xref: /freebsd/usr.bin/patch/inp.c (revision 3fe401a500cdfc73d8c066da3c577c4b9f0aa953)
1 /*-
2  * Copyright 1986, Larry Wall
3  *
4  * Redistribution and use in source and binary forms, with or without
5  * modification, are permitted provided that the following condition is met:
6  * 1. Redistributions of source code must retain the above copyright notice,
7  * this condition and the following disclaimer.
8  *
9  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND ANY
10  * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
11  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
12  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR
13  * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
14  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
15  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
16  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
17  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
18  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
19  * SUCH DAMAGE.
20  *
21  * patch - a program to apply diffs to original files
22  *
23  * -C option added in 1998, original code by Marc Espie, based on FreeBSD
24  * behaviour
25  *
26  * $OpenBSD: inp.c,v 1.36 2012/04/10 14:46:34 ajacoutot Exp $
27  * $FreeBSD$
28  */
29 
30 #include <sys/types.h>
31 #include <sys/file.h>
32 #include <sys/stat.h>
33 #include <sys/mman.h>
34 #include <sys/wait.h>
35 
36 #include <ctype.h>
37 #include <errno.h>
38 #include <libgen.h>
39 #include <stddef.h>
40 #include <stdint.h>
41 #include <stdio.h>
42 #include <stdlib.h>
43 #include <string.h>
44 #include <unistd.h>
45 
46 #include "common.h"
47 #include "util.h"
48 #include "pch.h"
49 #include "inp.h"
50 
51 
52 /* Input-file-with-indexable-lines abstract type */
53 
54 static size_t	i_size;		/* size of the input file */
55 static char	*i_womp;	/* plan a buffer for entire file */
56 static char	**i_ptr;	/* pointers to lines in i_womp */
57 static char	empty_line[] = { '\0' };
58 
59 static int	tifd = -1;	/* plan b virtual string array */
60 static char	*tibuf[2];	/* plan b buffers */
61 static LINENUM	tiline[2] = {-1, -1};	/* 1st line in each buffer */
62 static size_t	lines_per_buf;	/* how many lines per buffer */
63 static size_t	tibuflen;	/* plan b buffer length */
64 static size_t	tireclen;	/* length of records in tmp file */
65 
66 static bool	rev_in_string(const char *);
67 static bool	reallocate_lines(size_t *);
68 
69 /* returns false if insufficient memory */
70 static bool	plan_a(const char *);
71 
72 static void	plan_b(const char *);
73 
74 /* New patch--prepare to edit another file. */
75 
76 void
77 re_input(void)
78 {
79 	if (using_plan_a) {
80 		free(i_ptr);
81 		i_ptr = NULL;
82 		if (i_womp != NULL) {
83 			munmap(i_womp, i_size);
84 			i_womp = NULL;
85 		}
86 		i_size = 0;
87 	} else {
88 		using_plan_a = true;	/* maybe the next one is smaller */
89 		close(tifd);
90 		tifd = -1;
91 		free(tibuf[0]);
92 		free(tibuf[1]);
93 		tibuf[0] = tibuf[1] = NULL;
94 		tiline[0] = tiline[1] = -1;
95 		tireclen = 0;
96 	}
97 }
98 
99 /* Construct the line index, somehow or other. */
100 
101 void
102 scan_input(const char *filename)
103 {
104 	if (!plan_a(filename))
105 		plan_b(filename);
106 	if (verbose) {
107 		say("Patching file %s using Plan %s...\n", filename,
108 		    (using_plan_a ? "A" : "B"));
109 	}
110 }
111 
112 static bool
113 reallocate_lines(size_t *lines_allocated)
114 {
115 	char	**p;
116 	size_t	new_size;
117 
118 	new_size = *lines_allocated * 3 / 2;
119 	p = realloc(i_ptr, (new_size + 2) * sizeof(char *));
120 	if (p == NULL) {	/* shucks, it was a near thing */
121 		munmap(i_womp, i_size);
122 		i_womp = NULL;
123 		free(i_ptr);
124 		i_ptr = NULL;
125 		*lines_allocated = 0;
126 		return false;
127 	}
128 	*lines_allocated = new_size;
129 	i_ptr = p;
130 	return true;
131 }
132 
133 /* Try keeping everything in memory. */
134 
135 static bool
136 plan_a(const char *filename)
137 {
138 	int		ifd, statfailed, devnull, pstat;
139 	char		*p, *s, lbuf[INITLINELEN];
140 	struct stat	filestat;
141 	ptrdiff_t	sz;
142 	size_t		i;
143 	size_t		iline, lines_allocated;
144 	pid_t		pid;
145 	char		*argp[4] = {NULL};
146 
147 #ifdef DEBUGGING
148 	if (debug & 8)
149 		return false;
150 #endif
151 
152 	if (filename == NULL || *filename == '\0')
153 		return false;
154 
155 	statfailed = stat(filename, &filestat);
156 	if (statfailed && ok_to_create_file) {
157 		if (verbose)
158 			say("(Creating file %s...)\n", filename);
159 
160 		/*
161 		 * in check_patch case, we still display `Creating file' even
162 		 * though we're not. The rule is that -C should be as similar
163 		 * to normal patch behavior as possible
164 		 */
165 		if (check_only)
166 			return true;
167 		makedirs(filename, true);
168 		close(creat(filename, 0666));
169 		statfailed = stat(filename, &filestat);
170 	}
171 	if (statfailed && check_only)
172 		fatal("%s not found, -C mode, can't probe further\n", filename);
173 	/* For nonexistent or read-only files, look for RCS versions.  */
174 
175 	if (statfailed ||
176 	    /* No one can write to it.  */
177 	    (filestat.st_mode & 0222) == 0 ||
178 	    /* I can't write to it.  */
179 	    ((filestat.st_mode & 0022) == 0 && filestat.st_uid != getuid())) {
180 		char	*filebase, *filedir;
181 		struct stat	cstat;
182 		char *tmp_filename1, *tmp_filename2;
183 
184 		tmp_filename1 = strdup(filename);
185 		tmp_filename2 = strdup(filename);
186 		if (tmp_filename1 == NULL || tmp_filename2 == NULL)
187 			fatal("strdupping filename");
188 
189 		filebase = basename(tmp_filename1);
190 		filedir = dirname(tmp_filename2);
191 
192 #define try(f, a1, a2, a3) \
193 	(snprintf(lbuf, sizeof(lbuf), f, a1, a2, a3), stat(lbuf, &cstat) == 0)
194 
195 		/*
196 		 * else we can't write to it but it's not under a version
197 		 * control system, so just proceed.
198 		 */
199 		if (try("%s/RCS/%s%s", filedir, filebase, RCSSUFFIX) ||
200 		    try("%s/RCS/%s%s", filedir, filebase, "") ||
201 		    try("%s/%s%s", filedir, filebase, RCSSUFFIX)) {
202 			if (!statfailed) {
203 				if ((filestat.st_mode & 0222) != 0)
204 					/* The owner can write to it.  */
205 					fatal("file %s seems to be locked "
206 					    "by somebody else under RCS\n",
207 					    filename);
208 				/*
209 				 * It might be checked out unlocked.  See if
210 				 * it's safe to check out the default version
211 				 * locked.
212 				 */
213 				if (verbose)
214 					say("Comparing file %s to default "
215 					    "RCS version...\n", filename);
216 
217 				switch (pid = fork()) {
218 				case -1:
219 					fatal("can't fork: %s\n",
220 					    strerror(errno));
221 				case 0:
222 					devnull = open("/dev/null", O_RDONLY);
223 					if (devnull == -1) {
224 						fatal("can't open /dev/null: %s",
225 						    strerror(errno));
226 					}
227 					(void)dup2(devnull, STDOUT_FILENO);
228 					argp[0] = strdup(RCSDIFF);
229 					argp[1] = strdup(filename);
230 					execv(RCSDIFF, argp);
231 					exit(127);
232 				}
233 				pid = waitpid(pid, &pstat, 0);
234 				if (pid == -1 || WEXITSTATUS(pstat) != 0) {
235 					fatal("can't check out file %s: "
236 					    "differs from default RCS version\n",
237 					    filename);
238 				}
239 			}
240 
241 			if (verbose)
242 				say("Checking out file %s from RCS...\n",
243 				    filename);
244 
245 			switch (pid = fork()) {
246 			case -1:
247 				fatal("can't fork: %s\n", strerror(errno));
248 			case 0:
249 				argp[0] = strdup(CHECKOUT);
250 				argp[1] = strdup("-l");
251 				argp[2] = strdup(filename);
252 				execv(CHECKOUT, argp);
253 				exit(127);
254 			}
255 			pid = waitpid(pid, &pstat, 0);
256 			if (pid == -1 || WEXITSTATUS(pstat) != 0 ||
257 			    stat(filename, &filestat)) {
258 				fatal("can't check out file %s from RCS\n",
259 				    filename);
260 			}
261 		} else if (statfailed) {
262 			fatal("can't find %s\n", filename);
263 		}
264 		free(tmp_filename1);
265 		free(tmp_filename2);
266 	}
267 
268 	filemode = filestat.st_mode;
269 	if (!S_ISREG(filemode))
270 		fatal("%s is not a normal file--can't patch\n", filename);
271 	if ((uint64_t)filestat.st_size > SIZE_MAX) {
272 		say("block too large to mmap\n");
273 		return false;
274 	}
275 	i_size = (size_t)filestat.st_size;
276 	if (out_of_mem) {
277 		set_hunkmax();	/* make sure dynamic arrays are allocated */
278 		out_of_mem = false;
279 		return false;	/* force plan b because plan a bombed */
280 	}
281 	if ((ifd = open(filename, O_RDONLY)) < 0)
282 		pfatal("can't open file %s", filename);
283 
284 	if (i_size) {
285 		i_womp = mmap(NULL, i_size, PROT_READ, MAP_PRIVATE, ifd, 0);
286 		if (i_womp == MAP_FAILED) {
287 			perror("mmap failed");
288 			i_womp = NULL;
289 			close(ifd);
290 			return false;
291 		}
292 	} else {
293 		i_womp = NULL;
294 	}
295 
296 	close(ifd);
297 	if (i_size)
298 		madvise(i_womp, i_size, MADV_SEQUENTIAL);
299 
300 	/* estimate the number of lines */
301 	lines_allocated = i_size / 25;
302 	if (lines_allocated < 100)
303 		lines_allocated = 100;
304 
305 	if (!reallocate_lines(&lines_allocated))
306 		return false;
307 
308 	/* now scan the buffer and build pointer array */
309 	iline = 1;
310 	i_ptr[iline] = i_womp;
311 	/* test for NUL too, to maintain the behavior of the original code */
312 	for (s = i_womp, i = 0; i < i_size && *s != '\0'; s++, i++) {
313 		if (*s == '\n') {
314 			if (iline == lines_allocated) {
315 				if (!reallocate_lines(&lines_allocated))
316 					return false;
317 			}
318 			/* these are NOT NUL terminated */
319 			i_ptr[++iline] = s + 1;
320 		}
321 	}
322 	/* if the last line contains no EOL, append one */
323 	if (i_size > 0 && i_womp[i_size - 1] != '\n') {
324 		last_line_missing_eol = true;
325 		/* fix last line */
326 		sz = s - i_ptr[iline];
327 		p = malloc(sz + 1);
328 		if (p == NULL) {
329 			free(i_ptr);
330 			i_ptr = NULL;
331 			munmap(i_womp, i_size);
332 			i_womp = NULL;
333 			return false;
334 		}
335 
336 		memcpy(p, i_ptr[iline], sz);
337 		p[sz] = '\n';
338 		i_ptr[iline] = p;
339 		/* count the extra line and make it point to some valid mem */
340 		i_ptr[++iline] = empty_line;
341 	} else
342 		last_line_missing_eol = false;
343 
344 	input_lines = iline - 1;
345 
346 	/* now check for revision, if any */
347 
348 	if (revision != NULL) {
349 		if (i_womp == NULL || !rev_in_string(i_womp)) {
350 			if (force) {
351 				if (verbose)
352 					say("Warning: this file doesn't appear "
353 					    "to be the %s version--patching anyway.\n",
354 					    revision);
355 			} else if (batch) {
356 				fatal("this file doesn't appear to be the "
357 				    "%s version--aborting.\n",
358 				    revision);
359 			} else {
360 				ask("This file doesn't appear to be the "
361 				    "%s version--patch anyway? [n] ",
362 				    revision);
363 				if (*buf != 'y')
364 					fatal("aborted\n");
365 			}
366 		} else if (verbose)
367 			say("Good.  This file appears to be the %s version.\n",
368 			    revision);
369 	}
370 	return true;		/* plan a will work */
371 }
372 
373 /* Keep (virtually) nothing in memory. */
374 
375 static void
376 plan_b(const char *filename)
377 {
378 	FILE	*ifp;
379 	size_t	i = 0, j, len, maxlen = 1;
380 	char	*lbuf = NULL, *p;
381 	bool	found_revision = (revision == NULL);
382 
383 	using_plan_a = false;
384 	if ((ifp = fopen(filename, "r")) == NULL)
385 		pfatal("can't open file %s", filename);
386 	unlink(TMPINNAME);
387 	if ((tifd = open(TMPINNAME, O_EXCL | O_CREAT | O_WRONLY, 0666)) < 0)
388 		pfatal("can't open file %s", TMPINNAME);
389 	while ((p = fgetln(ifp, &len)) != NULL) {
390 		if (p[len - 1] == '\n')
391 			p[len - 1] = '\0';
392 		else {
393 			/* EOF without EOL, copy and add the NUL */
394 			if ((lbuf = malloc(len + 1)) == NULL)
395 				fatal("out of memory\n");
396 			memcpy(lbuf, p, len);
397 			lbuf[len] = '\0';
398 			p = lbuf;
399 
400 			last_line_missing_eol = true;
401 			len++;
402 		}
403 		if (revision != NULL && !found_revision && rev_in_string(p))
404 			found_revision = true;
405 		if (len > maxlen)
406 			maxlen = len;   /* find longest line */
407 	}
408 	free(lbuf);
409 	if (ferror(ifp))
410 		pfatal("can't read file %s", filename);
411 
412 	if (revision != NULL) {
413 		if (!found_revision) {
414 			if (force) {
415 				if (verbose)
416 					say("Warning: this file doesn't appear "
417 					    "to be the %s version--patching anyway.\n",
418 					    revision);
419 			} else if (batch) {
420 				fatal("this file doesn't appear to be the "
421 				    "%s version--aborting.\n",
422 				    revision);
423 			} else {
424 				ask("This file doesn't appear to be the %s "
425 				    "version--patch anyway? [n] ",
426 				    revision);
427 				if (*buf != 'y')
428 					fatal("aborted\n");
429 			}
430 		} else if (verbose)
431 			say("Good.  This file appears to be the %s version.\n",
432 			    revision);
433 	}
434 	fseek(ifp, 0L, SEEK_SET);	/* rewind file */
435 	tireclen = maxlen;
436 	tibuflen = maxlen > BUFFERSIZE ? maxlen : BUFFERSIZE;
437 	lines_per_buf = tibuflen / maxlen;
438 	tibuf[0] = malloc(tibuflen + 1);
439 	if (tibuf[0] == NULL)
440 		fatal("out of memory\n");
441 	tibuf[1] = malloc(tibuflen + 1);
442 	if (tibuf[1] == NULL)
443 		fatal("out of memory\n");
444 	for (i = 1;; i++) {
445 		p = tibuf[0] + maxlen * (i % lines_per_buf);
446 		if (i % lines_per_buf == 0)	/* new block */
447 			if (write(tifd, tibuf[0], tibuflen) !=
448 			    (ssize_t) tibuflen)
449 				pfatal("can't write temp file");
450 		if (fgets(p, maxlen + 1, ifp) == NULL) {
451 			input_lines = i - 1;
452 			if (i % lines_per_buf != 0)
453 				if (write(tifd, tibuf[0], tibuflen) !=
454 				    (ssize_t) tibuflen)
455 					pfatal("can't write temp file");
456 			break;
457 		}
458 		j = strlen(p);
459 		/* These are '\n' terminated strings, so no need to add a NUL */
460 		if (j == 0 || p[j - 1] != '\n')
461 			p[j] = '\n';
462 	}
463 	fclose(ifp);
464 	close(tifd);
465 	if ((tifd = open(TMPINNAME, O_RDONLY)) < 0)
466 		pfatal("can't reopen file %s", TMPINNAME);
467 }
468 
469 /*
470  * Fetch a line from the input file, \n terminated, not necessarily \0.
471  */
472 char *
473 ifetch(LINENUM line, int whichbuf)
474 {
475 	if (line < 1 || line > input_lines) {
476 		if (warn_on_invalid_line) {
477 			say("No such line %ld in input file, ignoring\n", line);
478 			warn_on_invalid_line = false;
479 		}
480 		return NULL;
481 	}
482 	if (using_plan_a)
483 		return i_ptr[line];
484 	else {
485 		LINENUM	offline = line % lines_per_buf;
486 		LINENUM	baseline = line - offline;
487 
488 		if (tiline[0] == baseline)
489 			whichbuf = 0;
490 		else if (tiline[1] == baseline)
491 			whichbuf = 1;
492 		else {
493 			tiline[whichbuf] = baseline;
494 
495 			if (lseek(tifd, (off_t) (baseline / lines_per_buf *
496 			    tibuflen), SEEK_SET) < 0)
497 				pfatal("cannot seek in the temporary input file");
498 
499 			if (read(tifd, tibuf[whichbuf], tibuflen) !=
500 			    (ssize_t) tibuflen)
501 				pfatal("error reading tmp file %s", TMPINNAME);
502 		}
503 		return tibuf[whichbuf] + (tireclen * offline);
504 	}
505 }
506 
507 /*
508  * True if the string argument contains the revision number we want.
509  */
510 static bool
511 rev_in_string(const char *string)
512 {
513 	const char	*s;
514 	size_t		patlen;
515 
516 	if (revision == NULL)
517 		return true;
518 	patlen = strlen(revision);
519 	if (strnEQ(string, revision, patlen) && isspace((unsigned char)string[patlen]))
520 		return true;
521 	for (s = string; *s; s++) {
522 		if (isspace((unsigned char)*s) && strnEQ(s + 1, revision, patlen) &&
523 		    isspace((unsigned char)s[patlen + 1])) {
524 			return true;
525 		}
526 	}
527 	return false;
528 }
529