| 1 | /* This is the Assembler Pre-Processor |
| 2 | Copyright (C) 1987, 1990, 1991, 1992, 1994 Free Software Foundation, Inc. |
| 3 | |
| 4 | This file is part of GAS, the GNU Assembler. |
| 5 | |
| 6 | GAS is free software; you can redistribute it and/or modify |
| 7 | it under the terms of the GNU General Public License as published by |
| 8 | the Free Software Foundation; either version 2, or (at your option) |
| 9 | any later version. |
| 10 | |
| 11 | GAS is distributed in the hope that it will be useful, |
| 12 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 14 | GNU General Public License for more details. |
| 15 | |
| 16 | You should have received a copy of the GNU General Public License |
| 17 | along with GAS; see the file COPYING. If not, write to |
| 18 | the Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ |
| 19 | |
| 20 | /* Modified by Allen Wirfs-Brock, Instantiations Inc 2/90 */ |
| 21 | /* App, the assembler pre-processor. This pre-processor strips out excess |
| 22 | spaces, turns single-quoted characters into a decimal constant, and turns |
| 23 | # <number> <filename> <garbage> into a .line <number>\n.file <filename> |
| 24 | pair. This needs better error-handling. */ |
| 25 | |
| 26 | #include <stdio.h> |
| 27 | #include "as.h" /* For BAD_CASE() only */ |
| 28 | |
| 29 | #if (__STDC__ != 1) |
| 30 | #ifndef const |
| 31 | #define const /* empty */ |
| 32 | #endif |
| 33 | #endif |
| 34 | |
| 35 | static char lex[256]; |
| 36 | static const char symbol_chars[] = |
| 37 | "$._ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789"; |
| 38 | |
| 39 | #define LEX_IS_SYMBOL_COMPONENT 1 |
| 40 | #define LEX_IS_WHITESPACE 2 |
| 41 | #define LEX_IS_LINE_SEPARATOR 3 |
| 42 | #define LEX_IS_COMMENT_START 4 |
| 43 | #define LEX_IS_LINE_COMMENT_START 5 |
| 44 | #define LEX_IS_TWOCHAR_COMMENT_1ST 6 |
| 45 | #define LEX_IS_TWOCHAR_COMMENT_2ND 7 |
| 46 | #define LEX_IS_STRINGQUOTE 8 |
| 47 | #define LEX_IS_COLON 9 |
| 48 | #define LEX_IS_NEWLINE 10 |
| 49 | #define LEX_IS_ONECHAR_QUOTE 11 |
| 50 | #define IS_SYMBOL_COMPONENT(c) (lex[c] == LEX_IS_SYMBOL_COMPONENT) |
| 51 | #define IS_WHITESPACE(c) (lex[c] == LEX_IS_WHITESPACE) |
| 52 | #define IS_LINE_SEPARATOR(c) (lex[c] == LEX_IS_LINE_SEPARATOR) |
| 53 | #define IS_COMMENT(c) (lex[c] == LEX_IS_COMMENT_START) |
| 54 | #define IS_LINE_COMMENT(c) (lex[c] == LEX_IS_LINE_COMMENT_START) |
| 55 | #define IS_NEWLINE(c) (lex[c] == LEX_IS_NEWLINE) |
| 56 | |
| 57 | static int process_escape PARAMS ((int)); |
| 58 | |
| 59 | /* FIXME-soon: The entire lexer/parser thingy should be |
| 60 | built statically at compile time rather than dynamically |
| 61 | each and every time the assembler is run. xoxorich. */ |
| 62 | |
| 63 | void |
| 64 | do_scrub_begin () |
| 65 | { |
| 66 | const char *p; |
| 67 | |
| 68 | lex[' '] = LEX_IS_WHITESPACE; |
| 69 | lex['\t'] = LEX_IS_WHITESPACE; |
| 70 | lex['\n'] = LEX_IS_NEWLINE; |
| 71 | lex[';'] = LEX_IS_LINE_SEPARATOR; |
| 72 | lex[':'] = LEX_IS_COLON; |
| 73 | |
| 74 | if (! flag_mri) |
| 75 | { |
| 76 | lex['"'] = LEX_IS_STRINGQUOTE; |
| 77 | |
| 78 | #ifndef TC_HPPA |
| 79 | lex['\''] = LEX_IS_ONECHAR_QUOTE; |
| 80 | #endif |
| 81 | |
| 82 | #ifdef SINGLE_QUOTE_STRINGS |
| 83 | lex['\''] = LEX_IS_STRINGQUOTE; |
| 84 | #endif |
| 85 | } |
| 86 | |
| 87 | /* Note: if any other character can be LEX_IS_STRINGQUOTE, the loop |
| 88 | in state 5 of do_scrub_chars must be changed. */ |
| 89 | |
| 90 | /* Note that these override the previous defaults, e.g. if ';' is a |
| 91 | comment char, then it isn't a line separator. */ |
| 92 | for (p = symbol_chars; *p; ++p) |
| 93 | { |
| 94 | lex[(unsigned char) *p] = LEX_IS_SYMBOL_COMPONENT; |
| 95 | } /* declare symbol characters */ |
| 96 | |
| 97 | for (p = comment_chars; *p; p++) |
| 98 | { |
| 99 | lex[(unsigned char) *p] = LEX_IS_COMMENT_START; |
| 100 | } /* declare comment chars */ |
| 101 | |
| 102 | for (p = line_comment_chars; *p; p++) |
| 103 | { |
| 104 | lex[(unsigned char) *p] = LEX_IS_LINE_COMMENT_START; |
| 105 | } /* declare line comment chars */ |
| 106 | |
| 107 | for (p = line_separator_chars; *p; p++) |
| 108 | { |
| 109 | lex[(unsigned char) *p] = LEX_IS_LINE_SEPARATOR; |
| 110 | } /* declare line separators */ |
| 111 | |
| 112 | /* Only allow slash-star comments if slash is not in use */ |
| 113 | if (lex['/'] == 0) |
| 114 | { |
| 115 | lex['/'] = LEX_IS_TWOCHAR_COMMENT_1ST; |
| 116 | } |
| 117 | /* FIXME-soon. This is a bad hack but otherwise, we can't do |
| 118 | c-style comments when '/' is a line comment char. xoxorich. */ |
| 119 | if (lex['*'] == 0) |
| 120 | { |
| 121 | lex['*'] = LEX_IS_TWOCHAR_COMMENT_2ND; |
| 122 | } |
| 123 | |
| 124 | if (flag_mri) |
| 125 | { |
| 126 | lex['\''] = LEX_IS_STRINGQUOTE; |
| 127 | lex[';'] = LEX_IS_COMMENT_START; |
| 128 | lex['*'] = LEX_IS_LINE_COMMENT_START; |
| 129 | /* The MRI documentation says '!' is LEX_IS_COMMENT_START, but |
| 130 | then it can't be used in an expression. */ |
| 131 | lex['!'] = LEX_IS_LINE_COMMENT_START; |
| 132 | } |
| 133 | } /* do_scrub_begin() */ |
| 134 | |
| 135 | /* Saved state of the scrubber */ |
| 136 | static int state; |
| 137 | static int old_state; |
| 138 | static char *out_string; |
| 139 | static char out_buf[20]; |
| 140 | static int add_newlines; |
| 141 | static char *saved_input; |
| 142 | static int saved_input_len; |
| 143 | |
| 144 | /* Data structure for saving the state of app across #include's. Note that |
| 145 | app is called asynchronously to the parsing of the .include's, so our |
| 146 | state at the time .include is interpreted is completely unrelated. |
| 147 | That's why we have to save it all. */ |
| 148 | |
| 149 | struct app_save |
| 150 | { |
| 151 | int state; |
| 152 | int old_state; |
| 153 | char *out_string; |
| 154 | char out_buf[sizeof (out_buf)]; |
| 155 | int add_newlines; |
| 156 | char *saved_input; |
| 157 | int saved_input_len; |
| 158 | }; |
| 159 | |
| 160 | char * |
| 161 | app_push () |
| 162 | { |
| 163 | register struct app_save *saved; |
| 164 | |
| 165 | saved = (struct app_save *) xmalloc (sizeof (*saved)); |
| 166 | saved->state = state; |
| 167 | saved->old_state = old_state; |
| 168 | saved->out_string = out_string; |
| 169 | memcpy (saved->out_buf, out_buf, sizeof (out_buf)); |
| 170 | saved->add_newlines = add_newlines; |
| 171 | saved->saved_input = saved_input; |
| 172 | saved->saved_input_len = saved_input_len; |
| 173 | |
| 174 | /* do_scrub_begin() is not useful, just wastes time. */ |
| 175 | |
| 176 | state = 0; |
| 177 | saved_input = NULL; |
| 178 | |
| 179 | return (char *) saved; |
| 180 | } |
| 181 | |
| 182 | void |
| 183 | app_pop (arg) |
| 184 | char *arg; |
| 185 | { |
| 186 | register struct app_save *saved = (struct app_save *) arg; |
| 187 | |
| 188 | /* There is no do_scrub_end (). */ |
| 189 | state = saved->state; |
| 190 | old_state = saved->old_state; |
| 191 | out_string = saved->out_string; |
| 192 | memcpy (out_buf, saved->out_buf, sizeof (out_buf)); |
| 193 | add_newlines = saved->add_newlines; |
| 194 | saved_input = saved->saved_input; |
| 195 | saved_input_len = saved->saved_input_len; |
| 196 | |
| 197 | free (arg); |
| 198 | } /* app_pop() */ |
| 199 | |
| 200 | /* @@ This assumes that \n &c are the same on host and target. This is not |
| 201 | necessarily true. */ |
| 202 | static int |
| 203 | process_escape (ch) |
| 204 | int ch; |
| 205 | { |
| 206 | switch (ch) |
| 207 | { |
| 208 | case 'b': |
| 209 | return '\b'; |
| 210 | case 'f': |
| 211 | return '\f'; |
| 212 | case 'n': |
| 213 | return '\n'; |
| 214 | case 'r': |
| 215 | return '\r'; |
| 216 | case 't': |
| 217 | return '\t'; |
| 218 | case '\'': |
| 219 | return '\''; |
| 220 | case '"': |
| 221 | return '\"'; |
| 222 | default: |
| 223 | return ch; |
| 224 | } |
| 225 | } |
| 226 | |
| 227 | /* This function is called to process input characters. The GET |
| 228 | parameter is used to retrieve more input characters. GET should |
| 229 | set its parameter to point to a buffer, and return the length of |
| 230 | the buffer; it should return 0 at end of file. The scrubbed output |
| 231 | characters are put into the buffer starting at TOSTART; the TOSTART |
| 232 | buffer is TOLEN bytes in length. The function returns the number |
| 233 | of scrubbed characters put into TOSTART. This will be TOLEN unless |
| 234 | end of file was seen. This function is arranged as a state |
| 235 | machine, and saves its state so that it may return at any point. |
| 236 | This is the way the old code used to work. */ |
| 237 | |
| 238 | int |
| 239 | do_scrub_chars (get, tostart, tolen) |
| 240 | int (*get) PARAMS ((char **)); |
| 241 | char *tostart; |
| 242 | int tolen; |
| 243 | { |
| 244 | char *to = tostart; |
| 245 | char *toend = tostart + tolen; |
| 246 | char *from; |
| 247 | char *fromend; |
| 248 | int fromlen; |
| 249 | register int ch, ch2 = 0; |
| 250 | int not_cpp_line = 0; |
| 251 | |
| 252 | /*State 0: beginning of normal line |
| 253 | 1: After first whitespace on line (flush more white) |
| 254 | 2: After first non-white (opcode) on line (keep 1white) |
| 255 | 3: after second white on line (into operands) (flush white) |
| 256 | 4: after putting out a .line, put out digits |
| 257 | 5: parsing a string, then go to old-state |
| 258 | 6: putting out \ escape in a "d string. |
| 259 | 7: After putting out a .appfile, put out string. |
| 260 | 8: After putting out a .appfile string, flush until newline. |
| 261 | 9: After seeing symbol char in state 3 (keep 1white after symchar) |
| 262 | 10: After seeing whitespace in state 9 (keep white before symchar) |
| 263 | 11: After seeing a symbol character in state 0 (eg a label definition) |
| 264 | -1: output string in out_string and go to the state in old_state |
| 265 | -2: flush text until a '*' '/' is seen, then go to state old_state |
| 266 | */ |
| 267 | |
| 268 | /* I added states 9 and 10 because the MIPS ECOFF assembler uses |
| 269 | constructs like ``.loc 1 20''. This was turning into ``.loc |
| 270 | 120''. States 9 and 10 ensure that a space is never dropped in |
| 271 | between characters which could appear in a identifier. Ian |
| 272 | Taylor, ian@cygnus.com. |
| 273 | |
| 274 | I added state 11 so that something like "Lfoo add %r25,%r26,%r27" works |
| 275 | correctly on the PA (and any other target where colons are optional). |
| 276 | Jeff Law, law@cs.utah.edu. */ |
| 277 | |
| 278 | /* This macro gets the next input character. */ |
| 279 | |
| 280 | #define GET() \ |
| 281 | (from < fromend \ |
| 282 | ? *from++ \ |
| 283 | : ((saved_input != NULL \ |
| 284 | ? (free (saved_input), \ |
| 285 | saved_input = NULL, \ |
| 286 | 0) \ |
| 287 | : 0), \ |
| 288 | fromlen = (*get) (&from), \ |
| 289 | fromend = from + fromlen, \ |
| 290 | (fromlen == 0 \ |
| 291 | ? EOF \ |
| 292 | : *from++))) |
| 293 | |
| 294 | /* This macro pushes a character back on the input stream. */ |
| 295 | |
| 296 | #define UNGET(uch) (*--from = (uch)) |
| 297 | |
| 298 | /* This macro puts a character into the output buffer. If this |
| 299 | character fills the output buffer, this macro jumps to the label |
| 300 | TOFULL. We use this rather ugly approach because we need to |
| 301 | handle two different termination conditions: EOF on the input |
| 302 | stream, and a full output buffer. It would be simpler if we |
| 303 | always read in the entire input stream before processing it, but |
| 304 | I don't want to make such a significant change to the assembler's |
| 305 | memory usage. */ |
| 306 | |
| 307 | #define PUT(pch) \ |
| 308 | do \ |
| 309 | { \ |
| 310 | *to++ = (pch); \ |
| 311 | if (to >= toend) \ |
| 312 | goto tofull; \ |
| 313 | } \ |
| 314 | while (0) |
| 315 | |
| 316 | if (saved_input != NULL) |
| 317 | { |
| 318 | from = saved_input; |
| 319 | fromend = from + saved_input_len; |
| 320 | } |
| 321 | else |
| 322 | { |
| 323 | fromlen = (*get) (&from); |
| 324 | if (fromlen == 0) |
| 325 | return 0; |
| 326 | fromend = from + fromlen; |
| 327 | } |
| 328 | |
| 329 | while (1) |
| 330 | { |
| 331 | /* The cases in this switch end with continue, in order to |
| 332 | branch back to the top of this while loop and generate the |
| 333 | next output character in the appropriate state. */ |
| 334 | switch (state) |
| 335 | { |
| 336 | case -1: |
| 337 | ch = *out_string++; |
| 338 | if (*out_string == '\0') |
| 339 | { |
| 340 | state = old_state; |
| 341 | old_state = 3; |
| 342 | } |
| 343 | PUT (ch); |
| 344 | continue; |
| 345 | |
| 346 | case -2: |
| 347 | for (;;) |
| 348 | { |
| 349 | do |
| 350 | { |
| 351 | ch = GET (); |
| 352 | |
| 353 | if (ch == EOF) |
| 354 | { |
| 355 | as_warn ("end of file in comment"); |
| 356 | goto fromeof; |
| 357 | } |
| 358 | |
| 359 | if (ch == '\n') |
| 360 | PUT ('\n'); |
| 361 | } |
| 362 | while (ch != '*'); |
| 363 | |
| 364 | while ((ch = GET ()) == '*') |
| 365 | ; |
| 366 | |
| 367 | if (ch == EOF) |
| 368 | { |
| 369 | as_warn ("end of file in comment"); |
| 370 | goto fromeof; |
| 371 | } |
| 372 | |
| 373 | if (ch == '/') |
| 374 | break; |
| 375 | |
| 376 | UNGET (ch); |
| 377 | } |
| 378 | |
| 379 | state = old_state; |
| 380 | PUT (' '); |
| 381 | continue; |
| 382 | |
| 383 | case 4: |
| 384 | ch = GET (); |
| 385 | if (ch == EOF) |
| 386 | goto fromeof; |
| 387 | else if (ch >= '0' && ch <= '9') |
| 388 | PUT (ch); |
| 389 | else |
| 390 | { |
| 391 | while (ch != EOF && IS_WHITESPACE (ch)) |
| 392 | ch = GET (); |
| 393 | if (ch == '"') |
| 394 | { |
| 395 | UNGET (ch); |
| 396 | out_string = "\n\t.appfile "; |
| 397 | old_state = 7; |
| 398 | state = -1; |
| 399 | PUT (*out_string++); |
| 400 | } |
| 401 | else |
| 402 | { |
| 403 | while (ch != EOF && ch != '\n') |
| 404 | ch = GET (); |
| 405 | state = 0; |
| 406 | PUT (ch); |
| 407 | } |
| 408 | } |
| 409 | continue; |
| 410 | |
| 411 | case 5: |
| 412 | /* We are going to copy everything up to a quote character, |
| 413 | with special handling for a backslash. We try to |
| 414 | optimize the copying in the simple case without using the |
| 415 | GET and PUT macros. */ |
| 416 | { |
| 417 | char *s; |
| 418 | int len; |
| 419 | |
| 420 | for (s = from; s < fromend; s++) |
| 421 | { |
| 422 | ch = *s; |
| 423 | /* This condition must be changed if the type of any |
| 424 | other character can be LEX_IS_STRINGQUOTE. */ |
| 425 | if (ch == '\\' |
| 426 | || ch == '"' |
| 427 | || ch == '\'' |
| 428 | || ch == '\n') |
| 429 | break; |
| 430 | } |
| 431 | len = s - from; |
| 432 | if (len > toend - to) |
| 433 | len = toend - to; |
| 434 | if (len > 0) |
| 435 | { |
| 436 | memcpy (to, from, len); |
| 437 | to += len; |
| 438 | from += len; |
| 439 | } |
| 440 | } |
| 441 | |
| 442 | ch = GET (); |
| 443 | if (ch == EOF) |
| 444 | { |
| 445 | as_warn ("end of file in string: inserted '\"'"); |
| 446 | state = old_state; |
| 447 | UNGET ('\n'); |
| 448 | PUT ('"'); |
| 449 | } |
| 450 | else if (lex[ch] == LEX_IS_STRINGQUOTE) |
| 451 | { |
| 452 | state = old_state; |
| 453 | PUT (ch); |
| 454 | } |
| 455 | #ifndef NO_STRING_ESCAPES |
| 456 | else if (ch == '\\') |
| 457 | { |
| 458 | state = 6; |
| 459 | PUT (ch); |
| 460 | } |
| 461 | #endif |
| 462 | else if (flag_mri && ch == '\n') |
| 463 | { |
| 464 | /* Just quietly terminate the string. This permits lines like |
| 465 | bne label loop if we haven't reach end yet |
| 466 | */ |
| 467 | state = old_state; |
| 468 | UNGET (ch); |
| 469 | PUT ('\''); |
| 470 | } |
| 471 | else |
| 472 | { |
| 473 | PUT (ch); |
| 474 | } |
| 475 | continue; |
| 476 | |
| 477 | case 6: |
| 478 | state = 5; |
| 479 | ch = GET (); |
| 480 | switch (ch) |
| 481 | { |
| 482 | /* Handle strings broken across lines, by turning '\n' into |
| 483 | '\\' and 'n'. */ |
| 484 | case '\n': |
| 485 | UNGET ('n'); |
| 486 | add_newlines++; |
| 487 | PUT ('\\'); |
| 488 | continue; |
| 489 | |
| 490 | case '"': |
| 491 | case '\\': |
| 492 | case 'b': |
| 493 | case 'f': |
| 494 | case 'n': |
| 495 | case 'r': |
| 496 | case 't': |
| 497 | case 'v': |
| 498 | case 'x': |
| 499 | case 'X': |
| 500 | case '0': |
| 501 | case '1': |
| 502 | case '2': |
| 503 | case '3': |
| 504 | case '4': |
| 505 | case '5': |
| 506 | case '6': |
| 507 | case '7': |
| 508 | break; |
| 509 | #if defined(IGNORE_NONSTANDARD_ESCAPES) | defined(ONLY_STANDARD_ESCAPES) |
| 510 | default: |
| 511 | as_warn ("Unknown escape '\\%c' in string: Ignored", ch); |
| 512 | break; |
| 513 | #else /* ONLY_STANDARD_ESCAPES */ |
| 514 | default: |
| 515 | /* Accept \x as x for any x */ |
| 516 | break; |
| 517 | #endif /* ONLY_STANDARD_ESCAPES */ |
| 518 | |
| 519 | case EOF: |
| 520 | as_warn ("End of file in string: '\"' inserted"); |
| 521 | PUT ('"'); |
| 522 | continue; |
| 523 | } |
| 524 | PUT (ch); |
| 525 | continue; |
| 526 | |
| 527 | case 7: |
| 528 | ch = GET (); |
| 529 | state = 5; |
| 530 | old_state = 8; |
| 531 | if (ch == EOF) |
| 532 | goto fromeof; |
| 533 | PUT (ch); |
| 534 | continue; |
| 535 | |
| 536 | case 8: |
| 537 | do |
| 538 | ch = GET (); |
| 539 | while (ch != '\n' && ch != EOF); |
| 540 | if (ch == EOF) |
| 541 | goto fromeof; |
| 542 | state = 0; |
| 543 | PUT (ch); |
| 544 | continue; |
| 545 | } |
| 546 | |
| 547 | /* OK, we are somewhere in states 0 through 4 or 9 through 11 */ |
| 548 | |
| 549 | /* flushchar: */ |
| 550 | ch = GET (); |
| 551 | recycle: |
| 552 | if (ch == EOF) |
| 553 | { |
| 554 | if (state != 0) |
| 555 | { |
| 556 | as_warn ("end of file not at end of a line; newline inserted"); |
| 557 | state = 0; |
| 558 | PUT ('\n'); |
| 559 | } |
| 560 | goto fromeof; |
| 561 | } |
| 562 | |
| 563 | switch (lex[ch]) |
| 564 | { |
| 565 | case LEX_IS_WHITESPACE: |
| 566 | if (state == 0) |
| 567 | { |
| 568 | /* Preserve a single whitespace character at the |
| 569 | beginning of a line. */ |
| 570 | state = 1; |
| 571 | PUT (ch); |
| 572 | break; |
| 573 | } |
| 574 | do |
| 575 | { |
| 576 | ch = GET (); |
| 577 | } |
| 578 | while (ch != EOF && IS_WHITESPACE (ch)); |
| 579 | if (ch == EOF) |
| 580 | goto fromeof; |
| 581 | |
| 582 | if (IS_COMMENT (ch) |
| 583 | || (state == 0 && IS_LINE_COMMENT (ch)) |
| 584 | || ch == '/' |
| 585 | || IS_LINE_SEPARATOR (ch)) |
| 586 | { |
| 587 | /* cpp never outputs a leading space before the #, so |
| 588 | try to avoid being confused. */ |
| 589 | not_cpp_line = 1; |
| 590 | if (flag_mri) |
| 591 | { |
| 592 | /* In MRI mode, we keep these spaces. */ |
| 593 | UNGET (ch); |
| 594 | PUT (' '); |
| 595 | break; |
| 596 | } |
| 597 | goto recycle; |
| 598 | } |
| 599 | |
| 600 | /* If we're in state 2 or 11, we've seen a non-white |
| 601 | character followed by whitespace. If the next character |
| 602 | is ':', this is whitespace after a label name which we |
| 603 | normally must ignore. In MRI mode, though, spaces are |
| 604 | not permitted between the label and the colon. */ |
| 605 | if ((state == 2 || state == 11) |
| 606 | && lex[ch] == LEX_IS_COLON |
| 607 | && ! flag_mri) |
| 608 | { |
| 609 | state = 1; |
| 610 | PUT (ch); |
| 611 | break; |
| 612 | } |
| 613 | |
| 614 | switch (state) |
| 615 | { |
| 616 | case 0: |
| 617 | state++; |
| 618 | goto recycle; /* Punted leading sp */ |
| 619 | case 1: |
| 620 | /* We can arrive here if we leave a leading whitespace |
| 621 | character at the beginning of a line. */ |
| 622 | goto recycle; |
| 623 | case 2: |
| 624 | state = 3; |
| 625 | if (to + 1 < toend) |
| 626 | { |
| 627 | /* Optimize common case by skipping UNGET/GET. */ |
| 628 | PUT (' '); /* Sp after opco */ |
| 629 | goto recycle; |
| 630 | } |
| 631 | UNGET (ch); |
| 632 | PUT (' '); |
| 633 | break; |
| 634 | case 3: |
| 635 | if (flag_mri) |
| 636 | { |
| 637 | /* In MRI mode, we keep these spaces. */ |
| 638 | UNGET (ch); |
| 639 | PUT (' '); |
| 640 | break; |
| 641 | } |
| 642 | goto recycle; /* Sp in operands */ |
| 643 | case 9: |
| 644 | case 10: |
| 645 | if (flag_mri) |
| 646 | { |
| 647 | /* In MRI mode, we keep these spaces. */ |
| 648 | state = 3; |
| 649 | UNGET (ch); |
| 650 | PUT (' '); |
| 651 | break; |
| 652 | } |
| 653 | state = 10; /* Sp after symbol char */ |
| 654 | goto recycle; |
| 655 | case 11: |
| 656 | state = 1; |
| 657 | UNGET (ch); |
| 658 | PUT (' '); /* Sp after label definition. */ |
| 659 | break; |
| 660 | default: |
| 661 | BAD_CASE (state); |
| 662 | } |
| 663 | break; |
| 664 | |
| 665 | case LEX_IS_TWOCHAR_COMMENT_1ST: |
| 666 | ch2 = GET (); |
| 667 | if (ch2 != EOF && lex[ch2] == LEX_IS_TWOCHAR_COMMENT_2ND) |
| 668 | { |
| 669 | for (;;) |
| 670 | { |
| 671 | do |
| 672 | { |
| 673 | ch2 = GET (); |
| 674 | if (ch2 != EOF && IS_NEWLINE (ch2)) |
| 675 | add_newlines++; |
| 676 | } |
| 677 | while (ch2 != EOF && |
| 678 | (lex[ch2] != LEX_IS_TWOCHAR_COMMENT_2ND)); |
| 679 | |
| 680 | while (ch2 != EOF && |
| 681 | (lex[ch2] == LEX_IS_TWOCHAR_COMMENT_2ND)) |
| 682 | { |
| 683 | ch2 = GET (); |
| 684 | } |
| 685 | |
| 686 | if (ch2 == EOF |
| 687 | || lex[ch2] == LEX_IS_TWOCHAR_COMMENT_1ST) |
| 688 | break; |
| 689 | UNGET (ch); |
| 690 | } |
| 691 | if (ch2 == EOF) |
| 692 | as_warn ("end of file in multiline comment"); |
| 693 | |
| 694 | ch = ' '; |
| 695 | goto recycle; |
| 696 | } |
| 697 | else |
| 698 | { |
| 699 | if (ch2 != EOF) |
| 700 | UNGET (ch2); |
| 701 | if (state == 9 || state == 10) |
| 702 | state = 3; |
| 703 | PUT (ch); |
| 704 | } |
| 705 | break; |
| 706 | |
| 707 | case LEX_IS_STRINGQUOTE: |
| 708 | if (state == 10) |
| 709 | { |
| 710 | /* Preserve the whitespace in foo "bar" */ |
| 711 | UNGET (ch); |
| 712 | state = 3; |
| 713 | PUT (' '); |
| 714 | |
| 715 | /* PUT didn't jump out. We could just break, but we |
| 716 | know what will happen, so optimize a bit. */ |
| 717 | ch = GET (); |
| 718 | old_state = 3; |
| 719 | } |
| 720 | else if (state == 9) |
| 721 | old_state = 3; |
| 722 | else |
| 723 | old_state = state; |
| 724 | state = 5; |
| 725 | PUT (ch); |
| 726 | break; |
| 727 | |
| 728 | #ifndef IEEE_STYLE |
| 729 | case LEX_IS_ONECHAR_QUOTE: |
| 730 | if (state == 10) |
| 731 | { |
| 732 | /* Preserve the whitespace in foo 'b' */ |
| 733 | UNGET (ch); |
| 734 | state = 3; |
| 735 | PUT (' '); |
| 736 | break; |
| 737 | } |
| 738 | ch = GET (); |
| 739 | if (ch == EOF) |
| 740 | { |
| 741 | as_warn ("end of file after a one-character quote; \\0 inserted"); |
| 742 | ch = 0; |
| 743 | } |
| 744 | if (ch == '\\') |
| 745 | { |
| 746 | ch = GET (); |
| 747 | if (ch == EOF) |
| 748 | { |
| 749 | as_warn ("end of file in escape character"); |
| 750 | ch = '\\'; |
| 751 | } |
| 752 | else |
| 753 | ch = process_escape (ch); |
| 754 | } |
| 755 | sprintf (out_buf, "%d", (int) (unsigned char) ch); |
| 756 | |
| 757 | /* None of these 'x constants for us. We want 'x'. */ |
| 758 | if ((ch = GET ()) != '\'') |
| 759 | { |
| 760 | #ifdef REQUIRE_CHAR_CLOSE_QUOTE |
| 761 | as_warn ("Missing close quote: (assumed)"); |
| 762 | #else |
| 763 | if (ch != EOF) |
| 764 | UNGET (ch); |
| 765 | #endif |
| 766 | } |
| 767 | if (strlen (out_buf) == 1) |
| 768 | { |
| 769 | PUT (out_buf[0]); |
| 770 | break; |
| 771 | } |
| 772 | if (state == 9) |
| 773 | old_state = 3; |
| 774 | else |
| 775 | old_state = state; |
| 776 | state = -1; |
| 777 | out_string = out_buf; |
| 778 | PUT (*out_string++); |
| 779 | break; |
| 780 | #endif |
| 781 | |
| 782 | case LEX_IS_COLON: |
| 783 | if (state == 9 || state == 10) |
| 784 | state = 3; |
| 785 | else if (state != 3) |
| 786 | state = 1; |
| 787 | PUT (ch); |
| 788 | break; |
| 789 | |
| 790 | case LEX_IS_NEWLINE: |
| 791 | /* Roll out a bunch of newlines from inside comments, etc. */ |
| 792 | if (add_newlines) |
| 793 | { |
| 794 | --add_newlines; |
| 795 | UNGET (ch); |
| 796 | } |
| 797 | /* fall thru into... */ |
| 798 | |
| 799 | case LEX_IS_LINE_SEPARATOR: |
| 800 | state = 0; |
| 801 | PUT (ch); |
| 802 | break; |
| 803 | |
| 804 | case LEX_IS_LINE_COMMENT_START: |
| 805 | if (state == 0) /* Only comment at start of line. */ |
| 806 | { |
| 807 | /* FIXME-someday: The two character comment stuff was |
| 808 | badly thought out. On i386, we want '/' as line |
| 809 | comment start AND we want C style comments. hence |
| 810 | this hack. The whole lexical process should be |
| 811 | reworked. xoxorich. */ |
| 812 | if (ch == '/') |
| 813 | { |
| 814 | ch2 = GET (); |
| 815 | if (ch2 == '*') |
| 816 | { |
| 817 | state = -2; |
| 818 | break; |
| 819 | } |
| 820 | else |
| 821 | { |
| 822 | UNGET (ch2); |
| 823 | } |
| 824 | } /* bad hack */ |
| 825 | |
| 826 | if (ch != '#') |
| 827 | not_cpp_line = 1; |
| 828 | |
| 829 | do |
| 830 | { |
| 831 | ch = GET (); |
| 832 | } |
| 833 | while (ch != EOF && IS_WHITESPACE (ch)); |
| 834 | if (ch == EOF) |
| 835 | { |
| 836 | as_warn ("end of file in comment; newline inserted"); |
| 837 | PUT ('\n'); |
| 838 | break; |
| 839 | } |
| 840 | if (ch < '0' || ch > '9' || not_cpp_line) |
| 841 | { |
| 842 | /* Non-numerics: Eat whole comment line */ |
| 843 | while (ch != EOF && !IS_NEWLINE (ch)) |
| 844 | ch = GET (); |
| 845 | if (ch == EOF) |
| 846 | as_warn ("EOF in Comment: Newline inserted"); |
| 847 | state = 0; |
| 848 | PUT ('\n'); |
| 849 | break; |
| 850 | } |
| 851 | /* Numerics begin comment. Perhaps CPP `# 123 "filename"' */ |
| 852 | UNGET (ch); |
| 853 | old_state = 4; |
| 854 | state = -1; |
| 855 | out_string = "\t.appline "; |
| 856 | PUT (*out_string++); |
| 857 | break; |
| 858 | } |
| 859 | |
| 860 | /* We have a line comment character which is not at the |
| 861 | start of a line. If this is also a normal comment |
| 862 | character, fall through. Otherwise treat it as a default |
| 863 | character. */ |
| 864 | if (strchr (comment_chars, ch) == NULL |
| 865 | && (! flag_mri |
| 866 | || (ch != '!' && ch != '*'))) |
| 867 | goto de_fault; |
| 868 | if (flag_mri |
| 869 | && (ch == '!' || ch == '*') |
| 870 | && state != 1 |
| 871 | && state != 10) |
| 872 | goto de_fault; |
| 873 | /* Fall through. */ |
| 874 | case LEX_IS_COMMENT_START: |
| 875 | do |
| 876 | { |
| 877 | ch = GET (); |
| 878 | } |
| 879 | while (ch != EOF && !IS_NEWLINE (ch)); |
| 880 | if (ch == EOF) |
| 881 | as_warn ("end of file in comment; newline inserted"); |
| 882 | state = 0; |
| 883 | PUT ('\n'); |
| 884 | break; |
| 885 | |
| 886 | case LEX_IS_SYMBOL_COMPONENT: |
| 887 | if (state == 10) |
| 888 | { |
| 889 | /* This is a symbol character following another symbol |
| 890 | character, with whitespace in between. We skipped |
| 891 | the whitespace earlier, so output it now. */ |
| 892 | UNGET (ch); |
| 893 | state = 3; |
| 894 | PUT (' '); |
| 895 | break; |
| 896 | } |
| 897 | |
| 898 | if (state == 3) |
| 899 | state = 9; |
| 900 | |
| 901 | /* This is a common case. Quickly copy CH and all the |
| 902 | following symbol component or normal characters. */ |
| 903 | if (to + 1 < toend) |
| 904 | { |
| 905 | char *s; |
| 906 | int len; |
| 907 | |
| 908 | for (s = from; s < fromend; s++) |
| 909 | { |
| 910 | int type; |
| 911 | |
| 912 | ch2 = *s; |
| 913 | type = lex[ch2]; |
| 914 | if (type != 0 |
| 915 | && type != LEX_IS_SYMBOL_COMPONENT) |
| 916 | break; |
| 917 | } |
| 918 | if (s > from) |
| 919 | { |
| 920 | /* Handle the last character normally, for |
| 921 | simplicity. */ |
| 922 | --s; |
| 923 | } |
| 924 | len = s - from; |
| 925 | if (len > (toend - to) - 1) |
| 926 | len = (toend - to) - 1; |
| 927 | if (len > 0) |
| 928 | { |
| 929 | PUT (ch); |
| 930 | if (len > 8) |
| 931 | { |
| 932 | memcpy (to, from, len); |
| 933 | to += len; |
| 934 | from += len; |
| 935 | } |
| 936 | else |
| 937 | { |
| 938 | switch (len) |
| 939 | { |
| 940 | case 8: *to++ = *from++; |
| 941 | case 7: *to++ = *from++; |
| 942 | case 6: *to++ = *from++; |
| 943 | case 5: *to++ = *from++; |
| 944 | case 4: *to++ = *from++; |
| 945 | case 3: *to++ = *from++; |
| 946 | case 2: *to++ = *from++; |
| 947 | case 1: *to++ = *from++; |
| 948 | } |
| 949 | } |
| 950 | ch = GET (); |
| 951 | } |
| 952 | } |
| 953 | |
| 954 | /* Fall through. */ |
| 955 | default: |
| 956 | de_fault: |
| 957 | /* Some relatively `normal' character. */ |
| 958 | if (state == 0) |
| 959 | { |
| 960 | state = 11; /* Now seeing label definition */ |
| 961 | } |
| 962 | else if (state == 1) |
| 963 | { |
| 964 | state = 2; /* Ditto */ |
| 965 | } |
| 966 | else if (state == 9) |
| 967 | { |
| 968 | if (lex[ch] != LEX_IS_SYMBOL_COMPONENT) |
| 969 | state = 3; |
| 970 | } |
| 971 | else if (state == 10) |
| 972 | { |
| 973 | state = 3; |
| 974 | } |
| 975 | PUT (ch); |
| 976 | break; |
| 977 | } |
| 978 | } |
| 979 | |
| 980 | /*NOTREACHED*/ |
| 981 | |
| 982 | fromeof: |
| 983 | /* We have reached the end of the input. */ |
| 984 | return to - tostart; |
| 985 | |
| 986 | tofull: |
| 987 | /* The output buffer is full. Save any input we have not yet |
| 988 | processed. */ |
| 989 | if (fromend > from) |
| 990 | { |
| 991 | char *save; |
| 992 | |
| 993 | save = (char *) xmalloc (fromend - from); |
| 994 | memcpy (save, from, fromend - from); |
| 995 | if (saved_input != NULL) |
| 996 | free (saved_input); |
| 997 | saved_input = save; |
| 998 | saved_input_len = fromend - from; |
| 999 | } |
| 1000 | else |
| 1001 | { |
| 1002 | if (saved_input != NULL) |
| 1003 | { |
| 1004 | free (saved_input); |
| 1005 | saved_input = NULL; |
| 1006 | } |
| 1007 | } |
| 1008 | return to - tostart; |
| 1009 | } |
| 1010 | |
| 1011 | /* end of app.c */ |