xref: /freebsd/usr.bin/grep/util.c (revision bf2fa8d9d11c9f2ceff09bacc406876fa37096be)
1 /*	$NetBSD: util.c,v 1.9 2011/02/27 17:33:37 joerg Exp $	*/
2 /*	$FreeBSD$	*/
3 /*	$OpenBSD: util.c,v 1.39 2010/07/02 22:18:03 tedu Exp $	*/
4 
5 /*-
6  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
7  *
8  * Copyright (c) 1999 James Howard and Dag-Erling Coïdan Smørgrav
9  * Copyright (C) 2008-2010 Gabor Kovesdan <gabor@FreeBSD.org>
10  * Copyright (C) 2017 Kyle Evans <kevans@FreeBSD.org>
11  * All rights reserved.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
37 
38 #include <sys/stat.h>
39 #include <sys/types.h>
40 
41 #include <ctype.h>
42 #include <err.h>
43 #include <errno.h>
44 #include <fnmatch.h>
45 #include <fts.h>
46 #include <libgen.h>
47 #include <stdbool.h>
48 #include <stdio.h>
49 #include <stdlib.h>
50 #include <string.h>
51 #include <unistd.h>
52 #include <wchar.h>
53 #include <wctype.h>
54 
55 #include "grep.h"
56 
57 static bool	 first_match = true;
58 
59 /*
60  * Match printing context
61  */
62 struct mprintc {
63 	long long	tail;		/* Number of trailing lines to record */
64 	int		last_outed;	/* Number of lines since last output */
65 	bool		doctx;		/* Printing context? */
66 	bool		printmatch;	/* Printing matches? */
67 	bool		same_file;	/* Same file as previously printed? */
68 };
69 
70 static void procmatch_match(struct mprintc *mc, struct parsec *pc);
71 static void procmatch_nomatch(struct mprintc *mc, struct parsec *pc);
72 static bool procmatches(struct mprintc *mc, struct parsec *pc, bool matched);
73 #ifdef WITH_INTERNAL_NOSPEC
74 static int litexec(const struct pat *pat, const char *string,
75     size_t nmatch, regmatch_t pmatch[]);
76 #endif
77 static bool procline(struct parsec *pc);
78 static void printline(struct parsec *pc, int sep);
79 static void printline_metadata(struct str *line, int sep);
80 
81 bool
82 file_matching(const char *fname)
83 {
84 	char *fname_base, *fname_buf;
85 	bool ret;
86 
87 	ret = finclude ? false : true;
88 	fname_buf = strdup(fname);
89 	if (fname_buf == NULL)
90 		err(2, "strdup");
91 	fname_base = basename(fname_buf);
92 
93 	for (unsigned int i = 0; i < fpatterns; ++i) {
94 		if (fnmatch(fpattern[i].pat, fname, 0) == 0 ||
95 		    fnmatch(fpattern[i].pat, fname_base, 0) == 0)
96 			/*
97 			 * The last pattern matched wins exclusion/inclusion
98 			 * rights, so we can't reasonably bail out early here.
99 			 */
100 			ret = (fpattern[i].mode != EXCL_PAT);
101 	}
102 	free(fname_buf);
103 	return (ret);
104 }
105 
106 static inline bool
107 dir_matching(const char *dname)
108 {
109 	bool ret;
110 
111 	ret = dinclude ? false : true;
112 
113 	for (unsigned int i = 0; i < dpatterns; ++i) {
114 		if (dname != NULL && fnmatch(dpattern[i].pat, dname, 0) == 0)
115 			/*
116 			 * The last pattern matched wins exclusion/inclusion
117 			 * rights, so we can't reasonably bail out early here.
118 			 */
119 			ret = (dpattern[i].mode != EXCL_PAT);
120 	}
121 	return (ret);
122 }
123 
124 /*
125  * Processes a directory when a recursive search is performed with
126  * the -R option.  Each appropriate file is passed to procfile().
127  */
128 bool
129 grep_tree(char **argv)
130 {
131 	FTS *fts;
132 	FTSENT *p;
133 	int fts_flags;
134 	bool matched, ok;
135 	const char *wd[] = { ".", NULL };
136 
137 	matched = false;
138 
139 	/* This switch effectively initializes 'fts_flags' */
140 	switch(linkbehave) {
141 	case LINK_EXPLICIT:
142 		fts_flags = FTS_COMFOLLOW;
143 		break;
144 	case LINK_SKIP:
145 		fts_flags = FTS_PHYSICAL;
146 		break;
147 	default:
148 		fts_flags = FTS_LOGICAL;
149 	}
150 
151 	fts_flags |= FTS_NOSTAT | FTS_NOCHDIR;
152 
153 	fts = fts_open((argv[0] == NULL) ?
154 	    __DECONST(char * const *, wd) : argv, fts_flags, NULL);
155 	if (fts == NULL)
156 		err(2, "fts_open");
157 	while (errno = 0, (p = fts_read(fts)) != NULL) {
158 		switch (p->fts_info) {
159 		case FTS_DNR:
160 			/* FALLTHROUGH */
161 		case FTS_ERR:
162 			file_err = true;
163 			if(!sflag)
164 				warnx("%s: %s", p->fts_path, strerror(p->fts_errno));
165 			break;
166 		case FTS_D:
167 			/* FALLTHROUGH */
168 		case FTS_DP:
169 			if (dexclude || dinclude)
170 				if (!dir_matching(p->fts_name) ||
171 				    !dir_matching(p->fts_path))
172 					fts_set(fts, p, FTS_SKIP);
173 			break;
174 		case FTS_DC:
175 			/* Print a warning for recursive directory loop */
176 			warnx("warning: %s: recursive directory loop",
177 			    p->fts_path);
178 			break;
179 		default:
180 			/* Check for file exclusion/inclusion */
181 			ok = true;
182 			if (fexclude || finclude)
183 				ok &= file_matching(p->fts_path);
184 
185 			if (ok && procfile(p->fts_path))
186 				matched = true;
187 			break;
188 		}
189 	}
190 	if (errno != 0)
191 		err(2, "fts_read");
192 
193 	fts_close(fts);
194 	return (matched);
195 }
196 
197 static void
198 procmatch_match(struct mprintc *mc, struct parsec *pc)
199 {
200 
201 	if (mc->doctx) {
202 		if (!first_match && (!mc->same_file || mc->last_outed > 0))
203 			printf("--\n");
204 		if (Bflag > 0)
205 			printqueue();
206 		mc->tail = Aflag;
207 	}
208 
209 	/* Print the matching line, but only if not quiet/binary */
210 	if (mc->printmatch) {
211 		printline(pc, ':');
212 		while (pc->matchidx >= MAX_MATCHES) {
213 			/* Reset matchidx and try again */
214 			pc->matchidx = 0;
215 			if (procline(pc) == !vflag)
216 				printline(pc, ':');
217 			else
218 				break;
219 		}
220 		first_match = false;
221 		mc->same_file = true;
222 		mc->last_outed = 0;
223 	}
224 }
225 
226 static void
227 procmatch_nomatch(struct mprintc *mc, struct parsec *pc)
228 {
229 
230 	/* Deal with any -A context as needed */
231 	if (mc->tail > 0) {
232 		grep_printline(&pc->ln, '-');
233 		mc->tail--;
234 		if (Bflag > 0)
235 			clearqueue();
236 	} else if (Bflag == 0 || (Bflag > 0 && enqueue(&pc->ln)))
237 		/*
238 		 * Enqueue non-matching lines for -B context. If we're not
239 		 * actually doing -B context or if the enqueue resulted in a
240 		 * line being rotated out, then go ahead and increment
241 		 * last_outed to signify a gap between context/match.
242 		 */
243 		++mc->last_outed;
244 }
245 
246 /*
247  * Process any matches in the current parsing context, return a boolean
248  * indicating whether we should halt any further processing or not. 'true' to
249  * continue processing, 'false' to halt.
250  */
251 static bool
252 procmatches(struct mprintc *mc, struct parsec *pc, bool matched)
253 {
254 
255 	if (mflag && mcount <= 0) {
256 		/*
257 		 * We already hit our match count, but we need to keep dumping
258 		 * lines until we've lost our tail.
259 		 */
260 		grep_printline(&pc->ln, '-');
261 		mc->tail--;
262 		return (mc->tail != 0);
263 	}
264 
265 	/*
266 	 * XXX TODO: This should loop over pc->matches and handle things on a
267 	 * line-by-line basis, setting up a `struct str` as needed.
268 	 */
269 	/* Deal with any -B context or context separators */
270 	if (matched) {
271 		procmatch_match(mc, pc);
272 
273 		/* Count the matches if we have a match limit */
274 		if (mflag) {
275 			/* XXX TODO: Decrement by number of matched lines */
276 			mcount -= 1;
277 			if (mcount <= 0)
278 				return (mc->tail != 0);
279 		}
280 	} else if (mc->doctx)
281 		procmatch_nomatch(mc, pc);
282 
283 	return (true);
284 }
285 
286 /*
287  * Opens a file and processes it.  Each file is processed line-by-line
288  * passing the lines to procline().
289  */
290 bool
291 procfile(const char *fn)
292 {
293 	struct parsec pc;
294 	struct mprintc mc;
295 	struct file *f;
296 	struct stat sb;
297 	mode_t s;
298 	int lines;
299 	bool line_matched;
300 
301 	if (strcmp(fn, "-") == 0) {
302 		fn = label != NULL ? label : errstr[1];
303 		f = grep_open(NULL);
304 	} else {
305 		if (stat(fn, &sb) == 0) {
306 			/* Check if we need to process the file */
307 			s = sb.st_mode & S_IFMT;
308 			if (dirbehave == DIR_SKIP && s == S_IFDIR)
309 				return (false);
310 			if (devbehave == DEV_SKIP && (s == S_IFIFO ||
311 			    s == S_IFCHR || s == S_IFBLK || s == S_IFSOCK))
312 				return (false);
313 		}
314 		f = grep_open(fn);
315 	}
316 	if (f == NULL) {
317 		file_err = true;
318 		if (!sflag)
319 			warn("%s", fn);
320 		return (false);
321 	}
322 
323 	pc.ln.file = grep_strdup(fn);
324 	pc.ln.line_no = 0;
325 	pc.ln.len = 0;
326 	pc.ln.boff = 0;
327 	pc.ln.off = -1;
328 	pc.binary = f->binary;
329 	pc.cntlines = false;
330 	memset(&mc, 0, sizeof(mc));
331 	mc.printmatch = true;
332 	if ((pc.binary && binbehave == BINFILE_BIN) || cflag || qflag ||
333 	    lflag || Lflag)
334 		mc.printmatch = false;
335 	if (mc.printmatch && (Aflag != 0 || Bflag != 0))
336 		mc.doctx = true;
337 	if (mc.printmatch && (Aflag != 0 || Bflag != 0 || mflag || nflag))
338 		pc.cntlines = true;
339 	mcount = mlimit;
340 
341 	for (lines = 0; lines == 0 || !(lflag || qflag); ) {
342 		/*
343 		 * XXX TODO: We need to revisit this in a chunking world. We're
344 		 * not going to be doing per-line statistics because of the
345 		 * overhead involved. procmatches can figure that stuff out as
346 		 * needed. */
347 		/* Reset per-line statistics */
348 		pc.printed = 0;
349 		pc.matchidx = 0;
350 		pc.lnstart = 0;
351 		pc.ln.boff = 0;
352 		pc.ln.off += pc.ln.len + 1;
353 		/* XXX TODO: Grab a chunk */
354 		if ((pc.ln.dat = grep_fgetln(f, &pc)) == NULL ||
355 		    pc.ln.len == 0)
356 			break;
357 
358 		if (pc.ln.len > 0 && pc.ln.dat[pc.ln.len - 1] == fileeol)
359 			--pc.ln.len;
360 		pc.ln.line_no++;
361 
362 		/* Return if we need to skip a binary file */
363 		if (pc.binary && binbehave == BINFILE_SKIP) {
364 			grep_close(f);
365 			free(pc.ln.file);
366 			free(f);
367 			return (0);
368 		}
369 
370 		if (mflag && mcount <= 0) {
371 			/*
372 			 * Short-circuit, already hit match count and now we're
373 			 * just picking up any remaining pieces.
374 			 */
375 			if (!procmatches(&mc, &pc, false))
376 				break;
377 			continue;
378 		}
379 		line_matched = procline(&pc) == !vflag;
380 		if (line_matched)
381 			++lines;
382 
383 		/* Halt processing if we hit our match limit */
384 		if (!procmatches(&mc, &pc, line_matched))
385 			break;
386 	}
387 	if (Bflag > 0)
388 		clearqueue();
389 	grep_close(f);
390 
391 	if (cflag && !qflag) {
392 		if (!hflag)
393 			printf("%s:", pc.ln.file);
394 		printf("%u\n", lines);
395 	}
396 	if (lflag && !qflag && lines != 0)
397 		printf("%s%c", fn, nullflag ? 0 : '\n');
398 	if (Lflag && !qflag && lines == 0)
399 		printf("%s%c", fn, nullflag ? 0 : '\n');
400 	if (lines != 0 && !cflag && !lflag && !Lflag &&
401 	    binbehave == BINFILE_BIN && f->binary && !qflag)
402 		printf(errstr[7], fn);
403 
404 	free(pc.ln.file);
405 	free(f);
406 	return (lines != 0);
407 }
408 
409 #ifdef WITH_INTERNAL_NOSPEC
410 /*
411  * Internal implementation of literal string search within a string, modeled
412  * after regexec(3), for use when the regex(3) implementation doesn't offer
413  * either REG_NOSPEC or REG_LITERAL. This does not apply in the default FreeBSD
414  * config, but in other scenarios such as building against libgnuregex or on
415  * some non-FreeBSD OSes.
416  */
417 static int
418 litexec(const struct pat *pat, const char *string, size_t nmatch,
419     regmatch_t pmatch[])
420 {
421 	char *(*strstr_fn)(const char *, const char *);
422 	char *sub, *subject;
423 	const char *search;
424 	size_t idx, n, ofs, stringlen;
425 
426 	if (cflags & REG_ICASE)
427 		strstr_fn = strcasestr;
428 	else
429 		strstr_fn = strstr;
430 	idx = 0;
431 	ofs = pmatch[0].rm_so;
432 	stringlen = pmatch[0].rm_eo;
433 	if (ofs >= stringlen)
434 		return (REG_NOMATCH);
435 	subject = strndup(string, stringlen);
436 	if (subject == NULL)
437 		return (REG_ESPACE);
438 	for (n = 0; ofs < stringlen;) {
439 		search = (subject + ofs);
440 		if ((unsigned long)pat->len > strlen(search))
441 			break;
442 		sub = strstr_fn(search, pat->pat);
443 		/*
444 		 * Ignoring the empty string possibility due to context: grep optimizes
445 		 * for empty patterns and will never reach this point.
446 		 */
447 		if (sub == NULL)
448 			break;
449 		++n;
450 		/* Fill in pmatch if necessary */
451 		if (nmatch > 0) {
452 			pmatch[idx].rm_so = ofs + (sub - search);
453 			pmatch[idx].rm_eo = pmatch[idx].rm_so + pat->len;
454 			if (++idx == nmatch)
455 				break;
456 			ofs = pmatch[idx].rm_so + 1;
457 		} else
458 			/* We only needed to know if we match or not */
459 			break;
460 	}
461 	free(subject);
462 	if (n > 0 && nmatch > 0)
463 		for (n = idx; n < nmatch; ++n)
464 			pmatch[n].rm_so = pmatch[n].rm_eo = -1;
465 
466 	return (n > 0 ? 0 : REG_NOMATCH);
467 }
468 #endif /* WITH_INTERNAL_NOSPEC */
469 
470 #define iswword(x)	(iswalnum((x)) || (x) == L'_')
471 
472 /*
473  * Processes a line comparing it with the specified patterns.  Each pattern
474  * is looped to be compared along with the full string, saving each and every
475  * match, which is necessary to colorize the output and to count the
476  * matches.  The matching lines are passed to printline() to display the
477  * appropriate output.
478  */
479 static bool
480 procline(struct parsec *pc)
481 {
482 	regmatch_t pmatch, lastmatch, chkmatch;
483 	wchar_t wbegin, wend;
484 	size_t st, nst;
485 	unsigned int i;
486 	int r = 0, leflags = eflags;
487 	size_t startm = 0, matchidx;
488 	unsigned int retry;
489 	bool lastmatched, matched;
490 
491 	matchidx = pc->matchidx;
492 
493 	/* Null pattern shortcuts. */
494 	if (matchall) {
495 		if (xflag && pc->ln.len == 0) {
496 			/* Matches empty lines (-x). */
497 			return (true);
498 		} else if (!wflag && !xflag) {
499 			/* Matches every line (no -w or -x). */
500 			return (true);
501 		}
502 
503 		/*
504 		 * If we only have the NULL pattern, whether we match or not
505 		 * depends on if we got here with -w or -x.  If either is set,
506 		 * the answer is no.  If we have other patterns, we'll defer
507 		 * to them.
508 		 */
509 		if (patterns == 0) {
510 			return (!(wflag || xflag));
511 		}
512 	} else if (patterns == 0) {
513 		/* Pattern file with no patterns. */
514 		return (false);
515 	}
516 
517 	matched = false;
518 	st = pc->lnstart;
519 	nst = 0;
520 	/* Initialize to avoid a false positive warning from GCC. */
521 	lastmatch.rm_so = lastmatch.rm_eo = 0;
522 
523 	/* Loop to process the whole line */
524 	while (st <= pc->ln.len) {
525 		lastmatched = false;
526 		startm = matchidx;
527 		retry = 0;
528 		if (st > 0 && pc->ln.dat[st - 1] != fileeol)
529 			leflags |= REG_NOTBOL;
530 		/* Loop to compare with all the patterns */
531 		for (i = 0; i < patterns; i++) {
532 			pmatch.rm_so = st;
533 			pmatch.rm_eo = pc->ln.len;
534 #ifdef WITH_INTERNAL_NOSPEC
535 			if (grepbehave == GREP_FIXED)
536 				r = litexec(&pattern[i], pc->ln.dat, 1, &pmatch);
537 			else
538 #endif
539 			r = regexec(&r_pattern[i], pc->ln.dat, 1, &pmatch,
540 			    leflags);
541 			if (r != 0)
542 				continue;
543 			/* Check for full match */
544 			if (xflag && (pmatch.rm_so != 0 ||
545 			    (size_t)pmatch.rm_eo != pc->ln.len))
546 				continue;
547 			/* Check for whole word match */
548 			if (wflag) {
549 				wbegin = wend = L' ';
550 				if (pmatch.rm_so != 0 &&
551 				    sscanf(&pc->ln.dat[pmatch.rm_so - 1],
552 				    "%lc", &wbegin) != 1)
553 					r = REG_NOMATCH;
554 				else if ((size_t)pmatch.rm_eo !=
555 				    pc->ln.len &&
556 				    sscanf(&pc->ln.dat[pmatch.rm_eo],
557 				    "%lc", &wend) != 1)
558 					r = REG_NOMATCH;
559 				else if (iswword(wbegin) ||
560 				    iswword(wend))
561 					r = REG_NOMATCH;
562 				/*
563 				 * If we're doing whole word matching and we
564 				 * matched once, then we should try the pattern
565 				 * again after advancing just past the start of
566 				 * the earliest match. This allows the pattern
567 				 * to  match later on in the line and possibly
568 				 * still match a whole word.
569 				 */
570 				if (r == REG_NOMATCH &&
571 				    (retry == pc->lnstart ||
572 				    (unsigned int)pmatch.rm_so + 1 < retry))
573 					retry = pmatch.rm_so + 1;
574 				if (r == REG_NOMATCH)
575 					continue;
576 			}
577 			lastmatched = true;
578 			lastmatch = pmatch;
579 
580 			if (matchidx == 0)
581 				matched = true;
582 
583 			/*
584 			 * Replace previous match if the new one is earlier
585 			 * and/or longer. This will lead to some amount of
586 			 * extra work if -o/--color are specified, but it's
587 			 * worth it from a correctness point of view.
588 			 */
589 			if (matchidx > startm) {
590 				chkmatch = pc->matches[matchidx - 1];
591 				if (pmatch.rm_so < chkmatch.rm_so ||
592 				    (pmatch.rm_so == chkmatch.rm_so &&
593 				    (pmatch.rm_eo - pmatch.rm_so) >
594 				    (chkmatch.rm_eo - chkmatch.rm_so))) {
595 					pc->matches[matchidx - 1] = pmatch;
596 					nst = pmatch.rm_eo;
597 				}
598 			} else {
599 				/* Advance as normal if not */
600 				pc->matches[matchidx++] = pmatch;
601 				nst = pmatch.rm_eo;
602 			}
603 			/* avoid excessive matching - skip further patterns */
604 			if ((color == NULL && !oflag) || qflag || lflag ||
605 			    matchidx >= MAX_MATCHES) {
606 				pc->lnstart = nst;
607 				lastmatched = false;
608 				break;
609 			}
610 		}
611 
612 		/*
613 		 * Advance to just past the start of the earliest match, try
614 		 * again just in case we still have a chance to match later in
615 		 * the string.
616 		 */
617 		if (!lastmatched && retry > pc->lnstart) {
618 			st = retry;
619 			continue;
620 		}
621 
622 		/* XXX TODO: We will need to keep going, since we're chunky */
623 		/* One pass if we are not recording matches */
624 		if (!wflag && ((color == NULL && !oflag) || qflag || lflag || Lflag))
625 			break;
626 
627 		/* If we didn't have any matches or REG_NOSUB set */
628 		if (!lastmatched || (cflags & REG_NOSUB))
629 			nst = pc->ln.len;
630 
631 		if (!lastmatched)
632 			/* No matches */
633 			break;
634 		else if (st == nst && lastmatch.rm_so == lastmatch.rm_eo)
635 			/* Zero-length match -- advance one more so we don't get stuck */
636 			nst++;
637 
638 		/* Advance st based on previous matches */
639 		st = nst;
640 		pc->lnstart = st;
641 	}
642 
643 	/* Reflect the new matchidx in the context */
644 	pc->matchidx = matchidx;
645 	return matched;
646 }
647 
648 /*
649  * Safe malloc() for internal use.
650  */
651 void *
652 grep_malloc(size_t size)
653 {
654 	void *ptr;
655 
656 	if ((ptr = malloc(size)) == NULL)
657 		err(2, "malloc");
658 	return (ptr);
659 }
660 
661 /*
662  * Safe calloc() for internal use.
663  */
664 void *
665 grep_calloc(size_t nmemb, size_t size)
666 {
667 	void *ptr;
668 
669 	if ((ptr = calloc(nmemb, size)) == NULL)
670 		err(2, "calloc");
671 	return (ptr);
672 }
673 
674 /*
675  * Safe realloc() for internal use.
676  */
677 void *
678 grep_realloc(void *ptr, size_t size)
679 {
680 
681 	if ((ptr = realloc(ptr, size)) == NULL)
682 		err(2, "realloc");
683 	return (ptr);
684 }
685 
686 /*
687  * Safe strdup() for internal use.
688  */
689 char *
690 grep_strdup(const char *str)
691 {
692 	char *ret;
693 
694 	if ((ret = strdup(str)) == NULL)
695 		err(2, "strdup");
696 	return (ret);
697 }
698 
699 /*
700  * Print an entire line as-is, there are no inline matches to consider. This is
701  * used for printing context.
702  */
703 void grep_printline(struct str *line, int sep) {
704 	printline_metadata(line, sep);
705 	fwrite(line->dat, line->len, 1, stdout);
706 	putchar(fileeol);
707 }
708 
709 static void
710 printline_metadata(struct str *line, int sep)
711 {
712 	bool printsep;
713 
714 	printsep = false;
715 	if (!hflag) {
716 		if (!nullflag) {
717 			fputs(line->file, stdout);
718 			printsep = true;
719 		} else {
720 			printf("%s", line->file);
721 			putchar(0);
722 		}
723 	}
724 	if (nflag) {
725 		if (printsep)
726 			putchar(sep);
727 		printf("%d", line->line_no);
728 		printsep = true;
729 	}
730 	if (bflag) {
731 		if (printsep)
732 			putchar(sep);
733 		printf("%lld", (long long)(line->off + line->boff));
734 		printsep = true;
735 	}
736 	if (printsep)
737 		putchar(sep);
738 }
739 
740 /*
741  * Prints a matching line according to the command line options.
742  */
743 static void
744 printline(struct parsec *pc, int sep)
745 {
746 	size_t a = 0;
747 	size_t i, matchidx;
748 	regmatch_t match;
749 
750 	/* If matchall, everything matches but don't actually print for -o */
751 	if (oflag && matchall)
752 		return;
753 
754 	matchidx = pc->matchidx;
755 
756 	/* --color and -o */
757 	if ((oflag || color) && matchidx > 0) {
758 		/* Only print metadata once per line if --color */
759 		if (!oflag && pc->printed == 0)
760 			printline_metadata(&pc->ln, sep);
761 		for (i = 0; i < matchidx; i++) {
762 			match = pc->matches[i];
763 			/* Don't output zero length matches */
764 			if (match.rm_so == match.rm_eo)
765 				continue;
766 			/*
767 			 * Metadata is printed on a per-line basis, so every
768 			 * match gets file metadata with the -o flag.
769 			 */
770 			if (oflag) {
771 				pc->ln.boff = match.rm_so;
772 				printline_metadata(&pc->ln, sep);
773 			} else
774 				fwrite(pc->ln.dat + a, match.rm_so - a, 1,
775 				    stdout);
776 			if (color)
777 				fprintf(stdout, "\33[%sm\33[K", color);
778 			fwrite(pc->ln.dat + match.rm_so,
779 			    match.rm_eo - match.rm_so, 1, stdout);
780 			if (color)
781 				fprintf(stdout, "\33[m\33[K");
782 			a = match.rm_eo;
783 			if (oflag)
784 				putchar('\n');
785 		}
786 		if (!oflag) {
787 			if (pc->ln.len - a > 0)
788 				fwrite(pc->ln.dat + a, pc->ln.len - a, 1,
789 				    stdout);
790 			putchar('\n');
791 		}
792 	} else
793 		grep_printline(&pc->ln, sep);
794 	pc->printed++;
795 }
796