app.c revision 33965
1/* This is the Assembler Pre-Processor 2 Copyright (C) 1987, 90, 91, 92, 93, 94, 95, 96, 1997 3 Free Software Foundation, Inc. 4 5 This file is part of GAS, the GNU Assembler. 6 7 GAS is free software; you can redistribute it and/or modify 8 it under the terms of the GNU General Public License as published by 9 the Free Software Foundation; either version 2, or (at your option) 10 any later version. 11 12 GAS is distributed in the hope that it will be useful, 13 but WITHOUT ANY WARRANTY; without even the implied warranty of 14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 15 GNU General Public License for more details. 16 17 You should have received a copy of the GNU General Public License 18 along with GAS; see the file COPYING. If not, write to the Free 19 Software Foundation, 59 Temple Place - Suite 330, Boston, MA 20 02111-1307, USA. */ 21 22/* Modified by Allen Wirfs-Brock, Instantiations Inc 2/90 */ 23/* App, the assembler pre-processor. This pre-processor strips out excess 24 spaces, turns single-quoted characters into a decimal constant, and turns 25 # <number> <filename> <garbage> into a .line <number>\n.file <filename> 26 pair. This needs better error-handling. */ 27 28#include <stdio.h> 29#include "as.h" /* For BAD_CASE() only */ 30 31#if (__STDC__ != 1) 32#ifndef const 33#define const /* empty */ 34#endif 35#endif 36 37/* Whether we are scrubbing in m68k MRI mode. This is different from 38 flag_m68k_mri, because the two flags will be affected by the .mri 39 pseudo-op at different times. */ 40static int scrub_m68k_mri; 41 42/* The pseudo-op which switches in and out of MRI mode. See the 43 comment in do_scrub_chars. */ 44static const char mri_pseudo[] = ".mri 0"; 45 46static char lex[256]; 47static const char symbol_chars[] = 48"$._ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789"; 49 50#define LEX_IS_SYMBOL_COMPONENT 1 51#define LEX_IS_WHITESPACE 2 52#define LEX_IS_LINE_SEPARATOR 3 53#define LEX_IS_COMMENT_START 4 54#define LEX_IS_LINE_COMMENT_START 5 55#define LEX_IS_TWOCHAR_COMMENT_1ST 6 56#define LEX_IS_STRINGQUOTE 8 57#define LEX_IS_COLON 9 58#define LEX_IS_NEWLINE 10 59#define LEX_IS_ONECHAR_QUOTE 11 60#define IS_SYMBOL_COMPONENT(c) (lex[c] == LEX_IS_SYMBOL_COMPONENT) 61#define IS_WHITESPACE(c) (lex[c] == LEX_IS_WHITESPACE) 62#define IS_LINE_SEPARATOR(c) (lex[c] == LEX_IS_LINE_SEPARATOR) 63#define IS_COMMENT(c) (lex[c] == LEX_IS_COMMENT_START) 64#define IS_LINE_COMMENT(c) (lex[c] == LEX_IS_LINE_COMMENT_START) 65#define IS_NEWLINE(c) (lex[c] == LEX_IS_NEWLINE) 66 67static int process_escape PARAMS ((int)); 68 69/* FIXME-soon: The entire lexer/parser thingy should be 70 built statically at compile time rather than dynamically 71 each and every time the assembler is run. xoxorich. */ 72 73void 74do_scrub_begin (m68k_mri) 75 int m68k_mri; 76{ 77 const char *p; 78 79 scrub_m68k_mri = m68k_mri; 80 81 lex[' '] = LEX_IS_WHITESPACE; 82 lex['\t'] = LEX_IS_WHITESPACE; 83 lex['\n'] = LEX_IS_NEWLINE; 84 lex[';'] = LEX_IS_LINE_SEPARATOR; 85 lex[':'] = LEX_IS_COLON; 86 87 if (! m68k_mri) 88 { 89 lex['"'] = LEX_IS_STRINGQUOTE; 90 91#ifndef TC_HPPA 92 lex['\''] = LEX_IS_ONECHAR_QUOTE; 93#endif 94 95#ifdef SINGLE_QUOTE_STRINGS 96 lex['\''] = LEX_IS_STRINGQUOTE; 97#endif 98 } 99 100 /* Note: if any other character can be LEX_IS_STRINGQUOTE, the loop 101 in state 5 of do_scrub_chars must be changed. */ 102 103 /* Note that these override the previous defaults, e.g. if ';' is a 104 comment char, then it isn't a line separator. */ 105 for (p = symbol_chars; *p; ++p) 106 { 107 lex[(unsigned char) *p] = LEX_IS_SYMBOL_COMPONENT; 108 } /* declare symbol characters */ 109 110 /* The m68k backend wants to be able to change comment_chars. */ 111#ifndef tc_comment_chars 112#define tc_comment_chars comment_chars 113#endif 114 for (p = tc_comment_chars; *p; p++) 115 { 116 lex[(unsigned char) *p] = LEX_IS_COMMENT_START; 117 } /* declare comment chars */ 118 119 for (p = line_comment_chars; *p; p++) 120 { 121 lex[(unsigned char) *p] = LEX_IS_LINE_COMMENT_START; 122 } /* declare line comment chars */ 123 124 for (p = line_separator_chars; *p; p++) 125 { 126 lex[(unsigned char) *p] = LEX_IS_LINE_SEPARATOR; 127 } /* declare line separators */ 128 129 /* Only allow slash-star comments if slash is not in use. 130 FIXME: This isn't right. We should always permit them. */ 131 if (lex['/'] == 0) 132 { 133 lex['/'] = LEX_IS_TWOCHAR_COMMENT_1ST; 134 } 135 136 if (m68k_mri) 137 { 138 lex['\''] = LEX_IS_STRINGQUOTE; 139 lex[';'] = LEX_IS_COMMENT_START; 140 lex['*'] = LEX_IS_LINE_COMMENT_START; 141 /* The MRI documentation says '!' is LEX_IS_COMMENT_START, but 142 then it can't be used in an expression. */ 143 lex['!'] = LEX_IS_LINE_COMMENT_START; 144 } 145} /* do_scrub_begin() */ 146 147/* Saved state of the scrubber */ 148static int state; 149static int old_state; 150static char *out_string; 151static char out_buf[20]; 152static int add_newlines; 153static char *saved_input; 154static int saved_input_len; 155static const char *mri_state; 156static char mri_last_ch; 157 158/* Data structure for saving the state of app across #include's. Note that 159 app is called asynchronously to the parsing of the .include's, so our 160 state at the time .include is interpreted is completely unrelated. 161 That's why we have to save it all. */ 162 163struct app_save 164 { 165 int state; 166 int old_state; 167 char *out_string; 168 char out_buf[sizeof (out_buf)]; 169 int add_newlines; 170 char *saved_input; 171 int saved_input_len; 172 int scrub_m68k_mri; 173 const char *mri_state; 174 char mri_last_ch; 175 }; 176 177char * 178app_push () 179{ 180 register struct app_save *saved; 181 182 saved = (struct app_save *) xmalloc (sizeof (*saved)); 183 saved->state = state; 184 saved->old_state = old_state; 185 saved->out_string = out_string; 186 memcpy (saved->out_buf, out_buf, sizeof (out_buf)); 187 saved->add_newlines = add_newlines; 188 saved->saved_input = saved_input; 189 saved->saved_input_len = saved_input_len; 190 saved->scrub_m68k_mri = scrub_m68k_mri; 191 saved->mri_state = mri_state; 192 saved->mri_last_ch = mri_last_ch; 193 194 /* do_scrub_begin() is not useful, just wastes time. */ 195 196 state = 0; 197 saved_input = NULL; 198 199 return (char *) saved; 200} 201 202void 203app_pop (arg) 204 char *arg; 205{ 206 register struct app_save *saved = (struct app_save *) arg; 207 208 /* There is no do_scrub_end (). */ 209 state = saved->state; 210 old_state = saved->old_state; 211 out_string = saved->out_string; 212 memcpy (out_buf, saved->out_buf, sizeof (out_buf)); 213 add_newlines = saved->add_newlines; 214 saved_input = saved->saved_input; 215 saved_input_len = saved->saved_input_len; 216 scrub_m68k_mri = saved->scrub_m68k_mri; 217 mri_state = saved->mri_state; 218 mri_last_ch = saved->mri_last_ch; 219 220 free (arg); 221} /* app_pop() */ 222 223/* @@ This assumes that \n &c are the same on host and target. This is not 224 necessarily true. */ 225static int 226process_escape (ch) 227 int ch; 228{ 229 switch (ch) 230 { 231 case 'b': 232 return '\b'; 233 case 'f': 234 return '\f'; 235 case 'n': 236 return '\n'; 237 case 'r': 238 return '\r'; 239 case 't': 240 return '\t'; 241 case '\'': 242 return '\''; 243 case '"': 244 return '\"'; 245 default: 246 return ch; 247 } 248} 249 250/* This function is called to process input characters. The GET 251 parameter is used to retrieve more input characters. GET should 252 set its parameter to point to a buffer, and return the length of 253 the buffer; it should return 0 at end of file. The scrubbed output 254 characters are put into the buffer starting at TOSTART; the TOSTART 255 buffer is TOLEN bytes in length. The function returns the number 256 of scrubbed characters put into TOSTART. This will be TOLEN unless 257 end of file was seen. This function is arranged as a state 258 machine, and saves its state so that it may return at any point. 259 This is the way the old code used to work. */ 260 261int 262do_scrub_chars (get, tostart, tolen) 263 int (*get) PARAMS ((char **)); 264 char *tostart; 265 int tolen; 266{ 267 char *to = tostart; 268 char *toend = tostart + tolen; 269 char *from; 270 char *fromend; 271 int fromlen; 272 register int ch, ch2 = 0; 273 274 /*State 0: beginning of normal line 275 1: After first whitespace on line (flush more white) 276 2: After first non-white (opcode) on line (keep 1white) 277 3: after second white on line (into operands) (flush white) 278 4: after putting out a .line, put out digits 279 5: parsing a string, then go to old-state 280 6: putting out \ escape in a "d string. 281 7: After putting out a .appfile, put out string. 282 8: After putting out a .appfile string, flush until newline. 283 9: After seeing symbol char in state 3 (keep 1white after symchar) 284 10: After seeing whitespace in state 9 (keep white before symchar) 285 11: After seeing a symbol character in state 0 (eg a label definition) 286 -1: output string in out_string and go to the state in old_state 287 -2: flush text until a '*' '/' is seen, then go to state old_state 288 */ 289 290 /* I added states 9 and 10 because the MIPS ECOFF assembler uses 291 constructs like ``.loc 1 20''. This was turning into ``.loc 292 120''. States 9 and 10 ensure that a space is never dropped in 293 between characters which could appear in a identifier. Ian 294 Taylor, ian@cygnus.com. 295 296 I added state 11 so that something like "Lfoo add %r25,%r26,%r27" works 297 correctly on the PA (and any other target where colons are optional). 298 Jeff Law, law@cs.utah.edu. */ 299 300 /* This macro gets the next input character. */ 301 302#define GET() \ 303 (from < fromend \ 304 ? *from++ \ 305 : ((saved_input != NULL \ 306 ? (free (saved_input), \ 307 saved_input = NULL, \ 308 0) \ 309 : 0), \ 310 fromlen = (*get) (&from), \ 311 fromend = from + fromlen, \ 312 (fromlen == 0 \ 313 ? EOF \ 314 : *from++))) 315 316 /* This macro pushes a character back on the input stream. */ 317 318#define UNGET(uch) (*--from = (uch)) 319 320 /* This macro puts a character into the output buffer. If this 321 character fills the output buffer, this macro jumps to the label 322 TOFULL. We use this rather ugly approach because we need to 323 handle two different termination conditions: EOF on the input 324 stream, and a full output buffer. It would be simpler if we 325 always read in the entire input stream before processing it, but 326 I don't want to make such a significant change to the assembler's 327 memory usage. */ 328 329#define PUT(pch) \ 330 do \ 331 { \ 332 *to++ = (pch); \ 333 if (to >= toend) \ 334 goto tofull; \ 335 } \ 336 while (0) 337 338 if (saved_input != NULL) 339 { 340 from = saved_input; 341 fromend = from + saved_input_len; 342 } 343 else 344 { 345 fromlen = (*get) (&from); 346 if (fromlen == 0) 347 return 0; 348 fromend = from + fromlen; 349 } 350 351 while (1) 352 { 353 /* The cases in this switch end with continue, in order to 354 branch back to the top of this while loop and generate the 355 next output character in the appropriate state. */ 356 switch (state) 357 { 358 case -1: 359 ch = *out_string++; 360 if (*out_string == '\0') 361 { 362 state = old_state; 363 old_state = 3; 364 } 365 PUT (ch); 366 continue; 367 368 case -2: 369 for (;;) 370 { 371 do 372 { 373 ch = GET (); 374 375 if (ch == EOF) 376 { 377 as_warn ("end of file in comment"); 378 goto fromeof; 379 } 380 381 if (ch == '\n') 382 PUT ('\n'); 383 } 384 while (ch != '*'); 385 386 while ((ch = GET ()) == '*') 387 ; 388 389 if (ch == EOF) 390 { 391 as_warn ("end of file in comment"); 392 goto fromeof; 393 } 394 395 if (ch == '/') 396 break; 397 398 UNGET (ch); 399 } 400 401 state = old_state; 402 UNGET (' '); 403 continue; 404 405 case 4: 406 ch = GET (); 407 if (ch == EOF) 408 goto fromeof; 409 else if (ch >= '0' && ch <= '9') 410 PUT (ch); 411 else 412 { 413 while (ch != EOF && IS_WHITESPACE (ch)) 414 ch = GET (); 415 if (ch == '"') 416 { 417 UNGET (ch); 418 if (scrub_m68k_mri) 419 out_string = "\n\tappfile "; 420 else 421 out_string = "\n\t.appfile "; 422 old_state = 7; 423 state = -1; 424 PUT (*out_string++); 425 } 426 else 427 { 428 while (ch != EOF && ch != '\n') 429 ch = GET (); 430 state = 0; 431 PUT (ch); 432 } 433 } 434 continue; 435 436 case 5: 437 /* We are going to copy everything up to a quote character, 438 with special handling for a backslash. We try to 439 optimize the copying in the simple case without using the 440 GET and PUT macros. */ 441 { 442 char *s; 443 int len; 444 445 for (s = from; s < fromend; s++) 446 { 447 ch = *s; 448 /* This condition must be changed if the type of any 449 other character can be LEX_IS_STRINGQUOTE. */ 450 if (ch == '\\' 451 || ch == '"' 452 || ch == '\'' 453 || ch == '\n') 454 break; 455 } 456 len = s - from; 457 if (len > toend - to) 458 len = toend - to; 459 if (len > 0) 460 { 461 memcpy (to, from, len); 462 to += len; 463 from += len; 464 } 465 } 466 467 ch = GET (); 468 if (ch == EOF) 469 { 470 as_warn ("end of file in string: inserted '\"'"); 471 state = old_state; 472 UNGET ('\n'); 473 PUT ('"'); 474 } 475 else if (lex[ch] == LEX_IS_STRINGQUOTE) 476 { 477 state = old_state; 478 PUT (ch); 479 } 480#ifndef NO_STRING_ESCAPES 481 else if (ch == '\\') 482 { 483 state = 6; 484 PUT (ch); 485 } 486#endif 487 else if (scrub_m68k_mri && ch == '\n') 488 { 489 /* Just quietly terminate the string. This permits lines like 490 bne label loop if we haven't reach end yet 491 */ 492 state = old_state; 493 UNGET (ch); 494 PUT ('\''); 495 } 496 else 497 { 498 PUT (ch); 499 } 500 continue; 501 502 case 6: 503 state = 5; 504 ch = GET (); 505 switch (ch) 506 { 507 /* Handle strings broken across lines, by turning '\n' into 508 '\\' and 'n'. */ 509 case '\n': 510 UNGET ('n'); 511 add_newlines++; 512 PUT ('\\'); 513 continue; 514 515 case '"': 516 case '\\': 517 case 'b': 518 case 'f': 519 case 'n': 520 case 'r': 521 case 't': 522 case 'v': 523 case 'x': 524 case 'X': 525 case '0': 526 case '1': 527 case '2': 528 case '3': 529 case '4': 530 case '5': 531 case '6': 532 case '7': 533 break; 534#if defined(IGNORE_NONSTANDARD_ESCAPES) | defined(ONLY_STANDARD_ESCAPES) 535 default: 536 as_warn ("Unknown escape '\\%c' in string: Ignored", ch); 537 break; 538#else /* ONLY_STANDARD_ESCAPES */ 539 default: 540 /* Accept \x as x for any x */ 541 break; 542#endif /* ONLY_STANDARD_ESCAPES */ 543 544 case EOF: 545 as_warn ("End of file in string: '\"' inserted"); 546 PUT ('"'); 547 continue; 548 } 549 PUT (ch); 550 continue; 551 552 case 7: 553 ch = GET (); 554 state = 5; 555 old_state = 8; 556 if (ch == EOF) 557 goto fromeof; 558 PUT (ch); 559 continue; 560 561 case 8: 562 do 563 ch = GET (); 564 while (ch != '\n' && ch != EOF); 565 if (ch == EOF) 566 goto fromeof; 567 state = 0; 568 PUT (ch); 569 continue; 570 } 571 572 /* OK, we are somewhere in states 0 through 4 or 9 through 11 */ 573 574 /* flushchar: */ 575 ch = GET (); 576 577 recycle: 578 579#ifdef TC_M68K 580 /* We want to have pseudo-ops which control whether we are in 581 MRI mode or not. Unfortunately, since m68k MRI mode affects 582 the scrubber, that means that we need a special purpose 583 recognizer here. */ 584 if (mri_state == NULL) 585 { 586 if ((state == 0 || state == 1) 587 && ch == mri_pseudo[0]) 588 mri_state = mri_pseudo + 1; 589 } 590 else 591 { 592 /* We advance to the next state if we find the right 593 character, or if we need a space character and we get any 594 whitespace character, or if we need a '0' and we get a 595 '1' (this is so that we only need one state to handle 596 ``.mri 0'' and ``.mri 1''). */ 597 if (ch != '\0' 598 && (*mri_state == ch 599 || (*mri_state == ' ' 600 && lex[ch] == LEX_IS_WHITESPACE) 601 || (*mri_state == '0' 602 && ch == '1'))) 603 { 604 mri_last_ch = ch; 605 ++mri_state; 606 } 607 else if (*mri_state != '\0' 608 || (lex[ch] != LEX_IS_WHITESPACE 609 && lex[ch] != LEX_IS_NEWLINE)) 610 { 611 /* We did not get the expected character, or we didn't 612 get a valid terminating character after seeing the 613 entire pseudo-op, so we must go back to the 614 beginning. */ 615 mri_state = NULL; 616 } 617 else 618 { 619 /* We've read the entire pseudo-op. mips_last_ch is 620 either '0' or '1' indicating whether to enter or 621 leave MRI mode. */ 622 do_scrub_begin (mri_last_ch == '1'); 623 624 /* We continue handling the character as usual. The 625 main gas reader must also handle the .mri pseudo-op 626 to control expression parsing and the like. */ 627 } 628 } 629#endif 630 631 if (ch == EOF) 632 { 633 if (state != 0) 634 { 635 as_warn ("end of file not at end of a line; newline inserted"); 636 state = 0; 637 PUT ('\n'); 638 } 639 goto fromeof; 640 } 641 642 switch (lex[ch]) 643 { 644 case LEX_IS_WHITESPACE: 645 do 646 { 647 ch = GET (); 648 } 649 while (ch != EOF && IS_WHITESPACE (ch)); 650 if (ch == EOF) 651 goto fromeof; 652 653 if (state == 0) 654 { 655 /* Preserve a single whitespace character at the 656 beginning of a line. */ 657 state = 1; 658 UNGET (ch); 659 PUT (' '); 660 break; 661 } 662 663 if (IS_COMMENT (ch) 664 || ch == '/' 665 || IS_LINE_SEPARATOR (ch)) 666 { 667 if (scrub_m68k_mri) 668 { 669 /* In MRI mode, we keep these spaces. */ 670 UNGET (ch); 671 PUT (' '); 672 break; 673 } 674 goto recycle; 675 } 676 677 /* If we're in state 2 or 11, we've seen a non-white 678 character followed by whitespace. If the next character 679 is ':', this is whitespace after a label name which we 680 normally must ignore. In MRI mode, though, spaces are 681 not permitted between the label and the colon. */ 682 if ((state == 2 || state == 11) 683 && lex[ch] == LEX_IS_COLON 684 && ! scrub_m68k_mri) 685 { 686 state = 1; 687 PUT (ch); 688 break; 689 } 690 691 switch (state) 692 { 693 case 0: 694 state++; 695 goto recycle; /* Punted leading sp */ 696 case 1: 697 /* We can arrive here if we leave a leading whitespace 698 character at the beginning of a line. */ 699 goto recycle; 700 case 2: 701 state = 3; 702 if (to + 1 < toend) 703 { 704 /* Optimize common case by skipping UNGET/GET. */ 705 PUT (' '); /* Sp after opco */ 706 goto recycle; 707 } 708 UNGET (ch); 709 PUT (' '); 710 break; 711 case 3: 712 if (scrub_m68k_mri) 713 { 714 /* In MRI mode, we keep these spaces. */ 715 UNGET (ch); 716 PUT (' '); 717 break; 718 } 719 goto recycle; /* Sp in operands */ 720 case 9: 721 case 10: 722 if (scrub_m68k_mri) 723 { 724 /* In MRI mode, we keep these spaces. */ 725 state = 3; 726 UNGET (ch); 727 PUT (' '); 728 break; 729 } 730 state = 10; /* Sp after symbol char */ 731 goto recycle; 732 case 11: 733 if (flag_m68k_mri 734#ifdef LABELS_WITHOUT_COLONS 735 || 1 736#endif 737 ) 738 state = 1; 739 else 740 { 741 /* We know that ch is not ':', since we tested that 742 case above. Therefore this is not a label, so it 743 must be the opcode, and we've just seen the 744 whitespace after it. */ 745 state = 3; 746 } 747 UNGET (ch); 748 PUT (' '); /* Sp after label definition. */ 749 break; 750 default: 751 BAD_CASE (state); 752 } 753 break; 754 755 case LEX_IS_TWOCHAR_COMMENT_1ST: 756 ch2 = GET (); 757 if (ch2 == '*') 758 { 759 for (;;) 760 { 761 do 762 { 763 ch2 = GET (); 764 if (ch2 != EOF && IS_NEWLINE (ch2)) 765 add_newlines++; 766 } 767 while (ch2 != EOF && ch2 != '*'); 768 769 while (ch2 == '*') 770 ch2 = GET (); 771 772 if (ch2 == EOF || ch2 == '/') 773 break; 774 775 /* This UNGET will ensure that we count newlines 776 correctly. */ 777 UNGET (ch2); 778 } 779 780 if (ch2 == EOF) 781 as_warn ("end of file in multiline comment"); 782 783 ch = ' '; 784 goto recycle; 785 } 786 else 787 { 788 if (ch2 != EOF) 789 UNGET (ch2); 790 if (state == 9 || state == 10) 791 state = 3; 792 PUT (ch); 793 } 794 break; 795 796 case LEX_IS_STRINGQUOTE: 797 if (state == 10) 798 { 799 /* Preserve the whitespace in foo "bar" */ 800 UNGET (ch); 801 state = 3; 802 PUT (' '); 803 804 /* PUT didn't jump out. We could just break, but we 805 know what will happen, so optimize a bit. */ 806 ch = GET (); 807 old_state = 3; 808 } 809 else if (state == 9) 810 old_state = 3; 811 else 812 old_state = state; 813 state = 5; 814 PUT (ch); 815 break; 816 817#ifndef IEEE_STYLE 818 case LEX_IS_ONECHAR_QUOTE: 819 if (state == 10) 820 { 821 /* Preserve the whitespace in foo 'b' */ 822 UNGET (ch); 823 state = 3; 824 PUT (' '); 825 break; 826 } 827 ch = GET (); 828 if (ch == EOF) 829 { 830 as_warn ("end of file after a one-character quote; \\0 inserted"); 831 ch = 0; 832 } 833 if (ch == '\\') 834 { 835 ch = GET (); 836 if (ch == EOF) 837 { 838 as_warn ("end of file in escape character"); 839 ch = '\\'; 840 } 841 else 842 ch = process_escape (ch); 843 } 844 sprintf (out_buf, "%d", (int) (unsigned char) ch); 845 846 /* None of these 'x constants for us. We want 'x'. */ 847 if ((ch = GET ()) != '\'') 848 { 849#ifdef REQUIRE_CHAR_CLOSE_QUOTE 850 as_warn ("Missing close quote: (assumed)"); 851#else 852 if (ch != EOF) 853 UNGET (ch); 854#endif 855 } 856 if (strlen (out_buf) == 1) 857 { 858 PUT (out_buf[0]); 859 break; 860 } 861 if (state == 9) 862 old_state = 3; 863 else 864 old_state = state; 865 state = -1; 866 out_string = out_buf; 867 PUT (*out_string++); 868 break; 869#endif 870 871 case LEX_IS_COLON: 872 if (state == 9 || state == 10) 873 state = 3; 874 else if (state != 3) 875 state = 1; 876 PUT (ch); 877 break; 878 879 case LEX_IS_NEWLINE: 880 /* Roll out a bunch of newlines from inside comments, etc. */ 881 if (add_newlines) 882 { 883 --add_newlines; 884 UNGET (ch); 885 } 886 /* fall thru into... */ 887 888 case LEX_IS_LINE_SEPARATOR: 889 state = 0; 890 PUT (ch); 891 break; 892 893 case LEX_IS_LINE_COMMENT_START: 894 /* FIXME-someday: The two character comment stuff was badly 895 thought out. On i386, we want '/' as line comment start 896 AND we want C style comments. hence this hack. The 897 whole lexical process should be reworked. xoxorich. */ 898 if (ch == '/') 899 { 900 ch2 = GET (); 901 if (ch2 == '*') 902 { 903 old_state = 3; 904 state = -2; 905 break; 906 } 907 else 908 { 909 UNGET (ch2); 910 } 911 } /* bad hack */ 912 913 if (state == 0 || state == 1) /* Only comment at start of line. */ 914 { 915 int startch; 916 917 startch = ch; 918 919 do 920 { 921 ch = GET (); 922 } 923 while (ch != EOF && IS_WHITESPACE (ch)); 924 if (ch == EOF) 925 { 926 as_warn ("end of file in comment; newline inserted"); 927 PUT ('\n'); 928 break; 929 } 930 if (ch < '0' || ch > '9' || state != 0 || startch != '#') 931 { 932 /* Not a cpp line. */ 933 while (ch != EOF && !IS_NEWLINE (ch)) 934 ch = GET (); 935 if (ch == EOF) 936 as_warn ("EOF in Comment: Newline inserted"); 937 state = 0; 938 PUT ('\n'); 939 break; 940 } 941 /* Loks like `# 123 "filename"' from cpp. */ 942 UNGET (ch); 943 old_state = 4; 944 state = -1; 945 if (scrub_m68k_mri) 946 out_string = "\tappline "; 947 else 948 out_string = "\t.appline "; 949 PUT (*out_string++); 950 break; 951 } 952 953 /* We have a line comment character which is not at the 954 start of a line. If this is also a normal comment 955 character, fall through. Otherwise treat it as a default 956 character. */ 957 if (strchr (tc_comment_chars, ch) == NULL 958 && (! scrub_m68k_mri 959 || (ch != '!' && ch != '*'))) 960 goto de_fault; 961 if (scrub_m68k_mri 962 && (ch == '!' || ch == '*' || ch == '#') 963 && state != 1 964 && state != 10) 965 goto de_fault; 966 /* Fall through. */ 967 case LEX_IS_COMMENT_START: 968 do 969 { 970 ch = GET (); 971 } 972 while (ch != EOF && !IS_NEWLINE (ch)); 973 if (ch == EOF) 974 as_warn ("end of file in comment; newline inserted"); 975 state = 0; 976 PUT ('\n'); 977 break; 978 979 case LEX_IS_SYMBOL_COMPONENT: 980 if (state == 10) 981 { 982 /* This is a symbol character following another symbol 983 character, with whitespace in between. We skipped 984 the whitespace earlier, so output it now. */ 985 UNGET (ch); 986 state = 3; 987 PUT (' '); 988 break; 989 } 990 991 if (state == 3) 992 state = 9; 993 994 /* This is a common case. Quickly copy CH and all the 995 following symbol component or normal characters. */ 996 if (to + 1 < toend && mri_state == NULL) 997 { 998 char *s; 999 int len; 1000 1001 for (s = from; s < fromend; s++) 1002 { 1003 int type; 1004 1005 ch2 = *s; 1006 type = lex[ch2]; 1007 if (type != 0 1008 && type != LEX_IS_SYMBOL_COMPONENT) 1009 break; 1010 } 1011 if (s > from) 1012 { 1013 /* Handle the last character normally, for 1014 simplicity. */ 1015 --s; 1016 } 1017 len = s - from; 1018 if (len > (toend - to) - 1) 1019 len = (toend - to) - 1; 1020 if (len > 0) 1021 { 1022 PUT (ch); 1023 if (len > 8) 1024 { 1025 memcpy (to, from, len); 1026 to += len; 1027 from += len; 1028 } 1029 else 1030 { 1031 switch (len) 1032 { 1033 case 8: *to++ = *from++; 1034 case 7: *to++ = *from++; 1035 case 6: *to++ = *from++; 1036 case 5: *to++ = *from++; 1037 case 4: *to++ = *from++; 1038 case 3: *to++ = *from++; 1039 case 2: *to++ = *from++; 1040 case 1: *to++ = *from++; 1041 } 1042 } 1043 ch = GET (); 1044 } 1045 } 1046 1047 /* Fall through. */ 1048 default: 1049 de_fault: 1050 /* Some relatively `normal' character. */ 1051 if (state == 0) 1052 { 1053 state = 11; /* Now seeing label definition */ 1054 } 1055 else if (state == 1) 1056 { 1057 state = 2; /* Ditto */ 1058 } 1059 else if (state == 9) 1060 { 1061 if (lex[ch] != LEX_IS_SYMBOL_COMPONENT) 1062 state = 3; 1063 } 1064 else if (state == 10) 1065 { 1066 state = 3; 1067 } 1068 PUT (ch); 1069 break; 1070 } 1071 } 1072 1073 /*NOTREACHED*/ 1074 1075 fromeof: 1076 /* We have reached the end of the input. */ 1077 return to - tostart; 1078 1079 tofull: 1080 /* The output buffer is full. Save any input we have not yet 1081 processed. */ 1082 if (fromend > from) 1083 { 1084 char *save; 1085 1086 save = (char *) xmalloc (fromend - from); 1087 memcpy (save, from, fromend - from); 1088 if (saved_input != NULL) 1089 free (saved_input); 1090 saved_input = save; 1091 saved_input_len = fromend - from; 1092 } 1093 else 1094 { 1095 if (saved_input != NULL) 1096 { 1097 free (saved_input); 1098 saved_input = NULL; 1099 } 1100 } 1101 return to - tostart; 1102} 1103 1104/* end of app.c */ 1105