xref: /netbsd-src/usr.bin/patch/inp.c (revision b5677b36047b601b9addaaa494a58ceae82c2a6c)
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 1.19 2008/09/19 18:33:34 joerg 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 1.19 2008/09/19 18:33:34 joerg Exp $");
35 
36 #include <sys/types.h>
37 #include <sys/file.h>
38 #include <sys/stat.h>
39 #include <sys/mman.h>
40 
41 #include <ctype.h>
42 #include <libgen.h>
43 #include <limits.h>
44 #include <stddef.h>
45 #include <stdio.h>
46 #include <stdlib.h>
47 #include <string.h>
48 #include <unistd.h>
49 
50 #include "common.h"
51 #include "util.h"
52 #include "pch.h"
53 #include "inp.h"
54 
55 
56 /* Input-file-with-indexable-lines abstract type */
57 
58 static off_t	i_size;		/* size of the input file */
59 static char	*i_womp;	/* plan a buffer for entire file */
60 static char	**i_ptr;	/* pointers to lines in i_womp */
61 static char	empty_line[] = { '\0' };
62 
63 static int	tifd = -1;	/* plan b virtual string array */
64 static char	*tibuf[2];	/* plan b buffers */
65 static LINENUM	tiline[2] = {-1, -1};	/* 1st line in each buffer */
66 static LINENUM	lines_per_buf;	/* how many lines per buffer */
67 static int	tireclen;	/* length of records in tmp file */
68 
69 static bool	rev_in_string(const char *);
70 static bool	reallocate_lines(size_t *);
71 
72 /* returns false if insufficient memory */
73 static bool	plan_a(const char *);
74 
75 static void	plan_b(const char *);
76 
77 /* New patch--prepare to edit another file. */
78 
79 void
80 re_input(void)
81 {
82 	if (using_plan_a) {
83 		i_size = 0;
84 		free(i_ptr);
85 		i_ptr = NULL;
86 		if (i_womp != NULL) {
87 			munmap(i_womp, i_size);
88 			i_womp = NULL;
89 		}
90 	} else {
91 		using_plan_a = true;	/* maybe the next one is smaller */
92 		close(tifd);
93 		tifd = -1;
94 		free(tibuf[0]);
95 		free(tibuf[1]);
96 		tibuf[0] = tibuf[1] = NULL;
97 		tiline[0] = tiline[1] = -1;
98 		tireclen = 0;
99 	}
100 }
101 
102 /* Construct the line index, somehow or other. */
103 
104 void
105 scan_input(const char *filename)
106 {
107 	if (!plan_a(filename))
108 		plan_b(filename);
109 	if (verbose) {
110 		say("Patching file %s using Plan %s...\n", filename,
111 		    (using_plan_a ? "A" : "B"));
112 	}
113 }
114 
115 static bool
116 reallocate_lines(size_t *lines_allocated)
117 {
118 	char	**p;
119 	size_t	new_size;
120 
121 	new_size = *lines_allocated * 3 / 2;
122 	p = realloc(i_ptr, (new_size + 2) * sizeof(char *));
123 	if (p == NULL) {	/* shucks, it was a near thing */
124 		munmap(i_womp, i_size);
125 		i_womp = NULL;
126 		free(i_ptr);
127 		i_ptr = NULL;
128 		*lines_allocated = 0;
129 		return false;
130 	}
131 	*lines_allocated = new_size;
132 	i_ptr = p;
133 	return true;
134 }
135 
136 /* Try keeping everything in memory. */
137 
138 static bool
139 plan_a(const char *filename)
140 {
141 	int		ifd, statfailed;
142 	char		*p, *s, lbuf[MAXLINELEN];
143 	struct stat	filestat;
144 	off_t		i;
145 	ptrdiff_t	sz;
146 	size_t		iline, lines_allocated;
147 
148 #ifdef DEBUGGING
149 	if (debug & 8)
150 		return false;
151 #endif
152 
153 	if (filename == NULL || *filename == '\0')
154 		return false;
155 
156 	statfailed = stat(filename, &filestat);
157 	if (statfailed && ok_to_create_file) {
158 		if (verbose)
159 			say("(Creating file %s...)\n", filename);
160 
161 		/*
162 		 * in check_patch case, we still display `Creating file' even
163 		 * though we're not. The rule is that -C should be as similar
164 		 * to normal patch behavior as possible
165 		 */
166 		if (check_only)
167 			return true;
168 		makedirs(filename, true);
169 		close(creat(filename, 0666));
170 		statfailed = stat(filename, &filestat);
171 	}
172 	if (statfailed && check_only)
173 		fatal("%s not found, -C mode, can't probe further\n", filename);
174 	/* For nonexistent or read-only files, look for RCS or SCCS versions.  */
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 		const char	*cs = NULL, *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 		filebase = basename(tmp_filename1);
189 		filedir = dirname(tmp_filename2);
190 
191 		/* Leave room in lbuf for the diff command.  */
192 		s = lbuf + 20;
193 
194 #define try(f, a1, a2, a3) \
195 	(snprintf(s, sizeof lbuf - 20, f, a1, a2, a3), stat(s, &cstat) == 0)
196 
197 		if (try("%s/RCS/%s%s", filedir, filebase, RCSSUFFIX) ||
198 		    try("%s/RCS/%s%s", filedir, filebase, "") ||
199 		    try("%s/%s%s", filedir, filebase, RCSSUFFIX)) {
200 			snprintf(buf, buf_len, CHECKOUT, filename);
201 			snprintf(lbuf, sizeof lbuf, RCSDIFF, filename);
202 			cs = "RCS";
203 		} else if (try("%s/SCCS/%s%s", filedir, SCCSPREFIX, filebase) ||
204 		    try("%s/%s%s", filedir, SCCSPREFIX, filebase)) {
205 			snprintf(buf, buf_len, GET, s);
206 			snprintf(lbuf, sizeof lbuf, SCCSDIFF, s, filename);
207 			cs = "SCCS";
208 		} else if (statfailed)
209 			fatal("can't find %s\n", filename);
210 
211 		free(tmp_filename1);
212 		free(tmp_filename2);
213 
214 		/*
215 		 * else we can't write to it but it's not under a version
216 		 * control system, so just proceed.
217 		 */
218 		if (cs) {
219 			if (!statfailed) {
220 				if ((filestat.st_mode & 0222) != 0)
221 					/* The owner can write to it.  */
222 					fatal("file %s seems to be locked "
223 					    "by somebody else under %s\n",
224 					    filename, cs);
225 				/*
226 				 * It might be checked out unlocked.  See if
227 				 * it's safe to check out the default version
228 				 * locked.
229 				 */
230 				if (verbose)
231 					say("Comparing file %s to default "
232 					    "%s version...\n",
233 					    filename, cs);
234 				if (system(lbuf))
235 					fatal("can't check out file %s: "
236 					    "differs from default %s version\n",
237 					    filename, cs);
238 			}
239 			if (verbose)
240 				say("Checking out file %s from %s...\n",
241 				    filename, cs);
242 			if (system(buf) || stat(filename, &filestat))
243 				fatal("can't check out file %s from %s\n",
244 				    filename, cs);
245 		}
246 	}
247 	filemode = filestat.st_mode;
248 	if (!S_ISREG(filemode))
249 		fatal("%s is not a normal file--can't patch\n", filename);
250 	i_size = filestat.st_size;
251 	if (out_of_mem) {
252 		set_hunkmax();	/* make sure dynamic arrays are allocated */
253 		out_of_mem = false;
254 		return false;	/* force plan b because plan a bombed */
255 	}
256 	if (i_size > SIZE_MAX) {
257 		say("block too large to mmap\n");
258 		return false;
259 	}
260 	if ((ifd = open(filename, O_RDONLY)) < 0)
261 		pfatal("can't open file %s", filename);
262 
263 	i_womp = mmap(NULL, i_size, PROT_READ, MAP_PRIVATE, ifd, 0);
264 	if (i_womp == MAP_FAILED) {
265 		perror("mmap failed");
266 		i_womp = NULL;
267 		close(ifd);
268 		return false;
269 	}
270 
271 	close(ifd);
272 	if (i_size)
273 		madvise(i_womp, i_size, MADV_SEQUENTIAL);
274 
275 	/* estimate the number of lines */
276 	lines_allocated = i_size / 25;
277 	if (lines_allocated < 100)
278 		lines_allocated = 100;
279 
280 	if (!reallocate_lines(&lines_allocated))
281 		return false;
282 
283 	/* now scan the buffer and build pointer array */
284 	iline = 1;
285 	i_ptr[iline] = i_womp;
286 	/* test for NUL too, to maintain the behavior of the original code */
287 	for (s = i_womp, i = 0; i < i_size && *s != '\0'; s++, i++) {
288 		if (*s == '\n') {
289 			if (iline == lines_allocated) {
290 				if (!reallocate_lines(&lines_allocated))
291 					return false;
292 			}
293 			/* these are NOT NUL terminated */
294 			i_ptr[++iline] = s + 1;
295 		}
296 	}
297 	/* if the last line contains no EOL, append one */
298 	if (i_size > 0 && i_womp[i_size - 1] != '\n') {
299 		last_line_missing_eol = true;
300 		/* fix last line */
301 		sz = s - i_ptr[iline];
302 		p = malloc(sz + 1);
303 		if (p == NULL) {
304 			free(i_ptr);
305 			i_ptr = NULL;
306 			munmap(i_womp, i_size);
307 			i_womp = NULL;
308 			return false;
309 		}
310 
311 		memcpy(p, i_ptr[iline], sz);
312 		p[sz] = '\n';
313 		i_ptr[iline] = p;
314 		/* count the extra line and make it point to some valid mem */
315 		i_ptr[++iline] = empty_line;
316 	} else
317 		last_line_missing_eol = false;
318 
319 	input_lines = iline - 1;
320 
321 	/* now check for revision, if any */
322 
323 	if (revision != NULL) {
324 		if (!rev_in_string(i_womp)) {
325 			if (force) {
326 				if (verbose)
327 					say("Warning: this file doesn't appear "
328 					    "to be the %s version--patching anyway.\n",
329 					    revision);
330 			} else if (batch) {
331 				fatal("this file doesn't appear to be the "
332 				    "%s version--aborting.\n",
333 				    revision);
334 			} else {
335 				ask("This file doesn't appear to be the "
336 				    "%s version--patch anyway? [n] ",
337 				    revision);
338 				if (*buf != 'y')
339 					fatal("aborted\n");
340 			}
341 		} else if (verbose)
342 			say("Good.  This file appears to be the %s version.\n",
343 			    revision);
344 	}
345 	return true;		/* plan a will work */
346 }
347 
348 /* Keep (virtually) nothing in memory. */
349 
350 static void
351 plan_b(const char *filename)
352 {
353 	FILE	*ifp;
354 	size_t	i = 0, j, maxlen = 1;
355 	char	*p;
356 	bool	found_revision = (revision == NULL);
357 
358 	using_plan_a = false;
359 	if ((ifp = fopen(filename, "r")) == NULL)
360 		pfatal("can't open file %s", filename);
361 	unlink(TMPINNAME);
362 	if ((tifd = open(TMPINNAME, O_EXCL | O_CREAT | O_WRONLY, 0666)) < 0)
363 		pfatal("can't open file %s", TMPINNAME);
364 	while (fgets(buf, buf_len, ifp) != NULL) {
365 		if (revision != NULL && !found_revision && rev_in_string(buf))
366 			found_revision = true;
367 		if ((i = strlen(buf)) > maxlen)
368 			maxlen = i;	/* find longest line */
369 	}
370 	last_line_missing_eol = i > 0 && buf[i - 1] != '\n';
371 	if (last_line_missing_eol && maxlen == i)
372 		maxlen++;
373 
374 	if (revision != NULL) {
375 		if (!found_revision) {
376 			if (force) {
377 				if (verbose)
378 					say("Warning: this file doesn't appear "
379 					    "to be the %s version--patching anyway.\n",
380 					    revision);
381 			} else if (batch) {
382 				fatal("this file doesn't appear to be the "
383 				    "%s version--aborting.\n",
384 				    revision);
385 			} else {
386 				ask("This file doesn't appear to be the %s "
387 				    "version--patch anyway? [n] ",
388 				    revision);
389 				if (*buf != 'y')
390 					fatal("aborted\n");
391 			}
392 		} else if (verbose)
393 			say("Good.  This file appears to be the %s version.\n",
394 			    revision);
395 	}
396 	fseek(ifp, 0L, SEEK_SET);	/* rewind file */
397 	lines_per_buf = BUFFERSIZE / maxlen;
398 	tireclen = maxlen;
399 	tibuf[0] = malloc(BUFFERSIZE + 1);
400 	if (tibuf[0] == NULL)
401 		fatal("out of memory\n");
402 	tibuf[1] = malloc(BUFFERSIZE + 1);
403 	if (tibuf[1] == NULL)
404 		fatal("out of memory\n");
405 	for (i = 1;; i++) {
406 		p = tibuf[0] + maxlen * (i % lines_per_buf);
407 		if (i % lines_per_buf == 0)	/* new block */
408 			if (write(tifd, tibuf[0], BUFFERSIZE) < BUFFERSIZE)
409 				pfatal("can't write temp file");
410 		if (fgets(p, maxlen + 1, ifp) == NULL) {
411 			input_lines = i - 1;
412 			if (i % lines_per_buf != 0)
413 				if (write(tifd, tibuf[0], BUFFERSIZE) < BUFFERSIZE)
414 					pfatal("can't write temp file");
415 			break;
416 		}
417 		j = strlen(p);
418 		/* These are '\n' terminated strings, so no need to add a NUL */
419 		if (j == 0 || p[j - 1] != '\n')
420 			p[j] = '\n';
421 	}
422 	fclose(ifp);
423 	close(tifd);
424 	if ((tifd = open(TMPINNAME, O_RDONLY)) < 0)
425 		pfatal("can't reopen file %s", TMPINNAME);
426 }
427 
428 /*
429  * Fetch a line from the input file, \n terminated, not necessarily \0.
430  */
431 char *
432 ifetch(LINENUM line, int whichbuf)
433 {
434 	if (line < 1 || line > input_lines) {
435 		if (warn_on_invalid_line) {
436 			say("No such line %ld in input file, ignoring\n", line);
437 			warn_on_invalid_line = false;
438 		}
439 		return NULL;
440 	}
441 	if (using_plan_a)
442 		return i_ptr[line];
443 	else {
444 		LINENUM	offline = line % lines_per_buf;
445 		LINENUM	baseline = line - offline;
446 
447 		if (tiline[0] == baseline)
448 			whichbuf = 0;
449 		else if (tiline[1] == baseline)
450 			whichbuf = 1;
451 		else {
452 			tiline[whichbuf] = baseline;
453 
454 			if (lseek(tifd, (off_t) (baseline / lines_per_buf *
455 			    BUFFERSIZE), SEEK_SET) < 0)
456 				pfatal("cannot seek in the temporary input file");
457 
458 			if (read(tifd, tibuf[whichbuf], BUFFERSIZE) < 0)
459 				pfatal("error reading tmp file %s", TMPINNAME);
460 		}
461 		return tibuf[whichbuf] + (tireclen * offline);
462 	}
463 }
464 
465 /*
466  * True if the string argument contains the revision number we want.
467  */
468 static bool
469 rev_in_string(const char *string)
470 {
471 	const char	*s;
472 	size_t		patlen;
473 
474 	if (revision == NULL)
475 		return true;
476 	patlen = strlen(revision);
477 	if (strnEQ(string, revision, patlen) && isspace((unsigned char)string[patlen]))
478 		return true;
479 	for (s = string; *s; s++) {
480 		if (isspace((unsigned char)*s) && strnEQ(s + 1, revision, patlen) &&
481 		    isspace((unsigned char)s[patlen + 1])) {
482 			return true;
483 		}
484 	}
485 	return false;
486 }
487