filter.c revision 250156
1/* filter - postprocessing of flex output through filters */ 2 3/* This file is part of flex. */ 4 5/* Redistribution and use in source and binary forms, with or without */ 6/* modification, are permitted provided that the following conditions */ 7/* are met: */ 8 9/* 1. Redistributions of source code must retain the above copyright */ 10/* notice, this list of conditions and the following disclaimer. */ 11/* 2. Redistributions in binary form must reproduce the above copyright */ 12/* notice, this list of conditions and the following disclaimer in the */ 13/* documentation and/or other materials provided with the distribution. */ 14 15/* Neither the name of the University nor the names of its contributors */ 16/* may be used to endorse or promote products derived from this software */ 17/* without specific prior written permission. */ 18 19/* THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR */ 20/* IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED */ 21/* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR */ 22/* PURPOSE. */ 23 24#include "flexdef.h" 25static const char * check_4_gnu_m4 = 26 "m4_dnl ifdef(`__gnu__', ," 27 "`errprint(Flex requires GNU M4. Set the PATH or set the M4 environment variable to its path name.)" 28 " m4exit(2)')\n"; 29 30 31/** global chain. */ 32struct filter *output_chain = NULL; 33 34/* Allocate and initialize an external filter. 35 * @param chain the current chain or NULL for new chain 36 * @param cmd the command to execute. 37 * @param ... a NULL terminated list of (const char*) arguments to command, 38 * not including argv[0]. 39 * @return newest filter in chain 40 */ 41struct filter *filter_create_ext (struct filter *chain, const char *cmd, 42 ...) 43{ 44 struct filter *f; 45 int max_args; 46 const char *s; 47 va_list ap; 48 49 /* allocate and initialize new filter */ 50 f = (struct filter *) flex_alloc (sizeof (struct filter)); 51 if (!f) 52 flexerror (_("flex_alloc failed (f) in filter_create_ext")); 53 memset (f, 0, sizeof (*f)); 54 f->filter_func = NULL; 55 f->extra = NULL; 56 f->next = NULL; 57 f->argc = 0; 58 59 if (chain != NULL) { 60 /* append f to end of chain */ 61 while (chain->next) 62 chain = chain->next; 63 chain->next = f; 64 } 65 66 67 /* allocate argv, and populate it with the argument list. */ 68 max_args = 8; 69 f->argv = 70 (const char **) flex_alloc (sizeof (char *) * 71 (max_args + 1)); 72 if (!f->argv) 73 flexerror (_("flex_alloc failed (f->argv) in filter_create_ext")); 74 f->argv[f->argc++] = cmd; 75 76 va_start (ap, cmd); 77 while ((s = va_arg (ap, const char *)) != NULL) { 78 if (f->argc >= max_args) { 79 max_args += 8; 80 f->argv = 81 (const char **) flex_realloc (f->argv, 82 sizeof (char 83 *) * 84 (max_args + 85 1)); 86 } 87 f->argv[f->argc++] = s; 88 } 89 f->argv[f->argc] = NULL; 90 91 va_end (ap); 92 return f; 93} 94 95/* Allocate and initialize an internal filter. 96 * @param chain the current chain or NULL for new chain 97 * @param filter_func The function that will perform the filtering. 98 * filter_func should return 0 if successful, and -1 99 * if an error occurs -- or it can simply exit(). 100 * @param extra optional user-defined data to pass to the filter. 101 * @return newest filter in chain 102 */ 103struct filter *filter_create_int (struct filter *chain, 104 int (*filter_func) (struct filter *), 105 void *extra) 106{ 107 struct filter *f; 108 109 /* allocate and initialize new filter */ 110 f = (struct filter *) flex_alloc (sizeof (struct filter)); 111 if (!f) 112 flexerror (_("flex_alloc failed in filter_create_int")); 113 memset (f, 0, sizeof (*f)); 114 f->next = NULL; 115 f->argc = 0; 116 f->argv = NULL; 117 118 f->filter_func = filter_func; 119 f->extra = extra; 120 121 if (chain != NULL) { 122 /* append f to end of chain */ 123 while (chain->next) 124 chain = chain->next; 125 chain->next = f; 126 } 127 128 return f; 129} 130 131/** Fork and exec entire filter chain. 132 * @param chain The head of the chain. 133 * @return true on success. 134 */ 135bool filter_apply_chain (struct filter * chain) 136{ 137 int pid, pipes[2]; 138 int r; 139 const int readsz = 512; 140 char *buf; 141 142 143 /* Tricky recursion, since we want to begin the chain 144 * at the END. Why? Because we need all the forked processes 145 * to be children of the main flex process. 146 */ 147 if (chain) 148 filter_apply_chain (chain->next); 149 else 150 return true; 151 152 /* Now we are the right-most unprocessed link in the chain. 153 */ 154 155 fflush (stdout); 156 fflush (stderr); 157 158 159 if (pipe (pipes) == -1) 160 flexerror (_("pipe failed")); 161 162 if ((pid = fork ()) == -1) 163 flexerror (_("fork failed")); 164 165 if (pid == 0) { 166 /* child */ 167 168 /* We need stdin (the FILE* stdin) to connect to this new pipe. 169 * There is no portable way to set stdin to a new file descriptor, 170 * as stdin is not an lvalue on some systems (BSD). 171 * So we dup the new pipe onto the stdin descriptor and use a no-op fseek 172 * to sync the stream. This is a Hail Mary situation. It seems to work. 173 */ 174 close (pipes[1]); 175clearerr(stdin); 176 if (dup2 (pipes[0], fileno (stdin)) == -1) 177 flexfatal (_("dup2(pipes[0],0)")); 178 close (pipes[0]); 179 fseek (stdin, 0, SEEK_CUR); 180 181 /* run as a filter, either internally or by exec */ 182 if (chain->filter_func) { 183 int r; 184 185 if ((r = chain->filter_func (chain)) == -1) 186 flexfatal (_("filter_func failed")); 187 exit (0); 188 } 189 else { 190 execvp (chain->argv[0], 191 (char **const) (chain->argv)); 192 lerrsf_fatal ( _("exec of %s failed"), 193 chain->argv[0]); 194 } 195 196 exit (1); 197 } 198 199 /* Parent */ 200 close (pipes[0]); 201 if (dup2 (pipes[1], fileno (stdout)) == -1) 202 flexfatal (_("dup2(pipes[1],1)")); 203 close (pipes[1]); 204 fseek (stdout, 0, SEEK_CUR); 205 206 return true; 207} 208 209/** Truncate the chain to max_len number of filters. 210 * @param chain the current chain. 211 * @param max_len the maximum length of the chain. 212 * @return the resulting length of the chain. 213 */ 214int filter_truncate (struct filter *chain, int max_len) 215{ 216 int len = 1; 217 218 if (!chain) 219 return 0; 220 221 while (chain->next && len < max_len) { 222 chain = chain->next; 223 ++len; 224 } 225 226 chain->next = NULL; 227 return len; 228} 229 230/** Splits the chain in order to write to a header file. 231 * Similar in spirit to the 'tee' program. 232 * The header file name is in extra. 233 * @return 0 (zero) on success, and -1 on failure. 234 */ 235int filter_tee_header (struct filter *chain) 236{ 237 /* This function reads from stdin and writes to both the C file and the 238 * header file at the same time. 239 */ 240 241 const int readsz = 512; 242 char *buf; 243 int to_cfd = -1; 244 FILE *to_c = NULL, *to_h = NULL; 245 bool write_header; 246 247 write_header = (chain->extra != NULL); 248 249 /* Store a copy of the stdout pipe, which is already piped to C file 250 * through the running chain. Then create a new pipe to the H file as 251 * stdout, and fork the rest of the chain again. 252 */ 253 254 if ((to_cfd = dup (1)) == -1) 255 flexfatal (_("dup(1) failed")); 256 to_c = fdopen (to_cfd, "w"); 257 258 if (write_header) { 259 if (freopen ((char *) chain->extra, "w", stdout) == NULL) 260 flexfatal (_("freopen(headerfilename) failed")); 261 262 filter_apply_chain (chain->next); 263 to_h = stdout; 264 } 265 266 /* Now to_c is a pipe to the C branch, and to_h is a pipe to the H branch. 267 */ 268 269 if (write_header) { 270 fputs (check_4_gnu_m4, to_h); 271 fputs ("m4_changecom`'m4_dnl\n", to_h); 272 fputs ("m4_changequote`'m4_dnl\n", to_h); 273 fputs ("m4_changequote([[,]])[[]]m4_dnl\n", to_h); 274 fputs ("m4_define([[M4_YY_NOOP]])[[]]m4_dnl\n", to_h); 275 fputs ("m4_define( [[M4_YY_IN_HEADER]],[[]])m4_dnl\n", 276 to_h); 277 fprintf (to_h, "#ifndef %sHEADER_H\n", prefix); 278 fprintf (to_h, "#define %sHEADER_H 1\n", prefix); 279 fprintf (to_h, "#define %sIN_HEADER 1\n\n", prefix); 280 fprintf (to_h, 281 "m4_define( [[M4_YY_OUTFILE_NAME]],[[%s]])m4_dnl\n", 282 headerfilename ? headerfilename : "<stdout>"); 283 284 } 285 286 fputs (check_4_gnu_m4, to_c); 287 fputs ("m4_changecom`'m4_dnl\n", to_c); 288 fputs ("m4_changequote`'m4_dnl\n", to_c); 289 fputs ("m4_changequote([[,]])[[]]m4_dnl\n", to_c); 290 fputs ("m4_define([[M4_YY_NOOP]])[[]]m4_dnl\n", to_c); 291 fprintf (to_c, "m4_define( [[M4_YY_OUTFILE_NAME]],[[%s]])m4_dnl\n", 292 outfilename ? outfilename : "<stdout>"); 293 294 buf = (char *) flex_alloc (readsz); 295 if (!buf) 296 flexerror (_("flex_alloc failed in filter_tee_header")); 297 while (fgets (buf, readsz, stdin)) { 298 fputs (buf, to_c); 299 if (write_header) 300 fputs (buf, to_h); 301 } 302 303 if (write_header) { 304 fprintf (to_h, "\n"); 305 306 /* write a fake line number. It will get fixed by the linedir filter. */ 307 fprintf (to_h, "#line 4000 \"M4_YY_OUTFILE_NAME\"\n"); 308 309 fprintf (to_h, "#undef %sIN_HEADER\n", prefix); 310 fprintf (to_h, "#endif /* %sHEADER_H */\n", prefix); 311 fputs ("m4_undefine( [[M4_YY_IN_HEADER]])m4_dnl\n", to_h); 312 313 fflush (to_h); 314 if (ferror (to_h)) 315 lerrsf (_("error writing output file %s"), 316 (char *) chain->extra); 317 318 else if (fclose (to_h)) 319 lerrsf (_("error closing output file %s"), 320 (char *) chain->extra); 321 } 322 323 fflush (to_c); 324 if (ferror (to_c)) 325 lerrsf (_("error writing output file %s"), 326 outfilename ? outfilename : "<stdout>"); 327 328 else if (fclose (to_c)) 329 lerrsf (_("error closing output file %s"), 330 outfilename ? outfilename : "<stdout>"); 331 332 while (wait (0) > 0) ; 333 334 exit (0); 335 return 0; 336} 337 338/** Adjust the line numbers in the #line directives of the generated scanner. 339 * After the m4 expansion, the line numbers are incorrect since the m4 macros 340 * can add or remove lines. This only adjusts line numbers for generated code, 341 * not user code. This also happens to be a good place to squeeze multiple 342 * blank lines into a single blank line. 343 */ 344int filter_fix_linedirs (struct filter *chain) 345{ 346 char *buf; 347 const int readsz = 512; 348 int lineno = 1; 349 bool in_gen = true; /* in generated code */ 350 bool last_was_blank = false; 351 352 if (!chain) 353 return 0; 354 355 buf = (char *) flex_alloc (readsz); 356 if (!buf) 357 flexerror (_("flex_alloc failed in filter_fix_linedirs")); 358 359 while (fgets (buf, readsz, stdin)) { 360 361 regmatch_t m[10]; 362 363 /* Check for #line directive. */ 364 if (buf[0] == '#' 365 && regexec (®ex_linedir, buf, 3, m, 0) == 0) { 366 367 int num; 368 char *fname; 369 370 /* extract the line number and filename */ 371 num = regmatch_strtol (&m[1], buf, NULL, 0); 372 fname = regmatch_dup (&m[2], buf); 373 374 if (strcmp (fname, 375 outfilename ? outfilename : "<stdout>") 376 == 0 377 || strcmp (fname, 378 headerfilename ? headerfilename : "<stdout>") 379 == 0) { 380 381 char *s1, *s2; 382 char filename[MAXLINE]; 383 384 s1 = fname; 385 s2 = filename; 386 387 while ((s2 - filename) < (MAXLINE - 1) && *s1) { 388 /* Escape the backslash */ 389 if (*s1 == '\\') 390 *s2++ = '\\'; 391 /* Escape the double quote */ 392 if (*s1 == '\"') 393 *s2++ = '\\'; 394 /* Copy the character as usual */ 395 *s2++ = *s1++; 396 } 397 398 *s2 = '\0'; 399 400 /* Adjust the line directives. */ 401 in_gen = true; 402 snprintf (buf, readsz, "#line %d \"%s\"\n", 403 lineno + 1, filename); 404 } 405 else { 406 /* it's a #line directive for code we didn't write */ 407 in_gen = false; 408 } 409 410 free (fname); 411 last_was_blank = false; 412 } 413 414 /* squeeze blank lines from generated code */ 415 else if (in_gen 416 && regexec (®ex_blank_line, buf, 0, NULL, 417 0) == 0) { 418 if (last_was_blank) 419 continue; 420 else 421 last_was_blank = true; 422 } 423 424 else { 425 /* it's a line of normal, non-empty code. */ 426 last_was_blank = false; 427 } 428 429 fputs (buf, stdout); 430 lineno++; 431 } 432 fflush (stdout); 433 if (ferror (stdout)) 434 lerrsf (_("error writing output file %s"), 435 outfilename ? outfilename : "<stdout>"); 436 437 else if (fclose (stdout)) 438 lerrsf (_("error closing output file %s"), 439 outfilename ? outfilename : "<stdout>"); 440 441 return 0; 442} 443 444/* vim:set expandtab cindent tabstop=4 softtabstop=4 shiftwidth=4 textwidth=0: */ 445