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]; 138 139 140 /* Tricky recursion, since we want to begin the chain 141 * at the END. Why? Because we need all the forked processes 142 * to be children of the main flex process. 143 */ 144 if (chain) 145 filter_apply_chain (chain->next); 146 else 147 return true; 148 149 /* Now we are the right-most unprocessed link in the chain. 150 */ 151 152 fflush (stdout); 153 fflush (stderr); 154 155 156 if (pipe (pipes) == -1) 157 flexerror (_("pipe failed")); 158 159 if ((pid = fork ()) == -1) 160 flexerror (_("fork failed")); 161 162 if (pid == 0) { 163 /* child */ 164 165 /* We need stdin (the FILE* stdin) to connect to this new pipe. 166 * There is no portable way to set stdin to a new file descriptor, 167 * as stdin is not an lvalue on some systems (BSD). 168 * So we dup the new pipe onto the stdin descriptor and use a no-op fseek 169 * to sync the stream. This is a Hail Mary situation. It seems to work. 170 */ 171 close (pipes[1]); 172clearerr(stdin); 173 if (dup2 (pipes[0], fileno (stdin)) == -1) 174 flexfatal (_("dup2(pipes[0],0)")); 175 close (pipes[0]); 176 fseek (stdin, 0, SEEK_CUR); 177 178 /* run as a filter, either internally or by exec */ 179 if (chain->filter_func) { 180 int r; 181 182 if ((r = chain->filter_func (chain)) == -1) 183 flexfatal (_("filter_func failed")); 184 exit (0); 185 } 186 else { 187 execvp (chain->argv[0], 188 (char **const) (chain->argv)); 189 lerrsf_fatal ( _("exec of %s failed"), 190 chain->argv[0]); 191 } 192 193 exit (1); 194 } 195 196 /* Parent */ 197 close (pipes[0]); 198 if (dup2 (pipes[1], fileno (stdout)) == -1) 199 flexfatal (_("dup2(pipes[1],1)")); 200 close (pipes[1]); 201 fseek (stdout, 0, SEEK_CUR); 202 203 return true; 204} 205 206/** Truncate the chain to max_len number of filters. 207 * @param chain the current chain. 208 * @param max_len the maximum length of the chain. 209 * @return the resulting length of the chain. 210 */ 211int filter_truncate (struct filter *chain, int max_len) 212{ 213 int len = 1; 214 215 if (!chain) 216 return 0; 217 218 while (chain->next && len < max_len) { 219 chain = chain->next; 220 ++len; 221 } 222 223 chain->next = NULL; 224 return len; 225} 226 227/** Splits the chain in order to write to a header file. 228 * Similar in spirit to the 'tee' program. 229 * The header file name is in extra. 230 * @return 0 (zero) on success, and -1 on failure. 231 */ 232int filter_tee_header (struct filter *chain) 233{ 234 /* This function reads from stdin and writes to both the C file and the 235 * header file at the same time. 236 */ 237 238 const int readsz = 512; 239 char *buf; 240 int to_cfd = -1; 241 FILE *to_c = NULL, *to_h = NULL; 242 bool write_header; 243 244 write_header = (chain->extra != NULL); 245 246 /* Store a copy of the stdout pipe, which is already piped to C file 247 * through the running chain. Then create a new pipe to the H file as 248 * stdout, and fork the rest of the chain again. 249 */ 250 251 if ((to_cfd = dup (1)) == -1) 252 flexfatal (_("dup(1) failed")); 253 to_c = fdopen (to_cfd, "w"); 254 255 if (write_header) { 256 if (freopen ((char *) chain->extra, "w", stdout) == NULL) 257 flexfatal (_("freopen(headerfilename) failed")); 258 259 filter_apply_chain (chain->next); 260 to_h = stdout; 261 } 262 263 /* Now to_c is a pipe to the C branch, and to_h is a pipe to the H branch. 264 */ 265 266 if (write_header) { 267 fputs (check_4_gnu_m4, to_h); 268 fputs ("m4_changecom`'m4_dnl\n", to_h); 269 fputs ("m4_changequote`'m4_dnl\n", to_h); 270 fputs ("m4_changequote([[,]])[[]]m4_dnl\n", to_h); 271 fputs ("m4_define([[M4_YY_NOOP]])[[]]m4_dnl\n", to_h); 272 fputs ("m4_define( [[M4_YY_IN_HEADER]],[[]])m4_dnl\n", 273 to_h); 274 fprintf (to_h, "#ifndef %sHEADER_H\n", prefix); 275 fprintf (to_h, "#define %sHEADER_H 1\n", prefix); 276 fprintf (to_h, "#define %sIN_HEADER 1\n\n", prefix); 277 fprintf (to_h, 278 "m4_define( [[M4_YY_OUTFILE_NAME]],[[%s]])m4_dnl\n", 279 headerfilename ? headerfilename : "<stdout>"); 280 281 } 282 283 fputs (check_4_gnu_m4, to_c); 284 fputs ("m4_changecom`'m4_dnl\n", to_c); 285 fputs ("m4_changequote`'m4_dnl\n", to_c); 286 fputs ("m4_changequote([[,]])[[]]m4_dnl\n", to_c); 287 fputs ("m4_define([[M4_YY_NOOP]])[[]]m4_dnl\n", to_c); 288 fprintf (to_c, "m4_define( [[M4_YY_OUTFILE_NAME]],[[%s]])m4_dnl\n", 289 outfilename ? outfilename : "<stdout>"); 290 291 buf = (char *) flex_alloc (readsz); 292 if (!buf) 293 flexerror (_("flex_alloc failed in filter_tee_header")); 294 while (fgets (buf, readsz, stdin)) { 295 fputs (buf, to_c); 296 if (write_header) 297 fputs (buf, to_h); 298 } 299 300 if (write_header) { 301 fprintf (to_h, "\n"); 302 303 /* write a fake line number. It will get fixed by the linedir filter. */ 304 fprintf (to_h, "#line 4000 \"M4_YY_OUTFILE_NAME\"\n"); 305 306 fprintf (to_h, "#undef %sIN_HEADER\n", prefix); 307 fprintf (to_h, "#endif /* %sHEADER_H */\n", prefix); 308 fputs ("m4_undefine( [[M4_YY_IN_HEADER]])m4_dnl\n", to_h); 309 310 fflush (to_h); 311 if (ferror (to_h)) 312 lerrsf (_("error writing output file %s"), 313 (char *) chain->extra); 314 315 else if (fclose (to_h)) 316 lerrsf (_("error closing output file %s"), 317 (char *) chain->extra); 318 } 319 320 fflush (to_c); 321 if (ferror (to_c)) 322 lerrsf (_("error writing output file %s"), 323 outfilename ? outfilename : "<stdout>"); 324 325 else if (fclose (to_c)) 326 lerrsf (_("error closing output file %s"), 327 outfilename ? outfilename : "<stdout>"); 328 329 while (wait (0) > 0) ; 330 331 exit (0); 332 return 0; 333} 334 335/** Adjust the line numbers in the #line directives of the generated scanner. 336 * After the m4 expansion, the line numbers are incorrect since the m4 macros 337 * can add or remove lines. This only adjusts line numbers for generated code, 338 * not user code. This also happens to be a good place to squeeze multiple 339 * blank lines into a single blank line. 340 */ 341int filter_fix_linedirs (struct filter *chain) 342{ 343 char *buf; 344 const int readsz = 512; 345 int lineno = 1; 346 bool in_gen = true; /* in generated code */ 347 bool last_was_blank = false; 348 349 if (!chain) 350 return 0; 351 352 buf = (char *) flex_alloc (readsz); 353 if (!buf) 354 flexerror (_("flex_alloc failed in filter_fix_linedirs")); 355 356 while (fgets (buf, readsz, stdin)) { 357 358 regmatch_t m[10]; 359 360 /* Check for #line directive. */ 361 if (buf[0] == '#' 362 && regexec (®ex_linedir, buf, 3, m, 0) == 0) { 363 364 int num; 365 char *fname; 366 367 /* extract the line number and filename */ 368 num = regmatch_strtol (&m[1], buf, NULL, 0); 369 fname = regmatch_dup (&m[2], buf); 370 371 if (strcmp (fname, 372 outfilename ? outfilename : "<stdout>") 373 == 0 374 || strcmp (fname, 375 headerfilename ? headerfilename : "<stdout>") 376 == 0) { 377 378 char *s1, *s2; 379 char filename[MAXLINE]; 380 381 s1 = fname; 382 s2 = filename; 383 384 while ((s2 - filename) < (MAXLINE - 1) && *s1) { 385 /* Escape the backslash */ 386 if (*s1 == '\\') 387 *s2++ = '\\'; 388 /* Escape the double quote */ 389 if (*s1 == '\"') 390 *s2++ = '\\'; 391 /* Copy the character as usual */ 392 *s2++ = *s1++; 393 } 394 395 *s2 = '\0'; 396 397 /* Adjust the line directives. */ 398 in_gen = true; 399 snprintf (buf, readsz, "#line %d \"%s\"\n", 400 lineno + 1, filename); 401 } 402 else { 403 /* it's a #line directive for code we didn't write */ 404 in_gen = false; 405 } 406 407 free (fname); 408 last_was_blank = false; 409 } 410 411 /* squeeze blank lines from generated code */ 412 else if (in_gen 413 && regexec (®ex_blank_line, buf, 0, NULL, 414 0) == 0) { 415 if (last_was_blank) 416 continue; 417 else 418 last_was_blank = true; 419 } 420 421 else { 422 /* it's a line of normal, non-empty code. */ 423 last_was_blank = false; 424 } 425 426 fputs (buf, stdout); 427 lineno++; 428 } 429 fflush (stdout); 430 if (ferror (stdout)) 431 lerrsf (_("error writing output file %s"), 432 outfilename ? outfilename : "<stdout>"); 433 434 else if (fclose (stdout)) 435 lerrsf (_("error closing output file %s"), 436 outfilename ? outfilename : "<stdout>"); 437 438 return 0; 439} 440 441/* vim:set expandtab cindent tabstop=4 softtabstop=4 shiftwidth=4 textwidth=0: */ 442