xref: /freebsd/contrib/bmake/arch.c (revision 2ccf8a827cce7e34e2063a3d33e5cce4b79799cc)
1 /*	$NetBSD: arch.c,v 1.107 2020/08/30 11:15:05 rillig Exp $	*/
2 
3 /*
4  * Copyright (c) 1988, 1989, 1990, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Adam de Boor.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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 /*
36  * Copyright (c) 1989 by Berkeley Softworks
37  * All rights reserved.
38  *
39  * This code is derived from software contributed to Berkeley by
40  * Adam de Boor.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  * 3. All advertising materials mentioning features or use of this software
51  *    must display the following acknowledgement:
52  *	This product includes software developed by the University of
53  *	California, Berkeley and its contributors.
54  * 4. Neither the name of the University nor the names of its contributors
55  *    may be used to endorse or promote products derived from this software
56  *    without specific prior written permission.
57  *
58  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
59  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
60  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
61  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
62  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
63  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
64  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
65  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
66  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
67  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
68  * SUCH DAMAGE.
69  */
70 
71 #ifndef MAKE_NATIVE
72 static char rcsid[] = "$NetBSD: arch.c,v 1.107 2020/08/30 11:15:05 rillig Exp $";
73 #else
74 #include <sys/cdefs.h>
75 #ifndef lint
76 #if 0
77 static char sccsid[] = "@(#)arch.c	8.2 (Berkeley) 1/2/94";
78 #else
79 __RCSID("$NetBSD: arch.c,v 1.107 2020/08/30 11:15:05 rillig Exp $");
80 #endif
81 #endif /* not lint */
82 #endif
83 
84 /*-
85  * arch.c --
86  *	Functions to manipulate libraries, archives and their members.
87  *
88  *	Once again, cacheing/hashing comes into play in the manipulation
89  * of archives. The first time an archive is referenced, all of its members'
90  * headers are read and hashed and the archive closed again. All hashed
91  * archives are kept on a list which is searched each time an archive member
92  * is referenced.
93  *
94  * The interface to this module is:
95  *	Arch_ParseArchive   	Given an archive specification, return a list
96  *	    	  	    	of GNode's, one for each member in the spec.
97  *	    	  	    	FALSE is returned if the specification is
98  *	    	  	    	invalid for some reason.
99  *
100  *	Arch_Touch	    	Alter the modification time of the archive
101  *	    	  	    	member described by the given node to be
102  *	    	  	    	the current time.
103  *
104  *	Arch_TouchLib	    	Update the modification time of the library
105  *	    	  	    	described by the given node. This is special
106  *	    	  	    	because it also updates the modification time
107  *	    	  	    	of the library's table of contents.
108  *
109  *	Arch_MTime	    	Find the modification time of a member of
110  *	    	  	    	an archive *in the archive*. The time is also
111  *	    	  	    	placed in the member's GNode. Returns the
112  *	    	  	    	modification time.
113  *
114  *	Arch_MemTime	    	Find the modification time of a member of
115  *	    	  	    	an archive. Called when the member doesn't
116  *	    	  	    	already exist. Looks in the archive for the
117  *	    	  	    	modification time. Returns the modification
118  *	    	  	    	time.
119  *
120  *	Arch_FindLib	    	Search for a library along a path. The
121  *	    	  	    	library name in the GNode should be in
122  *	    	  	    	-l<name> format.
123  *
124  *	Arch_LibOODate	    	Special function to decide if a library node
125  *	    	  	    	is out-of-date.
126  *
127  *	Arch_Init 	    	Initialize this module.
128  *
129  *	Arch_End 	    	Cleanup this module.
130  */
131 
132 #ifdef HAVE_CONFIG_H
133 # include "config.h"
134 #endif
135 #include    <sys/types.h>
136 #include    <sys/stat.h>
137 #include    <sys/time.h>
138 #include    <sys/param.h>
139 #ifdef HAVE_AR_H
140 #include    <ar.h>
141 #else
142 struct ar_hdr {
143         char ar_name[16];               /* name */
144         char ar_date[12];               /* modification time */
145         char ar_uid[6];                 /* user id */
146         char ar_gid[6];                 /* group id */
147         char ar_mode[8];                /* octal file permissions */
148         char ar_size[10];               /* size in bytes */
149 #ifndef ARFMAG
150 #define ARFMAG  "`\n"
151 #endif
152         char ar_fmag[2];                /* consistency check */
153 };
154 #endif
155 #if defined(HAVE_RANLIB_H) && !(defined(__ELF__) || defined(NO_RANLIB))
156 #include    <ranlib.h>
157 #endif
158 #include    <stdio.h>
159 #include    <stdlib.h>
160 #ifdef HAVE_UTIME_H
161 #include    <utime.h>
162 #endif
163 
164 #include    "make.h"
165 #include    "hash.h"
166 #include    "dir.h"
167 
168 #ifdef TARGET_MACHINE
169 #undef MAKE_MACHINE
170 #define MAKE_MACHINE TARGET_MACHINE
171 #endif
172 #ifdef TARGET_MACHINE_ARCH
173 #undef MAKE_MACHINE_ARCH
174 #define MAKE_MACHINE_ARCH TARGET_MACHINE_ARCH
175 #endif
176 
177 static Lst	  archives;   /* Lst of archives we've already examined */
178 
179 typedef struct Arch {
180     char	  *name;      /* Name of archive */
181     Hash_Table	  members;    /* All the members of the archive described
182 			       * by <name, struct ar_hdr *> key/value pairs */
183     char	  *fnametab;  /* Extended name table strings */
184     size_t	  fnamesize;  /* Size of the string table */
185 } Arch;
186 
187 static struct ar_hdr *ArchStatMember(const char *, const char *, Boolean);
188 static FILE *ArchFindMember(const char *, const char *,
189 			    struct ar_hdr *, const char *);
190 #if defined(__svr4__) || defined(__SVR4) || defined(__ELF__)
191 #define SVR4ARCHIVES
192 static int ArchSVR4Entry(Arch *, char *, size_t, FILE *);
193 #endif
194 
195 
196 #if defined(_AIX)
197 # define AR_NAME _ar_name.ar_name
198 # define AR_FMAG _ar_name.ar_fmag
199 # define SARMAG  SAIAMAG
200 # define ARMAG   AIAMAG
201 # define ARFMAG  AIAFMAG
202 #endif
203 #ifndef  AR_NAME
204 # define AR_NAME ar_name
205 #endif
206 #ifndef  AR_DATE
207 # define AR_DATE ar_date
208 #endif
209 #ifndef  AR_SIZE
210 # define AR_SIZE ar_size
211 #endif
212 #ifndef  AR_FMAG
213 # define AR_FMAG ar_fmag
214 #endif
215 #ifndef ARMAG
216 # define ARMAG	"!<arch>\n"
217 #endif
218 #ifndef SARMAG
219 # define SARMAG	8
220 #endif
221 
222 #define AR_MAX_NAME_LEN	    (sizeof(arh.AR_NAME)-1)
223 
224 #ifdef CLEANUP
225 static void
226 ArchFree(void *ap)
227 {
228     Arch *a = (Arch *)ap;
229     Hash_Search	  search;
230     Hash_Entry	  *entry;
231 
232     /* Free memory from hash entries */
233     for (entry = Hash_EnumFirst(&a->members, &search);
234 	 entry != NULL;
235 	 entry = Hash_EnumNext(&search))
236 	free(Hash_GetValue(entry));
237 
238     free(a->name);
239     free(a->fnametab);
240     Hash_DeleteTable(&a->members);
241     free(a);
242 }
243 #endif
244 
245 
246 /*-
247  *-----------------------------------------------------------------------
248  * Arch_ParseArchive --
249  *	Parse the archive specification in the given line and find/create
250  *	the nodes for the specified archive members, placing their nodes
251  *	on the given list.
252  *
253  * Input:
254  *	linePtr		Pointer to start of specification
255  *	nodeLst		Lst on which to place the nodes
256  *	ctxt		Context in which to expand variables
257  *
258  * Results:
259  *	TRUE if it was a valid specification. The linePtr is updated
260  *	to point to the first non-space after the archive spec. The
261  *	nodes for the members are placed on the given list.
262  *-----------------------------------------------------------------------
263  */
264 Boolean
265 Arch_ParseArchive(char **linePtr, Lst nodeLst, GNode *ctxt)
266 {
267     char	    *cp;	    /* Pointer into line */
268     GNode	    *gn;     	    /* New node */
269     char	    *libName;  	    /* Library-part of specification */
270     char	    *memName;  	    /* Member-part of specification */
271     char	    saveChar;  	    /* Ending delimiter of member-name */
272     Boolean 	    subLibName;	    /* TRUE if libName should have/had
273 				     * variable substitution performed on it */
274 
275     libName = *linePtr;
276 
277     subLibName = FALSE;
278 
279     for (cp = libName; *cp != '(' && *cp != '\0'; cp++) {
280 	if (*cp == '$') {
281 	    /*
282 	     * Variable spec, so call the Var module to parse the puppy
283 	     * so we can safely advance beyond it...
284 	     */
285 	    int 	length;
286 	    void	*result_freeIt;
287 	    const char	*result;
288 	    Boolean isError;
289 
290 	    result = Var_Parse(cp, ctxt, VARE_UNDEFERR|VARE_WANTRES,
291 			       &length, &result_freeIt);
292 	    isError = result == var_Error;
293 	    free(result_freeIt);
294 	    if (isError)
295 		return FALSE;
296 
297 	    subLibName = TRUE;
298 	    cp += length - 1;
299 	}
300     }
301 
302     *cp++ = '\0';
303     if (subLibName) {
304 	libName = Var_Subst(libName, ctxt, VARE_UNDEFERR|VARE_WANTRES);
305     }
306 
307 
308     for (;;) {
309 	/*
310 	 * First skip to the start of the member's name, mark that
311 	 * place and skip to the end of it (either white-space or
312 	 * a close paren).
313 	 */
314 	Boolean	doSubst = FALSE; /* TRUE if need to substitute in memName */
315 
316 	while (*cp != '\0' && *cp != ')' && isspace ((unsigned char)*cp)) {
317 	    cp++;
318 	}
319 	memName = cp;
320 	while (*cp != '\0' && *cp != ')' && !isspace ((unsigned char)*cp)) {
321 	    if (*cp == '$') {
322 		/*
323 		 * Variable spec, so call the Var module to parse the puppy
324 		 * so we can safely advance beyond it...
325 		 */
326 		int 	length;
327 		void	*freeIt;
328 		const char *result;
329 		Boolean isError;
330 
331 		result = Var_Parse(cp, ctxt, VARE_UNDEFERR|VARE_WANTRES,
332 				   &length, &freeIt);
333 		isError = result == var_Error;
334 		free(freeIt);
335 
336 		if (isError)
337 		    return FALSE;
338 
339 		doSubst = TRUE;
340 		cp += length;
341 	    } else {
342 		cp++;
343 	    }
344 	}
345 
346 	/*
347 	 * If the specification ends without a closing parenthesis,
348 	 * chances are there's something wrong (like a missing backslash),
349 	 * so it's better to return failure than allow such things to happen
350 	 */
351 	if (*cp == '\0') {
352 	    printf("No closing parenthesis in archive specification\n");
353 	    return FALSE;
354 	}
355 
356 	/*
357 	 * If we didn't move anywhere, we must be done
358 	 */
359 	if (cp == memName) {
360 	    break;
361 	}
362 
363 	saveChar = *cp;
364 	*cp = '\0';
365 
366 	/*
367 	 * XXX: This should be taken care of intelligently by
368 	 * SuffExpandChildren, both for the archive and the member portions.
369 	 */
370 	/*
371 	 * If member contains variables, try and substitute for them.
372 	 * This will slow down archive specs with dynamic sources, of course,
373 	 * since we'll be (non-)substituting them three times, but them's
374 	 * the breaks -- we need to do this since SuffExpandChildren calls
375 	 * us, otherwise we could assume the thing would be taken care of
376 	 * later.
377 	 */
378 	if (doSubst) {
379 	    char    *buf;
380 	    char    *sacrifice;
381 	    char    *oldMemName = memName;
382 
383 	    memName = Var_Subst(memName, ctxt, VARE_UNDEFERR | VARE_WANTRES);
384 
385 	    /*
386 	     * Now form an archive spec and recurse to deal with nested
387 	     * variables and multi-word variable values.... The results
388 	     * are just placed at the end of the nodeLst we're returning.
389 	     */
390 	    buf = sacrifice = str_concat4(libName, "(", memName, ")");
391 
392 	    if (strchr(memName, '$') && strcmp(memName, oldMemName) == 0) {
393 		/*
394 		 * Must contain dynamic sources, so we can't deal with it now.
395 		 * Just create an ARCHV node for the thing and let
396 		 * SuffExpandChildren handle it...
397 		 */
398 		gn = Targ_FindNode(buf, TARG_CREATE);
399 
400 		if (gn == NULL) {
401 		    free(buf);
402 		    return FALSE;
403 		} else {
404 		    gn->type |= OP_ARCHV;
405 		    Lst_Append(nodeLst, gn);
406 		}
407 	    } else if (!Arch_ParseArchive(&sacrifice, nodeLst, ctxt)) {
408 		/*
409 		 * Error in nested call -- free buffer and return FALSE
410 		 * ourselves.
411 		 */
412 		free(buf);
413 		return FALSE;
414 	    }
415 	    /*
416 	     * Free buffer and continue with our work.
417 	     */
418 	    free(buf);
419 	} else if (Dir_HasWildcards(memName)) {
420 	    Lst members = Lst_Init();
421 	    Buffer nameBuf;
422 
423 	    Buf_Init(&nameBuf, 0);
424 	    Dir_Expand(memName, dirSearchPath, members);
425 	    while (!Lst_IsEmpty(members)) {
426 		char *member = Lst_Dequeue(members);
427 
428 		Buf_Empty(&nameBuf);
429 		Buf_AddStr(&nameBuf, libName);
430 		Buf_AddStr(&nameBuf, "(");
431 		Buf_AddStr(&nameBuf, member);
432 		Buf_AddStr(&nameBuf, ")");
433 		free(member);
434 
435 		gn = Targ_FindNode(Buf_GetAll(&nameBuf, NULL), TARG_CREATE);
436 		if (gn == NULL) {
437 		    Buf_Destroy(&nameBuf, TRUE);
438 		    return FALSE;
439 		} else {
440 		    /*
441 		     * We've found the node, but have to make sure the rest of
442 		     * the world knows it's an archive member, without having
443 		     * to constantly check for parentheses, so we type the
444 		     * thing with the OP_ARCHV bit before we place it on the
445 		     * end of the provided list.
446 		     */
447 		    gn->type |= OP_ARCHV;
448 		    Lst_Append(nodeLst, gn);
449 		}
450 	    }
451 	    Lst_Free(members);
452 	    Buf_Destroy(&nameBuf, TRUE);
453 	} else {
454 	    Buffer nameBuf;
455 
456 	    Buf_Init(&nameBuf, 0);
457 	    Buf_AddStr(&nameBuf, libName);
458 	    Buf_AddStr(&nameBuf, "(");
459 	    Buf_AddStr(&nameBuf, memName);
460 	    Buf_AddStr(&nameBuf, ")");
461 
462 	    gn = Targ_FindNode(Buf_GetAll(&nameBuf, NULL), TARG_CREATE);
463 	    Buf_Destroy(&nameBuf, TRUE);
464 	    if (gn == NULL) {
465 		return FALSE;
466 	    } else {
467 		/*
468 		 * We've found the node, but have to make sure the rest of the
469 		 * world knows it's an archive member, without having to
470 		 * constantly check for parentheses, so we type the thing with
471 		 * the OP_ARCHV bit before we place it on the end of the
472 		 * provided list.
473 		 */
474 		gn->type |= OP_ARCHV;
475 		Lst_Append(nodeLst, gn);
476 	    }
477 	}
478 	if (doSubst) {
479 	    free(memName);
480 	}
481 
482 	*cp = saveChar;
483     }
484 
485     /*
486      * If substituted libName, free it now, since we need it no longer.
487      */
488     if (subLibName) {
489 	free(libName);
490     }
491 
492     /*
493      * We promised the pointer would be set up at the next non-space, so
494      * we must advance cp there before setting *linePtr... (note that on
495      * entrance to the loop, cp is guaranteed to point at a ')')
496      */
497     do {
498 	cp++;
499     } while (*cp != '\0' && isspace ((unsigned char)*cp));
500 
501     *linePtr = cp;
502     return TRUE;
503 }
504 
505 /* See if the given archive is the one we are looking for.
506  * Called via Lst_Find. */
507 static Boolean
508 ArchFindArchive(const void *ar, const void *desiredName)
509 {
510     return strcmp(((const Arch *)ar)->name, desiredName) == 0;
511 }
512 
513 /*-
514  *-----------------------------------------------------------------------
515  * ArchStatMember --
516  *	Locate a member of an archive, given the path of the archive and
517  *	the path of the desired member.
518  *
519  * Input:
520  *	archive		Path to the archive
521  *	member		Name of member. If it is a path, only the last
522  *			component is used.
523  *	hash		TRUE if archive should be hashed if not already so.
524  *
525  * Results:
526  *	A pointer to the current struct ar_hdr structure for the member. Note
527  *	That no position is returned, so this is not useful for touching
528  *	archive members. This is mostly because we have no assurances that
529  *	The archive will remain constant after we read all the headers, so
530  *	there's not much point in remembering the position...
531  *-----------------------------------------------------------------------
532  */
533 static struct ar_hdr *
534 ArchStatMember(const char *archive, const char *member, Boolean hash)
535 {
536     FILE *	  arch;	      /* Stream to archive */
537     size_t	  size;       /* Size of archive member */
538     char	  magic[SARMAG];
539     LstNode	  ln;	      /* Lst member containing archive descriptor */
540     Arch	  *ar;	      /* Archive descriptor */
541     Hash_Entry	  *he;	      /* Entry containing member's description */
542     struct ar_hdr arh;        /* archive-member header for reading archive */
543     char	  memName[MAXPATHLEN+1];
544 			    /* Current member name while hashing. */
545 
546     /*
547      * Because of space constraints and similar things, files are archived
548      * using their final path components, not the entire thing, so we need
549      * to point 'member' to the final component, if there is one, to make
550      * the comparisons easier...
551      */
552     const char *base = strrchr(member, '/');
553     if (base != NULL) {
554 	member = base + 1;
555     }
556 
557     ln = Lst_Find(archives, ArchFindArchive, archive);
558     if (ln != NULL) {
559 	ar = LstNode_Datum(ln);
560 
561 	he = Hash_FindEntry(&ar->members, member);
562 
563 	if (he != NULL) {
564 	    return (struct ar_hdr *)Hash_GetValue(he);
565 	} else {
566 	    /* Try truncated name */
567 	    char copy[AR_MAX_NAME_LEN+1];
568 	    size_t len = strlen(member);
569 
570 	    if (len > AR_MAX_NAME_LEN) {
571 		len = AR_MAX_NAME_LEN;
572 		snprintf(copy, sizeof copy, "%s", member);
573 	    }
574 	    if ((he = Hash_FindEntry(&ar->members, copy)) != NULL)
575 		return (struct ar_hdr *)Hash_GetValue(he);
576 	    return NULL;
577 	}
578     }
579 
580     if (!hash) {
581 	/*
582 	 * Caller doesn't want the thing hashed, just use ArchFindMember
583 	 * to read the header for the member out and close down the stream
584 	 * again. Since the archive is not to be hashed, we assume there's
585 	 * no need to allocate extra room for the header we're returning,
586 	 * so just declare it static.
587 	 */
588 	 static struct ar_hdr	sarh;
589 
590 	 arch = ArchFindMember(archive, member, &sarh, "r");
591 
592 	 if (arch == NULL) {
593 	    return NULL;
594 	} else {
595 	    fclose(arch);
596 	    return &sarh;
597 	}
598     }
599 
600     /*
601      * We don't have this archive on the list yet, so we want to find out
602      * everything that's in it and cache it so we can get at it quickly.
603      */
604     arch = fopen(archive, "r");
605     if (arch == NULL) {
606 	return NULL;
607     }
608 
609     /*
610      * We use the ARMAG string to make sure this is an archive we
611      * can handle...
612      */
613     if ((fread(magic, SARMAG, 1, arch) != 1) ||
614 	(strncmp(magic, ARMAG, SARMAG) != 0)) {
615 	    fclose(arch);
616 	    return NULL;
617     }
618 
619     ar = bmake_malloc(sizeof(Arch));
620     ar->name = bmake_strdup(archive);
621     ar->fnametab = NULL;
622     ar->fnamesize = 0;
623     Hash_InitTable(&ar->members, -1);
624     memName[AR_MAX_NAME_LEN] = '\0';
625 
626     while (fread((char *)&arh, sizeof(struct ar_hdr), 1, arch) == 1) {
627 	if (strncmp( arh.AR_FMAG, ARFMAG, sizeof(arh.AR_FMAG)) != 0) {
628 	    /*
629 	     * The header is bogus, so the archive is bad
630 	     * and there's no way we can recover...
631 	     */
632 	    goto badarch;
633 	} else {
634 	    char *nameend;
635 
636 	    /*
637 	     * We need to advance the stream's pointer to the start of the
638 	     * next header. Files are padded with newlines to an even-byte
639 	     * boundary, so we need to extract the size of the file from the
640 	     * 'size' field of the header and round it up during the seek.
641 	     */
642 	    arh.AR_SIZE[sizeof(arh.AR_SIZE)-1] = '\0';
643 	    size = (size_t)strtol(arh.ar_size, NULL, 10);
644 
645 	    memcpy(memName, arh.AR_NAME, sizeof(arh.AR_NAME));
646 	    nameend = memName + AR_MAX_NAME_LEN;
647 	    while (*nameend == ' ') {
648 		nameend--;
649 	    }
650 	    nameend[1] = '\0';
651 
652 #ifdef SVR4ARCHIVES
653 	    /*
654 	     * svr4 names are slash terminated. Also svr4 extended AR format.
655 	     */
656 	    if (memName[0] == '/') {
657 		/*
658 		 * svr4 magic mode; handle it
659 		 */
660 		switch (ArchSVR4Entry(ar, memName, size, arch)) {
661 		case -1:  /* Invalid data */
662 		    goto badarch;
663 		case 0:	  /* List of files entry */
664 		    continue;
665 		default:  /* Got the entry */
666 		    break;
667 		}
668 	    }
669 	    else {
670 		if (nameend[0] == '/')
671 		    nameend[0] = '\0';
672 	    }
673 #endif
674 
675 #ifdef AR_EFMT1
676 	    /*
677 	     * BSD 4.4 extended AR format: #1/<namelen>, with name as the
678 	     * first <namelen> bytes of the file
679 	     */
680 	    if (strncmp(memName, AR_EFMT1, sizeof(AR_EFMT1) - 1) == 0 &&
681 		isdigit((unsigned char)memName[sizeof(AR_EFMT1) - 1])) {
682 
683 		int elen = atoi(&memName[sizeof(AR_EFMT1)-1]);
684 
685 		if ((unsigned int)elen > MAXPATHLEN)
686 			goto badarch;
687 		if (fread(memName, (size_t)elen, 1, arch) != 1)
688 			goto badarch;
689 		memName[elen] = '\0';
690 		if (fseek(arch, -elen, SEEK_CUR) != 0)
691 			goto badarch;
692 		if (DEBUG(ARCH) || DEBUG(MAKE)) {
693 		    fprintf(debug_file, "ArchStat: Extended format entry for %s\n", memName);
694 		}
695 	    }
696 #endif
697 
698 	    he = Hash_CreateEntry(&ar->members, memName, NULL);
699 	    Hash_SetValue(he, bmake_malloc(sizeof(struct ar_hdr)));
700 	    memcpy(Hash_GetValue(he), &arh, sizeof(struct ar_hdr));
701 	}
702 	if (fseek(arch, ((long)size + 1) & ~1, SEEK_CUR) != 0)
703 	    goto badarch;
704     }
705 
706     fclose(arch);
707 
708     Lst_Append(archives, ar);
709 
710     /*
711      * Now that the archive has been read and cached, we can look into
712      * the hash table to find the desired member's header.
713      */
714     he = Hash_FindEntry(&ar->members, member);
715 
716     if (he != NULL) {
717 	return (struct ar_hdr *)Hash_GetValue(he);
718     } else {
719 	return NULL;
720     }
721 
722 badarch:
723     fclose(arch);
724     Hash_DeleteTable(&ar->members);
725     free(ar->fnametab);
726     free(ar);
727     return NULL;
728 }
729 
730 #ifdef SVR4ARCHIVES
731 /*-
732  *-----------------------------------------------------------------------
733  * ArchSVR4Entry --
734  *	Parse an SVR4 style entry that begins with a slash.
735  *	If it is "//", then load the table of filenames
736  *	If it is "/<offset>", then try to substitute the long file name
737  *	from offset of a table previously read.
738  *	If a table is read, the file pointer is moved to the next archive
739  *	member.
740  *
741  * Results:
742  *	-1: Bad data in archive
743  *	 0: A table was loaded from the file
744  *	 1: Name was successfully substituted from table
745  *	 2: Name was not successfully substituted from table
746  *-----------------------------------------------------------------------
747  */
748 static int
749 ArchSVR4Entry(Arch *ar, char *name, size_t size, FILE *arch)
750 {
751 #define ARLONGNAMES1 "//"
752 #define ARLONGNAMES2 "/ARFILENAMES"
753     size_t entry;
754     char *ptr, *eptr;
755 
756     if (strncmp(name, ARLONGNAMES1, sizeof(ARLONGNAMES1) - 1) == 0 ||
757 	strncmp(name, ARLONGNAMES2, sizeof(ARLONGNAMES2) - 1) == 0) {
758 
759 	if (ar->fnametab != NULL) {
760 	    if (DEBUG(ARCH)) {
761 		fprintf(debug_file, "Attempted to redefine an SVR4 name table\n");
762 	    }
763 	    return -1;
764 	}
765 
766 	/*
767 	 * This is a table of archive names, so we build one for
768 	 * ourselves
769 	 */
770 	ar->fnametab = bmake_malloc(size);
771 	ar->fnamesize = size;
772 
773 	if (fread(ar->fnametab, size, 1, arch) != 1) {
774 	    if (DEBUG(ARCH)) {
775 		fprintf(debug_file, "Reading an SVR4 name table failed\n");
776 	    }
777 	    return -1;
778 	}
779 	eptr = ar->fnametab + size;
780 	for (entry = 0, ptr = ar->fnametab; ptr < eptr; ptr++)
781 	    switch (*ptr) {
782 	    case '/':
783 		entry++;
784 		*ptr = '\0';
785 		break;
786 
787 	    case '\n':
788 		break;
789 
790 	    default:
791 		break;
792 	    }
793 	if (DEBUG(ARCH)) {
794 	    fprintf(debug_file, "Found svr4 archive name table with %lu entries\n",
795 		    (unsigned long)entry);
796 	}
797 	return 0;
798     }
799 
800     if (name[1] == ' ' || name[1] == '\0')
801 	return 2;
802 
803     entry = (size_t)strtol(&name[1], &eptr, 0);
804     if ((*eptr != ' ' && *eptr != '\0') || eptr == &name[1]) {
805 	if (DEBUG(ARCH)) {
806 	    fprintf(debug_file, "Could not parse SVR4 name %s\n", name);
807 	}
808 	return 2;
809     }
810     if (entry >= ar->fnamesize) {
811 	if (DEBUG(ARCH)) {
812 	    fprintf(debug_file, "SVR4 entry offset %s is greater than %lu\n",
813 		   name, (unsigned long)ar->fnamesize);
814 	}
815 	return 2;
816     }
817 
818     if (DEBUG(ARCH)) {
819 	fprintf(debug_file, "Replaced %s with %s\n", name, &ar->fnametab[entry]);
820     }
821 
822     snprintf(name, MAXPATHLEN + 1, "%s", &ar->fnametab[entry]);
823     return 1;
824 }
825 #endif
826 
827 
828 /*-
829  *-----------------------------------------------------------------------
830  * ArchFindMember --
831  *	Locate a member of an archive, given the path of the archive and
832  *	the path of the desired member. If the archive is to be modified,
833  *	the mode should be "r+", if not, it should be "r".
834  *	The passed struct ar_hdr structure is filled in.
835  *
836  * Input:
837  *	archive		Path to the archive
838  *	member		Name of member. If it is a path, only the last
839  *			component is used.
840  *	arhPtr		Pointer to header structure to be filled in
841  *	mode		The mode for opening the stream
842  *
843  * Results:
844  *	An FILE *, opened for reading and writing, positioned at the
845  *	start of the member's struct ar_hdr, or NULL if the member was
846  *	nonexistent. The current struct ar_hdr for member.
847  *-----------------------------------------------------------------------
848  */
849 static FILE *
850 ArchFindMember(const char *archive, const char *member, struct ar_hdr *arhPtr,
851     const char *mode)
852 {
853     FILE *	  arch;	      /* Stream to archive */
854     int		  size;       /* Size of archive member */
855     char	  magic[SARMAG];
856     size_t	  len, tlen;
857     const char *  base;
858 
859     arch = fopen(archive, mode);
860     if (arch == NULL) {
861 	return NULL;
862     }
863 
864     /*
865      * We use the ARMAG string to make sure this is an archive we
866      * can handle...
867      */
868     if ((fread(magic, SARMAG, 1, arch) != 1) ||
869 	(strncmp(magic, ARMAG, SARMAG) != 0)) {
870 	    fclose(arch);
871 	    return NULL;
872     }
873 
874     /*
875      * Because of space constraints and similar things, files are archived
876      * using their final path components, not the entire thing, so we need
877      * to point 'member' to the final component, if there is one, to make
878      * the comparisons easier...
879      */
880     base = strrchr(member, '/');
881     if (base != NULL) {
882 	member = base + 1;
883     }
884     len = tlen = strlen(member);
885     if (len > sizeof(arhPtr->AR_NAME)) {
886 	tlen = sizeof(arhPtr->AR_NAME);
887     }
888 
889     while (fread((char *)arhPtr, sizeof(struct ar_hdr), 1, arch) == 1) {
890 	if (strncmp(arhPtr->AR_FMAG, ARFMAG, sizeof(arhPtr->AR_FMAG) ) != 0) {
891 	     /*
892 	      * The header is bogus, so the archive is bad
893 	      * and there's no way we can recover...
894 	      */
895 	     fclose(arch);
896 	     return NULL;
897 	} else if (strncmp(member, arhPtr->AR_NAME, tlen) == 0) {
898 	    /*
899 	     * If the member's name doesn't take up the entire 'name' field,
900 	     * we have to be careful of matching prefixes. Names are space-
901 	     * padded to the right, so if the character in 'name' at the end
902 	     * of the matched string is anything but a space, this isn't the
903 	     * member we sought.
904 	     */
905 	    if (tlen != sizeof(arhPtr->AR_NAME) && arhPtr->AR_NAME[tlen] != ' '){
906 		goto skip;
907 	    } else {
908 		/*
909 		 * To make life easier, we reposition the file at the start
910 		 * of the header we just read before we return the stream.
911 		 * In a more general situation, it might be better to leave
912 		 * the file at the actual member, rather than its header, but
913 		 * not here...
914 		 */
915 		if (fseek(arch, -(long)sizeof(struct ar_hdr), SEEK_CUR) != 0) {
916 		    fclose(arch);
917 		    return NULL;
918 		}
919 		return arch;
920 	    }
921 	} else
922 #ifdef AR_EFMT1
923 		/*
924 		 * BSD 4.4 extended AR format: #1/<namelen>, with name as the
925 		 * first <namelen> bytes of the file
926 		 */
927 	    if (strncmp(arhPtr->AR_NAME, AR_EFMT1,
928 					sizeof(AR_EFMT1) - 1) == 0 &&
929 		isdigit((unsigned char)arhPtr->AR_NAME[sizeof(AR_EFMT1) - 1])) {
930 
931 		int elen = atoi(&arhPtr->AR_NAME[sizeof(AR_EFMT1)-1]);
932 		char ename[MAXPATHLEN + 1];
933 
934 		if ((unsigned int)elen > MAXPATHLEN) {
935 			fclose(arch);
936 			return NULL;
937 		}
938 		if (fread(ename, (size_t)elen, 1, arch) != 1) {
939 			fclose(arch);
940 			return NULL;
941 		}
942 		ename[elen] = '\0';
943 		if (DEBUG(ARCH) || DEBUG(MAKE)) {
944 		    fprintf(debug_file, "ArchFind: Extended format entry for %s\n", ename);
945 		}
946 		if (strncmp(ename, member, len) == 0) {
947 			/* Found as extended name */
948 			if (fseek(arch, -(long)sizeof(struct ar_hdr) - elen,
949 				SEEK_CUR) != 0) {
950 			    fclose(arch);
951 			    return NULL;
952 			}
953 			return arch;
954 		}
955 		if (fseek(arch, -elen, SEEK_CUR) != 0) {
956 		    fclose(arch);
957 		    return NULL;
958 		}
959 		goto skip;
960 	} else
961 #endif
962 	{
963 skip:
964 	    /*
965 	     * This isn't the member we're after, so we need to advance the
966 	     * stream's pointer to the start of the next header. Files are
967 	     * padded with newlines to an even-byte boundary, so we need to
968 	     * extract the size of the file from the 'size' field of the
969 	     * header and round it up during the seek.
970 	     */
971 	    arhPtr->AR_SIZE[sizeof(arhPtr->AR_SIZE)-1] = '\0';
972 	    size = (int)strtol(arhPtr->AR_SIZE, NULL, 10);
973 	    if (fseek(arch, (size + 1) & ~1, SEEK_CUR) != 0) {
974 		fclose(arch);
975 		return NULL;
976 	    }
977 	}
978     }
979 
980     /*
981      * We've looked everywhere, but the member is not to be found. Close the
982      * archive and return NULL -- an error.
983      */
984     fclose(arch);
985     return NULL;
986 }
987 
988 /*-
989  *-----------------------------------------------------------------------
990  * Arch_Touch --
991  *	Touch a member of an archive.
992  *	The modification time of the entire archive is also changed.
993  *	For a library, this could necessitate the re-ranlib'ing of the
994  *	whole thing.
995  *
996  * Input:
997  *	gn		Node of member to touch
998  *
999  * Results:
1000  *	The 'time' field of the member's header is updated.
1001  *-----------------------------------------------------------------------
1002  */
1003 void
1004 Arch_Touch(GNode *gn)
1005 {
1006     FILE *	  arch;	  /* Stream open to archive, positioned properly */
1007     struct ar_hdr arh;	  /* Current header describing member */
1008     char *p1, *p2;
1009 
1010     arch = ArchFindMember(Var_Value(ARCHIVE, gn, &p1),
1011 			  Var_Value(MEMBER, gn, &p2),
1012 			  &arh, "r+");
1013 
1014     bmake_free(p1);
1015     bmake_free(p2);
1016 
1017     snprintf(arh.AR_DATE, sizeof(arh.AR_DATE), "%-12ld", (long) now);
1018 
1019     if (arch != NULL) {
1020 	(void)fwrite((char *)&arh, sizeof(struct ar_hdr), 1, arch);
1021 	fclose(arch);
1022     }
1023 }
1024 
1025 /* Given a node which represents a library, touch the thing, making sure that
1026  * the table of contents also is touched.
1027  *
1028  * Both the modification time of the library and of the RANLIBMAG member are
1029  * set to 'now'.
1030  *
1031  * Input:
1032  *	gn		The node of the library to touch
1033  */
1034 void
1035 Arch_TouchLib(GNode *gn)
1036 {
1037 #ifdef RANLIBMAG
1038     FILE *	    arch;	/* Stream open to archive */
1039     struct ar_hdr   arh;      	/* Header describing table of contents */
1040     struct utimbuf  times;	/* Times for utime() call */
1041 
1042     arch = ArchFindMember(gn->path, RANLIBMAG, &arh, "r+");
1043     snprintf(arh.AR_DATE, sizeof(arh.AR_DATE), "%-12ld", (long) now);
1044 
1045     if (arch != NULL) {
1046 	(void)fwrite((char *)&arh, sizeof(struct ar_hdr), 1, arch);
1047 	fclose(arch);
1048 
1049 	times.actime = times.modtime = now;
1050 	utime(gn->path, &times);
1051     }
1052 #else
1053     (void)gn;
1054 #endif
1055 }
1056 
1057 /* Return the modification time of a member of an archive. The mtime field
1058  * of the given node is filled in with the value returned by the function.
1059  *
1060  * Input:
1061  *	gn		Node describing archive member
1062  */
1063 time_t
1064 Arch_MTime(GNode *gn)
1065 {
1066     struct ar_hdr *arhPtr;    /* Header of desired member */
1067     time_t	  modTime;    /* Modification time as an integer */
1068     char *p1, *p2;
1069 
1070     arhPtr = ArchStatMember(Var_Value(ARCHIVE, gn, &p1),
1071 			     Var_Value(MEMBER, gn, &p2),
1072 			     TRUE);
1073 
1074     bmake_free(p1);
1075     bmake_free(p2);
1076 
1077     if (arhPtr != NULL) {
1078 	modTime = (time_t)strtol(arhPtr->AR_DATE, NULL, 10);
1079     } else {
1080 	modTime = 0;
1081     }
1082 
1083     gn->mtime = modTime;
1084     return modTime;
1085 }
1086 
1087 /* Given a non-existent archive member's node, get its modification time from
1088  * its archived form, if it exists. gn->mtime is filled in as well. */
1089 time_t
1090 Arch_MemMTime(GNode *gn)
1091 {
1092     LstNode 	  ln;
1093     GNode   	  *pgn;
1094 
1095     Lst_Open(gn->parents);
1096     while ((ln = Lst_Next(gn->parents)) != NULL) {
1097 	pgn = LstNode_Datum(ln);
1098 
1099 	if (pgn->type & OP_ARCHV) {
1100 	    /*
1101 	     * If the parent is an archive specification and is being made
1102 	     * and its member's name matches the name of the node we were
1103 	     * given, record the modification time of the parent in the
1104 	     * child. We keep searching its parents in case some other
1105 	     * parent requires this child to exist...
1106 	     */
1107 	    const char *nameStart = strchr(pgn->name, '(') + 1;
1108 	    const char *nameEnd = strchr(nameStart, ')');
1109 	    size_t nameLen = (size_t)(nameEnd - nameStart);
1110 
1111 	    if ((pgn->flags & REMAKE) &&
1112 		strncmp(nameStart, gn->name, nameLen) == 0) {
1113 		gn->mtime = Arch_MTime(pgn);
1114 	    }
1115 	} else if (pgn->flags & REMAKE) {
1116 	    /*
1117 	     * Something which isn't a library depends on the existence of
1118 	     * this target, so it needs to exist.
1119 	     */
1120 	    gn->mtime = 0;
1121 	    break;
1122 	}
1123     }
1124 
1125     Lst_Close(gn->parents);
1126 
1127     return gn->mtime;
1128 }
1129 
1130 /* Search for a library along the given search path.
1131  *
1132  * The node's 'path' field is set to the found path (including the
1133  * actual file name, not -l...). If the system can handle the -L
1134  * flag when linking (or we cannot find the library), we assume that
1135  * the user has placed the .LIBS variable in the final linking
1136  * command (or the linker will know where to find it) and set the
1137  * TARGET variable for this node to be the node's name. Otherwise,
1138  * we set the TARGET variable to be the full path of the library,
1139  * as returned by Dir_FindFile.
1140  *
1141  * Input:
1142  *	gn		Node of library to find
1143  *	path		Search path
1144  */
1145 void
1146 Arch_FindLib(GNode *gn, Lst path)
1147 {
1148     char	    *libName;   /* file name for archive */
1149     size_t	     sz = strlen(gn->name) + 6 - 2;
1150 
1151     libName = bmake_malloc(sz);
1152     snprintf(libName, sz, "lib%s.a", &gn->name[2]);
1153 
1154     gn->path = Dir_FindFile(libName, path);
1155 
1156     free(libName);
1157 
1158 #ifdef LIBRARIES
1159     Var_Set(TARGET, gn->name, gn);
1160 #else
1161     Var_Set(TARGET, gn->path == NULL ? gn->name : gn->path, gn);
1162 #endif /* LIBRARIES */
1163 }
1164 
1165 /* Decide if a node with the OP_LIB attribute is out-of-date. Called from
1166  * Make_OODate to make its life easier.
1167  * The library will be hashed if it hasn't been already.
1168  *
1169  * There are several ways for a library to be out-of-date that are
1170  * not available to ordinary files. In addition, there are ways
1171  * that are open to regular files that are not available to
1172  * libraries. A library that is only used as a source is never
1173  * considered out-of-date by itself. This does not preclude the
1174  * library's modification time from making its parent be out-of-date.
1175  * A library will be considered out-of-date for any of these reasons,
1176  * given that it is a target on a dependency line somewhere:
1177  *
1178  *	Its modification time is less than that of one of its sources
1179  *	(gn->mtime < gn->cmgn->mtime).
1180  *
1181  *	Its modification time is greater than the time at which the make
1182  *	began (i.e. it's been modified in the course of the make, probably
1183  *	by archiving).
1184  *
1185  *	The modification time of one of its sources is greater than the one
1186  *	of its RANLIBMAG member (i.e. its table of contents is out-of-date).
1187  *	We don't compare of the archive time vs. TOC time because they can be
1188  *	too close. In my opinion we should not bother with the TOC at all
1189  *	since this is used by 'ar' rules that affect the data contents of the
1190  *	archive, not by ranlib rules, which affect the TOC.
1191  *
1192  * Input:
1193  *	gn		The library's graph node
1194  *
1195  * Results:
1196  *	TRUE if the library is out-of-date. FALSE otherwise.
1197  */
1198 Boolean
1199 Arch_LibOODate(GNode *gn)
1200 {
1201     Boolean 	  oodate;
1202 
1203     if (gn->type & OP_PHONY) {
1204 	oodate = TRUE;
1205     } else if (OP_NOP(gn->type) && Lst_IsEmpty(gn->children)) {
1206 	oodate = FALSE;
1207     } else if ((!Lst_IsEmpty(gn->children) && gn->cmgn == NULL) ||
1208 	       (gn->mtime > now) ||
1209 	       (gn->cmgn != NULL && gn->mtime < gn->cmgn->mtime)) {
1210 	oodate = TRUE;
1211     } else {
1212 #ifdef RANLIBMAG
1213 	struct ar_hdr  	*arhPtr;    /* Header for __.SYMDEF */
1214 	int 	  	modTimeTOC; /* The table-of-contents's mod time */
1215 
1216 	arhPtr = ArchStatMember(gn->path, RANLIBMAG, FALSE);
1217 
1218 	if (arhPtr != NULL) {
1219 	    modTimeTOC = (int)strtol(arhPtr->AR_DATE, NULL, 10);
1220 
1221 	    if (DEBUG(ARCH) || DEBUG(MAKE)) {
1222 		fprintf(debug_file, "%s modified %s...", RANLIBMAG, Targ_FmtTime(modTimeTOC));
1223 	    }
1224 	    oodate = (gn->cmgn == NULL || gn->cmgn->mtime > modTimeTOC);
1225 	} else {
1226 	    /*
1227 	     * A library w/o a table of contents is out-of-date
1228 	     */
1229 	    if (DEBUG(ARCH) || DEBUG(MAKE)) {
1230 		fprintf(debug_file, "No t.o.c....");
1231 	    }
1232 	    oodate = TRUE;
1233 	}
1234 #else
1235 	oodate = FALSE;
1236 #endif
1237     }
1238     return oodate;
1239 }
1240 
1241 /* Initialize things for this module. */
1242 void
1243 Arch_Init(void)
1244 {
1245     archives = Lst_Init();
1246 }
1247 
1248 /* Clean up things for this module. */
1249 void
1250 Arch_End(void)
1251 {
1252 #ifdef CLEANUP
1253     Lst_Destroy(archives, ArchFree);
1254 #endif
1255 }
1256 
1257 Boolean
1258 Arch_IsLib(GNode *gn)
1259 {
1260     static const char armag[] = "!<arch>\n";
1261     char buf[sizeof armag - 1];
1262     int fd;
1263 
1264     if ((fd = open(gn->path, O_RDONLY)) == -1)
1265 	return FALSE;
1266 
1267     if (read(fd, buf, sizeof buf) != sizeof buf) {
1268 	(void)close(fd);
1269 	return FALSE;
1270     }
1271 
1272     (void)close(fd);
1273 
1274     return memcmp(buf, armag, sizeof buf) == 0;
1275 }
1276