1 /* input_scrub.c - Break up input buffers into whole numbers of lines. 2 Copyright 1987, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 3 2000, 2001, 2003, 2006, 2007 4 Free Software Foundation, Inc. 5 6 This file is part of GAS, the GNU Assembler. 7 8 GAS is free software; you can redistribute it and/or modify 9 it under the terms of the GNU General Public License as published by 10 the Free Software Foundation; either version 3, or (at your option) 11 any later version. 12 13 GAS is distributed in the hope that it will be useful, 14 but WITHOUT ANY WARRANTY; without even the implied warranty of 15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 16 GNU General Public License for more details. 17 18 You should have received a copy of the GNU General Public License 19 along with GAS; see the file COPYING. If not, write to the Free 20 Software Foundation, 51 Franklin Street - Fifth Floor, Boston, MA 21 02110-1301, USA. */ 22 23 #include "as.h" 24 #include "input-file.h" 25 #include "sb.h" 26 #include "listing.h" 27 28 /* 29 * O/S independent module to supply buffers of sanitised source code 30 * to rest of assembler. We get sanitised input data of arbitrary length. 31 * We break these buffers on line boundaries, recombine pieces that 32 * were broken across buffers, and return a buffer of full lines to 33 * the caller. 34 * The last partial line begins the next buffer we build and return to caller. 35 * The buffer returned to caller is preceded by BEFORE_STRING and followed 36 * by AFTER_STRING, as sentinels. The last character before AFTER_STRING 37 * is a newline. 38 * Also looks after line numbers, for e.g. error messages. 39 */ 40 41 /* 42 * We don't care how filthy our buffers are, but our callers assume 43 * that the following sanitation has already been done. 44 * 45 * No comments, reduce a comment to a space. 46 * Reduce a tab to a space unless it is 1st char of line. 47 * All multiple tabs and spaces collapsed into 1 char. Tab only 48 * legal if 1st char of line. 49 * # line file statements converted to .line x;.file y; statements. 50 * Escaped newlines at end of line: remove them but add as many newlines 51 * to end of statement as you removed in the middle, to synch line numbers. 52 */ 53 54 #define BEFORE_STRING ("\n") 55 #define AFTER_STRING ("\0") /* memcpy of 0 chars might choke. */ 56 #define BEFORE_SIZE (1) 57 #define AFTER_SIZE (1) 58 59 #ifndef TC_EOL_IN_INSN 60 #define TC_EOL_IN_INSN(P) 0 61 #endif 62 63 static char *buffer_start; /*->1st char of full buffer area. */ 64 static char *partial_where; /*->after last full line in buffer. */ 65 static int partial_size; /* >=0. Number of chars in partial line in buffer. */ 66 67 /* Because we need AFTER_STRING just after last full line, it clobbers 68 1st part of partial line. So we preserve 1st part of partial line 69 here. */ 70 static char save_source[AFTER_SIZE]; 71 72 /* What is the largest size buffer that input_file_give_next_buffer() 73 could return to us? */ 74 static unsigned int buffer_length; 75 76 /* The index into an sb structure we are reading from. -1 if none. */ 77 static int sb_index = -1; 78 79 /* If we are reading from an sb structure, this is it. */ 80 static sb from_sb; 81 82 /* Should we do a conditional check on from_sb? */ 83 static int from_sb_is_expansion = 1; 84 85 /* The number of nested sb structures we have included. */ 86 int macro_nest; 87 88 /* We can have more than one source file open at once, though the info for all 89 but the latest one are saved off in a struct input_save. These files remain 90 open, so we are limited by the number of open files allowed by the 91 underlying OS. We may also sequentially read more than one source file in an 92 assembly. */ 93 94 /* We must track the physical file and line number for error messages. We also 95 track a "logical" file and line number corresponding to (C?) compiler 96 source line numbers. Whenever we open a file we must fill in 97 physical_input_file. So if it is NULL we have not opened any files yet. */ 98 99 static char *physical_input_file; 100 static char *logical_input_file; 101 102 typedef unsigned int line_numberT; /* 1-origin line number in a source file. */ 103 /* A line ends in '\n' or eof. */ 104 105 static line_numberT physical_input_line; 106 static int logical_input_line; 107 108 /* Struct used to save the state of the input handler during include files */ 109 struct input_save { 110 char * buffer_start; 111 char * partial_where; 112 int partial_size; 113 char save_source[AFTER_SIZE]; 114 unsigned int buffer_length; 115 char * physical_input_file; 116 char * logical_input_file; 117 line_numberT physical_input_line; 118 int logical_input_line; 119 int sb_index; 120 sb from_sb; 121 int from_sb_is_expansion; /* Should we do a conditional check? */ 122 struct input_save * next_saved_file; /* Chain of input_saves. */ 123 char * input_file_save; /* Saved state of input routines. */ 124 char * saved_position; /* Caller's saved position in buf. */ 125 }; 126 127 static struct input_save *input_scrub_push (char *saved_position); 128 static char *input_scrub_pop (struct input_save *arg); 129 130 /* Saved information about the file that .include'd this one. When we hit EOF, 131 we automatically pop to that file. */ 132 133 static struct input_save *next_saved_file; 134 135 /* Push the state of input reading and scrubbing so that we can #include. 136 The return value is a 'void *' (fudged for old compilers) to a save 137 area, which can be restored by passing it to input_scrub_pop(). */ 138 139 static struct input_save * 140 input_scrub_push (char *saved_position) 141 { 142 register struct input_save *saved; 143 144 saved = (struct input_save *) xmalloc (sizeof *saved); 145 146 saved->saved_position = saved_position; 147 saved->buffer_start = buffer_start; 148 saved->partial_where = partial_where; 149 saved->partial_size = partial_size; 150 saved->buffer_length = buffer_length; 151 saved->physical_input_file = physical_input_file; 152 saved->logical_input_file = logical_input_file; 153 saved->physical_input_line = physical_input_line; 154 saved->logical_input_line = logical_input_line; 155 saved->sb_index = sb_index; 156 saved->from_sb = from_sb; 157 saved->from_sb_is_expansion = from_sb_is_expansion; 158 memcpy (saved->save_source, save_source, sizeof (save_source)); 159 saved->next_saved_file = next_saved_file; 160 saved->input_file_save = input_file_push (); 161 162 input_file_begin (); /* Reinitialize! */ 163 logical_input_line = -1; 164 logical_input_file = (char *) NULL; 165 buffer_length = input_file_buffer_size (); 166 sb_index = -1; 167 168 buffer_start = xmalloc ((BEFORE_SIZE + buffer_length + buffer_length + AFTER_SIZE)); 169 memcpy (buffer_start, BEFORE_STRING, (int) BEFORE_SIZE); 170 171 return saved; 172 } 173 174 static char * 175 input_scrub_pop (struct input_save *saved) 176 { 177 char *saved_position; 178 179 input_scrub_end (); /* Finish off old buffer */ 180 181 input_file_pop (saved->input_file_save); 182 saved_position = saved->saved_position; 183 buffer_start = saved->buffer_start; 184 buffer_length = saved->buffer_length; 185 physical_input_file = saved->physical_input_file; 186 logical_input_file = saved->logical_input_file; 187 physical_input_line = saved->physical_input_line; 188 logical_input_line = saved->logical_input_line; 189 sb_index = saved->sb_index; 190 from_sb = saved->from_sb; 191 from_sb_is_expansion = saved->from_sb_is_expansion; 192 partial_where = saved->partial_where; 193 partial_size = saved->partial_size; 194 next_saved_file = saved->next_saved_file; 195 memcpy (save_source, saved->save_source, sizeof (save_source)); 196 197 free (saved); 198 return saved_position; 199 } 200 201 void 202 input_scrub_begin (void) 203 { 204 know (strlen (BEFORE_STRING) == BEFORE_SIZE); 205 know (strlen (AFTER_STRING) == AFTER_SIZE 206 || (AFTER_STRING[0] == '\0' && AFTER_SIZE == 1)); 207 208 input_file_begin (); 209 210 buffer_length = input_file_buffer_size (); 211 212 buffer_start = xmalloc ((BEFORE_SIZE + buffer_length + buffer_length + AFTER_SIZE)); 213 memcpy (buffer_start, BEFORE_STRING, (int) BEFORE_SIZE); 214 215 /* Line number things. */ 216 logical_input_line = -1; 217 logical_input_file = (char *) NULL; 218 physical_input_file = NULL; /* No file read yet. */ 219 next_saved_file = NULL; /* At EOF, don't pop to any other file */ 220 do_scrub_begin (flag_m68k_mri); 221 } 222 223 void 224 input_scrub_end (void) 225 { 226 if (buffer_start) 227 { 228 free (buffer_start); 229 buffer_start = 0; 230 input_file_end (); 231 } 232 } 233 234 /* Start reading input from a new file. 235 Return start of caller's part of buffer. */ 236 237 char * 238 input_scrub_new_file (char *filename) 239 { 240 input_file_open (filename, !flag_no_comments); 241 physical_input_file = filename[0] ? filename : _("{standard input}"); 242 physical_input_line = 0; 243 244 partial_size = 0; 245 return (buffer_start + BEFORE_SIZE); 246 } 247 248 /* Include a file from the current file. Save our state, cause it to 249 be restored on EOF, and begin handling a new file. Same result as 250 input_scrub_new_file. */ 251 252 char * 253 input_scrub_include_file (char *filename, char *position) 254 { 255 next_saved_file = input_scrub_push (position); 256 return input_scrub_new_file (filename); 257 } 258 259 /* Start getting input from an sb structure. This is used when 260 expanding a macro. */ 261 262 void 263 input_scrub_include_sb (sb *from, char *position, int is_expansion) 264 { 265 if (macro_nest > max_macro_nest) 266 as_fatal (_("macros nested too deeply")); 267 ++macro_nest; 268 269 #ifdef md_macro_start 270 if (is_expansion) 271 { 272 md_macro_start (); 273 } 274 #endif 275 276 next_saved_file = input_scrub_push (position); 277 278 sb_new (&from_sb); 279 from_sb_is_expansion = is_expansion; 280 if (from->len >= 1 && from->ptr[0] != '\n') 281 { 282 /* Add the sentinel required by read.c. */ 283 sb_add_char (&from_sb, '\n'); 284 } 285 sb_scrub_and_add_sb (&from_sb, from); 286 sb_index = 1; 287 288 /* These variables are reset by input_scrub_push. Restore them 289 since we are, after all, still at the same point in the file. */ 290 logical_input_line = next_saved_file->logical_input_line; 291 logical_input_file = next_saved_file->logical_input_file; 292 } 293 294 void 295 input_scrub_close (void) 296 { 297 input_file_close (); 298 } 299 300 char * 301 input_scrub_next_buffer (char **bufp) 302 { 303 register char *limit; /*->just after last char of buffer. */ 304 305 if (sb_index >= 0) 306 { 307 if (sb_index >= from_sb.len) 308 { 309 sb_kill (&from_sb); 310 if (from_sb_is_expansion 311 ) 312 { 313 cond_finish_check (macro_nest); 314 #ifdef md_macro_end 315 /* Allow the target to clean up per-macro expansion 316 data. */ 317 md_macro_end (); 318 #endif 319 } 320 --macro_nest; 321 partial_where = NULL; 322 if (next_saved_file != NULL) 323 *bufp = input_scrub_pop (next_saved_file); 324 return partial_where; 325 } 326 327 partial_where = from_sb.ptr + from_sb.len; 328 partial_size = 0; 329 *bufp = from_sb.ptr + sb_index; 330 sb_index = from_sb.len; 331 return partial_where; 332 } 333 334 *bufp = buffer_start + BEFORE_SIZE; 335 336 if (partial_size) 337 { 338 memcpy (buffer_start + BEFORE_SIZE, partial_where, 339 (unsigned int) partial_size); 340 memcpy (buffer_start + BEFORE_SIZE, save_source, AFTER_SIZE); 341 } 342 limit = input_file_give_next_buffer (buffer_start 343 + BEFORE_SIZE 344 + partial_size); 345 if (limit) 346 { 347 register char *p; /* Find last newline. */ 348 /* Terminate the buffer to avoid confusing TC_EOL_IN_INSN. */ 349 *limit = '\0'; 350 for (p = limit - 1; *p != '\n' || TC_EOL_IN_INSN (p); --p) 351 ; 352 ++p; 353 354 while (p <= buffer_start + BEFORE_SIZE) 355 { 356 int limoff; 357 358 limoff = limit - buffer_start; 359 buffer_length += input_file_buffer_size (); 360 buffer_start = xrealloc (buffer_start, 361 (BEFORE_SIZE 362 + 2 * buffer_length 363 + AFTER_SIZE)); 364 *bufp = buffer_start + BEFORE_SIZE; 365 limit = input_file_give_next_buffer (buffer_start + limoff); 366 367 if (limit == NULL) 368 { 369 as_warn (_("partial line at end of file ignored")); 370 partial_where = NULL; 371 if (next_saved_file) 372 *bufp = input_scrub_pop (next_saved_file); 373 return NULL; 374 } 375 376 /* Terminate the buffer to avoid confusing TC_EOL_IN_INSN. */ 377 *limit = '\0'; 378 for (p = limit - 1; *p != '\n' || TC_EOL_IN_INSN (p); --p) 379 ; 380 ++p; 381 } 382 383 partial_where = p; 384 partial_size = limit - p; 385 memcpy (save_source, partial_where, (int) AFTER_SIZE); 386 memcpy (partial_where, AFTER_STRING, (int) AFTER_SIZE); 387 } 388 else 389 { 390 partial_where = 0; 391 if (partial_size > 0) 392 { 393 as_warn (_("partial line at end of file ignored")); 394 } 395 396 /* Tell the listing we've finished the file. */ 397 LISTING_EOF (); 398 399 /* If we should pop to another file at EOF, do it. */ 400 if (next_saved_file) 401 { 402 *bufp = input_scrub_pop (next_saved_file); /* Pop state */ 403 /* partial_where is now correct to return, since we popped it. */ 404 } 405 } 406 return (partial_where); 407 } 408 409 /* The remaining part of this file deals with line numbers, error 410 messages and so on. Return TRUE if we opened any file. */ 411 412 int 413 seen_at_least_1_file (void) 414 { 415 return (physical_input_file != NULL); 416 } 417 418 void 419 bump_line_counters (void) 420 { 421 if (sb_index < 0) 422 { 423 ++physical_input_line; 424 if (logical_input_line >= 0) 425 ++logical_input_line; 426 } 427 } 428 429 /* Tells us what the new logical line number and file are. 430 If the line_number is -1, we don't change the current logical line 431 number. If it is -2, we decrement the logical line number (this is 432 to support the .appfile pseudo-op inserted into the stream by 433 do_scrub_chars). 434 If the fname is NULL, we don't change the current logical file name. 435 Returns nonzero if the filename actually changes. */ 436 437 int 438 new_logical_line_flags (char *fname, /* DON'T destroy it! We point to it! */ 439 int line_number, 440 int flags) 441 { 442 switch (flags) 443 { 444 case 0: 445 break; 446 case 1: 447 if (line_number != -1) 448 abort (); 449 break; 450 case 1 << 1: 451 case 1 << 2: 452 /* FIXME: we could check that include nesting is correct. */ 453 break; 454 default: 455 abort (); 456 } 457 458 if (line_number >= 0) 459 logical_input_line = line_number; 460 else if (line_number == -1 && fname && !*fname && (flags & (1 << 2))) 461 { 462 logical_input_file = physical_input_file; 463 logical_input_line = physical_input_line; 464 fname = NULL; 465 } 466 467 if (fname 468 && (logical_input_file == NULL 469 || strcmp (logical_input_file, fname))) 470 { 471 logical_input_file = fname; 472 return 1; 473 } 474 else 475 return 0; 476 } 477 478 int 479 new_logical_line (char *fname, int line_number) 480 { 481 return new_logical_line_flags (fname, line_number, 0); 482 } 483 484 485 /* Return the current file name and line number. 486 namep should be char * const *, but there are compilers which screw 487 up declarations like that, and it's easier to avoid it. */ 488 489 void 490 as_where (char **namep, unsigned int *linep) 491 { 492 if (logical_input_file != NULL 493 && (linep == NULL || logical_input_line >= 0)) 494 { 495 *namep = logical_input_file; 496 if (linep != NULL) 497 *linep = logical_input_line; 498 } 499 else if (physical_input_file != NULL) 500 { 501 *namep = physical_input_file; 502 if (linep != NULL) 503 *linep = physical_input_line; 504 } 505 else 506 { 507 *namep = 0; 508 if (linep != NULL) 509 *linep = 0; 510 } 511 } 512