xref: /illumos-gate/usr/src/cmd/xargs/xargs.c (revision 7fdd916c)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2012 DEY Storage Systems, Inc.  All rights reserved.
23  *
24  * Portions of this file developed by DEY Storage Systems, Inc. are licensed
25  * under the terms of the Common Development and Distribution License (CDDL)
26  * version 1.0 only.  The use of subsequent versions of the License are
27  * is specifically prohibited unless those terms are not in conflict with
28  * version 1.0 of the License.  You can find this license on-line at
29  * http://www.illumos.org/license/CDDL
30  */
31 /*
32  * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
33  * Use is subject to license terms.
34  */
35 
36 /*	Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T	*/
37 /*	  All Rights Reserved  	*/
38 
39 
40 #include <stdio.h>
41 #include <sys/types.h>
42 #include <sys/wait.h>
43 #include <unistd.h>
44 #include <fcntl.h>
45 #include <string.h>
46 #include <stdarg.h>
47 #include <stdlib.h>
48 #include <limits.h>
49 #include <wchar.h>
50 #include <locale.h>
51 #include <langinfo.h>
52 #include <stropts.h>
53 #include <poll.h>
54 #include <errno.h>
55 #include <stdarg.h>
56 #include "getresponse.h"
57 
58 #define	HEAD	0
59 #define	TAIL	1
60 #define	FALSE 0
61 #define	TRUE 1
62 #define	MAXSBUF 255
63 #define	MAXIBUF 512
64 #define	MAXINSERTS 5
65 #define	BUFSIZE LINE_MAX
66 #define	MAXARGS 255
67 #define	INSPAT_STR	"{}"	/* default replstr string for -[Ii]	*/
68 #define	FORK_RETRY	5
69 
70 #define	QBUF_STARTLEN 255  /* start size of growable string buffer */
71 #define	QBUF_INC 100	   /* how much to grow a growable string by */
72 
73 /* We use these macros to help make formatting look "consistent" */
74 #define	EMSG(s)		ermsg(gettext(s "\n"))
75 #define	EMSG2(s, a)	ermsg(gettext(s "\n"), a)
76 #define	PERR(s)		perror(gettext("xargs: " s))
77 
78 /* Some common error messages */
79 
80 #define	LIST2LONG	"Argument list too long"
81 #define	ARG2LONG	"A single argument was greater than %d bytes"
82 #define	MALLOCFAIL	"Memory allocation failure"
83 #define	CORRUPTFILE	"Corrupt input file"
84 #define	WAITFAIL	"Wait failure"
85 #define	CHILDSIG	"Child killed with signal %d"
86 #define	CHILDFAIL	"Command could not continue processing data"
87 #define	FORKFAIL	"Could not fork child"
88 #define	EXECFAIL	"Could not exec command"
89 #define	MISSQUOTE	"Missing quote"
90 #define	BADESCAPE	"Incomplete escape"
91 #define	IBUFOVERFLOW	"Insert buffer overflow"
92 
93 #define	_(x)	gettext(x)
94 
95 static wctype_t	blank;
96 static char	*arglist[MAXARGS+1];
97 static char	argbuf[BUFSIZE * 2 + 1];
98 static char	lastarg[BUFSIZE + 1];
99 static char	**ARGV = arglist;
100 static char	*LEOF = "_";
101 static char	*INSPAT = INSPAT_STR;
102 static char	ins_buf[MAXIBUF];
103 static char	*p_ibuf;
104 
105 static struct inserts {
106 	char	**p_ARGV;	/* where to put newarg ptr in arg list */
107 	char	*p_skel;	/* ptr to arg template */
108 } saveargv[MAXINSERTS];
109 
110 static int	PROMPT = -1;
111 static int	BUFLIM = BUFSIZE;
112 static int	N_ARGS = 0;
113 static int	N_args = 0;
114 static int	N_lines = 0;
115 static int	DASHX = FALSE;
116 static int	MORE = TRUE;
117 static int	PER_LINE = FALSE;
118 static int	ERR = FALSE;
119 static int	OK = TRUE;
120 static int	LEGAL = FALSE;
121 static int	TRACE = FALSE;
122 static int	INSERT = FALSE;
123 static int	ZERO = FALSE;
124 static int	linesize = 0;
125 static int	ibufsize = 0;
126 static int	exitstat = 0;	/* our exit status			*/
127 static int	mac;		/* modified argc, after parsing		*/
128 static char	**mav;		/* modified argv, after parsing		*/
129 static int	n_inserts;	/* # of insertions.			*/
130 
131 /* our usage message:							*/
132 #define	USAGEMSG "Usage: xargs: [-t] [-p] [-e[eofstr]] [-E eofstr] "\
133 	"[-I replstr] [-i[replstr]] [-L #] [-l[#]] [-n # [-x]] [-s size] "\
134 	"[cmd [args ...]]\n"
135 
136 static int	echoargs();
137 static wint_t	getwchr(char *, size_t *);
138 static int	lcall(char *sub, char **subargs);
139 static void	addibuf(struct inserts *p);
140 static void	ermsg(char *messages, ...);
141 static char	*addarg(char *arg);
142 static void	store_str(char **, char *, size_t);
143 static char	*getarg(char *);
144 static char	*insert(char *pattern, char *subst);
145 static void	usage();
146 static void	parseargs();
147 
148 int
149 main(int argc, char **argv)
150 {
151 	int	j;
152 	struct inserts *psave;
153 	int c;
154 	int	initsize;
155 	char	*cmdname, **initlist;
156 	char	*arg;
157 	char	*next;
158 
159 	/* initialization */
160 	blank = wctype("blank");
161 	n_inserts = 0;
162 	psave = saveargv;
163 	(void) setlocale(LC_ALL, "");
164 #if !defined(TEXT_DOMAIN)	/* Should be defined by cc -D 		*/
165 #define	TEXT_DOMAIN "SYS_TEST"	/* Use this only if it weren't 		*/
166 #endif
167 	(void) textdomain(TEXT_DOMAIN);
168 	if (init_yes() < 0) {
169 		ermsg(_(ERR_MSG_INIT_YES), strerror(errno));
170 		exit(1);
171 	}
172 
173 	parseargs(argc, argv);
174 
175 	/* handling all of xargs arguments:				*/
176 	while ((c = getopt(mac, mav, "0tpe:E:I:i:L:l:n:s:x")) != EOF) {
177 		switch (c) {
178 		case '0':
179 			ZERO = TRUE;
180 			break;
181 
182 		case 't':	/* -t: turn trace mode on		*/
183 			TRACE = TRUE;
184 			break;
185 
186 		case 'p':	/* -p: turn on prompt mode.		*/
187 			if ((PROMPT = open("/dev/tty", O_RDONLY)) == -1) {
188 				PERR("can't read from tty for -p");
189 			} else {
190 				TRACE = TRUE;
191 			}
192 			break;
193 
194 		case 'e':
195 			/*
196 			 * -e[eofstr]: set/disable end-of-file.
197 			 * N.B. that an argument *isn't* required here; but
198 			 * parseargs forced an argument if not was given.  The
199 			 * forced argument is the default...
200 			 */
201 			LEOF = optarg; /* can be empty */
202 			break;
203 
204 		case 'E':
205 			/*
206 			 * -E eofstr: change end-of-file string.
207 			 * eofstr *is* required here, but can be empty:
208 			 */
209 			LEOF = optarg;
210 			break;
211 
212 		case 'I':
213 			/* -I replstr: Insert mode. replstr *is* required. */
214 			INSERT = PER_LINE = LEGAL = TRUE;
215 			N_ARGS = 0;
216 			INSPAT = optarg;
217 			if (*optarg == '\0') {
218 				ermsg(_("Option requires an argument: -%c\n"),
219 				    c);
220 			}
221 			break;
222 
223 		case 'i':
224 			/*
225 			 * -i [replstr]: insert mode, with *optional* replstr.
226 			 * N.B. that an argument *isn't* required here; if
227 			 * it's not given, then the string INSPAT_STR will
228 			 * be assumed.
229 			 *
230 			 * Since getopts(3C) doesn't handle the case of an
231 			 * optional variable argument at all, we have to
232 			 * parse this by hand:
233 			 */
234 
235 			INSERT = PER_LINE = LEGAL = TRUE;
236 			N_ARGS = 0;
237 			if ((optarg != NULL) && (*optarg != '\0')) {
238 				INSPAT = optarg;
239 			} else {
240 				/*
241 				 * here, there is no next argument. so
242 				 * we reset INSPAT to the INSPAT_STR.
243 				 * we *have* to do this, as -i/I may have
244 				 * been given previously, and XCU4 requires
245 				 * that only "the last one specified takes
246 				 * effect".
247 				 */
248 				INSPAT = INSPAT_STR;
249 			}
250 			break;
251 
252 		case 'L':
253 			/*
254 			 * -L number: # of times cmd is executed
255 			 * number *is* required here:
256 			 */
257 			PER_LINE = TRUE;
258 			N_ARGS = 0;
259 			INSERT = FALSE;
260 			if ((PER_LINE = atoi(optarg)) <= 0) {
261 				ermsg(_("#lines must be positive int: %s\n"),
262 				    optarg);
263 			}
264 			break;
265 
266 		case 'l':
267 			/*
268 			 * -l [number]: # of times cmd is executed
269 			 * N.B. that an argument *isn't* required here; if
270 			 * it's not given, then 1 is assumed.
271 			 *
272 			 * parseargs handles the optional arg processing.
273 			 */
274 
275 			PER_LINE = LEGAL = TRUE;  /* initialization	*/
276 			N_ARGS = 0;
277 			INSERT = FALSE;
278 
279 			if ((optarg != NULL) && (*optarg != '\0')) {
280 				if ((PER_LINE = atoi(optarg)) <= 0)
281 					PER_LINE = 1;
282 			}
283 			break;
284 
285 		case 'n':	/* -n number: # stdin args		*/
286 			/*
287 			 * -n number: # stdin args.
288 			 * number *is* required here:
289 			 */
290 			if ((N_ARGS = atoi(optarg)) <= 0) {
291 				ermsg(_("#args must be positive int: %s\n"),
292 				    optarg);
293 			} else {
294 				LEGAL = DASHX || N_ARGS == 1;
295 				INSERT = PER_LINE = FALSE;
296 			}
297 			break;
298 
299 		case 's':	/* -s size: set max size of each arg list */
300 			BUFLIM = atoi(optarg);
301 			if (BUFLIM > BUFSIZE || BUFLIM <= 0) {
302 				ermsg(_("0 < max-cmd-line-size <= %d: %s\n"),
303 				    BUFSIZE, optarg);
304 			}
305 			break;
306 
307 		case 'x':	/* -x: terminate if args > size limit	*/
308 			DASHX = LEGAL = TRUE;
309 			break;
310 
311 		default:
312 			/*
313 			 * bad argument. complain and get ready to die.
314 			 */
315 			usage();
316 			exit(2);
317 			break;
318 		}
319 	}
320 
321 	/*
322 	 * if anything called ermsg(), something screwed up, so
323 	 * we exit early.
324 	 */
325 	if (OK == FALSE) {
326 		usage();
327 		exit(2);
328 	}
329 
330 	/*
331 	 * we're finished handling xargs's options, so now pick up
332 	 * the command name (if any), and it's options.
333 	 */
334 
335 
336 	mac -= optind;	/* dec arg count by what we've processed 	*/
337 	mav += optind;	/* inc to current mav				*/
338 
339 	if (mac <= 0) {	/* if there're no more args to process,	*/
340 		cmdname = "/usr/bin/echo";	/* our default command	*/
341 		*ARGV++ = addarg(cmdname);	/* use the default cmd.	*/
342 	} else {	/* otherwise keep parsing rest of the string.	*/
343 		/*
344 		 * note that we can't use getopts(3C), and *must* parse
345 		 * this by hand, as we don't know apriori what options the
346 		 * command will take.
347 		 */
348 		cmdname = *mav;	/* get the command name	*/
349 
350 
351 		/* pick up the remaining args from the command line:	*/
352 		while ((OK == TRUE) && (mac-- > 0)) {
353 			/*
354 			 * while we haven't crapped out, and there's
355 			 * work to do:
356 			 */
357 			if (INSERT && ! ERR) {
358 				if (strstr(*mav, INSPAT) != NULL) {
359 					if (++n_inserts > MAXINSERTS) {
360 						ermsg(_("too many args "
361 						    "with %s\n"), INSPAT);
362 						ERR = TRUE;
363 					}
364 					psave->p_ARGV = ARGV;
365 					(psave++)->p_skel = *mav;
366 				}
367 			}
368 			*ARGV++ = addarg(*mav++);
369 		}
370 	}
371 
372 	/* pick up args from standard input */
373 
374 	initlist = ARGV;
375 	initsize = linesize;
376 	lastarg[0] = '\0';
377 
378 	while (OK) {
379 		N_args = 0;
380 		N_lines = 0;
381 		ARGV = initlist;
382 		linesize = initsize;
383 		next = argbuf;
384 
385 		while (MORE || (lastarg[0] != '\0')) {
386 			int l;
387 
388 			if (*lastarg != '\0') {
389 				arg = strcpy(next, lastarg);
390 				*lastarg = '\0';
391 			} else if ((arg = getarg(next)) == NULL) {
392 				break;
393 			}
394 
395 			l = strlen(arg) + 1;
396 			linesize += l;
397 			next += l;
398 
399 			/* Inserts are handled specially later. */
400 			if ((n_inserts == 0) && (linesize >= BUFLIM)) {
401 				/*
402 				 * Legal indicates hard fail if the list is
403 				 * truncated due to size.  So fail, or if we
404 				 * cannot create any list because it would be
405 				 * too big.
406 				 */
407 				if (LEGAL || N_args == 0) {
408 					EMSG(LIST2LONG);
409 					exit(2);
410 					/* NOTREACHED */
411 				}
412 
413 				/*
414 				 * Otherwise just save argument for later.
415 				 */
416 				(void) strcpy(lastarg, arg);
417 				break;
418 			}
419 
420 			*ARGV++ = arg;
421 
422 			N_args++;
423 
424 			if ((PER_LINE && N_lines >= PER_LINE) ||
425 			    (N_ARGS && (N_args) >= N_ARGS)) {
426 				break;
427 			}
428 
429 
430 			if ((ARGV - arglist) == MAXARGS) {
431 				break;
432 			}
433 		}
434 
435 		*ARGV = NULL;
436 		if (N_args == 0) {
437 			/* Reached the end with no more work. */
438 			exit(exitstat);
439 		}
440 
441 		/* insert arg if requested */
442 
443 		if (!ERR && INSERT) {
444 
445 			p_ibuf = ins_buf;
446 			ARGV--;
447 			j = ibufsize = 0;
448 			for (psave = saveargv; ++j <= n_inserts; ++psave) {
449 				addibuf(psave);
450 				if (ERR)
451 					break;
452 			}
453 		}
454 		*ARGV = NULL;
455 
456 		if (n_inserts > 0) {
457 			/*
458 			 * if we've done any insertions, re-calculate the
459 			 * linesize. bomb out if we've exceeded our length.
460 			 */
461 			linesize = 0;
462 			for (ARGV = arglist; *ARGV != NULL; ARGV++) {
463 				linesize += strlen(*ARGV) + 1;
464 			}
465 			if (linesize >= BUFLIM) {
466 				EMSG(LIST2LONG);
467 				exit(2);
468 				/* NOTREACHED */
469 			}
470 		}
471 
472 		/* exec command */
473 
474 		if (!ERR) {
475 			if (!MORE &&
476 			    (PER_LINE && N_lines == 0 || N_ARGS && N_args == 0))
477 				exit(exitstat);
478 			OK = TRUE;
479 			j = TRACE ? echoargs() : TRUE;
480 			if (j) {
481 				/*
482 				 * for xcu4, all invocations of cmdname must
483 				 * return 0, in order for us to return 0.
484 				 * so if we have a non-zero status here,
485 				 * quit immediately.
486 				 */
487 				exitstat |= lcall(cmdname, arglist);
488 			}
489 		}
490 	}
491 
492 	if (OK)
493 		return (exitstat);
494 
495 	/*
496 	 * if exitstat was set, to match XCU4 complience,
497 	 * return that value, otherwise, return 1.
498 	 */
499 	return (exitstat ? exitstat : 1);
500 }
501 
502 static char *
503 addarg(char *arg)
504 {
505 	linesize += (strlen(arg) + 1);
506 	return (arg);
507 }
508 
509 
510 static void
511 store_str(char **buffer, char *str, size_t len)
512 {
513 	(void) memcpy(*buffer, str, len);
514 	(*buffer)[len] = '\0';
515 	*buffer += len;
516 }
517 
518 
519 static char *
520 getarg(char *arg)
521 {
522 	char	*xarg = arg;
523 	wchar_t	c;
524 	char	mbc[MB_LEN_MAX];
525 	size_t	len;
526 	int	escape = 0;
527 	int	inquote = 0;
528 
529 	arg[0] = '\0';
530 
531 	while (MORE) {
532 
533 		len = 0;
534 		c = getwchr(mbc, &len);
535 
536 		if (((arg - xarg) + len) > BUFLIM) {
537 			EMSG2(ARG2LONG, BUFLIM);
538 			exit(2);
539 			ERR = TRUE;
540 			return (NULL);
541 		}
542 
543 		switch (c) {
544 		case '\n':
545 			if (ZERO) {
546 				store_str(&arg, mbc, len);
547 				continue;
548 			}
549 			/* FALLTHRU */
550 
551 		case '\0':
552 		case WEOF:	/* Note WEOF == EOF */
553 
554 			if (escape) {
555 				EMSG(BADESCAPE);
556 				ERR = TRUE;
557 				return (NULL);
558 			}
559 			if (inquote) {
560 				EMSG(MISSQUOTE);
561 				ERR = TRUE;
562 				return (NULL);
563 			}
564 
565 			N_lines++;
566 			break;
567 
568 		case '"':
569 			if (ZERO || escape || (inquote == 1)) {
570 				/* treat it literally */
571 				escape = 0;
572 				store_str(&arg, mbc, len);
573 
574 			} else if (inquote == 2) {
575 				/* terminating double quote */
576 				inquote = 0;
577 
578 			} else {
579 				/* starting quoted string */
580 				inquote = 2;
581 			}
582 			continue;
583 
584 		case '\'':
585 			if (ZERO || escape || (inquote == 2)) {
586 				/* treat it literally */
587 				escape = 0;
588 				store_str(&arg, mbc, len);
589 
590 			} else if (inquote == 1) {
591 				/* terminating single quote */
592 				inquote = 0;
593 
594 			} else {
595 				/* starting quoted string */
596 				inquote = 1;
597 			}
598 			continue;
599 
600 		case '\\':
601 			/*
602 			 * Any unquoted character can be escaped by
603 			 * preceding it with a backslash.
604 			 */
605 			if (ZERO || inquote || escape) {
606 				escape = 0;
607 				store_str(&arg, mbc, len);
608 			} else {
609 				escape = 1;
610 			}
611 			continue;
612 
613 		default:
614 			/* most times we will just want to store it */
615 			if (inquote || escape || ZERO || !iswctype(c, blank)) {
616 				escape = 0;
617 				store_str(&arg, mbc, len);
618 				continue;
619 			}
620 			/* unquoted blank */
621 			break;
622 		}
623 
624 		/*
625 		 * At this point we are processing a complete argument.
626 		 */
627 		if (strcmp(xarg, LEOF) == 0 && *LEOF != '\0') {
628 			MORE = FALSE;
629 			return (NULL);
630 		}
631 		if (c == WEOF) {
632 			MORE = FALSE;
633 		}
634 		if (xarg[0] == '\0')
635 			continue;
636 		break;
637 	}
638 
639 	return (xarg[0] == '\0' ? NULL : xarg);
640 }
641 
642 /*
643  * ermsg():	print out an error message, and indicate failure globally.
644  *
645  *	Assumes that message has already been gettext()'d. It would be
646  *	nice if we could just do the gettext() here, but we can't, since
647  *	since xgettext(1M) wouldn't be able to pick up our error message.
648  */
649 /* PRINTFLIKE1 */
650 static void
651 ermsg(char *messages, ...)
652 {
653 	va_list	ap;
654 
655 	va_start(ap, messages);
656 
657 	(void) fprintf(stderr, "xargs: ");
658 	(void) vfprintf(stderr, messages, ap);
659 
660 	va_end(ap);
661 	OK = FALSE;
662 }
663 
664 static int
665 echoargs()
666 {
667 	char	**anarg;
668 	char	**tanarg;	/* tmp ptr			*/
669 	int		i;
670 	char		reply[LINE_MAX];
671 
672 	tanarg = anarg = arglist-1;
673 
674 	/*
675 	 * write out each argument, separated by a space. the tanarg
676 	 * nonsense is for xcu4 testsuite compliance - so that an
677 	 * extra space isn't echoed after the last argument.
678 	 */
679 	while (*++anarg) {		/* while there's an argument	*/
680 		++tanarg;		/* follow anarg			*/
681 		(void) write(2, *anarg, strlen(*anarg));
682 
683 		if (*++tanarg) {	/* if there's another argument:	*/
684 			(void) write(2, " ", 1); /* add a space		*/
685 			--tanarg;	/* reset back to anarg		*/
686 		}
687 	}
688 	if (PROMPT == -1) {
689 		(void) write(2, "\n", 1);
690 		return (TRUE);
691 	}
692 
693 	(void) write(2, "?...", 4);	/* ask the user for input	*/
694 
695 	for (i = 0; i < LINE_MAX && read(PROMPT, &reply[i], 1) > 0; i++) {
696 		if (reply[i] == '\n') {
697 			if (i == 0)
698 				return (FALSE);
699 			break;
700 		}
701 	}
702 	reply[i] = 0;
703 
704 	/* flush remainder of line if necessary */
705 	if (i == LINE_MAX) {
706 		char	bitbucket;
707 
708 		while ((read(PROMPT, &bitbucket, 1) > 0) && (bitbucket != '\n'))
709 			;
710 	}
711 
712 	return (yes_check(reply));
713 }
714 
715 
716 static char *
717 insert(char *pattern, char *subst)
718 {
719 	static char	buffer[MAXSBUF+1];
720 	int		len, ipatlen;
721 	char	*pat;
722 	char	*bufend;
723 	char	*pbuf;
724 
725 	len = strlen(subst);
726 	ipatlen = strlen(INSPAT) - 1;
727 	pat = pattern - 1;
728 	pbuf = buffer;
729 	bufend = &buffer[MAXSBUF];
730 
731 	while (*++pat) {
732 		if (strncmp(pat, INSPAT, ipatlen) == 0) {
733 			if (pbuf + len >= bufend) {
734 				break;
735 			} else {
736 				(void) strcpy(pbuf, subst);
737 				pat += ipatlen;
738 				pbuf += len;
739 			}
740 		} else {
741 			*pbuf++ = *pat;
742 			if (pbuf >= bufend)
743 				break;
744 		}
745 	}
746 
747 	if (!*pat) {
748 		*pbuf = '\0';
749 		return (buffer);
750 	} else {
751 		ermsg(gettext("Maximum argument size with insertion via %s's "
752 		    "exceeded\n"), INSPAT);
753 		ERR = TRUE;
754 		return (NULL);
755 	}
756 }
757 
758 
759 static void
760 addibuf(struct inserts	*p)
761 {
762 	char	*newarg, *skel, *sub;
763 	int		l;
764 
765 	skel = p->p_skel;
766 	sub = *ARGV;
767 	newarg = insert(skel, sub);
768 	if (ERR)
769 		return;
770 
771 	l = strlen(newarg) + 1;
772 	if ((ibufsize += l) > MAXIBUF) {
773 		EMSG(IBUFOVERFLOW);
774 		ERR = TRUE;
775 	}
776 	(void) strcpy(p_ibuf, newarg);
777 	*(p->p_ARGV) = p_ibuf;
778 	p_ibuf += l;
779 }
780 
781 
782 /*
783  * getwchr():	get the next wide character.
784  * description:
785  *	we get the next character from stdin.  This returns WEOF if no
786  *	character is present.  If ZERO is set, it gets a single byte instead
787  *	a wide character.
788  */
789 static wint_t
790 getwchr(char *mbc, size_t *sz)
791 {
792 	size_t		i;
793 	int		c;
794 	wchar_t		wch;
795 
796 	i = 0;
797 	while (i < MB_CUR_MAX) {
798 
799 		if ((c = fgetc(stdin)) == EOF) {
800 
801 			if (i == 0) {
802 				/* TRUE EOF has been reached */
803 				return (WEOF);
804 			}
805 
806 			/*
807 			 * We have some characters in our buffer still so it
808 			 * must be an invalid character right before EOF.
809 			 */
810 			break;
811 		}
812 		mbc[i++] = (char)c;
813 
814 		/* If this succeeds then we are done */
815 		if (ZERO) {
816 			*sz = i;
817 			return ((char)c);
818 		}
819 		if (mbtowc(&wch, mbc, i) != -1) {
820 			*sz = i;
821 			return ((wint_t)wch);
822 		}
823 	}
824 
825 	/*
826 	 * We have now encountered an illegal character sequence.
827 	 * There is nothing much we can do at this point but
828 	 * return an error.  If we attempt to recover we may in fact
829 	 * return garbage as arguments, from the customer's point
830 	 * of view.  After all what if they are feeding us a file
831 	 * generated in another locale?
832 	 */
833 	errno = EILSEQ;
834 	PERR(CORRUPTFILE);
835 	exit(1);
836 	/* NOTREACHED */
837 }
838 
839 
840 static int
841 lcall(char *sub, char **subargs)
842 {
843 	int retcode, retry = 0;
844 	pid_t iwait, child;
845 
846 	for (;;) {
847 		switch (child = fork()) {
848 		default:
849 			while ((iwait = wait(&retcode)) != child &&
850 			    iwait != (pid_t)-1)
851 				;
852 			if (iwait == (pid_t)-1) {
853 				PERR(WAITFAIL);
854 				exit(122);
855 				/* NOTREACHED */
856 			}
857 			if (WIFSIGNALED(retcode)) {
858 				EMSG2(CHILDSIG, WTERMSIG(retcode));
859 				exit(125);
860 				/* NOTREACHED */
861 			}
862 			if ((WEXITSTATUS(retcode) & 0377) == 0377) {
863 				EMSG(CHILDFAIL);
864 				exit(124);
865 				/* NOTREACHED */
866 			}
867 			return (WEXITSTATUS(retcode));
868 		case 0:
869 			(void) execvp(sub, subargs);
870 			PERR(EXECFAIL);
871 			if (errno == EACCES)
872 				exit(126);
873 			exit(127);
874 			/* NOTREACHED */
875 		case -1:
876 			if (errno != EAGAIN && retry++ < FORK_RETRY) {
877 				PERR(FORKFAIL);
878 				exit(123);
879 			}
880 			(void) sleep(1);
881 		}
882 	}
883 }
884 
885 
886 static void
887 usage()
888 {
889 	ermsg(_(USAGEMSG));
890 	OK = FALSE;
891 }
892 
893 
894 
895 /*
896  * parseargs():		modify the args
897  *	since the -e, -i and -l flags all take optional subarguments,
898  *	and getopts(3C) is clueless about this nonsense, we change the
899  *	our local argument count and strings to separate this out,
900  *	and make it easier to handle via getopts(3c).
901  *
902  *	-e	-> "-e ""
903  *	-e3	-> "-e "3"
904  *	-Estr	-> "-E "str"
905  *	-i	-> "-i "{}"
906  *	-irep	-> "-i "rep"
907  *	-l	-> "-i "1"
908  *	-l10	-> "-i "10"
909  *
910  *	since the -e, -i and -l flags all take optional subarguments,
911  */
912 static void
913 parseargs(int ac, char **av)
914 {
915 	int i;			/* current argument			*/
916 	int cflag;		/* 0 = not processing cmd arg		*/
917 
918 	if ((mav = malloc((ac * 2 + 1) * sizeof (char *))) == NULL) {
919 		PERR(MALLOCFAIL);
920 		exit(1);
921 	}
922 
923 	/* for each argument, see if we need to change things:		*/
924 	for (i = mac = cflag = 0; (av[i] != NULL) && i < ac; i++, mac++) {
925 		if ((mav[mac] = strdup(av[i])) == NULL) {
926 			PERR(MALLOCFAIL);
927 			exit(1);
928 		}
929 
930 		/* -- has been found or argument list is fully processes */
931 		if (cflag)
932 			continue;
933 
934 		/*
935 		 * if we're doing special processing, and we've got a flag
936 		 */
937 		else if ((av[i][0] == '-') && (av[i][1] != NULL)) {
938 			char	*def;
939 
940 			switch (av[i][1]) {
941 			case	'e':
942 				def = ""; /* -e with no arg turns off eof */
943 				goto process_special;
944 			case	'i':
945 				def = INSPAT_STR;
946 				goto process_special;
947 			case	'l':
948 				def = "1";
949 process_special:
950 				/*
951 				 * if there's no sub-option, we *must* add
952 				 * a default one. this is because xargs must
953 				 * be able to distinguish between a valid
954 				 * suboption, and a command name.
955 				 */
956 				if (av[i][2] == NULL) {
957 					mav[++mac] = strdup(def);
958 				} else {
959 					/* clear out our version: */
960 					mav[mac][2] = NULL;
961 					mav[++mac] = strdup(&av[i][2]);
962 				}
963 				if (mav[mac] == NULL) {
964 					PERR(MALLOCFAIL);
965 					exit(1);
966 				}
967 				break;
968 
969 			/* flags with required subarguments:		*/
970 
971 			/*
972 			 * there are two separate cases here. either the
973 			 * flag can have the normal XCU4 handling
974 			 * (of the form: -X subargument); or it can have
975 			 * the old solaris 2.[0-4] handling (of the
976 			 * form: -Xsubargument). in order to maintain
977 			 * backwards compatibility, we must support the
978 			 * latter case. we handle the latter possibility
979 			 * first so both the old solaris way of handling
980 			 * and the new XCU4 way of handling things are allowed.
981 			 */
982 			case	'n':	/* FALLTHROUGH			*/
983 			case	's':	/* FALLTHROUGH			*/
984 			case	'E':	/* FALLTHROUGH			*/
985 			case	'I':	/* FALLTHROUGH			*/
986 			case	'L':
987 				/*
988 				 * if the second character isn't null, then
989 				 * the user has specified the old syntax.
990 				 * we move the subargument into our
991 				 * mod'd argument list.
992 				 */
993 				if (av[i][2] != NULL) {
994 					/* first clean things up:	*/
995 					mav[mac][2] = NULL;
996 
997 					/* now add the separation:	*/
998 					++mac;	/* inc to next mod'd arg */
999 					if ((mav[mac] = strdup(&av[i][2])) ==
1000 					    NULL) {
1001 						PERR(MALLOCFAIL);
1002 						exit(1);
1003 					}
1004 					break;
1005 				}
1006 				i++;
1007 				mac++;
1008 
1009 				if (av[i] == NULL) {
1010 					mav[mac] = NULL;
1011 					return;
1012 				}
1013 				if ((mav[mac] = strdup(av[i])) == NULL) {
1014 					PERR(MALLOCFAIL);
1015 					exit(1);
1016 				}
1017 				break;
1018 
1019 			/* flags */
1020 			case 'p' :
1021 			case 't' :
1022 			case 'x' :
1023 				break;
1024 
1025 			case '-' :
1026 			default:
1027 				/*
1028 				 * here we've hit the cmd argument. so
1029 				 * we'll stop special processing, as the
1030 				 * cmd may have a "-i" etc., argument,
1031 				 * and we don't want to add a "" to it.
1032 				 */
1033 				cflag = 1;
1034 				break;
1035 			}
1036 		} else if (i > 0) {	/* if we're not the 1st arg	*/
1037 			/*
1038 			 * if it's not a flag, then it *must* be the cmd.
1039 			 * set cflag, so we don't mishandle the -[eil] flags.
1040 			 */
1041 			cflag = 1;
1042 		}
1043 	}
1044 
1045 	mav[mac] = NULL;
1046 }
1047