1 /* Support routines for GNU DIFF. 2 Copyright (C) 1988, 1989, 1992, 1993, 1994 Free Software Foundation, Inc. 3 4 This file is part of GNU DIFF. 5 6 GNU DIFF 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 GNU DIFF 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 GNU DIFF; see the file COPYING. If not, write to 18 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. */ 19 20 /* $FreeBSD: src/contrib/diff/util.c,v 1.2.6.2 2000/09/20 02:24:32 jkh Exp $ */ 21 22 #include "diff.h" 23 24 #ifndef PR_PROGRAM 25 #define PR_PROGRAM "/bin/pr" 26 #endif 27 28 /* Queue up one-line messages to be printed at the end, 29 when -l is specified. Each message is recorded with a `struct msg'. */ 30 31 struct msg 32 { 33 struct msg *next; 34 char const *format; 35 char const *arg1; 36 char const *arg2; 37 char const *arg3; 38 char const *arg4; 39 }; 40 41 /* Head of the chain of queues messages. */ 42 43 static struct msg *msg_chain; 44 45 /* Tail of the chain of queues messages. */ 46 47 static struct msg **msg_chain_end = &msg_chain; 48 49 /* Use when a system call returns non-zero status. 50 TEXT should normally be the file name. */ 51 52 void 53 perror_with_name (text) 54 char const *text; 55 { 56 int e = errno; 57 fprintf (stderr, "%s: ", program_name); 58 errno = e; 59 perror (text); 60 } 61 62 /* Use when a system call returns non-zero status and that is fatal. */ 63 64 void 65 pfatal_with_name (text) 66 char const *text; 67 { 68 int e = errno; 69 print_message_queue (); 70 fprintf (stderr, "%s: ", program_name); 71 errno = e; 72 perror (text); 73 exit (2); 74 } 75 76 /* Print an error message from the format-string FORMAT 77 with args ARG1 and ARG2. */ 78 79 void 80 error (format, arg, arg1) 81 char const *format, *arg, *arg1; 82 { 83 fprintf (stderr, "%s: ", program_name); 84 fprintf (stderr, format, arg, arg1); 85 fprintf (stderr, "\n"); 86 } 87 88 /* Print an error message containing the string TEXT, then exit. */ 89 90 void 91 fatal (m) 92 char const *m; 93 { 94 print_message_queue (); 95 error ("%s", m, 0); 96 exit (2); 97 } 98 99 /* Like printf, except if -l in effect then save the message and print later. 100 This is used for things like "binary files differ" and "Only in ...". */ 101 102 void 103 message (format, arg1, arg2) 104 char const *format, *arg1, *arg2; 105 { 106 message5 (format, arg1, arg2, 0, 0); 107 } 108 109 void 110 message5 (format, arg1, arg2, arg3, arg4) 111 char const *format, *arg1, *arg2, *arg3, *arg4; 112 { 113 if (paginate_flag) 114 { 115 struct msg *new = (struct msg *) xmalloc (sizeof (struct msg)); 116 new->format = format; 117 new->arg1 = concat (arg1, "", ""); 118 new->arg2 = concat (arg2, "", ""); 119 new->arg3 = arg3 ? concat (arg3, "", "") : 0; 120 new->arg4 = arg4 ? concat (arg4, "", "") : 0; 121 new->next = 0; 122 *msg_chain_end = new; 123 msg_chain_end = &new->next; 124 } 125 else 126 { 127 if (sdiff_help_sdiff) 128 putchar (' '); 129 printf (format, arg1, arg2, arg3, arg4); 130 } 131 } 132 133 /* Output all the messages that were saved up by calls to `message'. */ 134 135 void 136 print_message_queue () 137 { 138 struct msg *m; 139 140 for (m = msg_chain; m; m = m->next) 141 printf (m->format, m->arg1, m->arg2, m->arg3, m->arg4); 142 } 143 144 /* Call before outputting the results of comparing files NAME0 and NAME1 145 to set up OUTFILE, the stdio stream for the output to go to. 146 147 Usually, OUTFILE is just stdout. But when -l was specified 148 we fork off a `pr' and make OUTFILE a pipe to it. 149 `pr' then outputs to our stdout. */ 150 151 static char const *current_name0; 152 static char const *current_name1; 153 static int current_depth; 154 155 void 156 setup_output (name0, name1, depth) 157 char const *name0, *name1; 158 int depth; 159 { 160 current_name0 = name0; 161 current_name1 = name1; 162 current_depth = depth; 163 outfile = 0; 164 } 165 166 #if HAVE_FORK 167 static pid_t pr_pid; 168 #endif 169 170 void 171 begin_output () 172 { 173 char *name; 174 175 if (outfile != 0) 176 return; 177 178 /* Construct the header of this piece of diff. */ 179 name = xmalloc (strlen (current_name0) + strlen (current_name1) 180 + strlen (switch_string) + 7); 181 /* Posix.2 section 4.17.6.1.1 specifies this format. But there is a 182 bug in the first printing (IEEE Std 1003.2-1992 p 251 l 3304): 183 it says that we must print only the last component of the pathnames. 184 This requirement is silly and does not match historical practice. */ 185 sprintf (name, "diff%s %s %s", switch_string, current_name0, current_name1); 186 187 if (paginate_flag) 188 { 189 /* Make OUTFILE a pipe to a subsidiary `pr'. */ 190 191 #if HAVE_FORK 192 int pipes[2]; 193 194 if (pipe (pipes) != 0) 195 pfatal_with_name ("pipe"); 196 197 fflush (stdout); 198 199 pr_pid = fork (); 200 if (pr_pid < 0) 201 pfatal_with_name ("vfork"); 202 203 if (pr_pid == 0) 204 { 205 close (pipes[1]); 206 if (pipes[0] != STDIN_FILENO) 207 { 208 if (dup2 (pipes[0], STDIN_FILENO) < 0) 209 pfatal_with_name ("dup2"); 210 close (pipes[0]); 211 } 212 #ifdef __FreeBSD__ 213 execl (PR_PROGRAM, PR_PROGRAM, "-F", "-h", name, 0); 214 #else 215 execl (PR_PROGRAM, PR_PROGRAM, "-f", "-h", name, 0); 216 #endif 217 pfatal_with_name (PR_PROGRAM); 218 } 219 else 220 { 221 close (pipes[0]); 222 outfile = fdopen (pipes[1], "w"); 223 if (!outfile) 224 pfatal_with_name ("fdopen"); 225 } 226 #else /* ! HAVE_FORK */ 227 char *command = xmalloc (4 * strlen (name) + strlen (PR_PROGRAM) + 10); 228 char *p; 229 char const *a = name; 230 sprintf (command, "%s -f -h ", PR_PROGRAM); 231 p = command + strlen (command); 232 SYSTEM_QUOTE_ARG (p, a); 233 *p = 0; 234 outfile = popen (command, "w"); 235 if (!outfile) 236 pfatal_with_name (command); 237 free (command); 238 #endif /* ! HAVE_FORK */ 239 } 240 else 241 { 242 243 /* If -l was not specified, output the diff straight to `stdout'. */ 244 245 outfile = stdout; 246 247 /* If handling multiple files (because scanning a directory), 248 print which files the following output is about. */ 249 if (current_depth > 0) 250 printf ("%s\n", name); 251 } 252 253 free (name); 254 255 /* A special header is needed at the beginning of context output. */ 256 switch (output_style) 257 { 258 case OUTPUT_CONTEXT: 259 print_context_header (files, 0); 260 break; 261 262 case OUTPUT_UNIFIED: 263 print_context_header (files, 1); 264 break; 265 266 default: 267 break; 268 } 269 } 270 271 /* Call after the end of output of diffs for one file. 272 Close OUTFILE and get rid of the `pr' subfork. */ 273 274 void 275 finish_output () 276 { 277 if (outfile != 0 && outfile != stdout) 278 { 279 int wstatus; 280 if (ferror (outfile)) 281 fatal ("write error"); 282 #if ! HAVE_FORK 283 wstatus = pclose (outfile); 284 #else /* HAVE_FORK */ 285 if (fclose (outfile) != 0) 286 pfatal_with_name ("write error"); 287 if (waitpid (pr_pid, &wstatus, 0) < 0) 288 pfatal_with_name ("waitpid"); 289 #endif /* HAVE_FORK */ 290 if (wstatus != 0) 291 fatal ("subsidiary pr failed"); 292 } 293 294 outfile = 0; 295 } 296 297 /* Compare two lines (typically one from each input file) 298 according to the command line options. 299 For efficiency, this is invoked only when the lines do not match exactly 300 but an option like -i might cause us to ignore the difference. 301 Return nonzero if the lines differ. */ 302 303 int 304 line_cmp (s1, s2) 305 char const *s1, *s2; 306 { 307 register unsigned char const *t1 = (unsigned char const *) s1; 308 register unsigned char const *t2 = (unsigned char const *) s2; 309 310 while (1) 311 { 312 register unsigned char c1 = *t1++; 313 register unsigned char c2 = *t2++; 314 315 /* Test for exact char equality first, since it's a common case. */ 316 if (c1 != c2) 317 { 318 /* Ignore horizontal white space if -b or -w is specified. */ 319 320 if (ignore_all_space_flag) 321 { 322 /* For -w, just skip past any white space. */ 323 while (ISSPACE (c1) && c1 != '\n') c1 = *t1++; 324 while (ISSPACE (c2) && c2 != '\n') c2 = *t2++; 325 } 326 else if (ignore_space_change_flag) 327 { 328 /* For -b, advance past any sequence of white space in line 1 329 and consider it just one Space, or nothing at all 330 if it is at the end of the line. */ 331 if (ISSPACE (c1)) 332 { 333 while (c1 != '\n') 334 { 335 c1 = *t1++; 336 if (! ISSPACE (c1)) 337 { 338 --t1; 339 c1 = ' '; 340 break; 341 } 342 } 343 } 344 345 /* Likewise for line 2. */ 346 if (ISSPACE (c2)) 347 { 348 while (c2 != '\n') 349 { 350 c2 = *t2++; 351 if (! ISSPACE (c2)) 352 { 353 --t2; 354 c2 = ' '; 355 break; 356 } 357 } 358 } 359 360 if (c1 != c2) 361 { 362 /* If we went too far when doing the simple test 363 for equality, go back to the first non-white-space 364 character in both sides and try again. */ 365 if (c2 == ' ' && c1 != '\n' 366 && (unsigned char const *) s1 + 1 < t1 367 && ISSPACE(t1[-2])) 368 { 369 --t1; 370 continue; 371 } 372 if (c1 == ' ' && c2 != '\n' 373 && (unsigned char const *) s2 + 1 < t2 374 && ISSPACE(t2[-2])) 375 { 376 --t2; 377 continue; 378 } 379 } 380 } 381 382 /* Lowercase all letters if -i is specified. */ 383 384 if (ignore_case_flag) 385 { 386 if (ISUPPER (c1)) 387 c1 = tolower (c1); 388 if (ISUPPER (c2)) 389 c2 = tolower (c2); 390 } 391 392 if (c1 != c2) 393 break; 394 } 395 if (c1 == '\n') 396 return 0; 397 } 398 399 return (1); 400 } 401 402 /* Find the consecutive changes at the start of the script START. 403 Return the last link before the first gap. */ 404 405 struct change * 406 find_change (start) 407 struct change *start; 408 { 409 return start; 410 } 411 412 struct change * 413 find_reverse_change (start) 414 struct change *start; 415 { 416 return start; 417 } 418 419 /* Divide SCRIPT into pieces by calling HUNKFUN and 420 print each piece with PRINTFUN. 421 Both functions take one arg, an edit script. 422 423 HUNKFUN is called with the tail of the script 424 and returns the last link that belongs together with the start 425 of the tail. 426 427 PRINTFUN takes a subscript which belongs together (with a null 428 link at the end) and prints it. */ 429 430 void 431 print_script (script, hunkfun, printfun) 432 struct change *script; 433 struct change * (*hunkfun) PARAMS((struct change *)); 434 void (*printfun) PARAMS((struct change *)); 435 { 436 struct change *next = script; 437 438 while (next) 439 { 440 struct change *this, *end; 441 442 /* Find a set of changes that belong together. */ 443 this = next; 444 end = (*hunkfun) (next); 445 446 /* Disconnect them from the rest of the changes, 447 making them a hunk, and remember the rest for next iteration. */ 448 next = end->link; 449 end->link = 0; 450 #ifdef DEBUG 451 debug_script (this); 452 #endif 453 454 /* Print this hunk. */ 455 (*printfun) (this); 456 457 /* Reconnect the script so it will all be freed properly. */ 458 end->link = next; 459 } 460 } 461 462 /* Print the text of a single line LINE, 463 flagging it with the characters in LINE_FLAG (which say whether 464 the line is inserted, deleted, changed, etc.). */ 465 466 void 467 print_1_line (line_flag, line) 468 char const *line_flag; 469 char const * const *line; 470 { 471 char const *text = line[0], *limit = line[1]; /* Help the compiler. */ 472 FILE *out = outfile; /* Help the compiler some more. */ 473 char const *flag_format = 0; 474 475 /* If -T was specified, use a Tab between the line-flag and the text. 476 Otherwise use a Space (as Unix diff does). 477 Print neither space nor tab if line-flags are empty. */ 478 479 if (line_flag && *line_flag) 480 { 481 flag_format = tab_align_flag ? "%s\t" : "%s "; 482 fprintf (out, flag_format, line_flag); 483 } 484 485 output_1_line (text, limit, flag_format, line_flag); 486 487 if ((!line_flag || line_flag[0]) && limit[-1] != '\n') 488 fputc ('\n', out); 489 } 490 491 /* Output a line from TEXT up to LIMIT. Without -t, output verbatim. 492 With -t, expand white space characters to spaces, and if FLAG_FORMAT 493 is nonzero, output it with argument LINE_FLAG after every 494 internal carriage return, so that tab stops continue to line up. */ 495 496 void 497 output_1_line (text, limit, flag_format, line_flag) 498 char const *text, *limit, *flag_format, *line_flag; 499 { 500 if (!tab_expand_flag) 501 fwrite (text, sizeof (char), limit - text, outfile); 502 else 503 { 504 register FILE *out = outfile; 505 register unsigned char c; 506 register char const *t = text; 507 register unsigned column = 0; 508 509 while (t < limit) 510 switch ((c = *t++)) 511 { 512 case '\t': 513 { 514 unsigned spaces = TAB_WIDTH - column % TAB_WIDTH; 515 column += spaces; 516 do 517 putc (' ', out); 518 while (--spaces); 519 } 520 break; 521 522 case '\r': 523 putc (c, out); 524 if (flag_format && t < limit && *t != '\n') 525 fprintf (out, flag_format, line_flag); 526 column = 0; 527 break; 528 529 case '\b': 530 if (column == 0) 531 continue; 532 column--; 533 putc (c, out); 534 break; 535 536 default: 537 if (ISPRINT (c)) 538 column++; 539 putc (c, out); 540 break; 541 } 542 } 543 } 544 545 int 546 change_letter (inserts, deletes) 547 int inserts, deletes; 548 { 549 if (!inserts) 550 return 'd'; 551 else if (!deletes) 552 return 'a'; 553 else 554 return 'c'; 555 } 556 557 /* Translate an internal line number (an index into diff's table of lines) 558 into an actual line number in the input file. 559 The internal line number is LNUM. FILE points to the data on the file. 560 561 Internal line numbers count from 0 starting after the prefix. 562 Actual line numbers count from 1 within the entire file. */ 563 564 int 565 translate_line_number (file, lnum) 566 struct file_data const *file; 567 int lnum; 568 { 569 return lnum + file->prefix_lines + 1; 570 } 571 572 void 573 translate_range (file, a, b, aptr, bptr) 574 struct file_data const *file; 575 int a, b; 576 int *aptr, *bptr; 577 { 578 *aptr = translate_line_number (file, a - 1) + 1; 579 *bptr = translate_line_number (file, b + 1) - 1; 580 } 581 582 /* Print a pair of line numbers with SEPCHAR, translated for file FILE. 583 If the two numbers are identical, print just one number. 584 585 Args A and B are internal line numbers. 586 We print the translated (real) line numbers. */ 587 588 void 589 print_number_range (sepchar, file, a, b) 590 int sepchar; 591 struct file_data *file; 592 int a, b; 593 { 594 int trans_a, trans_b; 595 translate_range (file, a, b, &trans_a, &trans_b); 596 597 /* Note: we can have B < A in the case of a range of no lines. 598 In this case, we should print the line number before the range, 599 which is B. */ 600 if (trans_b > trans_a) 601 fprintf (outfile, "%d%c%d", trans_a, sepchar, trans_b); 602 else 603 fprintf (outfile, "%d", trans_b); 604 } 605 606 /* Look at a hunk of edit script and report the range of lines in each file 607 that it applies to. HUNK is the start of the hunk, which is a chain 608 of `struct change'. The first and last line numbers of file 0 are stored in 609 *FIRST0 and *LAST0, and likewise for file 1 in *FIRST1 and *LAST1. 610 Note that these are internal line numbers that count from 0. 611 612 If no lines from file 0 are deleted, then FIRST0 is LAST0+1. 613 614 Also set *DELETES nonzero if any lines of file 0 are deleted 615 and set *INSERTS nonzero if any lines of file 1 are inserted. 616 If only ignorable lines are inserted or deleted, both are 617 set to 0. */ 618 619 void 620 analyze_hunk (hunk, first0, last0, first1, last1, deletes, inserts) 621 struct change *hunk; 622 int *first0, *last0, *first1, *last1; 623 int *deletes, *inserts; 624 { 625 int l0, l1, show_from, show_to; 626 int i; 627 int trivial = ignore_blank_lines_flag || ignore_regexp_list; 628 struct change *next; 629 630 show_from = show_to = 0; 631 632 *first0 = hunk->line0; 633 *first1 = hunk->line1; 634 635 next = hunk; 636 do 637 { 638 l0 = next->line0 + next->deleted - 1; 639 l1 = next->line1 + next->inserted - 1; 640 show_from += next->deleted; 641 show_to += next->inserted; 642 643 for (i = next->line0; i <= l0 && trivial; i++) 644 if (!ignore_blank_lines_flag || files[0].linbuf[i][0] != '\n') 645 { 646 struct regexp_list *r; 647 char const *line = files[0].linbuf[i]; 648 int len = files[0].linbuf[i + 1] - line; 649 650 for (r = ignore_regexp_list; r; r = r->next) 651 if (0 <= re_search (&r->buf, line, len, 0, len, 0)) 652 break; /* Found a match. Ignore this line. */ 653 /* If we got all the way through the regexp list without 654 finding a match, then it's nontrivial. */ 655 if (!r) 656 trivial = 0; 657 } 658 659 for (i = next->line1; i <= l1 && trivial; i++) 660 if (!ignore_blank_lines_flag || files[1].linbuf[i][0] != '\n') 661 { 662 struct regexp_list *r; 663 char const *line = files[1].linbuf[i]; 664 int len = files[1].linbuf[i + 1] - line; 665 666 for (r = ignore_regexp_list; r; r = r->next) 667 if (0 <= re_search (&r->buf, line, len, 0, len, 0)) 668 break; /* Found a match. Ignore this line. */ 669 /* If we got all the way through the regexp list without 670 finding a match, then it's nontrivial. */ 671 if (!r) 672 trivial = 0; 673 } 674 } 675 while ((next = next->link) != 0); 676 677 *last0 = l0; 678 *last1 = l1; 679 680 /* If all inserted or deleted lines are ignorable, 681 tell the caller to ignore this hunk. */ 682 683 if (trivial) 684 show_from = show_to = 0; 685 686 *deletes = show_from; 687 *inserts = show_to; 688 } 689 690 /* malloc a block of memory, with fatal error message if we can't do it. */ 691 692 VOID * 693 xmalloc (size) 694 size_t size; 695 { 696 register VOID *value; 697 698 if (size == 0) 699 size = 1; 700 701 value = (VOID *) malloc (size); 702 703 if (!value) 704 fatal ("memory exhausted"); 705 return value; 706 } 707 708 /* realloc a block of memory, with fatal error message if we can't do it. */ 709 710 VOID * 711 xrealloc (old, size) 712 VOID *old; 713 size_t size; 714 { 715 register VOID *value; 716 717 if (size == 0) 718 size = 1; 719 720 value = (VOID *) realloc (old, size); 721 722 if (!value) 723 fatal ("memory exhausted"); 724 return value; 725 } 726 727 /* Concatenate three strings, returning a newly malloc'd string. */ 728 729 char * 730 concat (s1, s2, s3) 731 char const *s1, *s2, *s3; 732 { 733 size_t len = strlen (s1) + strlen (s2) + strlen (s3); 734 char *new = xmalloc (len + 1); 735 sprintf (new, "%s%s%s", s1, s2, s3); 736 return new; 737 } 738 739 /* Yield the newly malloc'd pathname 740 of the file in DIR whose filename is FILE. */ 741 742 char * 743 dir_file_pathname (dir, file) 744 char const *dir, *file; 745 { 746 char const *p = filename_lastdirchar (dir); 747 return concat (dir, "/" + (p && !p[1]), file); 748 } 749 750 void 751 debug_script (sp) 752 struct change *sp; 753 { 754 fflush (stdout); 755 for (; sp; sp = sp->link) 756 fprintf (stderr, "%3d %3d delete %d insert %d\n", 757 sp->line0, sp->line1, sp->deleted, sp->inserted); 758 fflush (stderr); 759 } 760