an input stack level for the main input file | |
888 | and copy the entire contents of the file into it. */ | |
889 | ||
890 | fp = &instack[++indepth]; | |
891 | ||
892 | /* JF check for stdin */ | |
893 | if (in_fname == NULL || *in_fname == 0) { | |
894 | in_fname = ""; | |
895 | f = 0; | |
896 | } else if ((f = open (in_fname, O_RDONLY, 0666)) < 0) | |
897 | goto perror; | |
898 | ||
899 | /* Either of two environment variables can specify output of deps. | |
900 | Its value is either "OUTPUT_FILE" or "OUTPUT_FILE DEPS_TARGET", | |
901 | where OUTPUT_FILE is the file to write deps info to | |
902 | and DEPS_TARGET is the target to mention in the deps. */ | |
903 | ||
904 | if (print_deps == 0 | |
905 | && (getenv ("SUNPRO_DEPENDENCIES") != 0 | |
906 | || getenv ("DEPENDENCIES_OUTPUT") != 0)) | |
907 | { | |
908 | char *spec = getenv ("DEPENDENCIES_OUTPUT"); | |
909 | char *s; | |
910 | char *output_file; | |
911 | ||
912 | if (spec == 0) | |
913 | { | |
914 | spec = getenv ("SUNPRO_DEPENDENCIES"); | |
915 | print_deps = 2; | |
916 | } | |
917 | else | |
918 | print_deps = 1; | |
919 | ||
920 | s = spec; | |
921 | /* Find the space before the DEPS_TARGET, if there is one. */ | |
922 | /* Don't use `index'; that causes trouble on USG. */ | |
923 | while (*s != 0 && *s != ' ') s++; | |
924 | if (*s != 0) | |
925 | { | |
926 | deps_target = s + 1; | |
927 | output_file = (char *) xmalloc (s - spec + 1); | |
928 | bcopy (spec, output_file, s - spec); | |
929 | output_file[s - spec] = 0; | |
930 | } | |
931 | else | |
932 | { | |
933 | deps_target = 0; | |
934 | output_file = spec; | |
935 | } | |
936 | ||
937 | deps_stream = fopen (output_file, "a"); | |
938 | if (deps_stream == 0) | |
939 | pfatal_with_name (output_file); | |
940 | } | |
941 | /* If the -M option was used, output the deps to standard output. */ | |
942 | else if (print_deps) | |
943 | deps_stream = stdout; | |
944 | ||
945 | /* For -M, print the expected object file name | |
946 | as the target of this Make-rule. */ | |
947 | if (print_deps) { | |
948 | deps_allocated_size = 200; | |
949 | deps_buffer = (char *) xmalloc (deps_allocated_size); | |
950 | deps_buffer[0] = 0; | |
951 | deps_size = 0; | |
952 | deps_column = 0; | |
953 | ||
954 | if (deps_target) { | |
955 | deps_output (deps_target, 0); | |
956 | deps_output (":", 0); | |
957 | } else if (*in_fname == 0) | |
958 | deps_output ("-: ", 0); | |
959 | else { | |
960 | int len; | |
961 | char *p = in_fname; | |
962 | char *p1 = p; | |
963 | /* Discard all directory prefixes from P. */ | |
964 | while (*p1) { | |
965 | if (*p1 == '/') | |
966 | p = p1 + 1; | |
967 | p1++; | |
968 | } | |
969 | /* Output P, but remove known suffixes. */ | |
970 | len = strlen (p); | |
971 | if (p[len - 2] == '.' | |
972 | && (p[len - 1] == 'c' || p[len - 1] == 'C' || p[len - 1] == 'S')) | |
973 | deps_output (p, len - 2); | |
974 | else if (p[len - 3] == '.' | |
975 | && p[len - 2] == 'c' | |
976 | && p[len - 1] == 'c') | |
977 | deps_output (p, len - 3); | |
978 | else | |
979 | deps_output (p, 0); | |
980 | /* Supply our own suffix. */ | |
981 | deps_output (".o : ", 0); | |
982 | deps_output (in_fname, 0); | |
983 | deps_output (" ", 0); | |
984 | } | |
985 | } | |
986 | ||
987 | file_size_and_mode (f, &st_mode, &st_size); | |
988 | fp->fname = in_fname; | |
989 | fp->lineno = 1; | |
990 | /* JF all this is mine about reading pipes and ttys */ | |
991 | if (!S_ISREG (st_mode)) { | |
992 | /* Read input from a file that is not a normal disk file. | |
993 | We cannot preallocate a buffer with the correct size, | |
994 | so we must read in the file a piece at the time and make it bigger. */ | |
995 | int size; | |
996 | int bsize; | |
997 | int cnt; | |
998 | U_CHAR *bufp; | |
999 | ||
1000 | bsize = 2000; | |
1001 | size = 0; | |
1002 | fp->buf = (U_CHAR *) xmalloc (bsize + 2); | |
1003 | bufp = fp->buf; | |
1004 | for (;;) { | |
1005 | cnt = read (f, bufp, bsize - size); | |
1006 | if (cnt < 0) goto perror; /* error! */ | |
1007 | if (cnt == 0) break; /* End of file */ | |
1008 | size += cnt; | |
1009 | bufp += cnt; | |
1010 | if (bsize == size) { /* Buffer is full! */ | |
1011 | bsize *= 2; | |
1012 | fp->buf = (U_CHAR *) xrealloc (fp->buf, bsize + 2); | |
1013 | bufp = fp->buf + size; /* May have moved */ | |
1014 | } | |
1015 | } | |
1016 | fp->length = size; | |
1017 | } else { | |
1018 | /* Read a file whose size we can determine in advance. | |
1019 | For the sake of VMS, st_size is just an upper bound. */ | |
1020 | long i; | |
1021 | fp->length = 0; | |
1022 | fp->buf = (U_CHAR *) xmalloc (st_size + 2); | |
1023 | ||
1024 | while (st_size > 0) { | |
1025 | i = read (f, fp->buf + fp->length, st_size); | |
1026 | if (i <= 0) { | |
1027 | if (i == 0) break; | |
1028 | goto perror; | |
1029 | } | |
1030 | fp->length += i; | |
1031 | st_size -= i; | |
1032 | } | |
1033 | } | |
1034 | fp->bufp = fp->buf; | |
1035 | fp->if_stack = if_stack; | |
1036 | ||
1037 | /* Unless inhibited, convert trigraphs in the input. */ | |
1038 | ||
1039 | if (!no_trigraphs) | |
1040 | trigraph_pcp (fp); | |
1041 | ||
1042 | /* Make sure data ends with a newline. And put a null after it. */ | |
1043 | ||
1044 | if (fp->length > 0 && fp->buf[fp->length-1] != '\n') | |
1045 | fp->buf[fp->length++] = '\n'; | |
1046 | fp->buf[fp->length] = '\0'; | |
1047 | ||
1048 | /* Now that we know the input file is valid, open the output. */ | |
1049 | ||
1050 | if (!out_fname || !strcmp (out_fname, "")) | |
1051 | out_fname = "stdout"; | |
1052 | else if (! freopen (out_fname, "w", stdout)) | |
1053 | pfatal_with_name (out_fname); | |
1054 | ||
1055 | output_line_command (fp, &outbuf, 0, same_file); | |
1056 | ||
1057 | /* Scan the input, processing macros and directives. */ | |
1058 | ||
1059 | rescan (&outbuf, 0); | |
1060 | ||
1061 | /* Now we have processed the entire input | |
1062 | Write whichever kind of output has been requested. */ | |
1063 | ||
1064 | ||
1065 | if (dump_macros) | |
1066 | dump_all_macros (); | |
1067 | else if (! inhibit_output && deps_stream != stdout) { | |
1068 | if (write (fileno (stdout), outbuf.buf, outbuf.bufp - outbuf.buf) < 0) | |
1069 | fatal ("I/O error on output"); | |
1070 | } | |
1071 | ||
1072 | if (print_deps) { | |
1073 | fputs (deps_buffer, deps_stream); | |
1074 | putc ('\n', deps_stream); | |
1075 | if (deps_stream != stdout) { | |
1076 | fclose (deps_stream); | |
1077 | if (ferror (deps_stream)) | |
1078 | fatal ("I/O error on output"); | |
1079 | } | |
1080 | } | |
1081 | ||
1082 | if (ferror (stdout)) | |
1083 | fatal ("I/O error on output"); | |
1084 | ||
1085 | if (errors) | |
1086 | exit (FATAL_EXIT_CODE); | |
1087 | exit (SUCCESS_EXIT_CODE); | |
1088 | ||
1089 | perror: | |
1090 | pfatal_with_name (in_fname); | |
1091 | } | |
1092 | \f | |
1093 | /* Pre-C-Preprocessor to translate ANSI trigraph idiocy in BUF | |
1094 | before main CCCP processing. Name `pcp' is also in honor of the | |
1095 | drugs the trigraph designers must have been on. | |
1096 | ||
1097 | Using an extra pass through the buffer takes a little extra time, | |
1098 | but is infinitely less hairy than trying to handle ??/" inside | |
1099 | strings, etc. everywhere, and also makes sure that trigraphs are | |
1100 | only translated in the top level of processing. */ | |
1101 | ||
1102 | trigraph_pcp (buf) | |
1103 | FILE_BUF *buf; | |
1104 | { | |
1105 | register U_CHAR c, *fptr, *bptr, *sptr; | |
1106 | int len; | |
1107 | ||
1108 | fptr = bptr = sptr = buf->buf; | |
1109 | while ((sptr = (U_CHAR *) index (sptr, '?')) != NULL) { | |
1110 | if (*++sptr != '?') | |
1111 | continue; | |
1112 | switch (*++sptr) { | |
1113 | case '=': | |
1114 | c = '#'; | |
1115 | break; | |
1116 | case '(': | |
1117 | c = '['; | |
1118 | break; | |
1119 | case '/': | |
1120 | c = '\\'; | |
1121 | break; | |
1122 | case ')': | |
1123 | c = ']'; | |
1124 | break; | |
1125 | case '\'': | |
1126 | c = '^'; | |
1127 | break; | |
1128 | case '<': | |
1129 | c = '{'; | |
1130 | break; | |
1131 | case '!': | |
1132 | c = '|'; | |
1133 | break; | |
1134 | case '>': | |
1135 | c = '}'; | |
1136 | break; | |
1137 | case '-': | |
1138 | c = '~'; | |
1139 | break; | |
1140 | case '?': | |
1141 | sptr--; | |
1142 | continue; | |
1143 | default: | |
1144 | continue; | |
1145 | } | |
1146 | len = sptr - fptr - 2; | |
1147 | if (bptr != fptr && len > 0) | |
1148 | bcopy (fptr, bptr, len); /* BSD doc says bcopy () works right | |
1149 | for overlapping strings. In ANSI | |
1150 | C, this will be memmove (). */ | |
1151 | bptr += len; | |
1152 | *bptr++ = c; | |
1153 | fptr = ++sptr; | |
1154 | } | |
1155 | len = buf->length - (fptr - buf->buf); | |
1156 | if (bptr != fptr && len > 0) | |
1157 | bcopy (fptr, bptr, len); | |
1158 | buf->length -= fptr - bptr; | |
1159 | buf->buf[buf->length] = '\0'; | |
1160 | if (warn_trigraphs && fptr != bptr) | |
1161 | warning ("%d trigraph(s) encountered", (fptr - bptr) / 2); | |
1162 | } | |
1163 | \f | |
1164 | /* Move all backslash-newline pairs out of embarrassing places. | |
1165 | Exchange all such pairs following BP | |
1166 | with any potentially-embarrasing characters that follow them. | |
1167 | Potentially-embarrassing characters are / and * | |
1168 | (because a backslash-newline inside a comment delimiter | |
1169 | would cause it not to be recognized). */ | |
1170 | ||
1171 | newline_fix (bp) | |
1172 | U_CHAR *bp; | |
1173 | { | |
1174 | register U_CHAR *p = bp; | |
1175 | register int count = 0; | |
1176 | ||
1177 | /* First count the backslash-newline pairs here. */ | |
1178 | ||
1179 | while (*p++ == '\\' && *p++ == '\n') | |
1180 | count++; | |
1181 | ||
1182 | p = bp + count * 2; | |
1183 | ||
1184 | /* Exit if what follows the backslash-newlines is not embarrassing. */ | |
1185 | ||
1186 | if (count == 0 || (*p != '/' && *p != '*')) | |
1187 | return; | |
1188 | ||
1189 | /* Copy all potentially embarrassing characters | |
1190 | that follow the backslash-newline pairs | |
1191 | down to where the pairs originally started. */ | |
1192 | ||
1193 | while (*p == '*' || *p == '/') | |
1194 | *bp++ = *p++; | |
1195 | ||
1196 | /* Now write the same number of pairs after the embarrassing chars. */ | |
1197 | while (count-- > 0) { | |
1198 | *bp++ = '\\'; | |
1199 | *bp++ = '\n'; | |
1200 | } | |
1201 | } | |
1202 | ||
1203 | /* Like newline_fix but for use within a directive-name. | |
1204 | Move any backslash-newlines up past any following symbol constituents. */ | |
1205 | ||
1206 | name_newline_fix (bp) | |
1207 | U_CHAR *bp; | |
1208 | { | |
1209 | register U_CHAR *p = bp; | |
1210 | register int count = 0; | |
1211 | ||
1212 | /* First count the backslash-newline pairs here. */ | |
1213 | ||
1214 | while (*p++ == '\\' && *p++ == '\n') | |
1215 | count++; | |
1216 | ||
1217 | p = bp + count * 2; | |
1218 | ||
1219 | /* What follows the backslash-newlines is not embarrassing. */ | |
1220 | ||
1221 | if (count == 0 || !is_idchar[*p]) | |
1222 | return; | |
1223 | ||
1224 | /* Copy all potentially embarrassing characters | |
1225 | that follow the backslash-newline pairs | |
1226 | down to where the pairs originally started. */ | |
1227 | ||
1228 | while (is_idchar[*p]) | |
1229 | *bp++ = *p++; | |
1230 | ||
1231 | /* Now write the same number of pairs after the embarrassing chars. */ | |
1232 | while (count-- > 0) { | |
1233 | *bp++ = '\\'; | |
1234 | *bp++ = '\n'; | |
1235 | } | |
1236 | } | |
1237 | \f | |
1238 | /* | |
1239 | * The main loop of the program. | |
1240 | * | |
1241 | * Read characters from the input stack, transferring them to the | |
1242 | * output buffer OP. | |
1243 | * | |
1244 | * Macros are expanded and push levels on the input stack. | |
1245 | * At the end of such a level it is popped off and we keep reading. | |
1246 | * At the end of any other kind of level, we return. | |
1247 | * #-directives are handled, except within macros. | |
1248 | * | |
1249 | * If OUTPUT_MARKS is nonzero, keep Newline markers found in the input | |
1250 | * and insert them when appropriate. This is set while scanning macro | |
1251 | * arguments before substitution. It is zero when scanning for final output. | |
1252 | * There are three types of Newline markers: | |
1253 | * * Newline - follows a macro name that was not expanded | |
1254 | * because it appeared inside an expansion of the same macro. | |
1255 | * This marker prevents future expansion of that identifier. | |
1256 | * When the input is rescanned into the final output, these are deleted. | |
1257 | * These are also deleted by ## concatenation. | |
1258 | * * Newline Space (or Newline and any other whitespace character) | |
1259 | * stands for a place that tokens must be separated or whitespace | |
1260 | * is otherwise desirable, but where the ANSI standard specifies there | |
1261 | * is no whitespace. This marker turns into a Space (or whichever other | |
1262 | * whitespace char appears in the marker) in the final output, | |
1263 | * but it turns into nothing in an argument that is stringified with #. | |
1264 | * Such stringified arguments are the only place where the ANSI standard | |
1265 | * specifies with precision that whitespace may not appear. | |
1266 | * | |
1267 | * During this function, IP->bufp is kept cached in IBP for speed of access. | |
1268 | * Likewise, OP->bufp is kept in OBP. Before calling a subroutine | |
1269 | * IBP, IP and OBP must be copied back to memory. IP and IBP are | |
1270 | * copied back with the RECACHE macro. OBP must be copied back from OP->bufp | |
1271 | * explicitly, and before RECACHE, since RECACHE uses OBP. | |
1272 | */ | |
1273 | ||
1274 | rescan (op, output_marks) | |
1275 | FILE_BUF *op; | |
1276 | int output_marks; | |
1277 | { | |
1278 | /* Character being scanned in main loop. */ | |
1279 | register U_CHAR c; | |
1280 | ||
1281 | /* Length of pending accumulated identifier. */ | |
1282 | register int ident_length = 0; | |
1283 | ||
1284 | /* Hash code of pending accumulated identifier. */ | |
1285 | register int hash = 0; | |
1286 | ||
1287 | /* Current input level (&instack[indepth]). */ | |
1288 | FILE_BUF *ip; | |
1289 | ||
1290 | /* Pointer for scanning input. */ | |
1291 | register U_CHAR *ibp; | |
1292 | ||
1293 | /* Pointer to end of input. End of scan is controlled by LIMIT. */ | |
1294 | register U_CHAR *limit; | |
1295 | ||
1296 | /* Pointer for storing output. */ | |
1297 | register U_CHAR *obp; | |
1298 | ||
1299 | /* REDO_CHAR is nonzero if we are processing an identifier | |
1300 | after backing up over the terminating character. | |
1301 | Sometimes we process an identifier without backing up over | |
1302 | the terminating character, if the terminating character | |
1303 | is not special. Backing up is done so that the terminating character | |
1304 | will be dispatched on again once the identifier is dealt with. */ | |
1305 | int redo_char = 0; | |
1306 | ||
1307 | /* 1 if within an identifier inside of which a concatenation | |
1308 | marker (Newline -) has been seen. */ | |
1309 | int concatenated = 0; | |
1310 | ||
1311 | /* While scanning a comment or a string constant, | |
1312 | this records the line it started on, for error messages. */ | |
1313 | int start_line; | |
1314 | ||
1315 | /* Record position of last `real' newline. */ | |
1316 | U_CHAR *beg_of_line; | |
1317 | ||
1318 | /* Pop the innermost input stack level, assuming it is a macro expansion. */ | |
1319 | ||
1320 | #define POPMACRO \ | |
1321 | do { ip->macro->type = T_MACRO; \ | |
1322 | if (ip->free_ptr) free (ip->free_ptr); \ | |
1323 | --indepth; } while (0) | |
1324 | ||
1325 | /* Reload `rescan's local variables that describe the current | |
1326 | level of the input stack. */ | |
1327 | ||
1328 | #define RECACHE \ | |
1329 | do { ip = &instack[indepth]; \ | |
1330 | ibp = ip->bufp; \ | |
1331 | limit = ip->buf + ip->length; \ | |
1332 | op->bufp = obp; \ | |
1333 | check_expand (op, limit - ibp); \ | |
1334 | beg_of_line = 0; \ | |
1335 | obp = op->bufp; } while (0) | |
1336 | ||
1337 | if (no_output && instack[indepth].fname != 0) | |
1338 | skip_if_group (&instack[indepth], 1); | |
1339 | ||
1340 | obp = op->bufp; | |
1341 | RECACHE; | |
1342 | beg_of_line = ibp; | |
1343 | ||
1344 | /* Our caller must always put a null after the end of | |
1345 | the input at each input stack level. */ | |
1346 | if (*limit != 0) | |
1347 | abort (); | |
1348 | ||
1349 | while (1) { | |
1350 | c = *ibp++; | |
1351 | *obp++ = c; | |
1352 | ||
1353 | switch (c) { | |
1354 | case '\\': | |
1355 | if (ibp >= limit) | |
1356 | break; | |
1357 | if (*ibp == '\n') { | |
1358 | /* Always merge lines ending with backslash-newline, | |
1359 | even in middle of identifier. */ | |
1360 | ++ibp; | |
1361 | ++ip->lineno; | |
1362 | --obp; /* remove backslash from obuf */ | |
1363 | break; | |
1364 | } | |
1365 | /* Otherwise, backslash suppresses specialness of following char, | |
1366 | so copy it here to prevent the switch from seeing it. | |
1367 | But first get any pending identifier processed. */ | |
1368 | if (ident_length > 0) | |
1369 | goto specialchar; | |
1370 | *obp++ = *ibp++; | |
1371 | break; | |
1372 | ||
1373 | case '#': | |
1374 | /* If this is expanding a macro definition, don't recognize | |
1375 | preprocessor directives. */ | |
1376 | if (ip->macro != 0) | |
1377 | goto randomchar; | |
1378 | if (ident_length) | |
1379 | goto specialchar; | |
1380 | ||
1381 | /* # keyword: a # must be first nonblank char on the line */ | |
1382 | if (beg_of_line == 0) | |
1383 | goto randomchar; | |
1384 | { | |
1385 | U_CHAR *bp; | |
1386 | ||
1387 | /* Scan from start of line, skipping whitespace, comments | |
1388 | and backslash-newlines, and see if we reach this #. | |
1389 | If not, this # is not special. */ | |
1390 | bp = beg_of_line; | |
1391 | while (1) { | |
1392 | if (is_hor_space[*bp]) | |
1393 | bp++; | |
1394 | else if (*bp == '\\' && bp[1] == '\n') | |
1395 | bp += 2; | |
1396 | else if (*bp == '/' && (newline_fix (bp + 1), bp[1]) == '*') { | |
1397 | bp += 2; | |
1398 | while (!(*bp == '*' && (newline_fix (bp + 1), bp[1]) == '/')) | |
1399 | bp++; | |
1400 | bp += 1; | |
1401 | } | |
1402 | else if (cplusplus && *bp == '/' && (newline_fix (bp + 1), bp[1]) == '/') { | |
1403 | bp += 2; | |
1404 | while (*bp++ != '\n') ; | |
1405 | } | |
1406 | else break; | |
1407 | } | |
1408 | if (bp + 1 != ibp) | |
1409 | goto randomchar; | |
1410 | } | |
1411 | ||
1412 | /* This # can start a directive. */ | |
1413 | ||
1414 | --obp; /* Don't copy the '#' */ | |
1415 | ||
1416 | ip->bufp = ibp; | |
1417 | op->bufp = obp; | |
1418 | if (! handle_directive (ip, op)) { | |
1419 | #ifdef USE_C_ALLOCA | |
1420 | alloca (0); | |
1421 | #endif | |
1422 | /* Not a known directive: treat it as ordinary text. | |
1423 | IP, OP, IBP, etc. have not been changed. */ | |
1424 | if (no_output && instack[indepth].fname) { | |
1425 | /* If not generating expanded output, | |
1426 | what we do with ordinary text is skip it. | |
1427 | Discard everything until next # directive. */ | |
1428 | skip_if_group (&instack[indepth], 1); | |
1429 | RECACHE; | |
1430 | beg_of_line = ibp; | |
1431 | break; | |
1432 | } | |
1433 | ++obp; /* Copy the '#' after all */ | |
1434 | goto randomchar; | |
1435 | } | |
1436 | #ifdef USE_C_ALLOCA | |
1437 | alloca (0); | |
1438 | #endif | |
1439 | /* A # directive has been successfully processed. */ | |
1440 | /* If not generating expanded output, ignore everything until | |
1441 | next # directive. */ | |
1442 | if (no_output && instack[indepth].fname) | |
1443 | skip_if_group (&instack[indepth], 1); | |
1444 | obp = op->bufp; | |
1445 | RECACHE; | |
1446 | beg_of_line = ibp; | |
1447 | break; | |
1448 | ||
1449 | case '\"': /* skip quoted string */ | |
1450 | case '\'': | |
1451 | /* A single quoted string is treated like a double -- some | |
1452 | programs (e.g., troff) are perverse this way */ | |
1453 | ||
1454 | if (ident_length) | |
1455 | goto specialchar; | |
1456 | ||
1457 | start_line = ip->lineno; | |
1458 | ||
1459 | /* Skip ahead to a matching quote. */ | |
1460 | ||
1461 | while (1) { | |
1462 | if (ibp >= limit) { | |
1463 | if (traditional) { | |
1464 | if (ip->macro != 0) { | |
1465 | /* try harder: this string crosses a macro expansion boundary */ | |
1466 | POPMACRO; | |
1467 | RECACHE; | |
1468 | continue; | |
1469 | } | |
1470 | } else | |
1471 | error_with_line (line_for_error (start_line), | |
1472 | "unterminated string or character constant"); | |
1473 | break; | |
1474 | } | |
1475 | *obp++ = *ibp; | |
1476 | switch (*ibp++) { | |
1477 | case '\n': | |
1478 | ++ip->lineno; | |
1479 | ++op->lineno; | |
1480 | /* Traditionally, end of line ends a string constant with no error. | |
1481 | So exit the loop and record the new line. */ | |
1482 | if (traditional) { | |
1483 | beg_of_line = ibp; | |
1484 | goto while2end; | |
1485 | } | |
1486 | if (pedantic || c == '\'') { | |
1487 | error_with_line (line_for_error (start_line), | |
1488 | "unterminated string or character constant"); | |
1489 | goto while2end; | |
1490 | } | |
1491 | break; | |
1492 | ||
1493 | case '\\': | |
1494 | if (ibp >= limit) | |
1495 | break; | |
1496 | if (*ibp == '\n') { | |
1497 | /* Backslash newline is replaced by nothing at all, | |
1498 | but keep the line counts correct. */ | |
1499 | --obp; | |
1500 | ++ibp; | |
1501 | ++ip->lineno; | |
1502 | } else { | |
1503 | /* ANSI stupidly requires that in \\ the second \ | |
1504 | is *not* prevented from combining with a newline. */ | |
1505 | while (*ibp == '\\' && ibp[1] == '\n') { | |
1506 | ibp += 2; | |
1507 | ++ip->lineno; | |
1508 | } | |
1509 | *obp++ = *ibp++; | |
1510 | } | |
1511 | break; | |
1512 | ||
1513 | case '\"': | |
1514 | case '\'': | |
1515 | if (ibp[-1] == c) | |
1516 | goto while2end; | |
1517 | break; | |
1518 | } | |
1519 | } | |
1520 | while2end: | |
1521 | break; | |
1522 | ||
1523 | case '/': | |
1524 | if (*ibp == '\\' && ibp[1] == '\n') | |
1525 | newline_fix (ibp); | |
1526 | /* Don't look for comments inside a macro definition. */ | |
1527 | if (ip->macro != 0) | |
1528 | goto randomchar; | |
1529 | /* A comment constitutes white space, so it can terminate an identifier. | |
1530 | Process the identifier, if any. */ | |
1531 | if (ident_length) | |
1532 | goto specialchar; | |
1533 | if (cplusplus && *ibp == '/') { | |
1534 | /* C++ style comment... */ | |
1535 | start_line = ip->lineno; | |
1536 | ||
1537 | --ibp; /* Back over the slash */ | |
1538 | --obp; | |
1539 | ||
1540 | /* Comments are equivalent to spaces. */ | |
1541 | if (! put_out_comments) | |
1542 | *obp++ = ' '; | |
1543 | else { | |
1544 | /* must fake up a comment here */ | |
1545 | *obp++ = '/'; | |
1546 | *obp++ = '/'; | |
1547 | } | |
1548 | { | |
1549 | U_CHAR *before_bp = ibp+2; | |
1550 | ||
1551 | while (ibp < limit) { | |
1552 | if (*ibp == '\\' && ibp[1] == '\n') { | |
1553 | ip->lineno++; | |
1554 | ibp += 2; | |
1555 | } else if (*ibp++ == '\n') { | |
1556 | ibp--; | |
1557 | if (put_out_comments) { | |
1558 | bcopy (before_bp, obp, ibp - before_bp); | |
1559 | obp += ibp - before_bp; | |
1560 | } | |
1561 | break; | |
1562 | } | |
1563 | } | |
1564 | break; | |
1565 | } | |
1566 | } | |
1567 | if (*ibp != '*') | |
1568 | goto randomchar; | |
1569 | ||
1570 | /* We have a comment. Skip it, optionally copying it to output. */ | |
1571 | ||
1572 | start_line = ip->lineno; | |
1573 | ||
1574 | ++ibp; /* Skip the star. */ | |
1575 | ||
1576 | /* Comments are equivalent to spaces. | |
1577 | Note that we already output the slash; we might not want it. | |
1578 | For -traditional, a comment is equivalent to nothing. */ | |
1579 | if (! put_out_comments) { | |
1580 | if (traditional) | |
1581 | obp--; | |
1582 | else | |
1583 | obp[-1] = ' '; | |
1584 | } | |
1585 | else | |
1586 | *obp++ = '*'; | |
1587 | ||
1588 | { | |
1589 | U_CHAR *before_bp = ibp; | |
1590 | ||
1591 | while (ibp < limit) { | |
1592 | switch (*ibp++) { | |
1593 | case '/': | |
1594 | if (warn_comments && ibp < limit && *ibp == '*') | |
1595 | warning("`/*' within comment"); | |
1596 | break; | |
1597 | case '*': | |
1598 | if (*ibp == '\\' && ibp[1] == '\n') | |
1599 | newline_fix (ibp); | |
1600 | if (ibp >= limit || *ibp == '/') | |
1601 | goto comment_end; | |
1602 | break; | |
1603 | case '\n': | |
1604 | ++ip->lineno; | |
1605 | /* Copy the newline into the output buffer, in order to | |
1606 | avoid the pain of a #line every time a multiline comment | |
1607 | is seen. */ | |
1608 | if (!put_out_comments) | |
1609 | *obp++ = '\n'; | |
1610 | ++op->lineno; | |
1611 | } | |
1612 | } | |
1613 | comment_end: | |
1614 | ||
1615 | if (ibp >= limit) | |
1616 | error_with_line (line_for_error (start_line), | |
1617 | "unterminated comment"); | |
1618 | else { | |
1619 | ibp++; | |
1620 | if (put_out_comments) { | |
1621 | bcopy (before_bp, obp, ibp - before_bp); | |
1622 | obp += ibp - before_bp; | |
1623 | } | |
1624 | } | |
1625 | } | |
1626 | break; | |
1627 | ||
1628 | case '$': | |
1629 | if (!dollars_in_ident) | |
1630 | goto randomchar; | |
1631 | goto letter; | |
1632 | ||
1633 | case '0': case '1': case '2': case '3': case '4': | |
1634 | case '5': case '6': case '7': case '8': case '9': | |
1635 | /* If digit is not part of identifier, it starts a number, | |
1636 | which means that following letters are not an identifier. | |
1637 | "0x5" does not refer to an identifier "x5". | |
1638 | So copy all alphanumerics that follow without accumulating | |
1639 | as an identifier. Periods also, for sake of "3.e7". */ | |
1640 | ||
1641 | if (ident_length == 0) { | |
1642 | while (ibp < limit) { | |
1643 | while (ibp < limit && ibp[0] == '\\' && ibp[1] == '\n') { | |
1644 | ++ip->lineno; | |
1645 | ibp += 2; | |
1646 | } | |
1647 | c = *ibp++; | |
1648 | if (!isalnum (c) && c != '.' && c != '_') { | |
1649 | --ibp; | |
1650 | break; | |
1651 | } | |
1652 | *obp++ = c; | |
1653 | /* A sign can be part of a preprocessing number | |
1654 | if it follows an e. */ | |
1655 | if (c == 'e' || c == 'E') { | |
1656 | while (ibp < limit && ibp[0] == '\\' && ibp[1] == '\n') { | |
1657 | ++ip->lineno; | |
1658 | ibp += 2; | |
1659 | } | |
1660 | if (ibp < limit && (*ibp == '+' || *ibp == '-')) { | |
1661 | *obp++ = *ibp++; | |
1662 | /* But traditional C does not let the token go past the sign. */ | |
1663 | if (traditional) | |
1664 | break; | |
1665 | } | |
1666 | } | |
1667 | } | |
1668 | break; | |
1669 | } | |
1670 | /* fall through */ | |
1671 | ||
1672 | case '_': | |
1673 | case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': | |
1674 | case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': | |
1675 | case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': | |
1676 | case 's': case 't': case 'u': case 'v': case 'w': case 'x': | |
1677 | case 'y': case 'z': | |
1678 | case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': | |
1679 | case 'G': case 'H': case 'I': case 'J': case 'K': case 'L': | |
1680 | case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': | |
1681 | case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': | |
1682 | case 'Y': case 'Z': | |
1683 | letter: | |
1684 | ident_length++; | |
1685 | /* Compute step of hash function, to avoid a proc call on every token */ | |
1686 | hash = HASHSTEP (hash, c); | |
1687 | break; | |
1688 | ||
1689 | case '\n': | |
1690 | /* If reprocessing a macro expansion, newline is a special marker. */ | |
1691 | if (ip->macro != 0) { | |
1692 | /* Newline White is a "funny space" to separate tokens that are | |
1693 | supposed to be separate but without space between. | |
1694 | Here White means any horizontal whitespace character. | |
1695 | Newline - marks a recursive macro use that is not | |
1696 | supposed to be expandable. */ | |
1697 | ||
1698 | if (*ibp == '-') { | |
1699 | /* Newline - inhibits expansion of preceding token. | |
1700 | If expanding a macro arg, we keep the newline -. | |
1701 | In final output, it is deleted. */ | |
1702 | if (! concatenated) { | |
1703 | ident_length = 0; | |
1704 | hash = 0; | |
1705 | } | |
1706 | ibp++; | |
1707 | if (!output_marks) { | |
1708 | obp--; | |
1709 | } else { | |
1710 | /* If expanding a macro arg, keep the newline -. */ | |
1711 | *obp++ = '-'; | |
1712 | } | |
1713 | } else if (is_space[*ibp]) { | |
1714 | /* Newline Space does not prevent expansion of preceding token | |
1715 | so expand the preceding token and then come back. */ | |
1716 | if (ident_length > 0) | |
1717 | goto specialchar; | |
1718 | ||
1719 | /* If generating final output, newline space makes a space. */ | |
1720 | if (!output_marks) { | |
1721 | obp[-1] = *ibp++; | |
1722 | /* And Newline Newline makes a newline, so count it. */ | |
1723 | if (obp[-1] == '\n') | |
1724 | op->lineno++; | |
1725 | } else { | |
1726 | /* If expanding a macro arg, keep the newline space. | |
1727 | If the arg gets stringified, newline space makes nothing. */ | |
1728 | *obp++ = *ibp++; | |
1729 | } | |
1730 | } else abort (); /* Newline followed by something random? */ | |
1731 | break; | |
1732 | } | |
1733 | ||
1734 | /* If there is a pending identifier, handle it and come back here. */ | |
1735 | if (ident_length > 0) | |
1736 | goto specialchar; | |
1737 | ||
1738 | beg_of_line = ibp; | |
1739 | ||
1740 | /* Update the line counts and output a #line if necessary. */ | |
1741 | ++ip->lineno; | |
1742 | ++op->lineno; | |
1743 | if (ip->lineno != op->lineno) { | |
1744 | op->bufp = obp; | |
1745 | output_line_command (ip, op, 1, same_file); | |
1746 | check_expand (op, ip->length - (ip->bufp - ip->buf)); | |
1747 | obp = op->bufp; | |
1748 | } | |
1749 | break; | |
1750 | ||
1751 | /* Come here either after (1) a null character that is part of the input | |
1752 | or (2) at the end of the input, because there is a null there. */ | |
1753 | case 0: | |
1754 | if (ibp <= limit) | |
1755 | /* Our input really contains a null character. */ | |
1756 | goto randomchar; | |
1757 | ||
1758 | /* At end of a macro-expansion level, pop it and read next level. */ | |
1759 | if (ip->macro != 0) { | |
1760 | obp--; | |
1761 | ibp--; | |
1762 | /* If traditional, and we have an identifier that ends here, | |
1763 | process it now, so we get the right error for recursion. */ | |
1764 | if (traditional && ident_length | |
1765 | && ! is_idchar[*instack[indepth - 1].bufp]) { | |
1766 | redo_char = 1; | |
1767 | goto randomchar; | |
1768 | } | |
1769 | POPMACRO; | |
1770 | RECACHE; | |
1771 | break; | |
1772 | } | |
1773 | ||
1774 | /* If we don't have a pending identifier, | |
1775 | return at end of input. */ | |
1776 | if (ident_length == 0) { | |
1777 | obp--; | |
1778 | ibp--; | |
1779 | op->bufp = obp; | |
1780 | ip->bufp = ibp; | |
1781 | goto ending; | |
1782 | } | |
1783 | ||
1784 | /* If we do have a pending identifier, just consider this null | |
1785 | a special character and arrange to dispatch on it again. | |
1786 | The second time, IDENT_LENGTH will be zero so we will return. */ | |
1787 | ||
1788 | /* Fall through */ | |
1789 | ||
1790 | specialchar: | |
1791 | ||
1792 | /* Handle the case of a character such as /, ', " or null | |
1793 | seen following an identifier. Back over it so that | |
1794 | after the identifier is processed the special char | |
1795 | will be dispatched on again. */ | |
1796 | ||
1797 | ibp--; | |
1798 | obp--; | |
1799 | redo_char = 1; | |
1800 | ||
1801 | default: | |
1802 | ||
1803 | randomchar: | |
1804 | ||
1805 | if (ident_length > 0) { | |
1806 | register HASHNODE *hp; | |
1807 | ||
1808 | /* We have just seen an identifier end. If it's a macro, expand it. | |
1809 | ||
1810 | IDENT_LENGTH is the length of the identifier | |
1811 | and HASH is its hash code. | |
1812 | ||
1813 | The identifier has already been copied to the output, | |
1814 | so if it is a macro we must remove it. | |
1815 | ||
1816 | If REDO_CHAR is 0, the char that terminated the identifier | |
1817 | has been skipped in the output and the input. | |
1818 | OBP-IDENT_LENGTH-1 points to the identifier. | |
1819 | If the identifier is a macro, we must back over the terminator. | |
1820 | ||
1821 | If REDO_CHAR is 1, the terminating char has already been | |
1822 | backed over. OBP-IDENT_LENGTH points to the identifier. */ | |
1823 | ||
1824 | for (hp = hashtab[MAKE_POS (hash) % HASHSIZE]; hp != NULL; | |
1825 | hp = hp->next) { | |
1826 | ||
1827 | if (hp->length == ident_length) { | |
1828 | U_CHAR *obufp_before_macroname; | |
1829 | int op_lineno_before_macroname; | |
1830 | register int i = ident_length; | |
1831 | register U_CHAR *p = hp->name; | |
1832 | register U_CHAR *q = obp - i; | |
1833 | int disabled; | |
1834 | ||
1835 | if (! redo_char) | |
1836 | q--; | |
1837 | ||
1838 | do { /* All this to avoid a strncmp () */ | |
1839 | if (*p++ != *q++) | |
1840 | goto hashcollision; | |
1841 | } while (--i); | |
1842 | ||
1843 | /* We found a use of a macro name. | |
1844 | see if the context shows it is a macro call. */ | |
1845 | ||
1846 | /* Back up over terminating character if not already done. */ | |
1847 | if (! redo_char) { | |
1848 | ibp--; | |
1849 | obp--; | |
1850 | } | |
1851 | ||
1852 | obufp_before_macroname = obp - ident_length; | |
1853 | op_lineno_before_macroname = op->lineno; | |
1854 | ||
1855 | /* Record whether the macro is disabled. */ | |
1856 | disabled = hp->type == T_DISABLED; | |
1857 | ||
1858 | /* This looks like a macro ref, but if the macro was disabled, | |
1859 | just copy its name and put in a marker if requested. */ | |
1860 | ||
1861 | if (disabled) { | |
1862 | #if 0 | |
1863 | /* This error check caught useful cases such as | |
1864 | #define foo(x,y) bar(x(y,0), y) | |
1865 | foo(foo, baz) */ | |
1866 | if (traditional) | |
1867 | error ("recursive use of macro `%s'", hp->name); | |
1868 | #endif | |
1869 | ||
1870 | if (output_marks) { | |
1871 | check_expand (op, limit - ibp + 2); | |
1872 | *obp++ = '\n'; | |
1873 | *obp++ = '-'; | |
1874 | } | |
1875 | break; | |
1876 | } | |
1877 | ||
1878 | /* If macro wants an arglist, verify that a '(' follows. | |
1879 | first skip all whitespace, copying it to the output | |
1880 | after the macro name. Then, if there is no '(', | |
1881 | decide this is not a macro call and leave things that way. */ | |
1882 | if ((hp->type == T_MACRO || hp->type == T_DISABLED) | |
1883 | && hp->value.defn->nargs >= 0) | |
1884 | { | |
1885 | while (1) { | |
1886 | /* Scan forward over whitespace, copying it to the output. */ | |
1887 | if (ibp == limit && ip->macro != 0) { | |
1888 | POPMACRO; | |
1889 | RECACHE; | |
1890 | } | |
1891 | /* A comment: copy it unchanged or discard it. */ | |
1892 | else if (*ibp == '/' && ibp+1 != limit && ibp[1] == '*') { | |
1893 | if (put_out_comments) { | |
1894 | *obp++ = '/'; | |
1895 | *obp++ = '*'; | |
1896 | } else if (! traditional) { | |
1897 | *obp++ = ' '; | |
1898 | } | |
1899 | ibp += 2; | |
1900 | while (ibp + 1 != limit | |
1901 | && !(ibp[0] == '*' && ibp[1] == '/')) { | |
1902 | /* We need not worry about newline-marks, | |
1903 | since they are never found in comments. */ | |
1904 | if (*ibp == '\n') { | |
1905 | /* Newline in a file. Count it. */ | |
1906 | ++ip->lineno; | |
1907 | ++op->lineno; | |
1908 | } | |
1909 | if (put_out_comments) | |
1910 | *obp++ = *ibp++; | |
1911 | else | |
1912 | ibp++; | |
1913 | } | |
1914 | ibp += 2; | |
1915 | if (put_out_comments) { | |
1916 | *obp++ = '*'; | |
1917 | *obp++ = '/'; | |
1918 | } | |
1919 | } | |
1920 | else if (is_space[*ibp]) { | |
1921 | *obp++ = *ibp++; | |
1922 | if (ibp[-1] == '\n') { | |
1923 | if (ip->macro == 0) { | |
1924 | /* Newline in a file. Count it. */ | |
1925 | ++ip->lineno; | |
1926 | ++op->lineno; | |
1927 | } else if (!output_marks) { | |
1928 | /* A newline mark, and we don't want marks | |
1929 | in the output. If it is newline-hyphen, | |
1930 | discard it entirely. Otherwise, it is | |
1931 | newline-whitechar, so keep the whitechar. */ | |
1932 | obp--; | |
1933 | if (*ibp == '-') | |
1934 | ibp++; | |
1935 | else { | |
1936 | if (*ibp == '\n') | |
1937 | ++op->lineno; | |
1938 | *obp++ = *ibp++; | |
1939 | } | |
1940 | } else { | |
1941 | /* A newline mark; copy both chars to the output. */ | |
1942 | *obp++ = *ibp++; | |
1943 | } | |
1944 | } | |
1945 | } | |
1946 | else break; | |
1947 | } | |
1948 | if (*ibp != '(') | |
1949 | break; | |
1950 | } | |
1951 | ||
1952 | /* This is now known to be a macro call. | |
1953 | Discard the macro name from the output, | |
1954 | along with any following whitespace just copied. */ | |
1955 | obp = obufp_before_macroname; | |
1956 | op->lineno = op_lineno_before_macroname; | |
1957 | ||
1958 | /* Expand the macro, reading arguments as needed, | |
1959 | and push the expansion on the input stack. */ | |
1960 | ip->bufp = ibp; | |
1961 | op->bufp = obp; | |
1962 | macroexpand (hp, op); | |
1963 | ||
1964 | /* Reexamine input stack, since macroexpand has pushed | |
1965 | a new level on it. */ | |
1966 | obp = op->bufp; | |
1967 | RECACHE; | |
1968 | break; | |
1969 | } | |
1970 | hashcollision: | |
1971 | ; | |
1972 | } /* End hash-table-search loop */ | |
1973 | ident_length = hash = 0; /* Stop collecting identifier */ | |
1974 | redo_char = 0; | |
1975 | concatenated = 0; | |
1976 | } /* End if (ident_length > 0) */ | |
1977 | } /* End switch */ | |
1978 | } /* End per-char loop */ | |
1979 | ||
1980 | /* Come here to return -- but first give an error message | |
1981 | if there was an unterminated successful conditional. */ | |
1982 | ending: | |
1983 | if (if_stack != ip->if_stack) { | |
1984 | char *str; | |
1985 | switch (if_stack->type) { | |
1986 | case T_IF: | |
1987 | str = "if"; | |
1988 | break; | |
1989 | case T_IFDEF: | |
1990 | str = "ifdef"; | |
1991 | break; | |
1992 | case T_IFNDEF: | |
1993 | str = "ifndef"; | |
1994 | break; | |
1995 | case T_ELSE: | |
1996 | str = "else"; | |
1997 | break; | |
1998 | case T_ELIF: | |
1999 | str = "elif"; | |
2000 | break; | |
2001 | } | |
2002 | error_with_line (line_for_error (if_stack->lineno), | |
2003 | "unterminated #%s conditional", str); | |
2004 | } | |
2005 | if_stack = ip->if_stack; | |
2006 | } | |
2007 | \f | |
2008 | /* | |
2009 | * Rescan a string into a temporary buffer and return the result | |
2010 | * as a FILE_BUF. Note this function returns a struct, not a pointer. | |
2011 | * | |
2012 | * OUTPUT_MARKS nonzero means keep Newline markers found in the input | |
2013 | * and insert such markers when appropriate. See `rescan' for details. | |
2014 | * OUTPUT_MARKS is 1 for macroexpanding a macro argument separately | |
2015 | * before substitution; it is 0 for other uses. | |
2016 | */ | |
2017 | FILE_BUF | |
2018 | expand_to_temp_buffer (buf, limit, output_marks) | |
2019 | U_CHAR *buf, *limit; | |
2020 | int output_marks; | |
2021 | { | |
2022 | register FILE_BUF *ip; | |
2023 | FILE_BUF obuf; | |
2024 | int length = limit - buf; | |
2025 | U_CHAR *buf1; | |
2026 | int odepth = indepth; | |
2027 | ||
2028 | if (length < 0) | |
2029 | abort (); | |
2030 | ||
2031 | /* Set up the input on the input stack. */ | |
2032 | ||
2033 | buf1 = (U_CHAR *) alloca (length + 1); | |
2034 | { | |
2035 | register U_CHAR *p1 = buf; | |
2036 | register U_CHAR *p2 = buf1; | |
2037 | ||
2038 | while (p1 != limit) | |
2039 | *p2++ = *p1++; | |
2040 | } | |
2041 | buf1[length] = 0; | |
2042 | ||
2043 | /* Set up to receive the output. */ | |
2044 | ||
2045 | obuf.length = length * 2 + 100; /* Usually enough. Why be stingy? */ | |
2046 | obuf.bufp = obuf.buf = (U_CHAR *) xmalloc (obuf.length); | |
2047 | obuf.fname = 0; | |
2048 | obuf.macro = 0; | |
2049 | obuf.free_ptr = 0; | |
2050 | ||
2051 | CHECK_DEPTH ({return obuf;}); | |
2052 | ||
2053 | ++indepth; | |
2054 | ||
2055 | ip = &instack[indepth]; | |
2056 | ip->fname = 0; | |
2057 | ip->macro = 0; | |
2058 | ip->free_ptr = 0; | |
2059 | ip->length = length; | |
2060 | ip->buf = ip->bufp = buf1; | |
2061 | ip->if_stack = if_stack; | |
2062 | ||
2063 | ip->lineno = obuf.lineno = 1; | |
2064 | ||
2065 | /* Scan the input, create the output. */ | |
2066 | ||
2067 | rescan (&obuf, output_marks); | |
2068 | ||
2069 | /* Pop input stack to original state. */ | |
2070 | --indepth; | |
2071 | ||
2072 | if (indepth != odepth) | |
2073 | abort (); | |
2074 | ||
2075 | /* Record the output. */ | |
2076 | obuf.length = obuf.bufp - obuf.buf; | |
2077 | ||
2078 | return obuf; | |
2079 | } | |
2080 | \f | |
2081 | /* | |
2082 | * Process a # directive. Expects IP->bufp to point to the '#', as in | |
2083 | * `#define foo bar'. Passes to the command handler | |
2084 | * (do_define, do_include, etc.): the addresses of the 1st and | |
2085 | * last chars of the command (starting immediately after the # | |
2086 | * keyword), plus op and the keyword table pointer. If the command | |
2087 | * contains comments it is copied into a temporary buffer sans comments | |
2088 | * and the temporary buffer is passed to the command handler instead. | |
2089 | * Likewise for backslash-newlines. | |
2090 | * | |
2091 | * Returns nonzero if this was a known # directive. | |
2092 | * Otherwise, returns zero, without advancing the input pointer. | |
2093 | */ | |
2094 | ||
2095 | int | |
2096 | handle_directive (ip, op) | |
2097 | FILE_BUF *ip, *op; | |
2098 | { | |
2099 | register U_CHAR *bp, *cp; | |
2100 | register struct directive *kt; | |
2101 | register int ident_length; | |
2102 | U_CHAR *resume_p; | |
2103 | ||
2104 | /* Nonzero means we must copy the entire command | |
2105 | to get rid of comments or backslash-newlines. */ | |
2106 | int copy_command = 0; | |
2107 | ||
2108 | U_CHAR *ident, *after_ident; | |
2109 | ||
2110 | bp = ip->bufp; | |
2111 | /* Skip whitespace and \-newline. */ | |
2112 | while (1) { | |
2113 | if (is_hor_space[*bp]) | |
2114 | bp++; | |
2115 | else if (*bp == '/' && (newline_fix (bp + 1), bp[1]) == '*') { | |
2116 | ip->bufp = bp; | |
2117 | skip_to_end_of_comment (ip, &ip->lineno); | |
2118 | bp = ip->bufp; | |
2119 | } else if (*bp == '\\' && bp[1] == '\n') { | |
2120 | bp += 2; ip->lineno++; | |
2121 | } else break; | |
2122 | } | |
2123 | ||
2124 | /* Now find end of directive name. | |
2125 | If we encounter a backslash-newline, exchange it with any following | |
2126 | symbol-constituents so that we end up with a contiguous name. */ | |
2127 | ||
2128 | cp = bp; | |
2129 | while (1) { | |
2130 | if (is_idchar[*cp]) | |
2131 | cp++; | |
2132 | else { | |
2133 | if (*cp == '\\' && cp[1] == '\n') | |
2134 | name_newline_fix (cp); | |
2135 | if (is_idchar[*cp]) | |
2136 | cp++; | |
2137 | else break; | |
2138 | } | |
2139 | } | |
2140 | ident_length = cp - bp; | |
2141 | ident = bp; | |
2142 | after_ident = cp; | |
2143 | ||
2144 | /* A line of just `#' becomes blank. */ | |
2145 | ||
2146 | if (ident_length == 0 && *after_ident == '\n') { | |
2147 | ip->bufp = after_ident; | |
2148 | return 1; | |
2149 | } | |
2150 | ||
2151 | /* | |
2152 | * Decode the keyword and call the appropriate expansion | |
2153 | * routine, after moving the input pointer up to the next line. | |
2154 | */ | |
2155 | for (kt = directive_table; kt->length > 0; kt++) { | |
2156 | if (kt->length == ident_length && !strncmp (kt->name, ident, ident_length)) { | |
2157 | register U_CHAR *buf; | |
2158 | register U_CHAR *limit = ip->buf + ip->length; | |
2159 | int unterminated = 0; | |
2160 | ||
2161 | /* Nonzero means do not delete comments within the directive. | |
2162 | #define needs this when -traditional. */ | |
2163 | int keep_comments = traditional && kt->traditional_comments; | |
2164 | ||
2165 | /* Find the end of this command (first newline not backslashed | |
2166 | and not in a string or comment). | |
2167 | Set COPY_COMMAND if the command must be copied | |
2168 | (it contains a backslash-newline or a comment). */ | |
2169 | ||
2170 | buf = bp = after_ident; | |
2171 | while (bp < limit) { | |
2172 | register U_CHAR c = *bp++; | |
2173 | switch (c) { | |
2174 | case '\\': | |
2175 | if (bp < limit) { | |
2176 | if (*bp == '\n') { | |
2177 | ip->lineno++; | |
2178 | copy_command = 1; | |
2179 | } | |
2180 | bp++; | |
2181 | } | |
2182 | break; | |
2183 | ||
2184 | case '\'': | |
2185 | case '\"': | |
2186 | bp = skip_quoted_string (bp - 1, limit, ip->lineno, &ip->lineno, ©_command, &unterminated); | |
2187 | /* Don't bother calling the directive if we already got an error | |
2188 | message due to unterminated string. Skip everything and pretend | |
2189 | we called the directive. */ | |
2190 | if (unterminated) { | |
2191 | if (traditional) { | |
2192 | /* Traditional preprocessing permits unterminated strings. */ | |
2193 | ip->bufp = bp; | |
2194 | goto endloop1; | |
2195 | } | |
2196 | ip->bufp = bp; | |
2197 | return 1; | |
2198 | } | |
2199 | break; | |
2200 | ||
2201 | /* <...> is special for #include. */ | |
2202 | case '<': | |
2203 | if (!kt->angle_brackets) | |
2204 | break; | |
2205 | while (*bp && *bp != '>') bp++; | |
2206 | break; | |
2207 | ||
2208 | case '/': | |
2209 | if (*bp == '\\' && bp[1] == '\n') | |
2210 | newline_fix (bp); | |
2211 | if (*bp == '*' | |
2212 | || (cplusplus && *bp == '/')) { | |
2213 | U_CHAR *obp = bp - 1; | |
2214 | ip->bufp = bp + 1; | |
2215 | skip_to_end_of_comment (ip, &ip->lineno); | |
2216 | bp = ip->bufp; | |
2217 | /* No need to copy the command because of a comment at the end; | |
2218 | just don't include the comment in the directive. */ | |
2219 | if (bp == limit || *bp == '\n') { | |
2220 | bp = obp; | |
2221 | goto endloop1; | |
2222 | } | |
2223 | /* Don't remove the comments if -traditional. */ | |
2224 | if (! keep_comments) | |
2225 | copy_command++; | |
2226 | } | |
2227 | break; | |
2228 | ||
2229 | case '\n': | |
2230 | --bp; /* Point to the newline */ | |
2231 | ip->bufp = bp; | |
2232 | goto endloop1; | |
2233 | } | |
2234 | } | |
2235 | ip->bufp = bp; | |
2236 | ||
2237 | endloop1: | |
2238 | resume_p = ip->bufp; | |
2239 | /* BP is the end of the directive. | |
2240 | RESUME_P is the next interesting data after the directive. | |
2241 | A comment may come between. */ | |
2242 | ||
2243 | if (copy_command) { | |
2244 | register U_CHAR *xp = buf; | |
2245 | /* Need to copy entire command into temp buffer before dispatching */ | |
2246 | ||
2247 | cp = (U_CHAR *) alloca (bp - buf + 5); /* room for cmd plus | |
2248 | some slop */ | |
2249 | buf = cp; | |
2250 | ||
2251 | /* Copy to the new buffer, deleting comments | |
2252 | and backslash-newlines (and whitespace surrounding the latter). */ | |
2253 | ||
2254 | while (xp < bp) { | |
2255 | register U_CHAR c = *xp++; | |
2256 | *cp++ = c; | |
2257 | ||
2258 | switch (c) { | |
2259 | case '\n': | |
2260 | break; | |
2261 | ||
2262 | /* <...> is special for #include. */ | |
2263 | case '<': | |
2264 | if (!kt->angle_brackets) | |
2265 | break; | |
2266 | while (xp < bp && c != '>') { | |
2267 | c = *xp++; | |
2268 | if (c == '\\' && xp < bp && *xp == '\n') | |
2269 | xp++, ip->lineno++; | |
2270 | else | |
2271 | *cp++ = c; | |
2272 | } | |
2273 | break; | |
2274 | ||
2275 | case '\\': | |
2276 | if (*xp == '\n') { | |
2277 | xp++; | |
2278 | cp--; | |
2279 | if (cp != buf && is_space[cp[-1]]) { | |
2280 | while (cp != buf && is_space[cp[-1]]) cp--; | |
2281 | cp++; | |
2282 | SKIP_WHITE_SPACE (xp); | |
2283 | } else if (is_space[*xp]) { | |
2284 | *cp++ = *xp++; | |
2285 | SKIP_WHITE_SPACE (xp); | |
2286 | } | |
2287 | } | |
2288 | break; | |
2289 | ||
2290 | case '\'': | |
2291 | case '\"': | |
2292 | { | |
2293 | register U_CHAR *bp1 | |
2294 | = skip_quoted_string (xp - 1, limit, ip->lineno, 0, 0, 0); | |
2295 | while (xp != bp1) | |
2296 | *cp++ = *xp++; | |
2297 | } | |
2298 | break; | |
2299 | ||
2300 | case '/': | |
2301 | if (*xp == '*' | |
2302 | || (cplusplus && *xp == '/')) { | |
2303 | ip->bufp = xp + 1; | |
2304 | skip_to_end_of_comment (ip, 0); | |
2305 | if (keep_comments) | |
2306 | while (xp != ip->bufp) | |
2307 | *cp++ = *xp++; | |
2308 | /* Delete or replace the slash. */ | |
2309 | else if (traditional) | |
2310 | cp--; | |
2311 | else | |
2312 | cp[-1] = ' '; | |
2313 | xp = ip->bufp; | |
2314 | } | |
2315 | } | |
2316 | } | |
2317 | ||
2318 | /* Null-terminate the copy. */ | |
2319 | ||
2320 | *cp = 0; | |
2321 | } | |
2322 | else | |
2323 | cp = bp; | |
2324 | ||
2325 | ip->bufp = resume_p; | |
2326 | ||
2327 | /* Some directives should be written out for cc1 to process, | |
2328 | just as if they were not defined. */ | |
2329 | ||
2330 | if (kt->pass_thru) { | |
2331 | int len; | |
2332 | ||
2333 | /* Output directive name. */ | |
2334 | check_expand (op, kt->length+1); | |
2335 | *op->bufp++ = '#'; | |
2336 | bcopy (kt->name, op->bufp, kt->length); | |
2337 | op->bufp += kt->length; | |
2338 | ||
2339 | /* Output arguments. */ | |
2340 | len = (cp - buf); | |
2341 | check_expand (op, len); | |
2342 | bcopy (buf, op->bufp, len); | |
2343 | op->bufp += len; | |
2344 | } | |
2345 | ||
2346 | /* Call the appropriate command handler. buf now points to | |
2347 | either the appropriate place in the input buffer, or to | |
2348 | the temp buffer if it was necessary to make one. cp | |
2349 | points to the first char after the contents of the (possibly | |
2350 | copied) command, in either case. */ | |
2351 | (*kt->func) (buf, cp, op, kt); | |
2352 | check_expand (op, ip->length - (ip->bufp - ip->buf)); | |
2353 | ||
2354 | return 1; | |
2355 | } | |
2356 | } | |
2357 | ||
2358 | return 0; | |
2359 | } | |
2360 | \f | |
2361 | static char *monthnames[] = {"Jan", "Feb", "Mar", "Apr", "May", "Jun", | |
2362 | "Jul", "Aug", "Sep", "Oct", "Nov", "Dec", | |
2363 | }; | |
2364 | ||
2365 | /* | |
2366 | * expand things like __FILE__. Place the expansion into the output | |
2367 | * buffer *without* rescanning. | |
2368 | */ | |
2369 | special_symbol (hp, op) | |
2370 | HASHNODE *hp; | |
2371 | FILE_BUF *op; | |
2372 | { | |
2373 | char *buf; | |
2374 | time_t t; | |
2375 | int i, len; | |
2376 | int true_indepth; | |
2377 | FILE_BUF *ip = NULL; | |
2378 | static struct tm *timebuf = NULL; | |
2379 | struct tm *localtime (); | |
2380 | ||
2381 | int paren = 0; /* For special `defined' keyword */ | |
2382 | ||
2383 | for (i = indepth; i >= 0; i--) | |
2384 | if (instack[i].fname != NULL) { | |
2385 | ip = &instack[i]; | |
2386 | break; | |
2387 | } | |
2388 | if (ip == NULL) { | |
2389 | error ("cccp error: not in any file?!"); | |
2390 | return; /* the show must go on */ | |
2391 | } | |
2392 | ||
2393 | switch (hp->type) { | |
2394 | case T_FILE: | |
2395 | case T_BASE_FILE: | |
2396 | { | |
2397 | char *string; | |
2398 | if (hp->type == T_FILE) | |
2399 | string = ip->fname; | |
2400 | else | |
2401 | string = instack[0].fname; | |
2402 | ||
2403 | if (string) | |
2404 | { | |
2405 | buf = (char *) alloca (3 + strlen (string)); | |
2406 | sprintf (buf, "\"%s\"", string); | |
2407 | } | |
2408 | else | |
2409 | buf = "\"\""; | |
2410 | ||
2411 | break; | |
2412 | } | |
2413 | ||
2414 | case T_INCLUDE_LEVEL: | |
2415 | true_indepth = 0; | |
2416 | for (i = indepth; i >= 0; i--) | |
2417 | if (instack[i].fname != NULL) | |
2418 | true_indepth++; | |
2419 | ||
2420 | buf = (char *) alloca (8); /* Eigth bytes ought to be more than enough */ | |
2421 | sprintf (buf, "%d", true_indepth - 1); | |
2422 | break; | |
2423 | ||
2424 | case T_VERSION: | |
2425 | buf = (char *) alloca (3 + strlen (version_string)); | |
2426 | sprintf (buf, "\"%s\"", version_string); | |
2427 | break; | |
2428 | ||
2429 | case T_CONST: | |
2430 | buf = (char *) alloca (4 * sizeof (int)); | |
2431 | sprintf (buf, "%d", hp->value.ival); | |
2432 | break; | |
2433 | ||
2434 | case T_SPECLINE: | |
2435 | buf = (char *) alloca (10); | |
2436 | sprintf (buf, "%d", ip->lineno); | |
2437 | break; | |
2438 | ||
2439 | case T_DATE: | |
2440 | case T_TIME: | |
2441 | if (timebuf == NULL) { | |
2442 | t = time (0); | |
2443 | timebuf = localtime (&t); | |
2444 | } | |
2445 | buf = (char *) alloca (20); | |
2446 | if (hp->type == T_DATE) | |
2447 | sprintf (buf, "\"%s %2d %4d\"", monthnames[timebuf->tm_mon], | |
2448 | timebuf->tm_mday, timebuf->tm_year + 1900); | |
2449 | else | |
2450 | sprintf (buf, "\"%02d:%02d:%02d\"", timebuf->tm_hour, timebuf->tm_min, | |
2451 | timebuf->tm_sec); | |
2452 | break; | |
2453 | ||
2454 | case T_SPEC_DEFINED: | |
2455 | buf = " 0 "; /* Assume symbol is not defined */ | |
2456 | ip = &instack[indepth]; | |
2457 | SKIP_WHITE_SPACE (ip->bufp); | |
2458 | if (*ip->bufp == '(') { | |
2459 | paren++; | |
2460 | ip->bufp++; /* Skip over the paren */ | |
2461 | SKIP_WHITE_SPACE (ip->bufp); | |
2462 | } | |
2463 | ||
2464 | if (!is_idstart[*ip->bufp]) | |
2465 | goto oops; | |
2466 | if (lookup (ip->bufp, -1, -1)) | |
2467 | buf = " 1 "; | |
2468 | while (is_idchar[*ip->bufp]) | |
2469 | ++ip->bufp; | |
2470 | SKIP_WHITE_SPACE (ip->bufp); | |
2471 | if (paren) { | |
2472 | if (*ip->bufp != ')') | |
2473 | goto oops; | |
2474 | ++ip->bufp; | |
2475 | } | |
2476 | break; | |
2477 | ||
2478 | oops: | |
2479 | ||
2480 | error ("`defined' must be followed by ident or (ident)"); | |
2481 | break; | |
2482 | ||
2483 | default: | |
2484 | error ("cccp error: invalid special hash type"); /* time for gdb */ | |
2485 | abort (); | |
2486 | } | |
2487 | len = strlen (buf); | |
2488 | check_expand (op, len); | |
2489 | bcopy (buf, op->bufp, len); | |
2490 | op->bufp += len; | |
2491 | ||
2492 | return; | |
2493 | } | |
2494 | ||
2495 | \f | |
2496 | /* Routines to handle #directives */ | |
2497 | ||
2498 | /* | |
2499 | * Process include file by reading it in and calling rescan. | |
2500 | * Expects to see "fname" or <fname> on the input. | |
2501 | */ | |
2502 | ||
2503 | do_include (buf, limit, op, keyword) | |
2504 | U_CHAR *buf, *limit; | |
2505 | FILE_BUF *op; | |
2506 | struct directive *keyword; | |
2507 | { | |
2508 | char *fname; /* Dynamically allocated fname buffer */ | |
2509 | U_CHAR *fbeg, *fend; /* Beginning and end of fname */ | |
2510 | ||
2511 | struct file_name_list *stackp = include; /* Chain of dirs to search */ | |
2512 | struct file_name_list dsp[1]; /* First in chain, if #include "..." */ | |
2513 | int flen; | |
2514 | ||
2515 | int f; /* file number */ | |
2516 | ||
2517 | int retried = 0; /* Have already tried macro | |
2518 | expanding the include line*/ | |
2519 | FILE_BUF trybuf; /* It got expanded into here */ | |
2520 | int system_header_p = 0; /* 0 for "...", 1 for <...> */ | |
2521 | ||
2522 | f= -1; /* JF we iz paranoid! */ | |
2523 | ||
2524 | get_filename: | |
2525 | ||
2526 | fbeg = buf; | |
2527 | SKIP_WHITE_SPACE (fbeg); | |
2528 | /* Discard trailing whitespace so we can easily see | |
2529 | if we have parsed all the significant chars we were given. */ | |
2530 | while (limit != fbeg && is_hor_space[limit[-1]]) limit--; | |
2531 | ||
2532 | switch (*fbeg++) { | |
2533 | case '\"': | |
2534 | fend = fbeg; | |
2535 | while (fend != limit && *fend != '\"') | |
2536 | fend++; | |
2537 | if (*fend == '\"' && fend + 1 == limit) { | |
2538 | FILE_BUF *fp; | |
2539 | ||
2540 | /* We have "filename". Figure out directory this source | |
2541 | file is coming from and put it on the front of the list. */ | |
2542 | ||
2543 | /* If -I- was specified, don't search current dir, only spec'd ones. */ | |
2544 | if (ignore_srcdir) break; | |
2545 | ||
2546 | for (fp = &instack[indepth]; fp >= instack; fp--) | |
2547 | { | |
2548 | int n; | |
2549 | char *ep,*nam; | |
2550 | extern char *rindex (); | |
2551 | ||
2552 | if ((nam = fp->fname) != NULL) { | |
2553 | /* Found a named file. Figure out dir of the file, | |
2554 | and put it in front of the search list. */ | |
2555 | dsp[0].next = stackp; | |
2556 | stackp = dsp; | |
2557 | #ifndef VMS | |
2558 | ep = rindex (nam, '/'); | |
2559 | #else /* VMS */ | |
2560 | ep = rindex (nam, ']'); | |
2561 | if (ep == NULL) ep = rindex (nam, '>'); | |
2562 | if (ep == NULL) ep = rindex (nam, ':'); | |
2563 | if (ep != NULL) ep++; | |
2564 | #endif /* VMS */ | |
2565 | if (ep != NULL) { | |
2566 | n = ep - nam; | |
2567 | dsp[0].fname = (char *) alloca (n + 1); | |
2568 | strncpy (dsp[0].fname, nam, n); | |
2569 | dsp[0].fname[n] = '\0'; | |
2570 | if (n > max_include_len) max_include_len = n; | |
2571 | } else { | |
2572 | dsp[0].fname = 0; /* Current directory */ | |
2573 | } | |
2574 | break; | |
2575 | } | |
2576 | } | |
2577 | break; | |
2578 | } | |
2579 | goto fail; | |
2580 | ||
2581 | case '<': | |
2582 | fend = fbeg; | |
2583 | while (fend != limit && *fend != '>') fend++; | |
2584 | if (*fend == '>' && fend + 1 == limit) { | |
2585 | system_header_p = 1; | |
2586 | /* If -I-, start with the first -I dir after the -I-. */ | |
2587 | if (first_bracket_include) | |
2588 | stackp = first_bracket_include; | |
2589 | break; | |
2590 | } | |
2591 | goto fail; | |
2592 | ||
2593 | default: | |
2594 | fail: | |
2595 | if (retried) { | |
2596 | error ("#include expects \"fname\" or <fname>"); | |
2597 | return; | |
2598 | } else { | |
2599 | trybuf = expand_to_temp_buffer (buf, limit, 0); | |
2600 | buf = (U_CHAR *) alloca (trybuf.bufp - trybuf.buf + 1); | |
2601 | bcopy (trybuf.buf, buf, trybuf.bufp - trybuf.buf); | |
2602 | limit = buf + (trybuf.bufp - trybuf.buf); | |
2603 | free (trybuf.buf); | |
2604 | retried++; | |
2605 | goto get_filename; | |
2606 | } | |
2607 | } | |
2608 | ||
2609 | flen = fend - fbeg; | |
2610 | fname = (char *) alloca (max_include_len + flen + 2); | |
2611 | /* + 2 above for slash and terminating null. */ | |
2612 | ||
2613 | /* If specified file name is absolute, just open it. */ | |
2614 | ||
2615 | if (*fbeg == '/') { | |
2616 | strncpy (fname, fbeg, flen); | |
2617 | fname[flen] = 0; | |
2618 | f = open (fname, O_RDONLY, 0666); | |
2619 | } else { | |
2620 | /* Search directory path, trying to open the file. | |
2621 | Copy each filename tried into FNAME. */ | |
2622 | ||
2623 | for (; stackp; stackp = stackp->next) { | |
2624 | if (stackp->fname) { | |
2625 | strcpy (fname, stackp->fname); | |
2626 | strcat (fname, "/"); | |
2627 | fname[strlen (fname) + flen] = 0; | |
2628 | } else { | |
2629 | fname[0] = 0; | |
2630 | } | |
2631 | strncat (fname, fbeg, flen); | |
2632 | #ifdef VMS | |
2633 | /* Change this 1/2 Unix 1/2 VMS file specification into a | |
2634 | full VMS file specification */ | |
2635 | if (stackp->fname && (stackp->fname[0] != 0)) { | |
2636 | /* Fix up the filename */ | |
2637 | hack_vms_include_specification (fname); | |
2638 | } else { | |
2639 | /* This is a normal VMS filespec, so use it unchanged. */ | |
2640 | strncpy (fname, fbeg, flen); | |
2641 | fname[flen] = 0; | |
2642 | } | |
2643 | #endif /* VMS */ | |
2644 | if ((f = open (fname, O_RDONLY, 0666)) >= 0) | |
2645 | break; | |
2646 | } | |
2647 | } | |
2648 | ||
2649 | if (f < 0) { | |
2650 | strncpy (fname, fbeg, flen); | |
2651 | fname[flen] = 0; | |
2652 | error_from_errno (fname); | |
2653 | ||
2654 | /* For -M, add this file to the dependencies. */ | |
2655 | if (print_deps > (system_header_p || (system_include_depth > 0))) { | |
2656 | if (system_header_p) | |
2657 | warning ("nonexistent file <%.*s> omitted from dependency output", | |
2658 | fend - fbeg, fbeg); | |
2659 | else | |
2660 | { | |
2661 | deps_output (fbeg, fend - fbeg); | |
2662 | deps_output (" ", 0); | |
2663 | } | |
2664 | } | |
2665 | } else { | |
2666 | ||
2667 | /* Check to see if this include file is a once-only include file. | |
2668 | If so, give up. */ | |
2669 | ||
2670 | struct file_name_list* ptr; | |
2671 | ||
2672 | for (ptr = dont_repeat_files; ptr; ptr = ptr->next) { | |
2673 | if (!strcmp (ptr->fname, fname)) { | |
2674 | close (f); | |
2675 | return; /* This file was once'd. */ | |
2676 | } | |
2677 | } | |
2678 | ||
2679 | for (ptr = all_include_files; ptr; ptr = ptr->next) { | |
2680 | if (!strcmp (ptr->fname, fname)) | |
2681 | break; /* This file was included before. */ | |
2682 | } | |
2683 | ||
2684 | if (ptr == 0) { | |
2685 | /* This is the first time for this file. */ | |
2686 | /* Add it to list of files included. */ | |
2687 | ||
2688 | ptr = (struct file_name_list *) xmalloc (sizeof (struct file_name_list)); | |
2689 | ptr->next = all_include_files; | |
2690 | all_include_files = ptr; | |
2691 | ptr->fname = savestring (fname); | |
2692 | ||
2693 | /* For -M, add this file to the dependencies. */ | |
2694 | if (print_deps > (system_header_p || (system_include_depth > 0))) { | |
2695 | deps_output (fname, strlen (fname)); | |
2696 | deps_output (" ", 0); | |
2697 | } | |
2698 | } | |
2699 | ||
2700 | if (system_header_p) | |
2701 | system_include_depth++; | |
2702 | ||
2703 | /* Actually process the file. */ | |
2704 | finclude (f, fname, op); | |
2705 | ||
2706 | if (system_header_p) | |
2707 | system_include_depth--; | |
2708 | ||
2709 | close (f); | |
2710 | } | |
2711 | } | |
2712 | ||
2713 | /* Process the contents of include file FNAME, already open on descriptor F, | |
2714 | with output to OP. */ | |
2715 | ||
2716 | finclude (f, fname, op) | |
2717 | int f; | |
2718 | char *fname; | |
2719 | FILE_BUF *op; | |
2720 | { | |
2721 | int st_mode; | |
2722 | long st_size; | |
2723 | long i; | |
2724 | FILE_BUF *fp; /* For input stack frame */ | |
2725 | int success = 0; | |
2726 | ||
2727 | CHECK_DEPTH (return;); | |
2728 | ||
2729 | if (file_size_and_mode (f, &st_mode, &st_size) < 0) | |
2730 | goto nope; /* Impossible? */ | |
2731 | ||
2732 | fp = &instack[indepth + 1]; | |
2733 | bzero (fp, sizeof (FILE_BUF)); | |
2734 | fp->fname = fname; | |
2735 | fp->length = 0; | |
2736 | fp->lineno = 1; | |
2737 | fp->if_stack = if_stack; | |
2738 | ||
2739 | if (S_ISREG (st_mode)) { | |
2740 | fp->buf = (U_CHAR *) alloca (st_size + 2); | |
2741 | fp->bufp = fp->buf; | |
2742 | ||
2743 | /* Read the file contents, knowing that st_size is an upper bound | |
2744 | on the number of bytes we can read. */ | |
2745 | while (st_size > 0) { | |
2746 | i = read (f, fp->buf + fp->length, st_size); | |
2747 | if (i <= 0) { | |
2748 | if (i == 0) break; | |
2749 | goto nope; | |
2750 | } | |
2751 | fp->length += i; | |
2752 | st_size -= i; | |
2753 | } | |
2754 | } | |
2755 | else { | |
2756 | /* Cannot count its file size before reading. | |
2757 | First read the entire file into heap and | |
2758 | copy them into buffer on stack. */ | |
2759 | ||
2760 | U_CHAR *bufp; | |
2761 | U_CHAR *basep; | |
2762 | int bsize = 2000; | |
2763 | ||
2764 | st_size = 0; | |
2765 | basep = (U_CHAR *) xmalloc (bsize + 2); | |
2766 | bufp = basep; | |
2767 | ||
2768 | for (;;) { | |
2769 | i = read (f, bufp, bsize - st_size); | |
2770 | if (i < 0) | |
2771 | goto nope; /* error! */ | |
2772 | if (i == 0) | |
2773 | break; /* End of file */ | |
2774 | st_size += i; | |
2775 | bufp += i; | |
2776 | if (bsize == st_size) { /* Buffer is full! */ | |
2777 | bsize *= 2; | |
2778 | basep = (U_CHAR *) xrealloc (basep, bsize + 2); | |
2779 | bufp = basep + st_size; /* May have moved */ | |
2780 | } | |
2781 | } | |
2782 | fp->buf = (U_CHAR *) alloca (st_size + 2); | |
2783 | fp->bufp = fp->buf; | |
2784 | bcopy (basep, fp->buf, st_size); | |
2785 | fp->length = st_size; | |
2786 | free (basep); | |
2787 | } | |
2788 | ||
2789 | if (!no_trigraphs) | |
2790 | trigraph_pcp (fp); | |
2791 | ||
2792 | if (fp->length > 0 && fp->buf[fp->length-1] != '\n') | |
2793 | fp->buf[fp->length++] = '\n'; | |
2794 | fp->buf[fp->length] = '\0'; | |
2795 | ||
2796 | success = 1; | |
2797 | indepth++; | |
2798 | ||
2799 | output_line_command (fp, op, 0, enter_file); | |
2800 | rescan (op, 0); | |
2801 | indepth--; | |
2802 | output_line_command (&instack[indepth], op, 0, leave_file); | |
2803 | ||
2804 | nope: | |
2805 | ||
2806 | if (!success) | |
2807 | perror_with_name (fname); | |
2808 | ||
2809 | close (f); | |
2810 | } | |
2811 | \f | |
2812 | /* The arglist structure is built by do_define to tell | |
2813 | collect_definition where the argument names begin. That | |
2814 | is, for a define like "#define f(x,y,z) foo+x-bar*y", the arglist | |
2815 | would contain pointers to the strings x, y, and z. | |
2816 | Collect_definition would then build a DEFINITION node, | |
2817 | with reflist nodes pointing to the places x, y, and z had | |
2818 | appeared. So the arglist is just convenience data passed | |
2819 | between these two routines. It is not kept around after | |
2820 | the current #define has been processed and entered into the | |
2821 | hash table. */ | |
2822 | ||
2823 | struct arglist { | |
2824 | struct arglist *next; | |
2825 | U_CHAR *name; | |
2826 | int length; | |
2827 | int argno; | |
2828 | }; | |
2829 | ||
2830 | /* Process a #define command. | |
2831 | BUF points to the contents of the #define command, as a continguous string. | |
2832 | LIMIT points to the first character past the end of the definition. | |
2833 | KEYWORD is the keyword-table entry for #define. */ | |
2834 | ||
2835 | do_define (buf, limit, op, keyword) | |
2836 | U_CHAR *buf, *limit; | |
2837 | FILE_BUF *op; | |
2838 | struct directive *keyword; | |
2839 | { | |
2840 | U_CHAR *bp; /* temp ptr into input buffer */ | |
2841 | U_CHAR *symname; /* remember where symbol name starts */ | |
2842 | int sym_length; /* and how long it is */ | |
2843 | ||
2844 | DEFINITION *defn; | |
2845 | int arglengths = 0; /* Accumulate lengths of arg names | |
2846 | plus number of args. */ | |
2847 | int hashcode; | |
2848 | ||
2849 | bp = buf; | |
2850 | ||
2851 | while (is_hor_space[*bp]) | |
2852 | bp++; | |
2853 | ||
2854 | symname = bp; /* remember where it starts */ | |
2855 | while (is_idchar[*bp] && bp < limit) { | |
2856 | bp++; | |
2857 | } | |
2858 | sym_length = bp - symname; | |
2859 | if (sym_length == 0) | |
2860 | error ("invalid macro name"); | |
2861 | else if (!is_idstart[*symname]) { | |
2862 | U_CHAR *msg; /* what pain... */ | |
2863 | msg = (U_CHAR *) alloca (sym_length + 1); | |
2864 | bcopy (symname, msg, sym_length); | |
2865 | msg[sym_length] = 0; | |
2866 | error ("invalid macro name `%s'", msg); | |
2867 | } else { | |
2868 | if (! strncmp (symname, "defined", 7) && sym_length == 7) | |
2869 | error ("defining `defined' as a macro"); | |
2870 | } | |
2871 | ||
2872 | /* lossage will occur if identifiers or control keywords are broken | |
2873 | across lines using backslash. This is not the right place to take | |
2874 | care of that. */ | |
2875 | ||
2876 | if (*bp == '(') { | |
2877 | struct arglist *arg_ptrs = NULL; | |
2878 | int argno = 0; | |
2879 | ||
2880 | bp++; /* skip '(' */ | |
2881 | SKIP_WHITE_SPACE (bp); | |
2882 | ||
2883 | /* Loop over macro argument names. */ | |
2884 | while (*bp != ')') { | |
2885 | struct arglist *temp; | |
2886 | ||
2887 | temp = (struct arglist *) alloca (sizeof (struct arglist)); | |
2888 | temp->name = bp; | |
2889 | temp->next = arg_ptrs; | |
2890 | temp->argno = argno++; | |
2891 | arg_ptrs = temp; | |
2892 | ||
2893 | if (!is_idstart[*bp]) | |
2894 | warning ("parameter name starts with a digit in #define"); | |
2895 | ||
2896 | /* Find the end of the arg name. */ | |
2897 | while (is_idchar[*bp]) { | |
2898 | bp++; | |
2899 | } | |
2900 | temp->length = bp - temp->name; | |
2901 | arglengths += temp->length + 2; | |
2902 | SKIP_WHITE_SPACE (bp); | |
2903 | if (temp->length == 0 || (*bp != ',' && *bp != ')')) { | |
2904 | error ("badly punctuated parameter list in #define"); | |
2905 | goto nope; | |
2906 | } | |
2907 | if (*bp == ',') { | |
2908 | bp++; | |
2909 | SKIP_WHITE_SPACE (bp); | |
2910 | } | |
2911 | if (bp >= limit) { | |
2912 | error ("unterminated parameter list in #define"); | |
2913 | goto nope; | |
2914 | } | |
2915 | } | |
2916 | ||
2917 | ++bp; /* skip paren */ | |
2918 | /* Skip exactly one space or tab if any. */ | |
2919 | if (bp < limit && (*bp == ' ' || *bp == '\t')) ++bp; | |
2920 | /* now everything from bp before limit is the definition. */ | |
2921 | defn = collect_expansion (bp, limit, argno, arg_ptrs); | |
2922 | ||
2923 | /* Now set defn->argnames to the result of concatenating | |
2924 | the argument names in reverse order | |
2925 | with comma-space between them. */ | |
2926 | defn->argnames = (U_CHAR *) xmalloc (arglengths + 1); | |
2927 | { | |
2928 | struct arglist *temp; | |
2929 | int i = 0; | |
2930 | for (temp = arg_ptrs; temp; temp = temp->next) { | |
2931 | bcopy (temp->name, &defn->argnames[i], temp->length); | |
2932 | i += temp->length; | |
2933 | if (temp->next != 0) { | |
2934 | defn->argnames[i++] = ','; | |
2935 | defn->argnames[i++] = ' '; | |
2936 | } | |
2937 | } | |
2938 | defn->argnames[i] = 0; | |
2939 | } | |
2940 | } else { | |
2941 | /* simple expansion or empty definition; gobble it */ | |
2942 | if (is_hor_space[*bp]) | |
2943 | ++bp; /* skip exactly one blank/tab char */ | |
2944 | /* now everything from bp before limit is the definition. */ | |
2945 | defn = collect_expansion (bp, limit, -1, 0); | |
2946 | defn->argnames = (U_CHAR *) ""; | |
2947 | } | |
2948 | ||
2949 | hashcode = hashf (symname, sym_length, HASHSIZE); | |
2950 | ||
2951 | { | |
2952 | HASHNODE *hp; | |
2953 | if ((hp = lookup (symname, sym_length, hashcode)) != NULL) { | |
2954 | if (hp->type != T_MACRO | |
2955 | || compare_defs (defn, hp->value.defn)) { | |
2956 | U_CHAR *msg; /* what pain... */ | |
2957 | msg = (U_CHAR *) alloca (sym_length + 20); | |
2958 | bcopy (symname, msg, sym_length); | |
2959 | strcpy ((char *) (msg + sym_length), " redefined"); | |
2960 | warning (msg); | |
2961 | } | |
2962 | /* Replace the old definition. */ | |
2963 | hp->type = T_MACRO; | |
2964 | hp->value.defn = defn; | |
2965 | } else | |
2966 | install (symname, sym_length, T_MACRO, defn, hashcode); | |
2967 | } | |
2968 | ||
2969 | return 0; | |
2970 | ||
2971 | nope: | |
2972 | ||
2973 | return 1; | |
2974 | } | |
2975 | ||
2976 | /* | |
2977 | * return zero if two DEFINITIONs are isomorphic | |
2978 | */ | |
2979 | int | |
2980 | compare_defs (d1, d2) | |
2981 | DEFINITION *d1, *d2; | |
2982 | { | |
2983 | register struct reflist *a1, *a2; | |
2984 | register U_CHAR *p1 = d1->expansion; | |
2985 | register U_CHAR *p2 = d2->expansion; | |
2986 | int first = 1; | |
2987 | ||
2988 | if (d1->nargs != d2->nargs) | |
2989 | return 1; | |
2990 | if (strcmp ((char *)d1->argnames, (char *)d2->argnames)) | |
2991 | return 1; | |
2992 | for (a1 = d1->pattern, a2 = d2->pattern; a1 && a2; | |
2993 | a1 = a1->next, a2 = a2->next) { | |
2994 | if (!((a1->nchars == a2->nchars && ! strncmp (p1, p2, a1->nchars)) | |
2995 | || ! comp_def_part (first, p1, a1->nchars, p2, a2->nchars, 0)) | |
2996 | || a1->argno != a2->argno | |
2997 | || a1->stringify != a2->stringify | |
2998 | || a1->raw_before != a2->raw_before | |
2999 | || a1->raw_after != a2->raw_after) | |
3000 | return 1; | |
3001 | first = 0; | |
3002 | p1 += a1->nchars; | |
3003 | p2 += a2->nchars; | |
3004 | } | |
3005 | if (a1 != a2) | |
3006 | return 1; | |
3007 | if (comp_def_part (first, p1, d1->length - (p1 - d1->expansion), | |
3008 | p2, d2->length - (p2 - d2->expansion), 1)) | |
3009 | return 1; | |
3010 | return 0; | |
3011 | } | |
3012 | ||
3013 | /* Return 1 if two parts of two macro definitions are effectively different. | |
3014 | One of the parts starts at BEG1 and has LEN1 chars; | |
3015 | the other has LEN2 chars at BEG2. | |
3016 | Any sequence of whitespace matches any other sequence of whitespace. | |
3017 | FIRST means these parts are the first of a macro definition; | |
3018 | so ignore leading whitespace entirely. | |
3019 | LAST means these parts are the last of a macro definition; | |
3020 | so ignore trailing whitespace entirely. */ | |
3021 | ||
3022 | comp_def_part (first, beg1, len1, beg2, len2, last) | |
3023 | int first; | |
3024 | U_CHAR *beg1, *beg2; | |
3025 | int len1, len2; | |
3026 | int last; | |
3027 | { | |
3028 | register U_CHAR *end1 = beg1 + len1; | |
3029 | register U_CHAR *end2 = beg2 + len2; | |
3030 | if (first) { | |
3031 | while (beg1 != end1 && is_space[*beg1]) beg1++; | |
3032 | while (beg2 != end2 && is_space[*beg2]) beg2++; | |
3033 | } | |
3034 | if (last) { | |
3035 | while (beg1 != end1 && is_space[end1[-1]]) end1--; | |
3036 | while (beg2 != end2 && is_space[end2[-1]]) end2--; | |
3037 | } | |
3038 | while (beg1 != end1 && beg2 != end2) { | |
3039 | if (is_space[*beg1] && is_space[*beg2]) { | |
3040 | while (beg1 != end1 && is_space[*beg1]) beg1++; | |
3041 | while (beg2 != end2 && is_space[*beg2]) beg2++; | |
3042 | } else if (*beg1 == *beg2) { | |
3043 | beg1++; beg2++; | |
3044 | } else break; | |
3045 | } | |
3046 | return (beg1 != end1) || (beg2 != end2); | |
3047 | } | |
3048 | ||
3049 | /* Read a replacement list for a macro with parameters. | |
3050 | Build the DEFINITION structure. | |
3051 | Reads characters of text starting at BUF until LIMIT. | |
3052 | ARGLIST specifies the formal parameters to look for | |
3053 | in the text of the definition; NARGS is the number of args | |
3054 | in that list, or -1 for a macro name that wants no argument list. | |
3055 | MACRONAME is the macro name itself (so we can avoid recursive expansion) | |
3056 | and NAMELEN is its length in characters. | |
3057 | ||
3058 | Note that comments and backslash-newlines have already been deleted | |
3059 | from the argument. */ | |
3060 | ||
3061 | /* Leading and trailing Space, Tab, etc. are converted to markers | |
3062 | Newline Space, Newline Tab, etc. | |
3063 | Newline Space makes a space in the final output | |
3064 | but is discarded if stringified. (Newline Tab is similar but | |
3065 | makes a Tab instead.) | |
3066 | ||
3067 | If there is no trailing whitespace, a Newline Space is added at the end | |
3068 | to prevent concatenation that would be contrary to the standard. */ | |
3069 | ||
3070 | DEFINITION * | |
3071 | collect_expansion (buf, end, nargs, arglist) | |
3072 | U_CHAR *buf, *end; | |
3073 | int nargs; | |
3074 | struct arglist *arglist; | |
3075 | { | |
3076 | DEFINITION *defn; | |
3077 | register U_CHAR *p, *limit, *lastp, *exp_p; | |
3078 | struct reflist *endpat = NULL; | |
3079 | /* Pointer to first nonspace after last ## seen. */ | |
3080 | U_CHAR *concat = 0; | |
3081 | /* Pointer to first nonspace after last single-# seen. */ | |
3082 | U_CHAR *stringify = 0; | |
3083 | int maxsize; | |
3084 | int expected_delimiter = '\0'; | |
3085 | ||
3086 | /* Scan thru the replacement list, ignoring comments and quoted | |
3087 | strings, picking up on the macro calls. It does a linear search | |
3088 | thru the arg list on every potential symbol. Profiling might say | |
3089 | that something smarter should happen. */ | |
3090 | ||
3091 | if (end < buf) | |
3092 | abort (); | |
3093 | ||
3094 | /* Find the beginning of the trailing whitespace. */ | |
3095 | /* Find end of leading whitespace. */ | |
3096 | limit = end; | |
3097 | p = buf; | |
3098 | while (p < limit && is_space[limit[-1]]) limit--; | |
3099 | while (p < limit && is_space[*p]) p++; | |
3100 | ||
3101 | /* Allocate space for the text in the macro definition. | |
3102 | Leading and trailing whitespace chars need 2 bytes each. | |
3103 | Each other input char may or may not need 1 byte, | |
3104 | so this is an upper bound. | |
3105 | The extra 2 are for invented trailing newline-marker and final null. */ | |
3106 | maxsize = (sizeof (DEFINITION) | |
3107 | + 2 * (end - limit) + 2 * (p - buf) | |
3108 | + (limit - p) + 3); | |
3109 | defn = (DEFINITION *) xcalloc (1, maxsize); | |
3110 | ||
3111 | defn->nargs = nargs; | |
3112 | exp_p = defn->expansion = (U_CHAR *) defn + sizeof (DEFINITION); | |
3113 | lastp = exp_p; | |
3114 | ||
3115 | p = buf; | |
3116 | ||
3117 | /* Convert leading whitespace to Newline-markers. */ | |
3118 | while (p < limit && is_space[*p]) { | |
3119 | *exp_p++ = '\n'; | |
3120 | *exp_p++ = *p++; | |
3121 | } | |
3122 | ||
3123 | if (p + 1 < limit && p[0] == '#' && p[1] == '#') { | |
3124 | error ("## operator at start of macro definition"); | |
3125 | p += 2; | |
3126 | } | |
3127 | ||
3128 | /* Process the main body of the definition. */ | |
3129 | while (p < limit) { | |
3130 | int skipped_arg = 0; | |
3131 | register U_CHAR c = *p++; | |
3132 | ||
3133 | *exp_p++ = c; | |
3134 | ||
3135 | if (!traditional) { | |
3136 | switch (c) { | |
3137 | case '\'': | |
3138 | case '\"': | |
3139 | for (; p < limit && *p != c; p++) { | |
3140 | *exp_p++ = *p; | |
3141 | if (*p == '\\') { | |
3142 | *exp_p++ = *++p; | |
3143 | } | |
3144 | } | |
3145 | *exp_p++ = *p++; | |
3146 | break; | |
3147 | ||
3148 | /* Special hack: if a \# is written in the #define | |
3149 | include a # in the definition. This is useless for C code | |
3150 | but useful for preprocessing other things. */ | |
3151 | ||
3152 | case '\\': | |
3153 | if (p < limit && *p == '#') { | |
3154 | /* Pass through this # */ | |
3155 | exp_p--; | |
3156 | *exp_p++ = *p++; | |
3157 | } else if (p < limit) { | |
3158 | /* Otherwise backslash goes through but makes next char ordinary. */ | |
3159 | *exp_p++ = *p++; | |
3160 | } | |
3161 | break; | |
3162 | ||
3163 | case '#': | |
3164 | if (p < limit && *p == '#') { | |
3165 | /* ##: concatenate preceding and following tokens. */ | |
3166 | /* Take out the first #, discard preceding whitespace. */ | |
3167 | exp_p--; | |
3168 | while (exp_p > lastp && is_hor_space[exp_p[-1]]) | |
3169 | --exp_p; | |
3170 | /* Skip the second #. */ | |
3171 | p++; | |
3172 | /* Discard following whitespace. */ | |
3173 | SKIP_WHITE_SPACE (p); | |
3174 | concat = p; | |
3175 | if (limit <= p) | |
3176 | error ("## operator at end of macro definition"); | |
3177 | } else { | |
3178 | /* Single #: stringify following argument ref. | |
3179 | Don't leave the # in the expansion. */ | |
3180 | exp_p--; | |
3181 | SKIP_WHITE_SPACE (p); | |
3182 | if (p == limit || ! is_idstart[*p] || nargs <= 0) | |
3183 | error ("# operator should be followed by a macro argument name"); | |
3184 | else | |
3185 | stringify = p; | |
3186 | } | |
3187 | break; | |
3188 | } | |
3189 | } else { | |
3190 | /* In -traditional mode, recognize arguments inside strings and | |
3191 | and character constants, and ignore special properties of #. | |
3192 | Arguments inside strings are considered "stringified", but no | |
3193 | extra quote marks are supplied. */ | |
3194 | switch (c) { | |
3195 | case '\'': | |
3196 | case '\"': | |
3197 | if (expected_delimiter != '\0') { | |
3198 | if (c == expected_delimiter) | |
3199 | expected_delimiter = '\0'; | |
3200 | } else | |
3201 | expected_delimiter = c; | |
3202 | break; | |
3203 | ||
3204 | case '\\': | |
3205 | /* Backslash quotes delimiters and itself, but not macro args. */ | |
3206 | if (expected_delimiter != 0 && p < limit | |
3207 | && (*p == expected_delimiter || *p == '\\')) { | |
3208 | *exp_p++ = *p++; | |
3209 | continue; | |
3210 | } | |
3211 | break; | |
3212 | ||
3213 | case '/': | |
3214 | if (expected_delimiter != '\0') /* No comments inside strings. */ | |
3215 | break; | |
3216 | if (*p == '*') { | |
3217 | /* If we find a comment that wasn't removed by handle_directive, | |
3218 | this must be -traditional. So replace the comment with | |
3219 | nothing at all. */ | |
3220 | exp_p--; | |
3221 | p += 1; | |
3222 | while (p < limit && !(p[-2] == '*' && p[-1] == '/')) | |
3223 | p++; | |
3224 | #if 0 | |
3225 | /* Mark this as a concatenation-point, as if it had been ##. */ | |
3226 | concat = p; | |
3227 | #endif | |
3228 | } | |
3229 | break; | |
3230 | } | |
3231 | } | |
3232 | ||
3233 | if (is_idchar[c] && nargs > 0) { | |
3234 | U_CHAR *id_beg = p - 1; | |
3235 | int id_len; | |
3236 | ||
3237 | --exp_p; | |
3238 | while (p != limit && is_idchar[*p]) p++; | |
3239 | id_len = p - id_beg; | |
3240 | ||
3241 | if (is_idstart[c]) { | |
3242 | register struct arglist *arg; | |
3243 | ||
3244 | for (arg = arglist; arg != NULL; arg = arg->next) { | |
3245 | struct reflist *tpat; | |
3246 | ||
3247 | if (arg->name[0] == c | |
3248 | && arg->length == id_len | |
3249 | && strncmp (arg->name, id_beg, id_len) == 0) { | |
3250 | /* make a pat node for this arg and append it to the end of | |
3251 | the pat list */ | |
3252 | tpat = (struct reflist *) xmalloc (sizeof (struct reflist)); | |
3253 | tpat->next = NULL; | |
3254 | tpat->raw_before = concat == id_beg; | |
3255 | tpat->raw_after = 0; | |
3256 | tpat->stringify = (traditional ? expected_delimiter != '\0' | |
3257 | : stringify == id_beg); | |
3258 | ||
3259 | if (endpat == NULL) | |
3260 | defn->pattern = tpat; | |
3261 | else | |
3262 | endpat->next = tpat; | |
3263 | endpat = tpat; | |
3264 | ||
3265 | tpat->argno = arg->argno; | |
3266 | tpat->nchars = exp_p - lastp; | |
3267 | { | |
3268 | register U_CHAR *p1 = p; | |
3269 | SKIP_WHITE_SPACE (p1); | |
3270 | if (p1 + 2 <= limit && p1[0] == '#' && p1[1] == '#') | |
3271 | tpat->raw_after = 1; | |
3272 | } | |
3273 | lastp = exp_p; /* place to start copying from next time */ | |
3274 | skipped_arg = 1; | |
3275 | break; | |
3276 | } | |
3277 | } | |
3278 | } | |
3279 | ||
3280 | /* If this was not a macro arg, copy it into the expansion. */ | |
3281 | if (! skipped_arg) { | |
3282 | register U_CHAR *lim1 = p; | |
3283 | p = id_beg; | |
3284 | while (p != lim1) | |
3285 | *exp_p++ = *p++; | |
3286 | if (stringify == id_beg) | |
3287 | error ("# operator should be followed by a macro argument name"); | |
3288 | } | |
3289 | } | |
3290 | } | |
3291 | ||
3292 | if (limit < end) { | |
3293 | /* Convert trailing whitespace to Newline-markers. */ | |
3294 | while (limit < end && is_space[*limit]) { | |
3295 | *exp_p++ = '\n'; | |
3296 | *exp_p++ = *limit++; | |
3297 | } | |
3298 | } else if (!traditional) { | |
3299 | /* There is no trailing whitespace, so invent some. */ | |
3300 | *exp_p++ = '\n'; | |
3301 | *exp_p++ = ' '; | |
3302 | } | |
3303 | ||
3304 | *exp_p = '\0'; | |
3305 | ||
3306 | defn->length = exp_p - defn->expansion; | |
3307 | ||
3308 | /* Crash now if we overrun the allocated size. */ | |
3309 | if (defn->length + 1 > maxsize) | |
3310 | abort (); | |
3311 | ||
3312 | #if 0 | |
3313 | /* This isn't worth the time it takes. */ | |
3314 | /* give back excess storage */ | |
3315 | defn->expansion = (U_CHAR *) xrealloc (defn->expansion, defn->length + 1); | |
3316 | #endif | |
3317 | ||
3318 | return defn; | |
3319 | } | |
3320 | \f | |
3321 | /* | |
3322 | * interpret #line command. Remembers previously seen fnames | |
3323 | * in its very own hash table. | |
3324 | */ | |
3325 | #define FNAME_HASHSIZE 37 | |
3326 | ||
3327 | do_line (buf, limit, op, keyword) | |
3328 | U_CHAR *buf, *limit; | |
3329 | FILE_BUF *op; | |
3330 | struct directive *keyword; | |
3331 | { | |
3332 | register U_CHAR *bp; | |
3333 | FILE_BUF *ip = &instack[indepth]; | |
3334 | FILE_BUF tem; | |
3335 | int new_lineno; | |
3336 | enum file_change_code file_change = same_file; | |
3337 | ||
3338 | /* Expand any macros. */ | |
3339 | tem = expand_to_temp_buffer (buf, limit, 0); | |
3340 | ||
3341 | /* Point to macroexpanded line, which is null-terminated now. */ | |
3342 | bp = tem.buf; | |
3343 | SKIP_WHITE_SPACE (bp); | |
3344 | ||
3345 | if (!isdigit (*bp)) { | |
3346 | error ("invalid format #line command"); | |
3347 | return; | |
3348 | } | |
3349 | ||
3350 | /* The Newline at the end of this line remains to be processed. | |
3351 | To put the next line at the specified line number, | |
3352 | we must store a line number now that is one less. */ | |
3353 | new_lineno = atoi (bp) - 1; | |
3354 | ||
3355 | /* skip over the line number. */ | |
3356 | while (isdigit (*bp)) | |
3357 | bp++; | |
3358 | ||
3359 | #if 0 /* #line 10"foo.c" is supposed to be allowed. */ | |
3360 | if (*bp && !is_space[*bp]) { | |
3361 | error ("invalid format #line command"); | |
3362 | return; | |
3363 | } | |
3364 | #endif | |
3365 | ||
3366 | SKIP_WHITE_SPACE (bp); | |
3367 | ||
3368 | if (*bp == '\"') { | |
3369 | static HASHNODE *fname_table[FNAME_HASHSIZE]; | |
3370 | HASHNODE *hp, **hash_bucket; | |
3371 | U_CHAR *fname; | |
3372 | int fname_length; | |
3373 | ||
3374 | fname = ++bp; | |
3375 | ||
3376 | while (*bp && *bp != '\"') | |
3377 | bp++; | |
3378 | if (*bp != '\"') { | |
3379 | error ("invalid format #line command"); | |
3380 | return; | |
3381 | } | |
3382 | ||
3383 | fname_length = bp - fname; | |
3384 | ||
3385 | bp++; | |
3386 | SKIP_WHITE_SPACE (bp); | |
3387 | if (*bp) { | |
3388 | if (*bp == '1') | |
3389 | file_change = enter_file; | |
3390 | else if (*bp == '2') | |
3391 | file_change = leave_file; | |
3392 | else { | |
3393 | error ("invalid format #line command"); | |
3394 | return; | |
3395 | } | |
3396 | ||
3397 | bp++; | |
3398 | SKIP_WHITE_SPACE (bp); | |
3399 | if (*bp) { | |
3400 | error ("invalid format #line command"); | |
3401 | return; | |
3402 | } | |
3403 | } | |
3404 | ||
3405 | hash_bucket = | |
3406 | &fname_table[hashf (fname, fname_length, FNAME_HASHSIZE)]; | |
3407 | for (hp = *hash_bucket; hp != NULL; hp = hp->next) | |
3408 | if (hp->length == fname_length && | |
3409 | strncmp (hp->value.cpval, fname, fname_length) == 0) { | |
3410 | ip->fname = hp->value.cpval; | |
3411 | break; | |
3412 | } | |
3413 | if (hp == 0) { | |
3414 | /* Didn't find it; cons up a new one. */ | |
3415 | hp = (HASHNODE *) xcalloc (1, sizeof (HASHNODE) + fname_length + 1); | |
3416 | hp->next = *hash_bucket; | |
3417 | *hash_bucket = hp; | |
3418 | ||
3419 | hp->length = fname_length; | |
3420 | ip->fname = hp->value.cpval = ((char *) hp) + sizeof (HASHNODE); | |
3421 | bcopy (fname, hp->value.cpval, fname_length); | |
3422 | } | |
3423 | } else if (*bp) { | |
3424 | error ("invalid format #line command"); | |
3425 | return; | |
3426 | } | |
3427 | ||
3428 | ip->lineno = new_lineno; | |
3429 | output_line_command (ip, op, 0, file_change); | |
3430 | check_expand (op, ip->length - (ip->bufp - ip->buf)); | |
3431 | } | |
3432 | ||
3433 | /* | |
3434 | * remove all definitions of symbol from symbol table. | |
3435 | * according to un*x /lib/cpp, it is not an error to undef | |
3436 | * something that has no definitions, so it isn't one here either. | |
3437 | */ | |
3438 | do_undef (buf, limit, op, keyword) | |
3439 | U_CHAR *buf, *limit; | |
3440 | FILE_BUF *op; | |
3441 | struct directive *keyword; | |
3442 | { | |
3443 | HASHNODE *hp; | |
3444 | ||
3445 | SKIP_WHITE_SPACE (buf); | |
3446 | ||
3447 | if (! strncmp (buf, "defined", 7) && ! is_idchar[buf[7]]) | |
3448 | warning ("undefining `defined'"); | |
3449 | ||
3450 | while ((hp = lookup (buf, -1, -1)) != NULL) { | |
3451 | if (hp->type != T_MACRO) | |
3452 | warning ("undefining `%s'", hp->name); | |
3453 | delete_macro (hp); | |
3454 | } | |
3455 | } | |
3456 | \f | |
3457 | /* | |
3458 | * Report a fatal error detected by the program we are processing. | |
3459 | * Use the text of the line in the error message, then terminate. | |
3460 | * (We use error() because it prints the filename & line#.) | |
3461 | */ | |
3462 | do_error (buf, limit, op, keyword) | |
3463 | U_CHAR *buf, *limit; | |
3464 | FILE_BUF *op; | |
3465 | struct directive *keyword; | |
3466 | { | |
3467 | int length = limit - buf; | |
3468 | char *copy = (char *) xmalloc (length + 1); | |
3469 | bcopy (buf, copy, length); | |
3470 | copy[length] = 0; | |
3471 | SKIP_WHITE_SPACE (copy); | |
3472 | error ("#error %s", copy); | |
3473 | exit (FATAL_EXIT_CODE); | |
3474 | } | |
3475 | ||
3476 | /* Remember the name of the current file being read from so that we can | |
3477 | avoid ever including it again. */ | |
3478 | ||
3479 | do_once () | |
3480 | { | |
3481 | int i; | |
3482 | FILE_BUF *ip = NULL; | |
3483 | ||
3484 | for (i = indepth; i >= 0; i--) | |
3485 | if (instack[i].fname != NULL) { | |
3486 | ip = &instack[i]; | |
3487 | break; | |
3488 | } | |
3489 | ||
3490 | if (ip != NULL) | |
3491 | { | |
3492 | struct file_name_list *new; | |
3493 | ||
3494 | new = (struct file_name_list *) xmalloc (sizeof (struct file_name_list)); | |
3495 | new->next = dont_repeat_files; | |
3496 | dont_repeat_files = new; | |
3497 | new->fname = savestring (ip->fname); | |
3498 | } | |
3499 | } | |
3500 | ||
3501 | /* #pragma and its argument line have already been copied to the output file. | |
3502 | Here just check for recognized pragmas. */ | |
3503 | ||
3504 | do_pragma (buf, limit) | |
3505 | U_CHAR *buf, *limit; | |
3506 | { | |
3507 | while (*buf == ' ' || *buf == '\t') | |
3508 | buf++; | |
3509 | if (!strncmp (buf, "once", 4)) | |
3510 | do_once (); | |
3511 | } | |
3512 | ||
3513 | #if 0 | |
3514 | /* This was a fun hack, but #pragma seems to start to be useful. | |
3515 | By failing to recognize it, we pass it through unchanged to cc1. */ | |
3516 | ||
3517 | /* | |
3518 | * the behavior of the #pragma directive is implementation defined. | |
3519 | * this implementation defines it as follows. | |
3520 | */ | |
3521 | do_pragma () | |
3522 | { | |
3523 | close (0); | |
3524 | if (open ("/dev/tty", O_RDONLY, 0666) != 0) | |
3525 | goto nope; | |
3526 | close (1); | |
3527 | if (open ("/dev/tty", O_WRONLY, 0666) != 1) | |
3528 | goto nope; | |
3529 | execl ("/usr/games/hack", "#pragma", 0); | |
3530 | execl ("/usr/games/rogue", "#pragma", 0); | |
3531 | execl ("/usr/new/emacs", "-f", "hanoi", "9", "-kill", 0); | |
3532 | execl ("/usr/local/emacs", "-f", "hanoi", "9", "-kill", 0); | |
3533 | nope: | |
3534 | fatal ("You are in a maze of twisty compiler features, all different"); | |
3535 | } | |
3536 | #endif | |
3537 | ||
3538 | /* Just ignore #sccs, on systems where we define it at all. */ | |
3539 | do_sccs () | |
3540 | { | |
3541 | if (pedantic) | |
3542 | error ("ANSI C does not allow #sccs"); | |
3543 | } | |
3544 | \f | |
3545 | /* | |
3546 | * handle #if command by | |
3547 | * 1) inserting special `defined' keyword into the hash table | |
3548 | * that gets turned into 0 or 1 by special_symbol (thus, | |
3549 | * if the luser has a symbol called `defined' already, it won't | |
3550 | * work inside the #if command) | |
3551 | * 2) rescan the input into a temporary output buffer | |
3552 | * 3) pass the output buffer to the yacc parser and collect a value | |
3553 | * 4) clean up the mess left from steps 1 and 2. | |
3554 | * 5) call conditional_skip to skip til the next #endif (etc.), | |
3555 | * or not, depending on the value from step 3. | |
3556 | */ | |
3557 | ||
3558 | do_if (buf, limit, op, keyword) | |
3559 | U_CHAR *buf, *limit; | |
3560 | FILE_BUF *op; | |
3561 | struct directive *keyword; | |
3562 | { | |
3563 | int value; | |
3564 | FILE_BUF *ip = &instack[indepth]; | |
3565 | ||
3566 | value = eval_if_expression (buf, limit - buf); | |
3567 | conditional_skip (ip, value == 0, T_IF); | |
3568 | } | |
3569 | ||
3570 | /* | |
3571 | * handle a #elif directive by not changing if_stack either. | |
3572 | * see the comment above do_else. | |
3573 | */ | |
3574 | ||
3575 | do_elif (buf, limit, op, keyword) | |
3576 | U_CHAR *buf, *limit; | |
3577 | FILE_BUF *op; | |
3578 | struct directive *keyword; | |
3579 | { | |
3580 | int value; | |
3581 | FILE_BUF *ip = &instack[indepth]; | |
3582 | ||
3583 | if (if_stack == instack[indepth].if_stack) { | |
3584 | error ("#elif not within a conditional"); | |
3585 | return; | |
3586 | } else { | |
3587 | if (if_stack->type != T_IF && if_stack->type != T_ELIF) { | |
3588 | error ("#elif after #else"); | |
3589 | fprintf (stderr, " (matches line %d", if_stack->lineno); | |
3590 | if (if_stack->fname != NULL && ip->fname != NULL && | |
3591 | strcmp (if_stack->fname, ip->fname) != 0) | |
3592 | fprintf (stderr, ", file %s", if_stack->fname); | |
3593 | fprintf (stderr, ")\n"); | |
3594 | } | |
3595 | if_stack->type = T_ELIF; | |
3596 | } | |
3597 | ||
3598 | if (if_stack->if_succeeded) | |
3599 | skip_if_group (ip, 0); | |
3600 | else { | |
3601 | value = eval_if_expression (buf, limit - buf); | |
3602 | if (value == 0) | |
3603 | skip_if_group (ip, 0); | |
3604 | else { | |
3605 | ++if_stack->if_succeeded; /* continue processing input */ | |
3606 | output_line_command (ip, op, 1, same_file); | |
3607 | } | |
3608 | } | |
3609 | } | |
3610 | ||
3611 | /* | |
3612 | * evaluate a #if expression in BUF, of length LENGTH, | |
3613 | * then parse the result as a C expression and return the value as an int. | |
3614 | */ | |
3615 | int | |
3616 | eval_if_expression (buf, length) | |
3617 | U_CHAR *buf; | |
3618 | int length; | |
3619 | { | |
3620 | FILE_BUF temp_obuf; | |
3621 | HASHNODE *save_defined; | |
3622 | int value; | |
3623 | ||
3624 | save_defined = install ("defined", -1, T_SPEC_DEFINED, 0, -1); | |
3625 | temp_obuf = expand_to_temp_buffer (buf, buf + length, 0); | |
3626 | delete_macro (save_defined); /* clean up special symbol */ | |
3627 | ||
3628 | value = parse_c_expression (temp_obuf.buf); | |
3629 | ||
3630 | free (temp_obuf.buf); | |
3631 | ||
3632 | return value; | |
3633 | } | |
3634 | ||
3635 | /* | |
3636 | * routine to handle ifdef/ifndef. Try to look up the symbol, | |
3637 | * then do or don't skip to the #endif/#else/#elif depending | |
3638 | * on what directive is actually being processed. | |
3639 | */ | |
3640 | do_xifdef (buf, limit, op, keyword) | |
3641 | U_CHAR *buf, *limit; | |
3642 | FILE_BUF *op; | |
3643 | struct directive *keyword; | |
3644 | { | |
3645 | int skip; | |
3646 | FILE_BUF *ip = &instack[indepth]; | |
3647 | U_CHAR *end; | |
3648 | ||
3649 | /* Discard leading and trailing whitespace. */ | |
3650 | SKIP_WHITE_SPACE (buf); | |
3651 | while (limit != buf && is_hor_space[limit[-1]]) limit--; | |
3652 | ||
3653 | /* Find the end of the identifier at the beginning. */ | |
3654 | for (end = buf; is_idchar[*end]; end++); | |
3655 | ||
3656 | if (end == buf) { | |
3657 | skip = (keyword->type == T_IFDEF); | |
3658 | if (! traditional) | |
3659 | warning (end == limit ? "#%s with no argument" | |
3660 | : "#%s argument starts with punctuation", | |
3661 | keyword->name); | |
3662 | } else { | |
3663 | if (pedantic && buf[0] >= '0' && buf[0] <= '9') | |
3664 | warning ("#%s argument starts with a digit", keyword->name); | |
3665 | else if (end != limit && !traditional) | |
3666 | warning ("garbage at end of #%s argument", keyword->name); | |
3667 | ||
3668 | skip = (lookup (buf, end-buf, -1) == NULL) ^ (keyword->type == T_IFNDEF); | |
3669 | } | |
3670 | ||
3671 | conditional_skip (ip, skip, T_IF); | |
3672 | } | |
3673 | ||
3674 | /* | |
3675 | * push TYPE on stack; then, if SKIP is nonzero, skip ahead. | |
3676 | */ | |
3677 | void | |
3678 | conditional_skip (ip, skip, type) | |
3679 | FILE_BUF *ip; | |
3680 | int skip; | |
3681 | enum node_type type; | |
3682 | { | |
3683 | IF_STACK_FRAME *temp; | |
3684 | ||
3685 | temp = (IF_STACK_FRAME *) xcalloc (1, sizeof (IF_STACK_FRAME)); | |
3686 | temp->fname = ip->fname; | |
3687 | temp->lineno = ip->lineno; | |
3688 | temp->next = if_stack; | |
3689 | if_stack = temp; | |
3690 | ||
3691 | if_stack->type = type; | |
3692 | ||
3693 | if (skip != 0) { | |
3694 | skip_if_group (ip, 0); | |
3695 | return; | |
3696 | } else { | |
3697 | ++if_stack->if_succeeded; | |
3698 | output_line_command (ip, &outbuf, 1, same_file); | |
3699 | } | |
3700 | } | |
3701 | ||
3702 | /* | |
3703 | * skip to #endif, #else, or #elif. adjust line numbers, etc. | |
3704 | * leaves input ptr at the sharp sign found. | |
3705 | * If ANY is nonzero, return at next directive of any sort. | |
3706 | */ | |
3707 | void | |
3708 | skip_if_group (ip, any) | |
3709 | FILE_BUF *ip; | |
3710 | int any; | |
3711 | { | |
3712 | register U_CHAR *bp = ip->bufp, *cp; | |
3713 | register U_CHAR *endb = ip->buf + ip->length; | |
3714 | struct directive *kt; | |
3715 | IF_STACK_FRAME *save_if_stack = if_stack; /* don't pop past here */ | |
3716 | U_CHAR *beg_of_line = bp; | |
3717 | ||
3718 | while (bp < endb) { | |
3719 | switch (*bp++) { | |
3720 | case '/': /* possible comment */ | |
3721 | if (*bp == '\\' && bp[1] == '\n') | |
3722 | newline_fix (bp); | |
3723 | if (*bp == '*' | |
3724 | || (cplusplus && *bp == '/')) { | |
3725 | ip->bufp = ++bp; | |
3726 | bp = skip_to_end_of_comment (ip, &ip->lineno); | |
3727 | } | |
3728 | break; | |
3729 | case '\"': | |
3730 | case '\'': | |
3731 | bp = skip_quoted_string (bp - 1, endb, ip->lineno, &ip->lineno, 0, 0); | |
3732 | break; | |
3733 | case '\\': | |
3734 | /* Char after backslash loses its special meaning. */ | |
3735 | if (bp < endb) { | |
3736 | if (*bp == '\n') | |
3737 | ++ip->lineno; /* But do update the line-count. */ | |
3738 | bp++; | |
3739 | } | |
3740 | break; | |
3741 | case '\n': | |
3742 | ++ip->lineno; | |
3743 | beg_of_line = bp; | |
3744 | break; | |
3745 | case '#': | |
3746 | ip->bufp = bp - 1; | |
3747 | ||
3748 | /* # keyword: a # must be first nonblank char on the line */ | |
3749 | if (beg_of_line == 0) | |
3750 | break; | |
3751 | /* Scan from start of line, skipping whitespace, comments | |
3752 | and backslash-newlines, and see if we reach this #. | |
3753 | If not, this # is not special. */ | |
3754 | bp = beg_of_line; | |
3755 | while (1) { | |
3756 | if (is_hor_space[*bp]) | |
3757 | bp++; | |
3758 | else if (*bp == '\\' && bp[1] == '\n') | |
3759 | bp += 2; | |
3760 | else if (*bp == '/' && bp[1] == '*') { | |
3761 | bp += 2; | |
3762 | while (!(*bp == '*' && bp[1] == '/')) | |
3763 | bp++; | |
3764 | bp += 2; | |
3765 | } | |
3766 | else if (cplusplus && *bp == '/' && bp[1] == '/') { | |
3767 | bp += 2; | |
3768 | while (*bp++ != '\n') ; | |
3769 | } | |
3770 | else break; | |
3771 | } | |
3772 | if (bp != ip->bufp) { | |
3773 | bp = ip->bufp + 1; /* Reset bp to after the #. */ | |
3774 | break; | |
3775 | } | |
3776 | ||
3777 | bp = ip->bufp + 1; /* point at '#' */ | |
3778 | ||
3779 | /* Skip whitespace and \-newline. */ | |
3780 | while (1) { | |
3781 | if (is_hor_space[*bp]) | |
3782 | bp++; | |
3783 | else if (*bp == '\\' && bp[1] == '\n') | |
3784 | bp += 2; | |
3785 | else break; | |
3786 | } | |
3787 | ||
3788 | cp = bp; | |
3789 | ||
3790 | /* Now find end of directive name. | |
3791 | If we encounter a backslash-newline, exchange it with any following | |
3792 | symbol-constituents so that we end up with a contiguous name. */ | |
3793 | ||
3794 | while (1) { | |
3795 | if (is_idchar[*bp]) | |
3796 | bp++; | |
3797 | else { | |
3798 | if (*bp == '\\' && bp[1] == '\n') | |
3799 | name_newline_fix (bp); | |
3800 | if (is_idchar[*bp]) | |
3801 | bp++; | |
3802 | else break; | |
3803 | } | |
3804 | } | |
3805 | ||
3806 | for (kt = directive_table; kt->length >= 0; kt++) { | |
3807 | IF_STACK_FRAME *temp; | |
3808 | if (strncmp (cp, kt->name, kt->length) == 0 | |
3809 | && !is_idchar[cp[kt->length]]) { | |
3810 | ||
3811 | /* If we are asked to return on next directive, | |
3812 | do so now. */ | |
3813 | if (any) | |
3814 | return; | |
3815 | ||
3816 | switch (kt->type) { | |
3817 | case T_IF: | |
3818 | case T_IFDEF: | |
3819 | case T_IFNDEF: | |
3820 | temp = (IF_STACK_FRAME *) xcalloc (1, sizeof (IF_STACK_FRAME)); | |
3821 | temp->next = if_stack; | |
3822 | if_stack = temp; | |
3823 | temp->lineno = ip->lineno; | |
3824 | temp->fname = ip->fname; | |
3825 | temp->type = kt->type; | |
3826 | break; | |
3827 | case T_ELSE: | |
3828 | case T_ENDIF: | |
3829 | if (pedantic && if_stack != save_if_stack) | |
3830 | validate_else (bp); | |
3831 | case T_ELIF: | |
3832 | if (if_stack == instack[indepth].if_stack) { | |
3833 | error ("#%s not within a conditional", kt->name); | |
3834 | break; | |
3835 | } | |
3836 | else if (if_stack == save_if_stack) | |
3837 | return; /* found what we came for */ | |
3838 | ||
3839 | if (kt->type != T_ENDIF) { | |
3840 | if (if_stack->type == T_ELSE) | |
3841 | error ("#else or #elif after #else"); | |
3842 | if_stack->type = kt->type; | |
3843 | break; | |
3844 | } | |
3845 | ||
3846 | temp = if_stack; | |
3847 | if_stack = if_stack->next; | |
3848 | free (temp); | |
3849 | break; | |
3850 | } | |
3851 | break; | |
3852 | } | |
3853 | } | |
3854 | } | |
3855 | } | |
3856 | ip->bufp = bp; | |
3857 | /* after this returns, rescan will exit because ip->bufp | |
3858 | now points to the end of the buffer. | |
3859 | rescan is responsible for the error message also. */ | |
3860 | } | |
3861 | ||
3862 | /* | |
3863 | * handle a #else directive. Do this by just continuing processing | |
3864 | * without changing if_stack ; this is so that the error message | |
3865 | * for missing #endif's etc. will point to the original #if. It | |
3866 | * is possible that something different would be better. | |
3867 | */ | |
3868 | do_else (buf, limit, op, keyword) | |
3869 | U_CHAR *buf, *limit; | |
3870 | FILE_BUF *op; | |
3871 | struct directive *keyword; | |
3872 | { | |
3873 | FILE_BUF *ip = &instack[indepth]; | |
3874 | ||
3875 | if (pedantic) { | |
3876 | SKIP_WHITE_SPACE (buf); | |
3877 | if (buf != limit) | |
3878 | warning ("text following #else violates ANSI standard"); | |
3879 | } | |
3880 | ||
3881 | if (if_stack == instack[indepth].if_stack) { | |
3882 | error ("#else not within a conditional"); | |
3883 | return; | |
3884 | } else { | |
3885 | if (if_stack->type != T_IF && if_stack->type != T_ELIF) { | |
3886 | error ("#else after #else"); | |
3887 | fprintf (stderr, " (matches line %d", if_stack->lineno); | |
3888 | if (strcmp (if_stack->fname, ip->fname) != 0) | |
3889 | fprintf (stderr, ", file %s", if_stack->fname); | |
3890 | fprintf (stderr, ")\n"); | |
3891 | } | |
3892 | if_stack->type = T_ELSE; | |
3893 | } | |
3894 | ||
3895 | if (if_stack->if_succeeded) | |
3896 | skip_if_group (ip, 0); | |
3897 | else { | |
3898 | ++if_stack->if_succeeded; /* continue processing input */ | |
3899 | output_line_command (ip, op, 1, same_file); | |
3900 | } | |
3901 | } | |
3902 | ||
3903 | /* | |
3904 | * unstack after #endif command | |
3905 | */ | |
3906 | do_endif (buf, limit, op, keyword) | |
3907 | U_CHAR *buf, *limit; | |
3908 | FILE_BUF *op; | |
3909 | struct directive *keyword; | |
3910 | { | |
3911 | if (pedantic) { | |
3912 | SKIP_WHITE_SPACE (buf); | |
3913 | if (buf != limit) | |
3914 | warning ("text following #endif violates ANSI standard"); | |
3915 | } | |
3916 | ||
3917 | if (if_stack == instack[indepth].if_stack) | |
3918 | error ("unbalanced #endif"); | |
3919 | else { | |
3920 | IF_STACK_FRAME *temp = if_stack; | |
3921 | if_stack = if_stack->next; | |
3922 | free (temp); | |
3923 | output_line_command (&instack[indepth], op, 1, same_file); | |
3924 | } | |
3925 | } | |
3926 | ||
3927 | /* When an #else or #endif is found while skipping failed conditional, | |
3928 | if -pedantic was specified, this is called to warn about text after | |
3929 | the command name. P points to the first char after the command name. */ | |
3930 | ||
3931 | validate_else (p) | |
3932 | register U_CHAR *p; | |
3933 | { | |
3934 | /* Advance P over whitespace and comments. */ | |
3935 | while (1) { | |
3936 | if (*p == '\\' && p[1] == '\n') | |
3937 | p += 2; | |
3938 | if (is_hor_space[*p]) | |
3939 | p++; | |
3940 | else if (*p == '/') { | |
3941 | if (p[1] == '\\' && p[2] == '\n') | |
3942 | newline_fix (p + 1); | |
3943 | if (p[1] == '*') { | |
3944 | p += 2; | |
3945 | /* Don't bother warning about unterminated comments | |
3946 | since that will happen later. Just be sure to exit. */ | |
3947 | while (*p) { | |
3948 | if (p[1] == '\\' && p[2] == '\n') | |
3949 | newline_fix (p + 1); | |
3950 | if (*p == '*' && p[1] == '/') { | |
3951 | p += 2; | |
3952 | break; | |
3953 | } | |
3954 | p++; | |
3955 | } | |
3956 | } | |
3957 | else if (cplusplus && p[1] == '/') { | |
3958 | p += 2; | |
3959 | while (*p && *p++ != '\n') ; | |
3960 | } | |
3961 | } else break; | |
3962 | } | |
3963 | if (*p && *p != '\n') | |
3964 | warning ("text following #else or #endif violates ANSI standard"); | |
3965 | } | |
3966 | \f | |
3967 | /* | |
3968 | * Skip a comment, assuming the input ptr immediately follows the | |
3969 | * initial slash-star. Bump line counter as necessary. | |
3970 | * (The canonical line counter is &ip->lineno). | |
3971 | * Don't use this routine (or the next one) if bumping the line | |
3972 | * counter is not sufficient to deal with newlines in the string. | |
3973 | */ | |
3974 | U_CHAR * | |
3975 | skip_to_end_of_comment (ip, line_counter) | |
3976 | register FILE_BUF *ip; | |
3977 | int *line_counter; /* place to remember newlines, or NULL */ | |
3978 | { | |
3979 | register U_CHAR *limit = ip->buf + ip->length; | |
3980 | register U_CHAR *bp = ip->bufp; | |
3981 | FILE_BUF *op = &outbuf; /* JF */ | |
3982 | int output = put_out_comments && !line_counter; | |
3983 | ||
3984 | /* JF this line_counter stuff is a crock to make sure the | |
3985 | comment is only put out once, no matter how many times | |
3986 | the comment is skipped. It almost works */ | |
3987 | if (output) { | |
3988 | *op->bufp++ = '/'; | |
3989 | *op->bufp++ = '*'; | |
3990 | } | |
3991 | if (cplusplus && bp[-1] == '/') { | |
3992 | if (output) { | |
3993 | while (bp < limit) | |
3994 | if ((*op->bufp++ = *bp++) == '\n') { | |
3995 | bp--; | |
3996 | break; | |
3997 | } | |
3998 | op->bufp[-1] = '*'; | |
3999 | *op->bufp++ = '/'; | |
4000 | *op->bufp++ = '\n'; | |
4001 | } else { | |
4002 | while (bp < limit) { | |
4003 | if (*bp++ == '\n') { | |
4004 | bp--; | |
4005 | break; | |
4006 | } | |
4007 | } | |
4008 | } | |
4009 | ip->bufp = bp; | |
4010 | return bp; | |
4011 | } | |
4012 | while (bp < limit) { | |
4013 | if (output) | |
4014 | *op->bufp++ = *bp; | |
4015 | switch (*bp++) { | |
4016 | case '\n': | |
4017 | if (line_counter != NULL) | |
4018 | ++*line_counter; | |
4019 | if (output) | |
4020 | ++op->lineno; | |
4021 | break; | |
4022 | case '*': | |
4023 | if (*bp == '\\' && bp[1] == '\n') | |
4024 | newline_fix (bp); | |
4025 | if (*bp == '/') { | |
4026 | if (output) | |
4027 | *op->bufp++ = '/'; | |
4028 | ip->bufp = ++bp; | |
4029 | return bp; | |
4030 | } | |
4031 | break; | |
4032 | } | |
4033 | } | |
4034 | ip->bufp = bp; | |
4035 | return bp; | |
4036 | } | |
4037 | ||
4038 | /* | |
4039 | * Skip over a quoted string. BP points to the opening quote. | |
4040 | * Returns a pointer after the closing quote. Don't go past LIMIT. | |
4041 | * START_LINE is the line number of the starting point (but it need | |
4042 | * not be valid if the starting point is inside a macro expansion). | |
4043 | * | |
4044 | * The input stack state is not changed. | |
4045 | * | |
4046 | * If COUNT_NEWLINES is nonzero, it points to an int to increment | |
4047 | * for each newline passed. | |
4048 | * | |
4049 | * If BACKSLASH_NEWLINES_P is nonzero, store 1 thru it | |
4050 | * if we pass a backslash-newline. | |
4051 | * | |
4052 | * If EOFP is nonzero, set *EOFP to 1 if the string is unterminated. | |
4053 | */ | |
4054 | U_CHAR * | |
4055 | skip_quoted_string (bp, limit, start_line, count_newlines, backslash_newlines_p, eofp) | |
4056 | register U_CHAR *bp; | |
4057 | register U_CHAR *limit; | |
4058 | int start_line; | |
4059 | int *count_newlines; | |
4060 | int *backslash_newlines_p; | |
4061 | int *eofp; | |
4062 | { | |
4063 | register U_CHAR c, match; | |
4064 | ||
4065 | match = *bp++; | |
4066 | while (1) { | |
4067 | if (bp >= limit) { | |
4068 | error_with_line (line_for_error (start_line), | |
4069 | "unterminated string or character constant"); | |
4070 | if (eofp) | |
4071 | *eofp = 1; | |
4072 | break; | |
4073 | } | |
4074 | c = *bp++; | |
4075 | if (c == '\\') { | |
4076 | while (*bp == '\\' && bp[1] == '\n') { | |
4077 | if (backslash_newlines_p) | |
4078 | *backslash_newlines_p = 1; | |
4079 | if (count_newlines) | |
4080 | ++*count_newlines; | |
4081 | bp += 2; | |
4082 | } | |
4083 | if (*bp == '\n' && count_newlines) { | |
4084 | if (backslash_newlines_p) | |
4085 | *backslash_newlines_p = 1; | |
4086 | ++*count_newlines; | |
4087 | } | |
4088 | bp++; | |
4089 | } else if (c == '\n') { | |
4090 | if (traditional) { | |
4091 | /* Unterminated strings and character constants are 'legal'. */ | |
4092 | bp--; /* Don't consume the newline. */ | |
4093 | if (eofp) | |
4094 | *eofp = 1; | |
4095 | break; | |
4096 | } | |
4097 | if (match == '\'') { | |
4098 | error_with_line (line_for_error (start_line), | |
4099 | "unterminated character constant"); | |
4100 | bp--; | |
4101 | if (eofp) | |
4102 | *eofp = 1; | |
4103 | break; | |
4104 | } | |
4105 | if (traditional) { /* Unterminated strings are 'legal'. */ | |
4106 | if (eofp) | |
4107 | *eofp = 1; | |
4108 | break; | |
4109 | } | |
4110 | /* If not traditional, then allow newlines inside strings. */ | |
4111 | if (count_newlines) | |
4112 | ++*count_newlines; | |
4113 | } else if (c == match) | |
4114 | break; | |
4115 | } | |
4116 | return bp; | |
4117 | } | |
4118 | \f | |
4119 | /* | |
4120 | * write out a #line command, for instance, after an #include file. | |
4121 | * If CONDITIONAL is nonzero, we can omit the #line if it would | |
4122 | * appear to be a no-op, and we can output a few newlines instead | |
4123 | * if we want to increase the line number by a small amount. | |
4124 | * FILE_CHANGE says whether we are entering a file, leaving, or neither. | |
4125 | */ | |
4126 | ||
4127 | void | |
4128 | output_line_command (ip, op, conditional, file_change) | |
4129 | FILE_BUF *ip, *op; | |
4130 | int conditional; | |
4131 | enum file_change_code file_change; | |
4132 | { | |
4133 | int len; | |
4134 | char line_cmd_buf[500]; | |
4135 | ||
4136 | if (no_line_commands | |
4137 | || ip->fname == NULL | |
4138 | || no_output) { | |
4139 | op->lineno = ip->lineno; | |
4140 | return; | |
4141 | } | |
4142 | ||
4143 | if (conditional) { | |
4144 | if (ip->lineno == op->lineno) | |
4145 | return; | |
4146 | ||
4147 | /* If the inherited line number is a little too small, | |
4148 | output some newlines instead of a #line command. */ | |
4149 | if (ip->lineno > op->lineno && ip->lineno < op->lineno + 8) { | |
4150 | check_expand (op, 10); | |
4151 | while (ip->lineno > op->lineno) { | |
4152 | *op->bufp++ = '\n'; | |
4153 | op->lineno++; | |
4154 | } | |
4155 | return; | |
4156 | } | |
4157 | } | |
4158 | ||
4159 | #ifdef OUTPUT_LINE_COMMANDS | |
4160 | sprintf (line_cmd_buf, "#line %d \"%s\"", ip->lineno, ip->fname); | |
4161 | #else | |
4162 | sprintf (line_cmd_buf, "# %d \"%s\"", ip->lineno, ip->fname); | |
4163 | #endif | |
4164 | if (file_change != same_file) | |
4165 | strcat (line_cmd_buf, file_change == enter_file ? " 1" : " 2"); | |
4166 | len = strlen (line_cmd_buf); | |
4167 | line_cmd_buf[len++] = '\n'; | |
4168 | check_expand (op, len + 1); | |
4169 | if (op->bufp > op->buf && op->bufp[-1] != '\n') | |
4170 | *op->bufp++ = '\n'; | |
4171 | bcopy (line_cmd_buf, op->bufp, len); | |
4172 | op->bufp += len; | |
4173 | op->lineno = ip->lineno; | |
4174 | } | |
4175 | \f | |
4176 | /* This structure represents one parsed argument in a macro call. | |
4177 | `raw' points to the argument text as written (`raw_length' is its length). | |
4178 | `expanded' points to the argument's macro-expansion | |
4179 | (its length is `expand_length'). | |
4180 | `stringified_length' is the length the argument would have | |
4181 | if stringified. | |
4182 | `free1' and `free2', if nonzero, point to blocks to be freed | |
4183 | when the macro argument data is no longer needed. */ | |
4184 | ||
4185 | struct argdata { | |
4186 | U_CHAR *raw, *expanded; | |
4187 | int raw_length, expand_length; | |
4188 | int stringified_length; | |
4189 | U_CHAR *free1, *free2; | |
4190 | char newlines; | |
4191 | char comments; | |
4192 | }; | |
4193 | ||
4194 | /* Expand a macro call. | |
4195 | HP points to the symbol that is the macro being called. | |
4196 | Put the result of expansion onto the input stack | |
4197 | so that subsequent input by our caller will use it. | |
4198 | ||
4199 | If macro wants arguments, caller has already verified that | |
4200 | an argument list follows; arguments come from the input stack. */ | |
4201 | ||
4202 | void | |
4203 | macroexpand (hp, op) | |
4204 | HASHNODE *hp; | |
4205 | FILE_BUF *op; | |
4206 | { | |
4207 | int nargs; | |
4208 | DEFINITION *defn = hp->value.defn; | |
4209 | register U_CHAR *xbuf; | |
4210 | int xbuf_len; | |
4211 | int start_line = instack[indepth].lineno; | |
4212 | ||
4213 | CHECK_DEPTH (return;); | |
4214 | ||
4215 | /* it might not actually be a macro. */ | |
4216 | if (hp->type != T_MACRO) { | |
4217 | special_symbol (hp, op); | |
4218 | return; | |
4219 | } | |
4220 | ||
4221 | nargs = defn->nargs; | |
4222 | ||
4223 | if (nargs >= 0) { | |
4224 | register int i; | |
4225 | struct argdata *args; | |
4226 | char *parse_error = 0; | |
4227 | ||
4228 | args = (struct argdata *) alloca ((nargs + 1) * sizeof (struct argdata)); | |
4229 | ||
4230 | for (i = 0; i < nargs; i++) { | |
4231 | args[i].raw = args[i].expanded = (U_CHAR *) ""; | |
4232 | args[i].raw_length = args[i].expand_length | |
4233 | = args[i].stringified_length = 0; | |
4234 | args[i].free1 = args[i].free2 = 0; | |
4235 | } | |
4236 | ||
4237 | /* Parse all the macro args that are supplied. I counts them. | |
4238 | The first NARGS args are stored in ARGS. | |
4239 | The rest are discarded. */ | |
4240 | i = 0; | |
4241 | do { | |
4242 | /* Discard the open-parenthesis or comma before the next arg. */ | |
4243 | ++instack[indepth].bufp; | |
4244 | parse_error | |
4245 | = macarg ((i < nargs || (nargs == 0 && i == 0)) ? &args[i] : 0); | |
4246 | if (parse_error) | |
4247 | { | |
4248 | error_with_line (line_for_error (start_line), parse_error); | |
4249 | break; | |
4250 | } | |
4251 | i++; | |
4252 | } while (*instack[indepth].bufp != ')'); | |
4253 | ||
4254 | /* If we got one arg but it was just whitespace, call that 0 args. */ | |
4255 | if (i == 1) { | |
4256 | register U_CHAR *bp = args[0].raw; | |
4257 | register U_CHAR *lim = bp + args[0].raw_length; | |
4258 | while (bp != lim && is_space[*bp]) bp++; | |
4259 | if (bp == lim) | |
4260 | i = 0; | |
4261 | } | |
4262 | ||
4263 | if (nargs == 0 && i > 0) | |
4264 | error ("arguments given to macro `%s'", hp->name); | |
4265 | else if (i < nargs) { | |
4266 | /* traditional C allows foo() if foo wants one argument. */ | |
4267 | if (nargs == 1 && i == 0 && traditional) | |
4268 | ; | |
4269 | else if (i == 0) | |
4270 | error ("no args to macro `%s'", hp->name); | |
4271 | else if (i == 1) | |
4272 | error ("only 1 arg to macro `%s'", hp->name); | |
4273 | else | |
4274 | error ("only %d args to macro `%s'", i, hp->name); | |
4275 | } else if (i > nargs) | |
4276 | error ("too many (%d) args to macro `%s'", i, hp->name); | |
4277 | ||
4278 | /* Swallow the closeparen. */ | |
4279 | ++instack[indepth].bufp; | |
4280 | ||
4281 | /* If macro wants zero args, we parsed the arglist for checking only. | |
4282 | Read directly from the macro definition. */ | |
4283 | if (nargs == 0) { | |
4284 | xbuf = defn->expansion; | |
4285 | xbuf_len = defn->length; | |
4286 | } else { | |
4287 | register U_CHAR *exp = defn->expansion; | |
4288 | register int offset; /* offset in expansion, | |
4289 | copied a piece at a time */ | |
4290 | register int totlen; /* total amount of exp buffer filled so far */ | |
4291 | ||
4292 | register struct reflist *ap; | |
4293 | ||
4294 | /* Macro really takes args. Compute the expansion of this call. */ | |
4295 | ||
4296 | /* Compute length in characters of the macro's expansion. */ | |
4297 | xbuf_len = defn->length; | |
4298 | for (ap = defn->pattern; ap != NULL; ap = ap->next) { | |
4299 | if (ap->stringify) | |
4300 | xbuf_len += args[ap->argno].stringified_length; | |
4301 | else if (ap->raw_before || ap->raw_after || traditional) | |
4302 | xbuf_len += args[ap->argno].raw_length; | |
4303 | else | |
4304 | xbuf_len += args[ap->argno].expand_length; | |
4305 | } | |
4306 | ||
4307 | xbuf = (U_CHAR *) xmalloc (xbuf_len + 1); | |
4308 | ||
4309 | /* Generate in XBUF the complete expansion | |
4310 | with arguments substituted in. | |
4311 | TOTLEN is the total size generated so far. | |
4312 | OFFSET is the index in the definition | |
4313 | of where we are copying from. */ | |
4314 | offset = totlen = 0; | |
4315 | for (ap = defn->pattern; ap != NULL; ap = ap->next) { | |
4316 | register struct argdata *arg = &args[ap->argno]; | |
4317 | ||
4318 | for (i = 0; i < ap->nchars; i++) | |
4319 | xbuf[totlen++] = exp[offset++]; | |
4320 | ||
4321 | if (ap->stringify != 0) { | |
4322 | int arglen = arg->raw_length; | |
4323 | int escaped = 0; | |
4324 | int in_string = 0; | |
4325 | int c; | |
4326 | i = 0; | |
4327 | while (i < arglen | |
4328 | && (c = arg->raw[i], is_space[c])) | |
4329 | i++; | |
4330 | while (i < arglen | |
4331 | && (c = arg->raw[arglen - 1], is_space[c])) | |
4332 | arglen--; | |
4333 | if (!traditional) | |
4334 | xbuf[totlen++] = '\"'; /* insert beginning quote */ | |
4335 | for (; i < arglen; i++) { | |
4336 | c = arg->raw[i]; | |
4337 | ||
4338 | /* Special markers Newline Space | |
4339 | generate nothing for a stringified argument. */ | |
4340 | if (c == '\n' && arg->raw[i+1] != '\n') { | |
4341 | i++; | |
4342 | continue; | |
4343 | } | |
4344 | ||
4345 | /* Internal sequences of whitespace are replaced by one space. */ | |
4346 | if (c == '\n' ? arg->raw[i+1] == '\n' : is_space[c]) { | |
4347 | while (1) { | |
4348 | /* Note that Newline Space does occur within whitespace | |
4349 | sequences; consider it part of the sequence. */ | |
4350 | if (c == '\n' && is_space[arg->raw[i+1]]) | |
4351 | i += 2; | |
4352 | else if (c != '\n' && is_space[c]) | |
4353 | i++; | |
4354 | else break; | |
4355 | c = arg->raw[i]; | |
4356 | } | |
4357 | i--; | |
4358 | c = ' '; | |
4359 | } | |
4360 | ||
4361 | if (escaped) | |
4362 | escaped = 0; | |
4363 | else { | |
4364 | if (c == '\\') | |
4365 | escaped = 1; | |
4366 | if (in_string) { | |
4367 | if (c == in_string) | |
4368 | in_string = 0; | |
4369 | } else if (c == '\"' || c == '\'') | |
4370 | in_string = c; | |
4371 | } | |
4372 | ||
4373 | /* Escape these chars */ | |
4374 | if (c == '\"' || (in_string && c == '\\')) | |
4375 | xbuf[totlen++] = '\\'; | |
4376 | if (isprint (c)) | |
4377 | xbuf[totlen++] = c; | |
4378 | else { | |
4379 | sprintf ((char *) &xbuf[totlen], "\\%03o", (unsigned int) c); | |
4380 | totlen += 4; | |
4381 | } | |
4382 | } | |
4383 | if (!traditional) | |
4384 | xbuf[totlen++] = '\"'; /* insert ending quote */ | |
4385 | } else if (ap->raw_before || ap->raw_after || traditional) { | |
4386 | U_CHAR *p1 = arg->raw; | |
4387 | U_CHAR *l1 = p1 + arg->raw_length; | |
4388 | ||
4389 | if (ap->raw_before) { | |
4390 | while (p1 != l1 && is_space[*p1]) p1++; | |
4391 | while (p1 != l1 && is_idchar[*p1]) | |
4392 | xbuf[totlen++] = *p1++; | |
4393 | /* Delete any no-reexpansion marker that follows | |
4394 | an identifier at the beginning of the argument | |
4395 | if the argument is concatenated with what precedes it. */ | |
4396 | if (p1[0] == '\n' && p1[1] == '-') | |
4397 | p1 += 2; | |
4398 | } | |
4399 | if (ap->raw_after) { | |
4400 | /* Arg is concatenated after: delete trailing whitespace, | |
4401 | whitespace markers, and no-reexpansion markers. */ | |
4402 | while (p1 != l1) { | |
4403 | if (is_space[l1[-1]]) l1--; | |
4404 | else if (l1[-1] == '-') { | |
4405 | U_CHAR *p2 = l1 - 1; | |
4406 | /* If a `-' is preceded by an odd number of newlines then it | |
4407 | and the last newline are a no-reexpansion marker. */ | |
4408 | while (p2 != p1 && p2[-1] == '\n') p2--; | |
4409 | if ((l1 - 1 - p2) & 1) { | |
4410 | l1 -= 2; | |
4411 | } | |
4412 | else break; | |
4413 | } | |
4414 | else break; | |
4415 | } | |
4416 | } | |
4417 | bcopy (p1, xbuf + totlen, l1 - p1); | |
4418 | totlen += l1 - p1; | |
4419 | } else { | |
4420 | bcopy (arg->expanded, xbuf + totlen, arg->expand_length); | |
4421 | totlen += arg->expand_length; | |
4422 | } | |
4423 | ||
4424 | if (totlen > xbuf_len) | |
4425 | abort (); | |
4426 | } | |
4427 | ||
4428 | /* if there is anything left of the definition | |
4429 | after handling the arg list, copy that in too. */ | |
4430 | ||
4431 | for (i = offset; i < defn->length; i++) | |
4432 | xbuf[totlen++] = exp[i]; | |
4433 | ||
4434 | xbuf[totlen] = 0; | |
4435 | xbuf_len = totlen; | |
4436 | ||
4437 | for (i = 0; i < nargs; i++) { | |
4438 | if (args[i].free1 != 0) | |
4439 | free (args[i].free1); | |
4440 | if (args[i].free2 != 0) | |
4441 | free (args[i].free2); | |
4442 | } | |
4443 | } | |
4444 | } else { | |
4445 | xbuf = defn->expansion; | |
4446 | xbuf_len = defn->length; | |
4447 | } | |
4448 | ||
4449 | /* Now put the expansion on the input stack | |
4450 | so our caller will commence reading from it. */ | |
4451 | { | |
4452 | register FILE_BUF *ip2; | |
4453 | ||
4454 | ip2 = &instack[++indepth]; | |
4455 | ||
4456 | ip2->fname = 0; | |
4457 | ip2->lineno = 0; | |
4458 | ip2->buf = xbuf; | |
4459 | ip2->length = xbuf_len; | |
4460 | ip2->bufp = xbuf; | |
4461 | ip2->free_ptr = (nargs > 0) ? xbuf : 0; | |
4462 | ip2->macro = hp; | |
4463 | ip2->if_stack = if_stack; | |
4464 | ||
4465 | /* Recursive macro use sometimes works traditionally. | |
4466 | #define foo(x,y) bar(x(y,0), y) | |
4467 | foo(foo, baz) */ | |
4468 | ||
4469 | if (!traditional) | |
4470 | hp->type = T_DISABLED; | |
4471 | } | |
4472 | } | |
4473 | \f | |
4474 | /* | |
4475 | * Parse a macro argument and store the info on it into *ARGPTR. | |
4476 | * Return nonzero to indicate a syntax error. | |
4477 | */ | |
4478 | ||
4479 | char * | |
4480 | macarg (argptr) | |
4481 | register struct argdata *argptr; | |
4482 | { | |
4483 | FILE_BUF *ip = &instack[indepth]; | |
4484 | int paren = 0; | |
4485 | int newlines = 0; | |
4486 | int comments = 0; | |
4487 | ||
4488 | /* Try to parse as much of the argument as exists at this | |
4489 | input stack level. */ | |
4490 | U_CHAR *bp = macarg1 (ip->bufp, ip->buf + ip->length, | |
4491 | &paren, &newlines, &comments); | |
4492 | ||
4493 | /* If we find the end of the argument at this level, | |
4494 | set up *ARGPTR to point at it in the input stack. */ | |
4495 | if (!(ip->fname != 0 && (newlines != 0 || comments != 0)) | |
4496 | && bp != ip->buf + ip->length) { | |
4497 | if (argptr != 0) { | |
4498 | argptr->raw = ip->bufp; | |
4499 | argptr->raw_length = bp - ip->bufp; | |
4500 | } | |
4501 | ip->bufp = bp; | |
4502 | } else { | |
4503 | /* This input stack level ends before the macro argument does. | |
4504 | We must pop levels and keep parsing. | |
4505 | Therefore, we must allocate a temporary buffer and copy | |
4506 | the macro argument into it. */ | |
4507 | int bufsize = bp - ip->bufp; | |
4508 | int extra = newlines; | |
4509 | U_CHAR *buffer = (U_CHAR *) xmalloc (bufsize + extra + 1); | |
4510 | int final_start = 0; | |
4511 | ||
4512 | bcopy (ip->bufp, buffer, bufsize); | |
4513 | ip->bufp = bp; | |
4514 | ip->lineno += newlines; | |
4515 | ||
4516 | while (bp == ip->buf + ip->length) { | |
4517 | if (instack[indepth].macro == 0) { | |
4518 | free (buffer); | |
4519 | return "unterminated macro call"; | |
4520 | } | |
4521 | ip->macro->type = T_MACRO; | |
4522 | if (ip->free_ptr) | |
4523 | free (ip->free_ptr); | |
4524 | ip = &instack[--indepth]; | |
4525 | newlines = 0; | |
4526 | comments = 0; | |
4527 | bp = macarg1 (ip->bufp, ip->buf + ip->length, &paren, | |
4528 | &newlines, &comments); | |
4529 | final_start = bufsize; | |
4530 | bufsize += bp - ip->bufp; | |
4531 | extra += newlines; | |
4532 | buffer = (U_CHAR *) xrealloc (buffer, bufsize + extra + 1); | |
4533 | bcopy (ip->bufp, buffer + bufsize - (bp - ip->bufp), bp - ip->bufp); | |
4534 | ip->bufp = bp; | |
4535 | ip->lineno += newlines; | |
4536 | } | |
4537 | ||
4538 | /* Now, if arg is actually wanted, record its raw form, | |
4539 | discarding comments and duplicating newlines in whatever | |
4540 | part of it did not come from a macro expansion. | |
4541 | EXTRA space has been preallocated for duplicating the newlines. | |
4542 | FINAL_START is the index of the start of that part. */ | |
4543 | if (argptr != 0) { | |
4544 | argptr->raw = buffer; | |
4545 | argptr->raw_length = bufsize; | |
4546 | argptr->free1 = buffer; | |
4547 | argptr->newlines = newlines; | |
4548 | argptr->comments = comments; | |
4549 | if ((newlines || comments) && ip->fname != 0) | |
4550 | argptr->raw_length | |
4551 | = final_start + | |
4552 | discard_comments (argptr->raw + final_start, | |
4553 | argptr->raw_length - final_start, | |
4554 | newlines); | |
4555 | argptr->raw[argptr->raw_length] = 0; | |
4556 | if (argptr->raw_length > bufsize + extra) | |
4557 | abort (); | |
4558 | } | |
4559 | } | |
4560 | ||
4561 | /* If we are not discarding this argument, | |
4562 | macroexpand it and compute its length as stringified. | |
4563 | All this info goes into *ARGPTR. */ | |
4564 | ||
4565 | if (argptr != 0) { | |
4566 | FILE_BUF obuf; | |
4567 | register U_CHAR *buf, *lim; | |
4568 | register int totlen; | |
4569 | ||
4570 | obuf = expand_to_temp_buffer (argptr->raw, | |
4571 | argptr->raw + argptr->raw_length, | |
4572 | 1); | |
4573 | ||
4574 | argptr->expanded = obuf.buf; | |
4575 | argptr->expand_length = obuf.length; | |
4576 | argptr->free2 = obuf.buf; | |
4577 | ||
4578 | buf = argptr->raw; | |
4579 | lim = buf + argptr->raw_length; | |
4580 | ||
4581 | /* If ANSI, discard leading and trailing space. */ | |
4582 | if (!traditional) { | |
4583 | while (buf != lim && is_space[*buf]) | |
4584 | buf++; | |
4585 | while (buf != lim && is_space[lim[-1]]) | |
4586 | lim--; | |
4587 | } | |
4588 | totlen = traditional ? 0 : 2; /* Count opening and closing quote. */ | |
4589 | while (buf != lim) { | |
4590 | register U_CHAR c = *buf++; | |
4591 | totlen++; | |
4592 | /* If ANSI, replace internal sequences of whitespace with one space. */ | |
4593 | if (is_space[c] && !traditional) | |
4594 | SKIP_ALL_WHITE_SPACE (buf); | |
4595 | else if (c == '\"' || c == '\\') /* escape these chars */ | |
4596 | totlen++; | |
4597 | else if (!isprint (c)) | |
4598 | totlen += 3; | |
4599 | } | |
4600 | argptr->stringified_length = totlen; | |
4601 | } | |
4602 | return 0; | |
4603 | } | |
4604 | ||
4605 | /* Scan text from START (inclusive) up to LIMIT (exclusive), | |
4606 | counting parens in *DEPTHPTR, | |
4607 | and return if reach LIMIT | |
4608 | or before a `)' that would make *DEPTHPTR negative | |
4609 | or before a comma when *DEPTHPTR is zero. | |
4610 | Single and double quotes are matched and termination | |
4611 | is inhibited within them. Comments also inhibit it. | |
4612 | Value returned is pointer to stopping place. | |
4613 | ||
4614 | Increment *NEWLINES each time a newline is passed. | |
4615 | Set *COMMENTS to 1 if a comment is seen. */ | |
4616 | ||
4617 | U_CHAR * | |
4618 | macarg1 (start, limit, depthptr, newlines, comments) | |
4619 | U_CHAR *start; | |
4620 | register U_CHAR *limit; | |
4621 | int *depthptr, *newlines, *comments; | |
4622 | { | |
4623 | register U_CHAR *bp = start; | |
4624 | ||
4625 | while (bp < limit) { | |
4626 | switch (*bp) { | |
4627 | case '(': | |
4628 | (*depthptr)++; | |
4629 | break; | |
4630 | case ')': | |
4631 | if (--(*depthptr) < 0) | |
4632 | return bp; | |
4633 | break; | |
4634 | case '\\': | |
4635 | /* Traditionally, backslash makes following char not special. */ | |
4636 | if (!traditional) | |
4637 | break; | |
4638 | if (bp + 1 < limit) | |
4639 | { | |
4640 | bp++; | |
4641 | /* But count source lines anyway. */ | |
4642 | if (*bp == '\n') | |
4643 | ++*newlines; | |
4644 | } | |
4645 | break; | |
4646 | case '\n': | |
4647 | ++*newlines; | |
4648 | break; | |
4649 | case '/': | |
4650 | if (bp[1] == '\\' && bp[2] == '\n') | |
4651 | newline_fix (bp + 1); | |
4652 | if (cplusplus && bp[1] == '/') { | |
4653 | *comments = 1; | |
4654 | bp += 2; | |
4655 | while (bp < limit && *bp++ != '\n') ; | |
4656 | ++*newlines; | |
4657 | break; | |
4658 | } | |
4659 | if (bp[1] != '*' || bp + 1 >= limit) | |
4660 | break; | |
4661 | *comments = 1; | |
4662 | bp += 2; | |
4663 | while (bp + 1 < limit) { | |
4664 | if (bp[0] == '*' | |
4665 | && bp[1] == '\\' && bp[2] == '\n') | |
4666 | newline_fix (bp + 1); | |
4667 | if (bp[0] == '*' && bp[1] == '/') | |
4668 | break; | |
4669 | if (*bp == '\n') ++*newlines; | |
4670 | bp++; | |
4671 | } | |
4672 | bp += 1; | |
4673 | break; | |
4674 | case '\'': | |
4675 | case '\"': | |
4676 | { | |
4677 | int quotec; | |
4678 | for (quotec = *bp++; bp + 1 < limit && *bp != quotec; bp++) { | |
4679 | if (*bp == '\\') { | |
4680 | bp++; | |
4681 | if (*bp == '\n') | |
4682 | ++*newlines; | |
4683 | while (*bp == '\\' && bp[1] == '\n') { | |
4684 | bp += 2; | |
4685 | } | |
4686 | } else if (*bp == '\n') { | |
4687 | ++*newlines; | |
4688 | if (quotec == '\'') | |
4689 | break; | |
4690 | } | |
4691 | } | |
4692 | } | |
4693 | break; | |
4694 | case ',': | |
4695 | if ((*depthptr) == 0) | |
4696 | return bp; | |
4697 | break; | |
4698 | } | |
4699 | bp++; | |
4700 | } | |
4701 | ||
4702 | return bp; | |
4703 | } | |
4704 | ||
4705 | /* Discard comments and duplicate newlines | |
4706 | in the string of length LENGTH at START, | |
4707 | except inside of string constants. | |
4708 | The string is copied into itself with its beginning staying fixed. | |
4709 | ||
4710 | NEWLINES is the number of newlines that must be duplicated. | |
4711 | We assume that that much extra space is available past the end | |
4712 | of the string. */ | |
4713 | ||
4714 | int | |
4715 | discard_comments (start, length, newlines) | |
4716 | U_CHAR *start; | |
4717 | int length; | |
4718 | int newlines; | |
4719 | { | |
4720 | register U_CHAR *ibp; | |
4721 | register U_CHAR *obp; | |
4722 | register U_CHAR *limit; | |
4723 | register int c; | |
4724 | ||
4725 | /* If we have newlines to duplicate, copy everything | |
4726 | that many characters up. Then, in the second part, | |
4727 | we will have room to insert the newlines | |
4728 | while copying down. | |
4729 | NEWLINES may actually be too large, because it counts | |
4730 | newlines in string constants, and we don't duplicate those. | |
4731 | But that does no harm. */ | |
4732 | if (newlines > 0) { | |
4733 | ibp = start + length; | |
4734 | obp = ibp + newlines; | |
4735 | limit = start; | |
4736 | while (limit != ibp) | |
4737 | *--obp = *--ibp; | |
4738 | } | |
4739 | ||
4740 | ibp = start + newlines; | |
4741 | limit = start + length + newlines; | |
4742 | obp = start; | |
4743 | ||
4744 | while (ibp < limit) { | |
4745 | *obp++ = c = *ibp++; | |
4746 | switch (c) { | |
4747 | case '\n': | |
4748 | /* Duplicate the newline. */ | |
4749 | *obp++ = '\n'; | |
4750 | break; | |
4751 | ||
4752 | case '\\': | |
4753 | if (*ibp == '\n') { | |
4754 | obp--; | |
4755 | ibp++; | |
4756 | } | |
4757 | break; | |
4758 | ||
4759 | case '/': | |
4760 | if (*ibp == '\\' && ibp[1] == '\n') | |
4761 | newline_fix (ibp); | |
4762 | /* Delete any comment. */ | |
4763 | if (cplusplus && ibp[0] == '/') { | |
4764 | obp--; | |
4765 | ibp++; | |
4766 | while (ibp < limit && *ibp++ != '\n') ; | |
4767 | break; | |
4768 | } | |
4769 | if (ibp[0] != '*' || ibp + 1 >= limit) | |
4770 | break; | |
4771 | obp--; | |
4772 | ibp++; | |
4773 | while (ibp + 1 < limit) { | |
4774 | if (ibp[0] == '*' | |
4775 | && ibp[1] == '\\' && ibp[2] == '\n') | |
4776 | newline_fix (ibp + 1); | |
4777 | if (ibp[0] == '*' && ibp[1] == '/') | |
4778 | break; | |
4779 | ibp++; | |
4780 | } | |
4781 | ibp += 2; | |
4782 | break; | |
4783 | ||
4784 | case '\'': | |
4785 | case '\"': | |
4786 | /* Notice and skip strings, so that we don't | |
4787 | think that comments start inside them, | |
4788 | and so we don't duplicate newlines in them. */ | |
4789 | { | |
4790 | int quotec = c; | |
4791 | while (ibp < limit) { | |
4792 | *obp++ = c = *ibp++; | |
4793 | if (c == quotec) | |
4794 | break; | |
4795 | if (c == '\n' && quotec == '\'') | |
4796 | break; | |
4797 | if (c == '\\' && ibp < limit) { | |
4798 | while (*ibp == '\\' && ibp[1] == '\n') | |
4799 | ibp += 2; | |
4800 | *obp++ = *ibp++; | |
4801 | } | |
4802 | } | |
4803 | } | |
4804 | break; | |
4805 | } | |
4806 | } | |
4807 | ||
4808 | return obp - start; | |
4809 | } | |
4810 | \f | |
4811 | /* | |
4812 | * error - print error message and increment count of errors. | |
4813 | */ | |
4814 | error (msg, arg1, arg2, arg3) | |
4815 | char *msg; | |
4816 | { | |
4817 | int i; | |
4818 | FILE_BUF *ip = NULL; | |
4819 | ||
4820 | for (i = indepth; i >= 0; i--) | |
4821 | if (instack[i].fname != NULL) { | |
4822 | ip = &instack[i]; | |
4823 | break; | |
4824 | } | |
4825 | ||
4826 | if (ip != NULL) | |
4827 | fprintf (stderr, "%s:%d: ", ip->fname, ip->lineno); | |
4828 | fprintf (stderr, msg, arg1, arg2, arg3); | |
4829 | fprintf (stderr, "\n"); | |
4830 | errors++; | |
4831 | return 0; | |
4832 | } | |
4833 | ||
4834 | /* Error including a message from `errno'. */ | |
4835 | ||
4836 | error_from_errno (name) | |
4837 | char *name; | |
4838 | { | |
4839 | int i; | |
4840 | FILE_BUF *ip = NULL; | |
4841 | extern int errno, sys_nerr; | |
4842 | extern char *sys_errlist[]; | |
4843 | ||
4844 | for (i = indepth; i >= 0; i--) | |
4845 | if (instack[i].fname != NULL) { | |
4846 | ip = &instack[i]; | |
4847 | break; | |
4848 | } | |
4849 | ||
4850 | if (ip != NULL) | |
4851 | fprintf (stderr, "%s:%d: ", ip->fname, ip->lineno); | |
4852 | ||
4853 | if (errno < sys_nerr) | |
4854 | fprintf (stderr, "%s: %s\n", name, sys_errlist[errno]); | |
4855 | else | |
4856 | fprintf (stderr, "%s: undocumented I/O error\n", name); | |
4857 | ||
4858 | errors++; | |
4859 | return 0; | |
4860 | } | |
4861 | ||
4862 | /* Print error message but don't count it. */ | |
4863 | ||
4864 | warning (msg, arg1, arg2, arg3) | |
4865 | char *msg; | |
4866 | { | |
4867 | int i; | |
4868 | FILE_BUF *ip = NULL; | |
4869 | ||
4870 | if (inhibit_warnings) | |
4871 | return 0; | |
4872 | ||
4873 | for (i = indepth; i >= 0; i--) | |
4874 | if (instack[i].fname != NULL) { | |
4875 | ip = &instack[i]; | |
4876 | break; | |
4877 | } | |
4878 | ||
4879 | if (ip != NULL) | |
4880 | fprintf (stderr, "%s:%d: ", ip->fname, ip->lineno); | |
4881 | fprintf (stderr, "warning: "); | |
4882 | fprintf (stderr, msg, arg1, arg2, arg3); | |
4883 | fprintf (stderr, "\n"); | |
4884 | return 0; | |
4885 | } | |
4886 | ||
4887 | error_with_line (line, msg, arg1, arg2, arg3) | |
4888 | int line; | |
4889 | char *msg; | |
4890 | { | |
4891 | int i; | |
4892 | FILE_BUF *ip = NULL; | |
4893 | ||
4894 | for (i = indepth; i >= 0; i--) | |
4895 | if (instack[i].fname != NULL) { | |
4896 | ip = &instack[i]; | |
4897 | break; | |
4898 | } | |
4899 | ||
4900 | if (ip != NULL) | |
4901 | fprintf (stderr, "%s:%d: ", ip->fname, line); | |
4902 | fprintf (stderr, msg, arg1, arg2, arg3); | |
4903 | fprintf (stderr, "\n"); | |
4904 | errors++; | |
4905 | return 0; | |
4906 | } | |
4907 | ||
4908 | /* Return the line at which an error occurred. | |
4909 | The error is not necessarily associated with the current spot | |
4910 | in the input stack, so LINE says where. LINE will have been | |
4911 | copied from ip->lineno for the current input level. | |
4912 | If the current level is for a file, we return LINE. | |
4913 | But if the current level is not for a file, LINE is meaningless. | |
4914 | In that case, we return the lineno of the innermost file. */ | |
4915 | int | |
4916 | line_for_error (line) | |
4917 | int line; | |
4918 | { | |
4919 | int i; | |
4920 | int line1 = line; | |
4921 | ||
4922 | for (i = indepth; i >= 0; ) { | |
4923 | if (instack[i].fname != 0) | |
4924 | return line1; | |
4925 | i--; | |
4926 | if (i < 0) | |
4927 | return 0; | |
4928 | line1 = instack[i].lineno; | |
4929 | } | |
4930 | } | |
4931 | ||
4932 | /* | |
4933 | * If OBUF doesn't have NEEDED bytes after OPTR, make it bigger. | |
4934 | * | |
4935 | * As things stand, nothing is ever placed in the output buffer to be | |
4936 | * removed again except when it's KNOWN to be part of an identifier, | |
4937 | * so flushing and moving down everything left, instead of expanding, | |
4938 | * should work ok. | |
4939 | */ | |
4940 | ||
4941 | int | |
4942 | grow_outbuf (obuf, needed) | |
4943 | register FILE_BUF *obuf; | |
4944 | register int needed; | |
4945 | { | |
4946 | register U_CHAR *p; | |
4947 | int minsize; | |
4948 | ||
4949 | if (obuf->length - (obuf->bufp - obuf->buf) > needed) | |
4950 | return; | |
4951 | ||
4952 | /* Make it at least twice as big as it is now. */ | |
4953 | obuf->length *= 2; | |
4954 | /* Make it have at least 150% of the free space we will need. */ | |
4955 | minsize = (3 * needed) / 2 + (obuf->bufp - obuf->buf); | |
4956 | if (minsize > obuf->length) | |
4957 | obuf->length = minsize; | |
4958 | ||
4959 | if ((p = (U_CHAR *) xrealloc (obuf->buf, obuf->length)) == NULL) | |
4960 | memory_full (); | |
4961 | ||
4962 | obuf->bufp = p + (obuf->bufp - obuf->buf); | |
4963 | obuf->buf = p; | |
4964 | } | |
4965 | \f | |
4966 | /* Symbol table for macro names and special symbols */ | |
4967 | ||
4968 | /* | |
4969 | * install a name in the main hash table, even if it is already there. | |
4970 | * name stops with first non alphanumeric, except leading '#'. | |
4971 | * caller must check against redefinition if that is desired. | |
4972 | * delete_macro () removes things installed by install () in fifo order. | |
4973 | * this is important because of the `defined' special symbol used | |
4974 | * in #if, and also if pushdef/popdef directives are ever implemented. | |
4975 | * | |
4976 | * If LEN is >= 0, it is the length of the name. | |
4977 | * Otherwise, compute the length by scanning the entire name. | |
4978 | * | |
4979 | * If HASH is >= 0, it is the precomputed hash code. | |
4980 | * Otherwise, compute the hash code. | |
4981 | */ | |
4982 | HASHNODE * | |
4983 | install (name, len, type, value, hash) | |
4984 | U_CHAR *name; | |
4985 | int len; | |
4986 | enum node_type type; | |
4987 | int value; | |
4988 | int hash; | |
4989 | /* watch out here if sizeof (U_CHAR *) != sizeof (int) */ | |
4990 | { | |
4991 | register HASHNODE *hp; | |
4992 | register int i, bucket; | |
4993 | register U_CHAR *p, *q; | |
4994 | ||
4995 | if (len < 0) { | |
4996 | p = name; | |
4997 | while (is_idchar[*p]) | |
4998 | p++; | |
4999 | len = p - name; | |
5000 | } | |
5001 | ||
5002 | if (hash < 0) | |
5003 | hash = hashf (name, len, HASHSIZE); | |
5004 | ||
5005 | i = sizeof (HASHNODE) + len + 1; | |
5006 | hp = (HASHNODE *) xmalloc (i); | |
5007 | bucket = hash; | |
5008 | hp->bucket_hdr = &hashtab[bucket]; | |
5009 | hp->next = hashtab[bucket]; | |
5010 | hashtab[bucket] = hp; | |
5011 | hp->prev = NULL; | |
5012 | if (hp->next != NULL) | |
5013 | hp->next->prev = hp; | |
5014 | hp->type = type; | |
5015 | hp->length = len; | |
5016 | hp->value.ival = value; | |
5017 | hp->name = ((U_CHAR *) hp) + sizeof (HASHNODE); | |
5018 | p = hp->name; | |
5019 | q = name; | |
5020 | for (i = 0; i < len; i++) | |
5021 | *p++ = *q++; | |
5022 | hp->name[len] = 0; | |
5023 | return hp; | |
5024 | } | |
5025 | ||
5026 | /* | |
5027 | * find the most recent hash node for name name (ending with first | |
5028 | * non-identifier char) installed by install | |
5029 | * | |
5030 | * If LEN is >= 0, it is the length of the name. | |
5031 | * Otherwise, compute the length by scanning the entire name. | |
5032 | * | |
5033 | * If HASH is >= 0, it is the precomputed hash code. | |
5034 | * Otherwise, compute the hash code. | |
5035 | */ | |
5036 | HASHNODE * | |
5037 | lookup (name, len, hash) | |
5038 | U_CHAR *name; | |
5039 | int len; | |
5040 | int hash; | |
5041 | { | |
5042 | register U_CHAR *bp; | |
5043 | register HASHNODE *bucket; | |
5044 | ||
5045 | if (len < 0) { | |
5046 | for (bp = name; is_idchar[*bp]; bp++) ; | |
5047 | len = bp - name; | |
5048 | } | |
5049 | ||
5050 | if (hash < 0) | |
5051 | hash = hashf (name, len, HASHSIZE); | |
5052 | ||
5053 | bucket = hashtab[hash]; | |
5054 | while (bucket) { | |
5055 | if (bucket->length == len && strncmp (bucket->name, name, len) == 0) | |
5056 | return bucket; | |
5057 | bucket = bucket->next; | |
5058 | } | |
5059 | return NULL; | |
5060 | } | |
5061 | ||
5062 | /* | |
5063 | * Delete a hash node. Some weirdness to free junk from macros. | |
5064 | * More such weirdness will have to be added if you define more hash | |
5065 | * types that need it. | |
5066 | */ | |
5067 | ||
5068 | /* Note that the DEFINITION of a macro is removed from the hash table | |
5069 | but its storage is not freed. This would be a storage leak | |
5070 | except that it is not reasonable to keep undefining and redefining | |
5071 | large numbers of macros many times. | |
5072 | In any case, this is necessary, because a macro can be #undef'd | |
5073 | in the middle of reading the arguments to a call to it. | |
5074 | If #undef freed the DEFINITION, that would crash. */ | |
5075 | ||
5076 | delete_macro (hp) | |
5077 | HASHNODE *hp; | |
5078 | { | |
5079 | ||
5080 | if (hp->prev != NULL) | |
5081 | hp->prev->next = hp->next; | |
5082 | if (hp->next != NULL) | |
5083 | hp->next->prev = hp->prev; | |
5084 | ||
5085 | /* make sure that the bucket chain header that | |
5086 | the deleted guy was on points to the right thing afterwards. */ | |
5087 | if (hp == *hp->bucket_hdr) | |
5088 | *hp->bucket_hdr = hp->next; | |
5089 | ||
5090 | #if 0 | |
5091 | if (hp->type == T_MACRO) { | |
5092 | DEFINITION *d = hp->value.defn; | |
5093 | struct reflist *ap, *nextap; | |
5094 | ||
5095 | for (ap = d->pattern; ap != NULL; ap = nextap) { | |
5096 | nextap = ap->next; | |
5097 | free (ap); | |
5098 | } | |
5099 | free (d); | |
5100 | } | |
5101 | #endif | |
5102 | free (hp); | |
5103 | } | |
5104 | ||
5105 | /* | |
5106 | * return hash function on name. must be compatible with the one | |
5107 | * computed a step at a time, elsewhere | |
5108 | */ | |
5109 | int | |
5110 | hashf (name, len, hashsize) | |
5111 | register U_CHAR *name; | |
5112 | register int len; | |
5113 | int hashsize; | |
5114 | { | |
5115 | register int r = 0; | |
5116 | ||
5117 | while (len--) | |
5118 | r = HASHSTEP (r, *name++); | |
5119 | ||
5120 | return MAKE_POS (r) % hashsize; | |
5121 | } | |
5122 | \f | |
5123 | /* Dump all macro definitions as #defines to stdout. */ | |
5124 | ||
5125 | void | |
5126 | dump_all_macros () | |
5127 | { | |
5128 | int bucket; | |
5129 | ||
5130 | for (bucket = 0; bucket < HASHSIZE; bucket++) { | |
5131 | register HASHNODE *hp; | |
5132 | ||
5133 | for (hp = hashtab[bucket]; hp; hp= hp->next) { | |
5134 | if (hp->type == T_MACRO) { | |
5135 | register DEFINITION *defn = hp->value.defn; | |
5136 | struct reflist *ap; | |
5137 | int offset; | |
5138 | int concat; | |
5139 | ||
5140 | ||
5141 | /* Print the definition of the macro HP. */ | |
5142 | ||
5143 | printf ("#define %s", hp->name); | |
5144 | if (defn->nargs >= 0) { | |
5145 | int i; | |
5146 | ||
5147 | printf ("("); | |
5148 | for (i = 0; i < defn->nargs; i++) { | |
5149 | dump_arg_n (defn, i); | |
5150 | if (i + 1 < defn->nargs) | |
5151 | printf (", "); | |
5152 | } | |
5153 | printf (")"); | |
5154 | } | |
5155 | ||
5156 | printf (" "); | |
5157 | ||
5158 | offset = 0; | |
5159 | concat = 0; | |
5160 | for (ap = defn->pattern; ap != NULL; ap = ap->next) { | |
5161 | dump_defn_1 (defn->expansion, offset, ap->nchars); | |
5162 | if (ap->nchars != 0) | |
5163 | concat = 0; | |
5164 | offset += ap->nchars; | |
5165 | if (ap->stringify) | |
5166 | printf (" #"); | |
5167 | if (ap->raw_before && !concat) | |
5168 | printf (" ## "); | |
5169 | concat = 0; | |
5170 | dump_arg_n (defn, ap->argno); | |
5171 | if (ap->raw_after) { | |
5172 | printf (" ## "); | |
5173 | concat = 1; | |
5174 | } | |
5175 | } | |
5176 | dump_defn_1 (defn->expansion, offset, defn->length - offset); | |
5177 | printf ("\n"); | |
5178 | } | |
5179 | } | |
5180 | } | |
5181 | } | |
5182 | ||
5183 | /* Output to stdout a substring of a macro definition. | |
5184 | BASE is the beginning of the definition. | |
5185 | Output characters START thru LENGTH. | |
5186 | Discard newlines outside of strings, thus | |
5187 | converting funny-space markers to ordinary spaces. */ | |
5188 | ||
5189 | dump_defn_1 (base, start, length) | |
5190 | U_CHAR *base; | |
5191 | int start; | |
5192 | int length; | |
5193 | { | |
5194 | U_CHAR *p = base + start; | |
5195 | U_CHAR *limit = base + start + length; | |
5196 | ||
5197 | while (p < limit) { | |
5198 | if (*p != '\n') | |
5199 | putchar (*p); | |
5200 | else if (*p == '\"' || *p =='\'') { | |
5201 | U_CHAR *p1 = skip_quoted_string (p, limit, 0, 0, 0, 0); | |
5202 | fwrite (p, p1 - p, 1, stdout); | |
5203 | p = p1 - 1; | |
5204 | } | |
5205 | p++; | |
5206 | } | |
5207 | } | |
5208 | ||
5209 | /* Print the name of argument number ARGNUM of macro definition DEFN. | |
5210 | Recall that DEFN->argnames contains all the arg names | |
5211 | concatenated in reverse order with comma-space in between. */ | |
5212 | ||
5213 | dump_arg_n (defn, argnum) | |
5214 | DEFINITION *defn; | |
5215 | int argnum; | |
5216 | { | |
5217 | register U_CHAR *p = defn->argnames; | |
5218 | while (argnum + 1 < defn->nargs) { | |
5219 | p = (U_CHAR *) index (p, ' ') + 1; | |
5220 | argnum++; | |
5221 | } | |
5222 | ||
5223 | while (*p && *p != ',') { | |
5224 | putchar (*p); | |
5225 | p++; | |
5226 | } | |
5227 | } | |
5228 | \f | |
5229 | /* Initialize syntactic classifications of characters. */ | |
5230 | ||
5231 | initialize_char_syntax () | |
5232 | { | |
5233 | register int i; | |
5234 | ||
5235 | /* | |
5236 | * Set up is_idchar and is_idstart tables. These should be | |
5237 | * faster than saying (is_alpha (c) || c == '_'), etc. | |
5238 | * Must do set up these things before calling any routines tthat | |
5239 | * refer to them. | |
5240 | */ | |
5241 | for (i = 'a'; i <= 'z'; i++) { | |
5242 | is_idchar[i - 'a' + 'A'] = 1; | |
5243 | is_idchar[i] = 1; | |
5244 | is_idstart[i - 'a' + 'A'] = 1; | |
5245 | is_idstart[i] = 1; | |
5246 | } | |
5247 | for (i = '0'; i <= '9'; i++) | |
5248 | is_idchar[i] = 1; | |
5249 | is_idchar['_'] = 1; | |
5250 | is_idstart['_'] = 1; | |
5251 | is_idchar['$'] = dollars_in_ident; | |
5252 | is_idstart['$'] = dollars_in_ident; | |
5253 | ||
5254 | /* horizontal space table */ | |
5255 | is_hor_space[' '] = 1; | |
5256 | is_hor_space['\t'] = 1; | |
5257 | is_hor_space['\v'] = 1; | |
5258 | is_hor_space['\f'] = 1; | |
5259 | is_hor_space['\r'] = 1; | |
5260 | ||
5261 | is_space[' '] = 1; | |
5262 | is_space['\t'] = 1; | |
5263 | is_space['\v'] = 1; | |
5264 | is_space['\f'] = 1; | |
5265 | is_space['\n'] = 1; | |
5266 | is_space['\r'] = 1; | |
5267 | } | |
5268 | ||
5269 | /* Initialize the built-in macros. */ | |
5270 | ||
5271 | initialize_builtins () | |
5272 | { | |
5273 | install ("__LINE__", -1, T_SPECLINE, 0, -1); | |
5274 | install ("__DATE__", -1, T_DATE, 0, -1); | |
5275 | install ("__FILE__", -1, T_FILE, 0, -1); | |
5276 | install ("__BASE_FILE__", -1, T_BASE_FILE, 0, -1); | |
5277 | install ("__INCLUDE_LEVEL__", -1, T_INCLUDE_LEVEL, 0, -1); | |
5278 | install ("__VERSION__", -1, T_VERSION, 0, -1); | |
5279 | install ("__TIME__", -1, T_TIME, 0, -1); | |
5280 | if (!traditional) | |
5281 | install ("__STDC__", -1, T_CONST, STDC_VALUE, -1); | |
5282 | /* install ("__GNU__", -1, T_CONST, 1, -1); */ | |
5283 | /* This is supplied using a -D by the compiler driver | |
5284 | so that it is present only when truly compiling with GNU C. */ | |
5285 | } | |
5286 | \f | |
5287 | /* | |
5288 | * process a given definition string, for initialization | |
5289 | * If STR is just an identifier, define it with value 1. | |
5290 | * If STR has anything after the identifier, then it should | |
5291 | * be identifier-space-definition. | |
5292 | */ | |
5293 | make_definition (str) | |
5294 | U_CHAR *str; | |
5295 | { | |
5296 | FILE_BUF *ip; | |
5297 | struct directive *kt; | |
5298 | U_CHAR *buf, *p; | |
5299 | ||
5300 | buf = str; | |
5301 | p = str; | |
5302 | while (is_idchar[*p]) p++; | |
5303 | if (p == str) { | |
5304 | error ("malformed option `-D %s'", str); | |
5305 | return; | |
5306 | } | |
5307 | if (*p == 0) { | |
5308 | buf = (U_CHAR *) alloca (p - buf + 4); | |
5309 | strcpy ((char *)buf, str); | |
5310 | strcat ((char *)buf, " 1"); | |
5311 | } else if (*p != ' ') { | |
5312 | error ("malformed option `-D %s'", str); | |
5313 | return; | |
5314 | } else { | |
5315 | U_CHAR *q; | |
5316 | /* Copy the entire option so we can modify it. */ | |
5317 | buf = (U_CHAR *) alloca (2 * strlen (str) + 1); | |
5318 | strncpy (buf, str, p - str); | |
5319 | /* Change the = to a space. */ | |
5320 | buf[p - str] = ' '; | |
5321 | /* Scan for any backslash-newline and remove it. */ | |
5322 | p++; | |
5323 | q = &buf[p - str]; | |
5324 | while (*p) { | |
5325 | if (*p == '\\' && p[1] == '\n') | |
5326 | p += 2; | |
5327 | /* Change newline chars into newline-markers. */ | |
5328 | else if (*p == '\n') | |
5329 | { | |
5330 | *q++ = '\n'; | |
5331 | *q++ = '\n'; | |
5332 | p++; | |
5333 | } | |
5334 | else | |
5335 | *q++ = *p++; | |
5336 | } | |
5337 | *q = 0; | |
5338 | } | |
5339 | ||
5340 | ip = &instack[++indepth]; | |
5341 | ip->fname = "*Initialization*"; | |
5342 | ||
5343 | ip->buf = ip->bufp = buf; | |
5344 | ip->length = strlen (buf); | |
5345 | ip->lineno = 1; | |
5346 | ip->macro = 0; | |
5347 | ip->free_ptr = 0; | |
5348 | ip->if_stack = if_stack; | |
5349 | ||
5350 | for (kt = directive_table; kt->type != T_DEFINE; kt++) | |
5351 | ; | |
5352 | ||
5353 | /* pass NULL as output ptr to do_define since we KNOW it never | |
5354 | does any output.... */ | |
5355 | do_define (buf, buf + strlen (buf) , NULL, kt); | |
5356 | --indepth; | |
5357 | } | |
5358 | ||
5359 | /* JF, this does the work for the -U option */ | |
5360 | make_undef (str) | |
5361 | U_CHAR *str; | |
5362 | { | |
5363 | FILE_BUF *ip; | |
5364 | struct directive *kt; | |
5365 | ||
5366 | ip = &instack[++indepth]; | |
5367 | ip->fname = "*undef*"; | |
5368 | ||
5369 | ip->buf = ip->bufp = str; | |
5370 | ip->length = strlen (str); | |
5371 | ip->lineno = 1; | |
5372 | ip->macro = 0; | |
5373 | ip->free_ptr = 0; | |
5374 | ip->if_stack = if_stack; | |
5375 | ||
5376 | for (kt = directive_table; kt->type != T_UNDEF; kt++) | |
5377 | ; | |
5378 | ||
5379 | do_undef (str,str + strlen (str) - 1, NULL, kt); | |
5380 | --indepth; | |
5381 | } | |
5382 | \f | |
5383 | /* Add output to `deps_buffer' for the -M switch. | |
5384 | STRING points to the text to be output. | |
5385 | SIZE is the number of bytes, or 0 meaning output until a null. | |
5386 | If SIZE is nonzero, we break the line first, if it is long enough. */ | |
5387 | ||
5388 | deps_output (string, size) | |
5389 | char *string; | |
5390 | int size; | |
5391 | { | |
5392 | #ifndef MAX_OUTPUT_COLUMNS | |
5393 | #define MAX_OUTPUT_COLUMNS 75 | |
5394 | #endif | |
5395 | if (size != 0 && deps_column != 0 | |
5396 | && size + deps_column > MAX_OUTPUT_COLUMNS) { | |
5397 | deps_output ("\\\n ", 0); | |
5398 | deps_column = 0; | |
5399 | } | |
5400 | ||
5401 | if (size == 0) | |
5402 | size = strlen (string); | |
5403 | ||
5404 | if (deps_size + size + 1 > deps_allocated_size) { | |
5405 | deps_allocated_size = deps_size + size + 50; | |
5406 | deps_allocated_size *= 2; | |
5407 | deps_buffer = (char *) xrealloc (deps_buffer, deps_allocated_size); | |
5408 | } | |
5409 | bcopy (string, &deps_buffer[deps_size], size); | |
5410 | deps_size += size; | |
5411 | deps_column += size; | |
5412 | deps_buffer[deps_size] = 0; | |
5413 | } | |
5414 | \f | |
5415 | #ifndef BSD | |
5416 | #ifndef BSTRING | |
5417 | ||
5418 | void | |
5419 | bzero (b, length) | |
5420 | register char *b; | |
5421 | register int length; | |
5422 | { | |
5423 | #ifdef VMS | |
5424 | short zero = 0; | |
5425 | long max_str = 65535; | |
5426 | ||
5427 | while (length > max_str) { | |
5428 | (void) LIB$MOVC5 (&zero, &zero, &zero, &max_str, b); | |
5429 | length -= max_str; | |
5430 | b += max_str; | |
5431 | } | |
5432 | (void) LIB$MOVC5 (&zero, &zero, &zero, &length, b); | |
5433 | #else | |
5434 | while (length-- > 0) | |
5435 | *b++ = 0; | |
5436 | #endif /* not VMS */ | |
5437 | } | |
5438 | ||
5439 | void | |
5440 | bcopy (b1, b2, length) | |
5441 | register char *b1; | |
5442 | register char *b2; | |
5443 | register int length; | |
5444 | { | |
5445 | #ifdef VMS | |
5446 | long max_str = 65535; | |
5447 | ||
5448 | while (length > max_str) { | |
5449 | (void) LIB$MOVC3 (&max_str, b1, b2); | |
5450 | length -= max_str; | |
5451 | b1 += max_str; | |
5452 | b2 += max_str; | |
5453 | } | |
5454 | (void) LIB$MOVC3 (&length, b1, b2); | |
5455 | #else | |
5456 | while (length-- > 0) | |
5457 | *b2++ = *b1++; | |
5458 | #endif /* not VMS */ | |
5459 | } | |
5460 | ||
5461 | int | |
5462 | bcmp (b1, b2, length) /* This could be a macro! */ | |
5463 | register char *b1; | |
5464 | register char *b2; | |
5465 | register int length; | |
5466 | { | |
5467 | #ifdef VMS | |
5468 | struct dsc$descriptor_s src1 = {length, DSC$K_DTYPE_T, DSC$K_CLASS_S, b1}; | |
5469 | struct dsc$descriptor_s src2 = {length, DSC$K_DTYPE_T, DSC$K_CLASS_S, b2}; | |
5470 | ||
5471 | return STR$COMPARE (&src1, &src2); | |
5472 | #else | |
5473 | while (length-- > 0) | |
5474 | if (*b1++ != *b2++) | |
5475 | return 1; | |
5476 | ||
5477 | return 0; | |
5478 | #endif /* not VMS */ | |
5479 | } | |
5480 | #endif /* not BSTRING */ | |
5481 | #endif /* not BSD */ | |
5482 | ||
5483 | \f | |
5484 | void | |
5485 | fatal (str, arg) | |
5486 | char *str, *arg; | |
5487 | { | |
5488 | fprintf (stderr, "%s: ", progname); | |
5489 | fprintf (stderr, str, arg); | |
5490 | fprintf (stderr, "\n"); | |
5491 | exit (FATAL_EXIT_CODE); | |
5492 | } | |
5493 | ||
5494 | /* More 'friendly' abort that prints the line and file. | |
5495 | config.h can #define abort fancy_abort if you like that sort of thing. */ | |
5496 | ||
5497 | void | |
5498 | fancy_abort () | |
5499 | { | |
5500 | fatal ("Internal gcc abort."); | |
5501 | } | |
5502 | ||
5503 | void | |
5504 | perror_with_name (name) | |
5505 | char *name; | |
5506 | { | |
5507 | extern int errno, sys_nerr; | |
5508 | extern char *sys_errlist[]; | |
5509 | ||
5510 | fprintf (stderr, "%s: ", progname); | |
5511 | if (errno < sys_nerr) | |
5512 | fprintf (stderr, "%s: %s\n", name, sys_errlist[errno]); | |
5513 | else | |
5514 | fprintf (stderr, "%s: undocumented I/O error\n", name); | |
5515 | errors++; | |
5516 | } | |
5517 | ||
5518 | void | |
5519 | pfatal_with_name (name) | |
5520 | char *name; | |
5521 | { | |
5522 | perror_with_name (name); | |
5523 | #ifdef VMS | |
5524 | exit (vaxc$errno); | |
5525 | #else | |
5526 | exit (FATAL_EXIT_CODE); | |
5527 | #endif | |
5528 | } | |
5529 | ||
5530 | \f | |
5531 | void | |
5532 | memory_full () | |
5533 | { | |
5534 | fatal ("Memory exhausted."); | |
5535 | } | |
5536 | ||
5537 | ||
5538 | char * | |
5539 | xmalloc (size) | |
5540 | int size; | |
5541 | { | |
5542 | extern char *malloc (); | |
5543 | register char *ptr = malloc (size); | |
5544 | if (ptr != 0) return (ptr); | |
5545 | memory_full (); | |
5546 | /*NOTREACHED*/ | |
5547 | } | |
5548 | ||
5549 | char * | |
5550 | xrealloc (old, size) | |
5551 | char *old; | |
5552 | int size; | |
5553 | { | |
5554 | extern char *realloc (); | |
5555 | register char *ptr = realloc (old, size); | |
5556 | if (ptr != 0) return (ptr); | |
5557 | memory_full (); | |
5558 | /*NOTREACHED*/ | |
5559 | } | |
5560 | ||
5561 | char * | |
5562 | xcalloc (number, size) | |
5563 | int number, size; | |
5564 | { | |
5565 | extern char *malloc (); | |
5566 | register int total = number * size; | |
5567 | register char *ptr = malloc (total); | |
5568 | if (ptr != 0) { | |
5569 | if (total > 100) | |
5570 | bzero (ptr, total); | |
5571 | else { | |
5572 | /* It's not too long, so loop, zeroing by longs. | |
5573 | It must be safe because malloc values are always well aligned. */ | |
5574 | register long *zp = (long *) ptr; | |
5575 | register long *zl = (long *) (ptr + total - 4); | |
5576 | register int i = total - 4; | |
5577 | while (zp < zl) | |
5578 | *zp++ = 0; | |
5579 | if (i < 0) | |
5580 | i = 0; | |
5581 | while (i < total) | |
5582 | ptr[i++] = 0; | |
5583 | } | |
5584 | return ptr; | |
5585 | } | |
5586 | memory_full (); | |
5587 | /*NOTREACHED*/ | |
5588 | } | |
5589 | ||
5590 | char * | |
5591 | savestring (input) | |
5592 | char *input; | |
5593 | { | |
5594 | int size = strlen (input); | |
5595 | char *output = xmalloc (size + 1); | |
5596 | strcpy (output, input); | |
5597 | return output; | |
5598 | } | |
5599 | \f | |
5600 | /* Get the file-mode and data size of the file open on FD | |
5601 | and store them in *MODE_POINTER and *SIZE_POINTER. */ | |
5602 | ||
5603 | int | |
5604 | file_size_and_mode (fd, mode_pointer, size_pointer) | |
5605 | int fd; | |
5606 | int *mode_pointer; | |
5607 | long int *size_pointer; | |
5608 | { | |
5609 | struct stat sbuf; | |
5610 | ||
5611 | if (fstat (fd, &sbuf) < 0) return (-1); | |
5612 | if (mode_pointer) *mode_pointer = sbuf.st_mode; | |
5613 | if (size_pointer) *size_pointer = sbuf.st_size; | |
5614 | return 0; | |
5615 | } | |
5616 | \f | |
5617 | #ifdef VMS | |
5618 | ||
5619 | /* Under VMS we need to fix up the "include" specification | |
5620 | filename so that everything following the 1st slash is | |
5621 | changed into its correct VMS file specification. */ | |
5622 | ||
5623 | hack_vms_include_specification (fname) | |
5624 | char *fname; | |
5625 | { | |
5626 | register char *cp, *cp1, *cp2; | |
5627 | char Local[512]; | |
5628 | extern char *index (), *rindex (); | |
5629 | ||
5630 | /* Ignore leading "./"s */ | |
5631 | while (fname[0] == '.' && fname[1] == '/') | |
5632 | strcpy (fname, fname+2); | |
5633 | /* Look for the boundary between the VMS and UNIX filespecs */ | |
5634 | cp = rindex (fname, ']'); /* Look for end of dirspec. */ | |
5635 | if (cp == 0) cp == rindex (fname, '>'); /* ... Ditto */ | |
5636 | if (cp == 0) cp == rindex (fname, ':'); /* Look for end of devspec. */ | |
5637 | if (cp) { | |
5638 | cp++; | |
5639 | } else { | |
5640 | cp = index (fname, '/'); /* Look for the "/" */ | |
5641 | } | |
5642 | /* See if we found that 1st slash */ | |
5643 | if (cp == 0) return; /* Nothing to do!!! */ | |
5644 | if (*cp != '/') return; /* Nothing to do!!! */ | |
5645 | /* Point to the UNIX filename part (which needs to be fixed!) */ | |
5646 | cp1 = cp+1; | |
5647 | /* If the directory spec is not rooted, we can just copy | |
5648 | the UNIX filename part and we are done */ | |
5649 | if (((cp - fname) > 2) | |
5650 | && ((cp[-1] == ']') || (cp[-1] == '>')) | |
5651 | && (cp[-2] != '.')) { | |
5652 | strcpy (cp, cp1); | |
5653 | return; | |
5654 | } | |
5655 | /* If there are no other slashes then the filename will be | |
5656 | in the "root" directory. Otherwise, we need to add | |
5657 | directory specifications. */ | |
5658 | if (index (cp1, '/') == 0) { | |
5659 | /* Just add "[000000]" as the directory string */ | |
5660 | strcpy (Local, "[000000]"); | |
5661 | cp2 = Local + strlen (Local); | |
5662 | } else { | |
5663 | /* Open the directory specification */ | |
5664 | cp2 = Local; | |
5665 | *cp2++ = '['; | |
5666 | /* As long as there are still subdirectories to add, do them. */ | |
5667 | while (index (cp1, '/') != 0) { | |
5668 | /* If this token is "." we can ignore it */ | |
5669 | if ((cp1[0] == '.') && (cp1[1] == '/')) { | |
5670 | cp1 += 2; | |
5671 | continue; | |
5672 | } | |
5673 | /* Add a subdirectory spec. */ | |
5674 | if (cp2 != Local+1) *cp2++ = '.'; | |
5675 | /* If this is ".." then the spec becomes "-" */ | |
5676 | if ((cp1[0] == '.') && (cp1[1] == '.') && (cp[2] == '/')) { | |
5677 | /* Add "-" and skip the ".." */ | |
5678 | *cp2++ = '-'; | |
5679 | cp1 += 3; | |
5680 | continue; | |
5681 | } | |
5682 | /* Copy the subdirectory */ | |
5683 | while (*cp1 != '/') *cp2++= *cp1++; | |
5684 | cp1++; /* Skip the "/" */ | |
5685 | } | |
5686 | /* Close the directory specification */ | |
5687 | *cp2++ = ']'; | |
5688 | } | |
5689 | /* Now add the filename */ | |
5690 | while (*cp1) *cp2++ = *cp1++; | |
5691 | *cp2 = 0; | |
5692 | /* Now append it to the original VMS spec. */ | |
5693 | strcpy (cp, Local); | |
5694 | return; | |
5695 | } | |
5696 | #endif /* VMS */ | |
5697 | \f | |
5698 | #ifdef VMS | |
5699 | ||
5700 | /* These are the read/write replacement routines for | |
5701 | VAX-11 "C". They make read/write behave enough | |
5702 | like their UNIX counterparts that CCCP will work */ | |
5703 | ||
5704 | int | |
5705 | read (fd, buf, size) | |
5706 | int fd; | |
5707 | char *buf; | |
5708 | int size; | |
5709 | { | |
5710 | #undef read /* Get back the REAL read routine */ | |
5711 | register int i; | |
5712 | register int total = 0; | |
5713 | ||
5714 | /* Read until the buffer is exhausted */ | |
5715 | while (size > 0) { | |
5716 | /* Limit each read to 32KB */ | |
5717 | i = (size > (32*1024)) ? (32*1024) : size; | |
5718 | i = read (fd, buf, i); | |
5719 | if (i <= 0) { | |
5720 | if (i == 0) return (total); | |
5721 | return(i); | |
5722 | } | |
5723 | /* Account for this read */ | |
5724 | total += i; | |
5725 | buf += i; | |
5726 | size -= i; | |
5727 | } | |
5728 | return (total); | |
5729 | } | |
5730 | ||
5731 | int | |
5732 | write (fd, buf, size) | |
5733 | int fd; | |
5734 | char *buf; | |
5735 | int size; | |
5736 | { | |
5737 | #undef write /* Get back the REAL write routine */ | |
5738 | int i; | |
5739 | int j; | |
5740 | ||
5741 | /* Limit individual writes to 32Kb */ | |
5742 | i = size; | |
5743 | while (i > 0) { | |
5744 | j = (i > (32*1024)) ? (32*1024) : i; | |
5745 | if (write (fd, buf, j) < 0) return (-1); | |
5746 | /* Account for the data written */ | |
5747 | buf += j; | |
5748 | i -= j; | |
5749 | } | |
5750 | return (size); | |
5751 | } | |
5752 | ||
5753 | #endif /* VMS */ |