1 /*
2 * $OpenBSD: inp.c,v 1.34 2006/03/11 19:41:30 otto Exp $
3 * $DragonFly: src/usr.bin/patch/inp.c,v 1.6 2007/09/29 23:11:10 swildner Exp $
4 * $NetBSD: inp.c,v 1.30 2023/06/16 23:36:26 wiz Exp $
5 */
6
7 /*
8 * patch - a program to apply diffs to original files
9 *
10 * Copyright 1986, Larry Wall
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following condition is met:
14 * 1. Redistributions of source code must retain the above copyright notice,
15 * this condition and the following disclaimer.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND ANY
18 * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
19 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
20 * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR
21 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
23 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
24 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * -C option added in 1998, original code by Marc Espie, based on FreeBSD
30 * behaviour
31 */
32
33 #include <sys/cdefs.h>
34 __RCSID("$NetBSD: inp.c,v 1.30 2023/06/16 23:36:26 wiz Exp $");
35
36 #include <sys/types.h>
37 #include <sys/file.h>
38 #include <sys/stat.h>
39 #include <sys/mman.h>
40 #include <sys/wait.h>
41
42 #include <ctype.h>
43 #include <errno.h>
44 #include <fcntl.h>
45 #include <libgen.h>
46 #include <limits.h>
47 #include <stddef.h>
48 #include <stdio.h>
49 #include <stdlib.h>
50 #include <string.h>
51 #include <unistd.h>
52
53 #include "common.h"
54 #include "util.h"
55 #include "pch.h"
56 #include "inp.h"
57
58
59 /* Input-file-with-indexable-lines abstract type */
60
61 static off_t i_size; /* size of the input file */
62 static char *i_womp; /* plan a buffer for entire file */
63 static char **i_ptr; /* pointers to lines in i_womp */
64 static char empty_line[] = { '\0' };
65
66 static int tifd = -1; /* plan b virtual string array */
67 static char *tibuf[2]; /* plan b buffers */
68 static LINENUM tiline[2] = {-1, -1}; /* 1st line in each buffer */
69 static LINENUM lines_per_buf; /* how many lines per buffer */
70 static int tireclen; /* length of records in tmp file */
71
72 static bool rev_in_string(const char *);
73 static bool reallocate_lines(size_t *);
74
75 /* returns false if insufficient memory */
76 static bool plan_a(const char *);
77
78 static void plan_b(const char *);
79
80 /* New patch--prepare to edit another file. */
81
82 void
re_input(void)83 re_input(void)
84 {
85 if (using_plan_a) {
86 i_size = 0;
87 free(i_ptr);
88 i_ptr = NULL;
89 if (i_womp != NULL) {
90 munmap(i_womp, i_size);
91 i_womp = NULL;
92 }
93 } else {
94 using_plan_a = true; /* maybe the next one is smaller */
95 close(tifd);
96 tifd = -1;
97 free(tibuf[0]);
98 free(tibuf[1]);
99 tibuf[0] = tibuf[1] = NULL;
100 tiline[0] = tiline[1] = -1;
101 tireclen = 0;
102 }
103 }
104
105 /* Construct the line index, somehow or other. */
106
107 void
scan_input(const char * filename)108 scan_input(const char *filename)
109 {
110 if (!plan_a(filename))
111 plan_b(filename);
112 if (verbose) {
113 say("Patching file %s using Plan %s...\n", filename,
114 (using_plan_a ? "A" : "B"));
115 }
116 }
117
118 static bool
reallocate_lines(size_t * lines_allocated)119 reallocate_lines(size_t *lines_allocated)
120 {
121 char **p;
122 size_t new_size;
123
124 new_size = *lines_allocated * 3 / 2;
125 p = pch_realloc(i_ptr, new_size + 2, sizeof(char *));
126 if (p == NULL) { /* shucks, it was a near thing */
127 munmap(i_womp, i_size);
128 i_womp = NULL;
129 free(i_ptr);
130 i_ptr = NULL;
131 *lines_allocated = 0;
132 return false;
133 }
134 *lines_allocated = new_size;
135 i_ptr = p;
136 return true;
137 }
138
139 /* Try keeping everything in memory. */
140
141 static bool
plan_a(const char * filename)142 plan_a(const char *filename)
143 {
144 int ifd, statfailed, devnull, pstat;
145 char *p, *s, *lbuf;
146 struct stat filestat;
147 off_t i;
148 ptrdiff_t sz;
149 size_t iline, lines_allocated, lbufsz;
150 pid_t pid;
151 char *argp[4] = {NULL};
152
153 #ifdef DEBUGGING
154 if (debug & 8)
155 return false;
156 #endif
157
158 if (filename == NULL || *filename == '\0')
159 return false;
160
161 statfailed = stat(filename, &filestat);
162 if (statfailed && ok_to_create_file) {
163 if (verbose)
164 say("(Creating file %s...)\n", filename);
165
166 /*
167 * in check_patch case, we still display `Creating file' even
168 * though we're not. The rule is that -C should be as similar
169 * to normal patch behavior as possible
170 */
171 if (check_only)
172 return true;
173 makedirs(filename, true);
174 close(creat(filename, 0666));
175 statfailed = stat(filename, &filestat);
176 }
177 if (statfailed && check_only)
178 fatal("%s not found, -C mode, can't probe further\n", filename);
179 /* For nonexistent or read-only files, look for RCS versions. */
180 if (statfailed ||
181 /* No one can write to it. */
182 (filestat.st_mode & 0222) == 0 ||
183 /* I can't write to it. */
184 ((filestat.st_mode & 0022) == 0 && filestat.st_uid != getuid())) {
185 char *filebase, *filedir;
186 struct stat cstat;
187 char *tmp_filename1, *tmp_filename2;
188
189 tmp_filename1 = strdup(filename);
190 tmp_filename2 = strdup(filename);
191 if (tmp_filename1 == NULL || tmp_filename2 == NULL)
192 fatal("strdupping filename");
193
194 filebase = basename(tmp_filename1);
195 filedir = dirname(tmp_filename2);
196
197 lbufsz = INITLINELEN;
198 if ((lbuf = malloc(bufsz)) == NULL)
199 pfatal("allocating line buffer");
200 lbuf[0] = '\0';
201
202 #define try(f, a1, a2, a3) \
203 (snprintf(lbuf, lbufsz, f, a1, a2, a3), stat(lbuf, &cstat) == 0)
204
205 /*
206 * else we can't write to it but it's not under a version
207 * control system, so just proceed.
208 */
209 if (try("%s/RCS/%s%s", filedir, filebase, RCSSUFFIX) ||
210 try("%s/RCS/%s%s", filedir, filebase, "") ||
211 try("%s/%s%s", filedir, filebase, RCSSUFFIX)) {
212 if (!statfailed) {
213 if ((filestat.st_mode & 0222) != 0)
214 /* The owner can write to it. */
215 fatal("file %s seems to be locked "
216 "by somebody else under RCS\n",
217 filename);
218 /*
219 * It might be checked out unlocked. See if
220 * it's safe to check out the default version
221 * locked.
222 */
223 if (verbose)
224 say("Comparing file %s to default "
225 "RCS version...\n", filename);
226
227 switch (pid = fork()) {
228 case -1:
229 fatal("can't fork: %s\n",
230 strerror(errno));
231 case 0:
232 devnull = open("/dev/null", O_RDONLY);
233 if (devnull == -1) {
234 fatal("can't open /dev/null: %s",
235 strerror(errno));
236 }
237 (void)dup2(devnull, STDOUT_FILENO);
238 argp[0] = __UNCONST(RCSDIFF);
239 argp[1] = __UNCONST(filename);
240 execv(RCSDIFF, argp);
241 exit(127);
242 }
243 pid = waitpid(pid, &pstat, 0);
244 if (pid == -1 || WEXITSTATUS(pstat) != 0) {
245 fatal("can't check out file %s: "
246 "differs from default RCS version\n",
247 filename);
248 }
249 }
250
251 if (verbose)
252 say("Checking out file %s from RCS...\n",
253 filename);
254
255 switch (pid = fork()) {
256 case -1:
257 fatal("can't fork: %s\n", strerror(errno));
258 case 0:
259 argp[0] = __UNCONST(CHECKOUT);
260 argp[1] = __UNCONST("-l");
261 argp[2] = __UNCONST(filename);
262 execv(CHECKOUT, argp);
263 exit(127);
264 }
265 pid = waitpid(pid, &pstat, 0);
266 if (pid == -1 || WEXITSTATUS(pstat) != 0 ||
267 stat(filename, &filestat)) {
268 fatal("can't check out file %s from RCS\n",
269 filename);
270 }
271 } else if (statfailed) {
272 fatal("can't find %s\n", filename);
273 }
274 free(lbuf);
275 free(tmp_filename1);
276 free(tmp_filename2);
277 }
278
279 filemode = filestat.st_mode;
280 if (!S_ISREG(filemode))
281 fatal("%s is not a normal file--can't patch\n", filename);
282 i_size = filestat.st_size;
283 if (out_of_mem) {
284 set_hunkmax(); /* make sure dynamic arrays are allocated */
285 out_of_mem = false;
286 return false; /* force plan b because plan a bombed */
287 }
288 if ((uintmax_t)i_size > (uintmax_t)SIZE_MAX) {
289 say("block too large to mmap\n");
290 return false;
291 }
292 if ((ifd = open(filename, O_RDONLY)) < 0)
293 pfatal("can't open file %s", filename);
294
295 if (i_size) {
296 i_womp = mmap(NULL, i_size, PROT_READ, MAP_PRIVATE, ifd, 0);
297 if (i_womp == MAP_FAILED) {
298 perror("mmap failed");
299 i_womp = NULL;
300 close(ifd);
301 return false;
302 }
303 } else {
304 i_womp = NULL;
305 }
306
307 close(ifd);
308 if (i_size)
309 madvise(i_womp, i_size, MADV_SEQUENTIAL);
310
311 /* estimate the number of lines */
312 lines_allocated = i_size / 25;
313 if (lines_allocated < 100)
314 lines_allocated = 100;
315
316 if (!reallocate_lines(&lines_allocated))
317 return false;
318
319 /* now scan the buffer and build pointer array */
320 iline = 1;
321 i_ptr[iline] = i_womp;
322 /* test for NUL too, to maintain the behavior of the original code */
323 for (s = i_womp, i = 0; i < i_size && *s != '\0'; s++, i++) {
324 if (*s == '\n') {
325 if (iline == lines_allocated) {
326 if (!reallocate_lines(&lines_allocated))
327 return false;
328 }
329 /* these are NOT NUL terminated */
330 i_ptr[++iline] = s + 1;
331 }
332 }
333 /* if the last line contains no EOL, append one */
334 if (i_size > 0 && i_womp[i_size - 1] != '\n') {
335 last_line_missing_eol = true;
336 /* fix last line */
337 sz = s - i_ptr[iline];
338 p = malloc(sz + 1);
339 if (p == NULL) {
340 free(i_ptr);
341 i_ptr = NULL;
342 munmap(i_womp, i_size);
343 i_womp = NULL;
344 return false;
345 }
346
347 memcpy(p, i_ptr[iline], sz);
348 p[sz] = '\n';
349 i_ptr[iline] = p;
350 /* count the extra line and make it point to some valid mem */
351 i_ptr[++iline] = empty_line;
352 } else
353 last_line_missing_eol = false;
354
355 input_lines = iline - 1;
356
357 /* now check for revision, if any */
358
359 if (revision != NULL) {
360 if (!rev_in_string(i_womp)) {
361 if (force) {
362 if (verbose)
363 say("Warning: this file doesn't appear "
364 "to be the %s version--patching anyway.\n",
365 revision);
366 } else if (batch) {
367 fatal("this file doesn't appear to be the "
368 "%s version--aborting.\n",
369 revision);
370 } else {
371 ask("This file doesn't appear to be the "
372 "%s version--patch anyway? [n] ",
373 revision);
374 if (*buf != 'y')
375 fatal("aborted\n");
376 }
377 } else if (verbose)
378 say("Good. This file appears to be the %s version.\n",
379 revision);
380 }
381 return true; /* plan a will work */
382 }
383
384 /* Keep (virtually) nothing in memory. */
385
386 static void
plan_b(const char * filename)387 plan_b(const char *filename)
388 {
389 FILE *ifp;
390 size_t i = 0, j, maxlen = 1;
391 char *p;
392 bool found_revision = (revision == NULL);
393
394 using_plan_a = false;
395 if ((ifp = fopen(filename, "r")) == NULL)
396 pfatal("can't open file %s", filename);
397 unlink(TMPINNAME);
398 if ((tifd = open(TMPINNAME, O_EXCL | O_CREAT | O_WRONLY, 0666)) < 0)
399 pfatal("can't open file %s", TMPINNAME);
400 while (getline(&buf, &bufsz, ifp) != -1) {
401 if (revision != NULL && !found_revision && rev_in_string(buf))
402 found_revision = true;
403 if ((i = strlen(buf)) > maxlen)
404 maxlen = i; /* find longest line */
405 }
406 last_line_missing_eol = i > 0 && buf[i - 1] != '\n';
407 if (last_line_missing_eol && maxlen == i)
408 maxlen++;
409
410 if (revision != NULL) {
411 if (!found_revision) {
412 if (force) {
413 if (verbose)
414 say("Warning: this file doesn't appear "
415 "to be the %s version--patching anyway.\n",
416 revision);
417 } else if (batch) {
418 fatal("this file doesn't appear to be the "
419 "%s version--aborting.\n",
420 revision);
421 } else {
422 ask("This file doesn't appear to be the %s "
423 "version--patch anyway? [n] ",
424 revision);
425 if (*buf != 'y')
426 fatal("aborted\n");
427 }
428 } else if (verbose)
429 say("Good. This file appears to be the %s version.\n",
430 revision);
431 }
432 fseek(ifp, 0L, SEEK_SET); /* rewind file */
433 lines_per_buf = BUFFERSIZE / maxlen;
434 tireclen = maxlen;
435 tibuf[0] = malloc(BUFFERSIZE + 1);
436 if (tibuf[0] == NULL)
437 fatal("out of memory\n");
438 tibuf[1] = malloc(BUFFERSIZE + 1);
439 if (tibuf[1] == NULL)
440 fatal("out of memory\n");
441 for (i = 1;; i++) {
442 p = tibuf[0] + maxlen * (i % lines_per_buf);
443 if (i % lines_per_buf == 0) /* new block */
444 if (write(tifd, tibuf[0], BUFFERSIZE) < BUFFERSIZE)
445 pfatal("can't write temp file");
446 if (fgets(p, maxlen + 1, ifp) == NULL) {
447 input_lines = i - 1;
448 if (i % lines_per_buf != 0)
449 if (write(tifd, tibuf[0], BUFFERSIZE) < BUFFERSIZE)
450 pfatal("can't write temp file");
451 break;
452 }
453 j = strlen(p);
454 /* These are '\n' terminated strings, so no need to add a NUL */
455 if (j == 0 || p[j - 1] != '\n')
456 p[j] = '\n';
457 }
458 fclose(ifp);
459 close(tifd);
460 if ((tifd = open(TMPINNAME, O_RDONLY)) < 0)
461 pfatal("can't reopen file %s", TMPINNAME);
462 }
463
464 /*
465 * Fetch a line from the input file, \n terminated, not necessarily \0.
466 */
467 char *
ifetch(LINENUM line,int whichbuf)468 ifetch(LINENUM line, int whichbuf)
469 {
470 if (line < 1 || line > input_lines) {
471 if (warn_on_invalid_line) {
472 say("No such line %ld in input file, ignoring\n", line);
473 warn_on_invalid_line = false;
474 }
475 return NULL;
476 }
477 if (using_plan_a)
478 return i_ptr[line];
479 else {
480 LINENUM offline = line % lines_per_buf;
481 LINENUM baseline = line - offline;
482
483 if (tiline[0] == baseline)
484 whichbuf = 0;
485 else if (tiline[1] == baseline)
486 whichbuf = 1;
487 else {
488 tiline[whichbuf] = baseline;
489
490 if (lseek(tifd, (off_t) (baseline / lines_per_buf *
491 BUFFERSIZE), SEEK_SET) < 0)
492 pfatal("cannot seek in the temporary input file");
493
494 if (read(tifd, tibuf[whichbuf], BUFFERSIZE) < 0)
495 pfatal("error reading tmp file %s", TMPINNAME);
496 }
497 return tibuf[whichbuf] + (tireclen * offline);
498 }
499 }
500
501 /*
502 * True if the string argument contains the revision number we want.
503 */
504 static bool
rev_in_string(const char * string)505 rev_in_string(const char *string)
506 {
507 const char *s;
508 size_t patlen;
509
510 if (revision == NULL)
511 return true;
512 patlen = strlen(revision);
513 if (strnEQ(string, revision, patlen) && isspace((unsigned char)string[patlen]))
514 return true;
515 for (s = string; *s; s++) {
516 if (isspace((unsigned char)*s) && strnEQ(s + 1, revision, patlen) &&
517 isspace((unsigned char)s[patlen + 1])) {
518 return true;
519 }
520 }
521 return false;
522 }
523