var.c revision 1.604 1 /* $NetBSD: var.c,v 1.604 2020/10/30 16:54:38 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 /*
72 * Handling of variables and the expressions formed from them.
73 *
74 * Variables are set using lines of the form VAR=value. Both the variable
75 * name and the value can contain references to other variables, by using
76 * expressions like ${VAR}, ${VAR:Modifiers}, ${${VARNAME}} or ${VAR:${MODS}}.
77 *
78 * Interface:
79 * Var_Init Initialize this module.
80 *
81 * Var_End Clean up the module.
82 *
83 * Var_Set Set the value of the variable, creating it if
84 * necessary.
85 *
86 * Var_Append Append more characters to the variable, creating it if
87 * necessary. A space is placed between the old value and
88 * the new one.
89 *
90 * Var_Exists See if a variable exists.
91 *
92 * Var_Value Return the unexpanded value of a variable, or NULL if
93 * the variable is undefined.
94 *
95 * Var_Subst Substitute all variable expressions in a string.
96 *
97 * Var_Parse Parse a variable expression such as ${VAR:Mpattern}.
98 *
99 * Var_Delete Delete a variable.
100 *
101 * Var_ExportVars Export some or even all variables to the environment
102 * of this process and its child processes.
103 *
104 * Var_Export Export the variable to the environment of this process
105 * and its child processes.
106 *
107 * Var_UnExport Don't export the variable anymore.
108 *
109 * Debugging:
110 * Var_Stats Print out hashing statistics if in -dh mode.
111 *
112 * Var_Dump Print out all variables defined in the given context.
113 *
114 * XXX: There's a lot of duplication in these functions.
115 */
116
117 #include <sys/stat.h>
118 #ifndef NO_REGEX
119 #include <sys/types.h>
120 #include <regex.h>
121 #endif
122 #include <inttypes.h>
123 #include <limits.h>
124 #include <time.h>
125
126 #include "make.h"
127 #include "dir.h"
128 #include "job.h"
129 #include "metachar.h"
130
131 /* "@(#)var.c 8.3 (Berkeley) 3/19/94" */
132 MAKE_RCSID("$NetBSD: var.c,v 1.604 2020/10/30 16:54:38 rillig Exp $");
133
134 #define VAR_DEBUG1(fmt, arg1) DEBUG1(VAR, fmt, arg1)
135 #define VAR_DEBUG2(fmt, arg1, arg2) DEBUG2(VAR, fmt, arg1, arg2)
136 #define VAR_DEBUG3(fmt, arg1, arg2, arg3) DEBUG3(VAR, fmt, arg1, arg2, arg3)
137 #define VAR_DEBUG4(fmt, arg1, arg2, arg3, arg4) DEBUG4(VAR, fmt, arg1, arg2, arg3, arg4)
138
139 ENUM_FLAGS_RTTI_3(VarEvalFlags,
140 VARE_UNDEFERR, VARE_WANTRES, VARE_ASSIGN);
141
142 /*
143 * This lets us tell if we have replaced the original environ
144 * (which we cannot free).
145 */
146 char **savedEnv = NULL;
147
148 /* Special return value for Var_Parse, indicating a parse error. It may be
149 * caused by an undefined variable, a syntax error in a modifier or
150 * something entirely different. */
151 char var_Error[] = "";
152
153 /* Special return value for Var_Parse, indicating an undefined variable in
154 * a case where VARE_UNDEFERR is not set. This undefined variable is
155 * typically a dynamic variable such as ${.TARGET}, whose expansion needs to
156 * be deferred until it is defined in an actual target. */
157 static char varUndefined[] = "";
158
159 /* Special return value for Var_Parse, just to avoid allocating empty strings.
160 * In contrast to var_Error and varUndefined, this is not an error marker but
161 * just an ordinary successful return value. */
162 static char emptyString[] = "";
163
164 /*
165 * Traditionally this make consumed $$ during := like any other expansion.
166 * Other make's do not, and this make follows straight since 2016-01-09.
167 *
168 * This knob allows controlling the behavior.
169 * FALSE to consume $$ during := assignment.
170 * TRUE to preserve $$ during := assignment.
171 */
172 #define MAKE_SAVE_DOLLARS ".MAKE.SAVE_DOLLARS"
173 static Boolean save_dollars = TRUE;
174
175 /*
176 * Internally, variables are contained in four different contexts.
177 * 1) the environment. They cannot be changed. If an environment
178 * variable is appended to, the result is placed in the global
179 * context.
180 * 2) the global context. Variables set in the makefiles are located
181 * here.
182 * 3) the command-line context. All variables set on the command line
183 * are placed in this context.
184 * 4) the local context. Each target has associated with it a context
185 * list. On this list are located the structures describing such
186 * local variables as $(@) and $(*)
187 * The four contexts are searched in the reverse order from which they are
188 * listed (but see opts.checkEnvFirst).
189 */
190 GNode *VAR_INTERNAL; /* variables from make itself */
191 GNode *VAR_GLOBAL; /* variables from the makefile */
192 GNode *VAR_CMDLINE; /* variables defined on the command-line */
193
194 typedef enum VarFlags {
195
196 /* The variable's value is currently being used by Var_Parse or Var_Subst.
197 * This marker is used to avoid endless recursion. */
198 VAR_IN_USE = 0x01,
199
200 /* The variable comes from the environment.
201 * These variables are not registered in any GNode, therefore they must
202 * be freed as soon as they are not used anymore. */
203 VAR_FROM_ENV = 0x02,
204
205 /* The variable is exported to the environment, to be used by child
206 * processes. */
207 VAR_EXPORTED = 0x10,
208
209 /* At the point where this variable was exported, it contained an
210 * unresolved reference to another variable. Before any child process is
211 * started, it needs to be exported again, in the hope that the referenced
212 * variable can then be resolved. */
213 VAR_REEXPORT = 0x20,
214
215 /* The variable came from the command line. */
216 VAR_FROM_CMD = 0x40,
217
218 /* The variable value cannot be changed anymore, and the variable cannot
219 * be deleted. Any attempts to do so are ignored. */
220 VAR_READONLY = 0x80
221 } VarFlags;
222
223 ENUM_FLAGS_RTTI_6(VarFlags,
224 VAR_IN_USE, VAR_FROM_ENV,
225 VAR_EXPORTED, VAR_REEXPORT, VAR_FROM_CMD, VAR_READONLY);
226
227 /* Variables are defined using one of the VAR=value assignments. Their
228 * value can be queried by expressions such as $V, ${VAR}, or with modifiers
229 * such as ${VAR:S,from,to,g:Q}.
230 *
231 * There are 3 kinds of variables: context variables, environment variables,
232 * undefined variables.
233 *
234 * Context variables are stored in a GNode.context. The only way to undefine
235 * a context variable is using the .undef directive. In particular, it must
236 * not be possible to undefine a variable during the evaluation of an
237 * expression, or Var.name might point nowhere.
238 *
239 * Environment variables are temporary. They are returned by VarFind, and
240 * after using them, they must be freed using VarFreeEnv.
241 *
242 * Undefined variables occur during evaluation of variable expressions such
243 * as ${UNDEF:Ufallback} in Var_Parse and ApplyModifiers.
244 */
245 typedef struct Var {
246 /* The name of the variable, once set, doesn't change anymore.
247 * For context variables, it aliases the corresponding HashEntry name.
248 * For environment and undefined variables, it is allocated. */
249 const char *name;
250 void *name_freeIt;
251
252 Buffer val; /* its value */
253 VarFlags flags; /* miscellaneous status flags */
254 } Var;
255
256 /*
257 * Exporting vars is expensive so skip it if we can
258 */
259 typedef enum VarExportedMode {
260 VAR_EXPORTED_NONE,
261 VAR_EXPORTED_SOME,
262 VAR_EXPORTED_ALL
263 } VarExportedMode;
264
265 static VarExportedMode var_exportedVars = VAR_EXPORTED_NONE;
266
267 typedef enum VarExportFlags {
268 /*
269 * We pass this to Var_Export when doing the initial export
270 * or after updating an exported var.
271 */
272 VAR_EXPORT_PARENT = 0x01,
273 /*
274 * We pass this to Var_Export1 to tell it to leave the value alone.
275 */
276 VAR_EXPORT_LITERAL = 0x02
277 } VarExportFlags;
278
279 /* Flags for pattern matching in the :S and :C modifiers */
280 typedef enum VarPatternFlags {
281 VARP_SUB_GLOBAL = 0x01, /* Replace as often as possible ('g') */
282 VARP_SUB_ONE = 0x02, /* Replace only once ('1') */
283 VARP_ANCHOR_START = 0x04, /* Match at start of word ('^') */
284 VARP_ANCHOR_END = 0x08 /* Match at end of word ('$') */
285 } VarPatternFlags;
286
287 static Var *
288 VarNew(const char *name, void *name_freeIt, const char *value, VarFlags flags)
289 {
290 size_t value_len = strlen(value);
291 Var *var = bmake_malloc(sizeof *var);
292 var->name = name;
293 var->name_freeIt = name_freeIt;
294 Buf_Init(&var->val, value_len + 1);
295 Buf_AddBytes(&var->val, value, value_len);
296 var->flags = flags;
297 return var;
298 }
299
300 static const char *
301 CanonicalVarname(const char *name)
302 {
303 if (*name == '.' && ch_isupper(name[1])) {
304 switch (name[1]) {
305 case 'A':
306 if (strcmp(name, ".ALLSRC") == 0)
307 name = ALLSRC;
308 if (strcmp(name, ".ARCHIVE") == 0)
309 name = ARCHIVE;
310 break;
311 case 'I':
312 if (strcmp(name, ".IMPSRC") == 0)
313 name = IMPSRC;
314 break;
315 case 'M':
316 if (strcmp(name, ".MEMBER") == 0)
317 name = MEMBER;
318 break;
319 case 'O':
320 if (strcmp(name, ".OODATE") == 0)
321 name = OODATE;
322 break;
323 case 'P':
324 if (strcmp(name, ".PREFIX") == 0)
325 name = PREFIX;
326 break;
327 case 'S':
328 if (strcmp(name, ".SHELL") == 0) {
329 if (!shellPath)
330 Shell_Init();
331 }
332 break;
333 case 'T':
334 if (strcmp(name, ".TARGET") == 0)
335 name = TARGET;
336 break;
337 }
338 }
339
340 /* GNU make has an additional alias $^ == ${.ALLSRC}. */
341
342 return name;
343 }
344
345 static Var *
346 GNode_FindVar(GNode *ctxt, const char *varname, unsigned int hash)
347 {
348 return HashTable_FindValueHash(&ctxt->context, varname, hash);
349 }
350
351 /*-
352 *-----------------------------------------------------------------------
353 * VarFind --
354 * Find the given variable in the given context and any other contexts
355 * indicated.
356 *
357 * Input:
358 * name name to find
359 * ctxt context in which to find it
360 * elsewhere to look in other contexts as well
361 *
362 * Results:
363 * A pointer to the structure describing the desired variable or
364 * NULL if the variable does not exist.
365 *-----------------------------------------------------------------------
366 */
367 static Var *
368 VarFind(const char *name, GNode *ctxt, Boolean elsewhere)
369 {
370 Var *var;
371 unsigned int nameHash;
372
373 /*
374 * If the variable name begins with a '.', it could very well be one of
375 * the local ones. We check the name against all the local variables
376 * and substitute the short version in for 'name' if it matches one of
377 * them.
378 */
379 name = CanonicalVarname(name);
380 nameHash = Hash_Hash(name);
381
382 /*
383 * First look for the variable in the given context. If it's not there,
384 * look for it in VAR_CMDLINE, VAR_GLOBAL and the environment, in that order,
385 * depending on the FIND_* flags in 'flags'
386 */
387 var = GNode_FindVar(ctxt, name, nameHash);
388 if (!elsewhere)
389 return var;
390
391 if (var == NULL && ctxt != VAR_CMDLINE)
392 var = GNode_FindVar(VAR_CMDLINE, name, nameHash);
393
394 if (!opts.checkEnvFirst && var == NULL && ctxt != VAR_GLOBAL) {
395 var = GNode_FindVar(VAR_GLOBAL, name, nameHash);
396 if (var == NULL && ctxt != VAR_INTERNAL) {
397 /* VAR_INTERNAL is subordinate to VAR_GLOBAL */
398 var = GNode_FindVar(VAR_INTERNAL, name, nameHash);
399 }
400 }
401
402 if (var == NULL) {
403 char *env;
404
405 if ((env = getenv(name)) != NULL) {
406 char *varname = bmake_strdup(name);
407 return VarNew(varname, varname, env, VAR_FROM_ENV);
408 }
409
410 if (opts.checkEnvFirst && ctxt != VAR_GLOBAL) {
411 var = GNode_FindVar(VAR_GLOBAL, name, nameHash);
412 if (var == NULL && ctxt != VAR_INTERNAL)
413 var = GNode_FindVar(VAR_INTERNAL, name, nameHash);
414 return var;
415 }
416
417 return NULL;
418 }
419
420 return var;
421 }
422
423 /*-
424 *-----------------------------------------------------------------------
425 * VarFreeEnv --
426 * If the variable is an environment variable, free it
427 *
428 * Input:
429 * v the variable
430 * destroy true if the value buffer should be destroyed.
431 *
432 * Results:
433 * TRUE if it is an environment variable, FALSE otherwise.
434 *-----------------------------------------------------------------------
435 */
436 static Boolean
437 VarFreeEnv(Var *v, Boolean destroy)
438 {
439 if (!(v->flags & VAR_FROM_ENV))
440 return FALSE;
441 free(v->name_freeIt);
442 Buf_Destroy(&v->val, destroy);
443 free(v);
444 return TRUE;
445 }
446
447 /* Add a new variable of the given name and value to the given context.
448 * The name and val arguments are duplicated so they may safely be freed. */
449 static void
450 VarAdd(const char *name, const char *val, GNode *ctxt, VarSet_Flags flags)
451 {
452 HashEntry *he = HashTable_CreateEntry(&ctxt->context, name, NULL);
453 Var *v = VarNew(he->key, NULL, val,
454 flags & VAR_SET_READONLY ? VAR_READONLY : 0);
455 HashEntry_Set(he, v);
456 if (!(ctxt->flags & INTERNAL)) {
457 VAR_DEBUG3("%s:%s = %s\n", ctxt->name, name, val);
458 }
459 }
460
461 /* Remove a variable from a context, freeing the Var structure as well. */
462 void
463 Var_Delete(const char *name, GNode *ctxt)
464 {
465 char *name_freeIt = NULL;
466 HashEntry *he;
467
468 if (strchr(name, '$') != NULL) {
469 (void)Var_Subst(name, VAR_GLOBAL, VARE_WANTRES, &name_freeIt);
470 /* TODO: handle errors */
471 name = name_freeIt;
472 }
473 he = HashTable_FindEntry(&ctxt->context, name);
474 VAR_DEBUG3("%s:delete %s%s\n",
475 ctxt->name, name, he != NULL ? "" : " (not found)");
476 free(name_freeIt);
477
478 if (he != NULL) {
479 Var *v = HashEntry_Get(he);
480 if (v->flags & VAR_EXPORTED)
481 unsetenv(v->name);
482 if (strcmp(v->name, MAKE_EXPORTED) == 0)
483 var_exportedVars = VAR_EXPORTED_NONE;
484 assert(v->name_freeIt == NULL);
485 HashTable_DeleteEntry(&ctxt->context, he);
486 Buf_Destroy(&v->val, TRUE);
487 free(v);
488 }
489 }
490
491 static Boolean
492 MayExport(const char *name)
493 {
494 if (name[0] == '.')
495 return FALSE; /* skip internals */
496 if (name[0] == '-')
497 return FALSE; /* skip misnamed variables */
498 if (name[1] == '\0') {
499 /*
500 * A single char.
501 * If it is one of the vars that should only appear in
502 * local context, skip it, else we can get Var_Subst
503 * into a loop.
504 */
505 switch (name[0]) {
506 case '@':
507 case '%':
508 case '*':
509 case '!':
510 return FALSE;
511 }
512 }
513 return TRUE;
514 }
515
516 /*
517 * Export a single variable.
518 * We ignore make internal variables (those which start with '.').
519 * Also we jump through some hoops to avoid calling setenv
520 * more than necessary since it can leak.
521 * We only manipulate flags of vars if 'parent' is set.
522 */
523 static Boolean
524 Var_Export1(const char *name, VarExportFlags flags)
525 {
526 VarExportFlags parent = flags & VAR_EXPORT_PARENT;
527 Var *v;
528 char *val;
529
530 if (!MayExport(name))
531 return FALSE;
532
533 v = VarFind(name, VAR_GLOBAL, 0);
534 if (v == NULL)
535 return FALSE;
536
537 if (!parent && (v->flags & VAR_EXPORTED) && !(v->flags & VAR_REEXPORT))
538 return FALSE; /* nothing to do */
539
540 val = Buf_GetAll(&v->val, NULL);
541 if (!(flags & VAR_EXPORT_LITERAL) && strchr(val, '$') != NULL) {
542 char *expr;
543
544 if (parent) {
545 /*
546 * Flag this as something we need to re-export.
547 * No point actually exporting it now though,
548 * the child can do it at the last minute.
549 */
550 v->flags |= VAR_EXPORTED | VAR_REEXPORT;
551 return TRUE;
552 }
553 if (v->flags & VAR_IN_USE) {
554 /*
555 * We recursed while exporting in a child.
556 * This isn't going to end well, just skip it.
557 */
558 return FALSE;
559 }
560
561 expr = str_concat3("${", name, "}");
562 (void)Var_Subst(expr, VAR_GLOBAL, VARE_WANTRES, &val);
563 /* TODO: handle errors */
564 setenv(name, val, 1);
565 free(val);
566 free(expr);
567 } else {
568 if (parent)
569 v->flags &= ~(unsigned)VAR_REEXPORT; /* once will do */
570 if (parent || !(v->flags & VAR_EXPORTED))
571 setenv(name, val, 1);
572 }
573 /*
574 * This is so Var_Set knows to call Var_Export again...
575 */
576 if (parent) {
577 v->flags |= VAR_EXPORTED;
578 }
579 return TRUE;
580 }
581
582 /*
583 * This gets called from our children.
584 */
585 void
586 Var_ExportVars(void)
587 {
588 char *val;
589
590 /*
591 * Several make's support this sort of mechanism for tracking
592 * recursion - but each uses a different name.
593 * We allow the makefiles to update MAKELEVEL and ensure
594 * children see a correctly incremented value.
595 */
596 char tmp[BUFSIZ];
597 snprintf(tmp, sizeof(tmp), "%d", makelevel + 1);
598 setenv(MAKE_LEVEL_ENV, tmp, 1);
599
600 if (var_exportedVars == VAR_EXPORTED_NONE)
601 return;
602
603 if (var_exportedVars == VAR_EXPORTED_ALL) {
604 HashIter hi;
605
606 /* Ouch! Exporting all variables at once is crazy... */
607 HashIter_Init(&hi, &VAR_GLOBAL->context);
608 while (HashIter_Next(&hi) != NULL) {
609 Var *var = hi.entry->value;
610 Var_Export1(var->name, 0);
611 }
612 return;
613 }
614
615 (void)Var_Subst("${" MAKE_EXPORTED ":O:u}", VAR_GLOBAL, VARE_WANTRES, &val);
616 /* TODO: handle errors */
617 if (*val) {
618 Words words = Str_Words(val, FALSE);
619 size_t i;
620
621 for (i = 0; i < words.len; i++)
622 Var_Export1(words.words[i], 0);
623 Words_Free(words);
624 }
625 free(val);
626 }
627
628 /*
629 * This is called when .export is seen or .MAKE.EXPORTED is modified.
630 *
631 * It is also called when any exported variable is modified.
632 * XXX: Is it really?
633 *
634 * str has the format "[-env|-literal] varname...".
635 */
636 void
637 Var_Export(const char *str, Boolean isExport)
638 {
639 VarExportFlags flags;
640 char *val;
641
642 if (isExport && str[0] == '\0') {
643 var_exportedVars = VAR_EXPORTED_ALL; /* use with caution! */
644 return;
645 }
646
647 if (isExport && strncmp(str, "-env", 4) == 0) {
648 str += 4;
649 flags = 0;
650 } else if (isExport && strncmp(str, "-literal", 8) == 0) {
651 str += 8;
652 flags = VAR_EXPORT_LITERAL;
653 } else {
654 flags = VAR_EXPORT_PARENT;
655 }
656
657 (void)Var_Subst(str, VAR_GLOBAL, VARE_WANTRES, &val);
658 /* TODO: handle errors */
659 if (val[0] != '\0') {
660 Words words = Str_Words(val, FALSE);
661
662 size_t i;
663 for (i = 0; i < words.len; i++) {
664 const char *name = words.words[i];
665 if (Var_Export1(name, flags)) {
666 if (var_exportedVars == VAR_EXPORTED_NONE)
667 var_exportedVars = VAR_EXPORTED_SOME;
668 if (isExport && (flags & VAR_EXPORT_PARENT)) {
669 Var_Append(MAKE_EXPORTED, name, VAR_GLOBAL);
670 }
671 }
672 }
673 Words_Free(words);
674 }
675 free(val);
676 }
677
678
679 extern char **environ;
680
681 /*
682 * This is called when .unexport[-env] is seen.
683 *
684 * str must have the form "unexport[-env] varname...".
685 */
686 void
687 Var_UnExport(const char *str)
688 {
689 const char *varnames;
690 char *varnames_freeIt;
691 Boolean unexport_env;
692
693 varnames = NULL;
694 varnames_freeIt = NULL;
695
696 str += strlen("unexport");
697 unexport_env = strncmp(str, "-env", 4) == 0;
698 if (unexport_env) {
699 const char *cp;
700 char **newenv;
701
702 cp = getenv(MAKE_LEVEL_ENV); /* we should preserve this */
703 if (environ == savedEnv) {
704 /* we have been here before! */
705 newenv = bmake_realloc(environ, 2 * sizeof(char *));
706 } else {
707 if (savedEnv) {
708 free(savedEnv);
709 savedEnv = NULL;
710 }
711 newenv = bmake_malloc(2 * sizeof(char *));
712 }
713
714 /* Note: we cannot safely free() the original environ. */
715 environ = savedEnv = newenv;
716 newenv[0] = NULL;
717 newenv[1] = NULL;
718 if (cp && *cp)
719 setenv(MAKE_LEVEL_ENV, cp, 1);
720 } else {
721 cpp_skip_whitespace(&str);
722 if (str[0] != '\0')
723 varnames = str;
724 }
725
726 if (varnames == NULL) {
727 /* Using .MAKE.EXPORTED */
728 (void)Var_Subst("${" MAKE_EXPORTED ":O:u}", VAR_GLOBAL, VARE_WANTRES,
729 &varnames_freeIt);
730 /* TODO: handle errors */
731 varnames = varnames_freeIt;
732 }
733
734 {
735 Var *v;
736 size_t i;
737
738 Words words = Str_Words(varnames, FALSE);
739 for (i = 0; i < words.len; i++) {
740 const char *varname = words.words[i];
741 v = VarFind(varname, VAR_GLOBAL, 0);
742 if (v == NULL) {
743 VAR_DEBUG1("Not unexporting \"%s\" (not found)\n", varname);
744 continue;
745 }
746
747 VAR_DEBUG1("Unexporting \"%s\"\n", varname);
748 if (!unexport_env && (v->flags & VAR_EXPORTED) &&
749 !(v->flags & VAR_REEXPORT))
750 unsetenv(v->name);
751 v->flags &= ~(unsigned)(VAR_EXPORTED | VAR_REEXPORT);
752
753 /*
754 * If we are unexporting a list,
755 * remove each one from .MAKE.EXPORTED.
756 * If we are removing them all,
757 * just delete .MAKE.EXPORTED below.
758 */
759 if (varnames == str) {
760 char *expr = str_concat3("${" MAKE_EXPORTED ":N", v->name, "}");
761 char *cp;
762 (void)Var_Subst(expr, VAR_GLOBAL, VARE_WANTRES, &cp);
763 /* TODO: handle errors */
764 Var_Set(MAKE_EXPORTED, cp, VAR_GLOBAL);
765 free(cp);
766 free(expr);
767 }
768 }
769 Words_Free(words);
770 if (varnames != str) {
771 Var_Delete(MAKE_EXPORTED, VAR_GLOBAL);
772 free(varnames_freeIt);
773 }
774 }
775 }
776
777 /* See Var_Set for documentation. */
778 void
779 Var_Set_with_flags(const char *name, const char *val, GNode *ctxt,
780 VarSet_Flags flags)
781 {
782 const char *unexpanded_name = name;
783 char *name_freeIt = NULL;
784 Var *v;
785
786 assert(val != NULL);
787
788 if (strchr(name, '$') != NULL) {
789 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
790 /* TODO: handle errors */
791 name = name_freeIt;
792 }
793
794 if (name[0] == '\0') {
795 VAR_DEBUG2("Var_Set(\"%s\", \"%s\", ...) "
796 "name expands to empty string - ignored\n",
797 unexpanded_name, val);
798 free(name_freeIt);
799 return;
800 }
801
802 if (ctxt == VAR_GLOBAL) {
803 v = VarFind(name, VAR_CMDLINE, 0);
804 if (v != NULL) {
805 if (v->flags & VAR_FROM_CMD) {
806 VAR_DEBUG3("%s:%s = %s ignored!\n", ctxt->name, name, val);
807 goto out;
808 }
809 VarFreeEnv(v, TRUE);
810 }
811 }
812
813 /*
814 * We only look for a variable in the given context since anything set
815 * here will override anything in a lower context, so there's not much
816 * point in searching them all just to save a bit of memory...
817 */
818 v = VarFind(name, ctxt, 0);
819 if (v == NULL) {
820 if (ctxt == VAR_CMDLINE && !(flags & VAR_NO_EXPORT)) {
821 /*
822 * This var would normally prevent the same name being added
823 * to VAR_GLOBAL, so delete it from there if needed.
824 * Otherwise -V name may show the wrong value.
825 */
826 Var_Delete(name, VAR_GLOBAL);
827 }
828 VarAdd(name, val, ctxt, flags);
829 } else {
830 if ((v->flags & VAR_READONLY) && !(flags & VAR_SET_READONLY)) {
831 VAR_DEBUG3("%s:%s = %s ignored (read-only)\n",
832 ctxt->name, name, val);
833 goto out;
834 }
835 Buf_Empty(&v->val);
836 if (val)
837 Buf_AddStr(&v->val, val);
838
839 VAR_DEBUG3("%s:%s = %s\n", ctxt->name, name, val);
840 if (v->flags & VAR_EXPORTED) {
841 Var_Export1(name, VAR_EXPORT_PARENT);
842 }
843 }
844 /*
845 * Any variables given on the command line are automatically exported
846 * to the environment (as per POSIX standard)
847 * Other than internals.
848 */
849 if (ctxt == VAR_CMDLINE && !(flags & VAR_NO_EXPORT) && name[0] != '.') {
850 if (v == NULL) {
851 /* we just added it */
852 v = VarFind(name, ctxt, 0);
853 }
854 if (v != NULL)
855 v->flags |= VAR_FROM_CMD;
856 /*
857 * If requested, don't export these in the environment
858 * individually. We still put them in MAKEOVERRIDES so
859 * that the command-line settings continue to override
860 * Makefile settings.
861 */
862 if (!opts.varNoExportEnv)
863 setenv(name, val ? val : "", 1);
864
865 Var_Append(MAKEOVERRIDES, name, VAR_GLOBAL);
866 }
867 if (name[0] == '.' && strcmp(name, MAKE_SAVE_DOLLARS) == 0)
868 save_dollars = s2Boolean(val, save_dollars);
869
870 out:
871 free(name_freeIt);
872 if (v != NULL)
873 VarFreeEnv(v, TRUE);
874 }
875
876 /*-
877 *-----------------------------------------------------------------------
878 * Var_Set --
879 * Set the variable name to the value val in the given context.
880 *
881 * If the variable doesn't yet exist, it is created.
882 * Otherwise the new value overwrites and replaces the old value.
883 *
884 * Input:
885 * name name of variable to set
886 * val value to give to the variable
887 * ctxt context in which to set it
888 *
889 * Notes:
890 * The variable is searched for only in its context before being
891 * created in that context. I.e. if the context is VAR_GLOBAL,
892 * only VAR_GLOBAL->context is searched. Likewise if it is VAR_CMDLINE,
893 * only VAR_CMDLINE->context is searched. This is done to avoid the
894 * literally thousands of unnecessary strcmp's that used to be done to
895 * set, say, $(@) or $(<).
896 * If the context is VAR_GLOBAL though, we check if the variable
897 * was set in VAR_CMDLINE from the command line and skip it if so.
898 *-----------------------------------------------------------------------
899 */
900 void
901 Var_Set(const char *name, const char *val, GNode *ctxt)
902 {
903 Var_Set_with_flags(name, val, ctxt, 0);
904 }
905
906 /*-
907 *-----------------------------------------------------------------------
908 * Var_Append --
909 * The variable of the given name has the given value appended to it in
910 * the given context.
911 *
912 * If the variable doesn't exist, it is created. Otherwise the strings
913 * are concatenated, with a space in between.
914 *
915 * Input:
916 * name name of variable to modify
917 * val string to append to it
918 * ctxt context in which this should occur
919 *
920 * Notes:
921 * Only if the variable is being sought in the global context is the
922 * environment searched.
923 * XXX: Knows its calling circumstances in that if called with ctxt
924 * an actual target, it will only search that context since only
925 * a local variable could be being appended to. This is actually
926 * a big win and must be tolerated.
927 *-----------------------------------------------------------------------
928 */
929 void
930 Var_Append(const char *name, const char *val, GNode *ctxt)
931 {
932 char *name_freeIt = NULL;
933 Var *v;
934
935 assert(val != NULL);
936
937 if (strchr(name, '$') != NULL) {
938 const char *unexpanded_name = name;
939 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
940 /* TODO: handle errors */
941 name = name_freeIt;
942 if (name[0] == '\0') {
943 VAR_DEBUG2("Var_Append(\"%s\", \"%s\", ...) "
944 "name expands to empty string - ignored\n",
945 unexpanded_name, val);
946 free(name_freeIt);
947 return;
948 }
949 }
950
951 v = VarFind(name, ctxt, ctxt == VAR_GLOBAL);
952
953 if (v == NULL) {
954 Var_Set(name, val, ctxt);
955 } else if (v->flags & VAR_READONLY) {
956 VAR_DEBUG1("Ignoring append to %s since it is read-only\n", name);
957 } else if (ctxt == VAR_CMDLINE || !(v->flags & VAR_FROM_CMD)) {
958 Buf_AddByte(&v->val, ' ');
959 Buf_AddStr(&v->val, val);
960
961 VAR_DEBUG3("%s:%s = %s\n",
962 ctxt->name, name, Buf_GetAll(&v->val, NULL));
963
964 if (v->flags & VAR_FROM_ENV) {
965 HashEntry *h;
966
967 /*
968 * If the original variable came from the environment, we
969 * have to install it in the global context (we could place
970 * it in the environment, but then we should provide a way to
971 * export other variables...)
972 */
973 v->flags &= ~(unsigned)VAR_FROM_ENV;
974 h = HashTable_CreateEntry(&ctxt->context, name, NULL);
975 HashEntry_Set(h, v);
976 }
977 }
978 free(name_freeIt);
979 }
980
981 /* See if the given variable exists, in the given context or in other
982 * fallback contexts.
983 *
984 * Input:
985 * name Variable to find
986 * ctxt Context in which to start search
987 */
988 Boolean
989 Var_Exists(const char *name, GNode *ctxt)
990 {
991 char *name_freeIt = NULL;
992 Var *v;
993
994 if (strchr(name, '$') != NULL) {
995 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
996 /* TODO: handle errors */
997 name = name_freeIt;
998 }
999
1000 v = VarFind(name, ctxt, TRUE);
1001 free(name_freeIt);
1002 if (v == NULL)
1003 return FALSE;
1004
1005 (void)VarFreeEnv(v, TRUE);
1006 return TRUE;
1007 }
1008
1009 /*-
1010 *-----------------------------------------------------------------------
1011 * Var_Value --
1012 * Return the unexpanded value of the given variable in the given
1013 * context, or the usual contexts.
1014 *
1015 * Input:
1016 * name name to find
1017 * ctxt context in which to search for it
1018 *
1019 * Results:
1020 * The value if the variable exists, NULL if it doesn't.
1021 * If the returned value is not NULL, the caller must free *freeIt
1022 * as soon as the returned value is no longer needed.
1023 *-----------------------------------------------------------------------
1024 */
1025 const char *
1026 Var_Value(const char *name, GNode *ctxt, char **freeIt)
1027 {
1028 Var *v = VarFind(name, ctxt, TRUE);
1029 char *p;
1030
1031 *freeIt = NULL;
1032 if (v == NULL)
1033 return NULL;
1034
1035 p = Buf_GetAll(&v->val, NULL);
1036 if (VarFreeEnv(v, FALSE))
1037 *freeIt = p;
1038 return p;
1039 }
1040
1041
1042 /* SepBuf is a string being built from "words", interleaved with separators. */
1043 typedef struct SepBuf {
1044 Buffer buf;
1045 Boolean needSep;
1046 char sep; /* usually ' ', but see the :ts modifier */
1047 } SepBuf;
1048
1049 static void
1050 SepBuf_Init(SepBuf *buf, char sep)
1051 {
1052 Buf_Init(&buf->buf, 32 /* bytes */);
1053 buf->needSep = FALSE;
1054 buf->sep = sep;
1055 }
1056
1057 static void
1058 SepBuf_Sep(SepBuf *buf)
1059 {
1060 buf->needSep = TRUE;
1061 }
1062
1063 static void
1064 SepBuf_AddBytes(SepBuf *buf, const char *mem, size_t mem_size)
1065 {
1066 if (mem_size == 0)
1067 return;
1068 if (buf->needSep && buf->sep != '\0') {
1069 Buf_AddByte(&buf->buf, buf->sep);
1070 buf->needSep = FALSE;
1071 }
1072 Buf_AddBytes(&buf->buf, mem, mem_size);
1073 }
1074
1075 static void
1076 SepBuf_AddBytesBetween(SepBuf *buf, const char *start, const char *end)
1077 {
1078 SepBuf_AddBytes(buf, start, (size_t)(end - start));
1079 }
1080
1081 static void
1082 SepBuf_AddStr(SepBuf *buf, const char *str)
1083 {
1084 SepBuf_AddBytes(buf, str, strlen(str));
1085 }
1086
1087 static char *
1088 SepBuf_Destroy(SepBuf *buf, Boolean free_buf)
1089 {
1090 return Buf_Destroy(&buf->buf, free_buf);
1091 }
1092
1093
1094 /* This callback for ModifyWords gets a single word from an expression and
1095 * typically adds a modification of this word to the buffer. It may also do
1096 * nothing or add several words. */
1097 typedef void (*ModifyWordsCallback)(const char *word, SepBuf *buf, void *data);
1098
1099
1100 /* Callback for ModifyWords to implement the :H modifier.
1101 * Add the dirname of the given word to the buffer. */
1102 static void
1103 ModifyWord_Head(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1104 {
1105 const char *slash = strrchr(word, '/');
1106 if (slash != NULL)
1107 SepBuf_AddBytesBetween(buf, word, slash);
1108 else
1109 SepBuf_AddStr(buf, ".");
1110 }
1111
1112 /* Callback for ModifyWords to implement the :T modifier.
1113 * Add the basename of the given word to the buffer. */
1114 static void
1115 ModifyWord_Tail(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1116 {
1117 const char *slash = strrchr(word, '/');
1118 const char *base = slash != NULL ? slash + 1 : word;
1119 SepBuf_AddStr(buf, base);
1120 }
1121
1122 /* Callback for ModifyWords to implement the :E modifier.
1123 * Add the filename suffix of the given word to the buffer, if it exists. */
1124 static void
1125 ModifyWord_Suffix(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1126 {
1127 const char *dot = strrchr(word, '.');
1128 if (dot != NULL)
1129 SepBuf_AddStr(buf, dot + 1);
1130 }
1131
1132 /* Callback for ModifyWords to implement the :R modifier.
1133 * Add the basename of the given word to the buffer. */
1134 static void
1135 ModifyWord_Root(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1136 {
1137 const char *dot = strrchr(word, '.');
1138 size_t len = dot != NULL ? (size_t)(dot - word) : strlen(word);
1139 SepBuf_AddBytes(buf, word, len);
1140 }
1141
1142 /* Callback for ModifyWords to implement the :M modifier.
1143 * Place the word in the buffer if it matches the given pattern. */
1144 static void
1145 ModifyWord_Match(const char *word, SepBuf *buf, void *data)
1146 {
1147 const char *pattern = data;
1148 VAR_DEBUG2("VarMatch [%s] [%s]\n", word, pattern);
1149 if (Str_Match(word, pattern))
1150 SepBuf_AddStr(buf, word);
1151 }
1152
1153 /* Callback for ModifyWords to implement the :N modifier.
1154 * Place the word in the buffer if it doesn't match the given pattern. */
1155 static void
1156 ModifyWord_NoMatch(const char *word, SepBuf *buf, void *data)
1157 {
1158 const char *pattern = data;
1159 if (!Str_Match(word, pattern))
1160 SepBuf_AddStr(buf, word);
1161 }
1162
1163 #ifdef SYSVVARSUB
1164 /*-
1165 *-----------------------------------------------------------------------
1166 * Str_SYSVMatch --
1167 * Check word against pattern for a match (% is wild),
1168 *
1169 * Input:
1170 * word Word to examine
1171 * pattern Pattern to examine against
1172 *
1173 * Results:
1174 * Returns the start of the match, or NULL.
1175 * *match_len returns the length of the match, if any.
1176 * *hasPercent returns whether the pattern contains a percent.
1177 *-----------------------------------------------------------------------
1178 */
1179 static const char *
1180 Str_SYSVMatch(const char *word, const char *pattern, size_t *match_len,
1181 Boolean *hasPercent)
1182 {
1183 const char *p = pattern;
1184 const char *w = word;
1185 const char *percent;
1186 size_t w_len;
1187 size_t p_len;
1188 const char *w_tail;
1189
1190 *hasPercent = FALSE;
1191 if (*p == '\0') { /* ${VAR:=suffix} */
1192 *match_len = strlen(w); /* Null pattern is the whole string */
1193 return w;
1194 }
1195
1196 percent = strchr(p, '%');
1197 if (percent != NULL) { /* ${VAR:...%...=...} */
1198 *hasPercent = TRUE;
1199 if (*w == '\0')
1200 return NULL; /* empty word does not match pattern */
1201
1202 /* check that the prefix matches */
1203 for (; p != percent && *w != '\0' && *w == *p; w++, p++)
1204 continue;
1205 if (p != percent)
1206 return NULL; /* No match */
1207
1208 p++; /* Skip the percent */
1209 if (*p == '\0') {
1210 /* No more pattern, return the rest of the string */
1211 *match_len = strlen(w);
1212 return w;
1213 }
1214 }
1215
1216 /* Test whether the tail matches */
1217 w_len = strlen(w);
1218 p_len = strlen(p);
1219 if (w_len < p_len)
1220 return NULL;
1221
1222 w_tail = w + w_len - p_len;
1223 if (memcmp(p, w_tail, p_len) != 0)
1224 return NULL;
1225
1226 *match_len = (size_t)(w_tail - w);
1227 return w;
1228 }
1229
1230 struct ModifyWord_SYSVSubstArgs {
1231 GNode *ctx;
1232 const char *lhs;
1233 const char *rhs;
1234 };
1235
1236 /* Callback for ModifyWords to implement the :%.from=%.to modifier. */
1237 static void
1238 ModifyWord_SYSVSubst(const char *word, SepBuf *buf, void *data)
1239 {
1240 const struct ModifyWord_SYSVSubstArgs *args = data;
1241 char *rhs_expanded;
1242 const char *rhs;
1243 const char *percent;
1244
1245 size_t match_len;
1246 Boolean lhsPercent;
1247 const char *match = Str_SYSVMatch(word, args->lhs, &match_len, &lhsPercent);
1248 if (match == NULL) {
1249 SepBuf_AddStr(buf, word);
1250 return;
1251 }
1252
1253 /* Append rhs to the buffer, substituting the first '%' with the
1254 * match, but only if the lhs had a '%' as well. */
1255
1256 (void)Var_Subst(args->rhs, args->ctx, VARE_WANTRES, &rhs_expanded);
1257 /* TODO: handle errors */
1258
1259 rhs = rhs_expanded;
1260 percent = strchr(rhs, '%');
1261
1262 if (percent != NULL && lhsPercent) {
1263 /* Copy the prefix of the replacement pattern */
1264 SepBuf_AddBytesBetween(buf, rhs, percent);
1265 rhs = percent + 1;
1266 }
1267 if (percent != NULL || !lhsPercent)
1268 SepBuf_AddBytes(buf, match, match_len);
1269
1270 /* Append the suffix of the replacement pattern */
1271 SepBuf_AddStr(buf, rhs);
1272
1273 free(rhs_expanded);
1274 }
1275 #endif
1276
1277
1278 struct ModifyWord_SubstArgs {
1279 const char *lhs;
1280 size_t lhsLen;
1281 const char *rhs;
1282 size_t rhsLen;
1283 VarPatternFlags pflags;
1284 Boolean matched;
1285 };
1286
1287 /* Callback for ModifyWords to implement the :S,from,to, modifier.
1288 * Perform a string substitution on the given word. */
1289 static void
1290 ModifyWord_Subst(const char *word, SepBuf *buf, void *data)
1291 {
1292 size_t wordLen = strlen(word);
1293 struct ModifyWord_SubstArgs *args = data;
1294 const char *match;
1295
1296 if ((args->pflags & VARP_SUB_ONE) && args->matched)
1297 goto nosub;
1298
1299 if (args->pflags & VARP_ANCHOR_START) {
1300 if (wordLen < args->lhsLen ||
1301 memcmp(word, args->lhs, args->lhsLen) != 0)
1302 goto nosub;
1303
1304 if (args->pflags & VARP_ANCHOR_END) {
1305 if (wordLen != args->lhsLen)
1306 goto nosub;
1307
1308 /* :S,^whole$,replacement, */
1309 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1310 args->matched = TRUE;
1311 } else {
1312 /* :S,^prefix,replacement, */
1313 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1314 SepBuf_AddBytes(buf, word + args->lhsLen, wordLen - args->lhsLen);
1315 args->matched = TRUE;
1316 }
1317 return;
1318 }
1319
1320 if (args->pflags & VARP_ANCHOR_END) {
1321 const char *start;
1322
1323 if (wordLen < args->lhsLen)
1324 goto nosub;
1325
1326 start = word + (wordLen - args->lhsLen);
1327 if (memcmp(start, args->lhs, args->lhsLen) != 0)
1328 goto nosub;
1329
1330 /* :S,suffix$,replacement, */
1331 SepBuf_AddBytesBetween(buf, word, start);
1332 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1333 args->matched = TRUE;
1334 return;
1335 }
1336
1337 if (args->lhs[0] == '\0')
1338 goto nosub;
1339
1340 /* unanchored case, may match more than once */
1341 while ((match = strstr(word, args->lhs)) != NULL) {
1342 SepBuf_AddBytesBetween(buf, word, match);
1343 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1344 args->matched = TRUE;
1345 wordLen -= (size_t)(match - word) + args->lhsLen;
1346 word += (size_t)(match - word) + args->lhsLen;
1347 if (wordLen == 0 || !(args->pflags & VARP_SUB_GLOBAL))
1348 break;
1349 }
1350 nosub:
1351 SepBuf_AddBytes(buf, word, wordLen);
1352 }
1353
1354 #ifndef NO_REGEX
1355 /* Print the error caused by a regcomp or regexec call. */
1356 static void
1357 VarREError(int reerr, regex_t *pat, const char *str)
1358 {
1359 size_t errlen = regerror(reerr, pat, 0, 0);
1360 char *errbuf = bmake_malloc(errlen);
1361 regerror(reerr, pat, errbuf, errlen);
1362 Error("%s: %s", str, errbuf);
1363 free(errbuf);
1364 }
1365
1366 struct ModifyWord_SubstRegexArgs {
1367 regex_t re;
1368 size_t nsub;
1369 char *replace;
1370 VarPatternFlags pflags;
1371 Boolean matched;
1372 };
1373
1374 /* Callback for ModifyWords to implement the :C/from/to/ modifier.
1375 * Perform a regex substitution on the given word. */
1376 static void
1377 ModifyWord_SubstRegex(const char *word, SepBuf *buf, void *data)
1378 {
1379 struct ModifyWord_SubstRegexArgs *args = data;
1380 int xrv;
1381 const char *wp = word;
1382 char *rp;
1383 int flags = 0;
1384 regmatch_t m[10];
1385
1386 if ((args->pflags & VARP_SUB_ONE) && args->matched)
1387 goto nosub;
1388
1389 tryagain:
1390 xrv = regexec(&args->re, wp, args->nsub, m, flags);
1391
1392 switch (xrv) {
1393 case 0:
1394 args->matched = TRUE;
1395 SepBuf_AddBytes(buf, wp, (size_t)m[0].rm_so);
1396
1397 for (rp = args->replace; *rp; rp++) {
1398 if (*rp == '\\' && (rp[1] == '&' || rp[1] == '\\')) {
1399 SepBuf_AddBytes(buf, rp + 1, 1);
1400 rp++;
1401 continue;
1402 }
1403
1404 if (*rp == '&') {
1405 SepBuf_AddBytesBetween(buf, wp + m[0].rm_so, wp + m[0].rm_eo);
1406 continue;
1407 }
1408
1409 if (*rp != '\\' || !ch_isdigit(rp[1])) {
1410 SepBuf_AddBytes(buf, rp, 1);
1411 continue;
1412 }
1413
1414 { /* \0 to \9 backreference */
1415 size_t n = (size_t)(rp[1] - '0');
1416 rp++;
1417
1418 if (n >= args->nsub) {
1419 Error("No subexpression \\%zu", n);
1420 } else if (m[n].rm_so == -1 && m[n].rm_eo == -1) {
1421 Error("No match for subexpression \\%zu", n);
1422 } else {
1423 SepBuf_AddBytesBetween(buf, wp + m[n].rm_so,
1424 wp + m[n].rm_eo);
1425 }
1426 }
1427 }
1428
1429 wp += m[0].rm_eo;
1430 if (args->pflags & VARP_SUB_GLOBAL) {
1431 flags |= REG_NOTBOL;
1432 if (m[0].rm_so == 0 && m[0].rm_eo == 0) {
1433 SepBuf_AddBytes(buf, wp, 1);
1434 wp++;
1435 }
1436 if (*wp)
1437 goto tryagain;
1438 }
1439 if (*wp) {
1440 SepBuf_AddStr(buf, wp);
1441 }
1442 break;
1443 default:
1444 VarREError(xrv, &args->re, "Unexpected regex error");
1445 /* FALLTHROUGH */
1446 case REG_NOMATCH:
1447 nosub:
1448 SepBuf_AddStr(buf, wp);
1449 break;
1450 }
1451 }
1452 #endif
1453
1454
1455 struct ModifyWord_LoopArgs {
1456 GNode *ctx;
1457 char *tvar; /* name of temporary variable */
1458 char *str; /* string to expand */
1459 VarEvalFlags eflags;
1460 };
1461
1462 /* Callback for ModifyWords to implement the :@var (at) ...@ modifier of ODE make. */
1463 static void
1464 ModifyWord_Loop(const char *word, SepBuf *buf, void *data)
1465 {
1466 const struct ModifyWord_LoopArgs *args;
1467 char *s;
1468
1469 if (word[0] == '\0')
1470 return;
1471
1472 args = data;
1473 Var_Set_with_flags(args->tvar, word, args->ctx, VAR_NO_EXPORT);
1474 (void)Var_Subst(args->str, args->ctx, args->eflags, &s);
1475 /* TODO: handle errors */
1476
1477 VAR_DEBUG4("ModifyWord_Loop: in \"%s\", replace \"%s\" with \"%s\" "
1478 "to \"%s\"\n",
1479 word, args->tvar, args->str, s);
1480
1481 if (s[0] == '\n' || Buf_EndsWith(&buf->buf, '\n'))
1482 buf->needSep = FALSE;
1483 SepBuf_AddStr(buf, s);
1484 free(s);
1485 }
1486
1487
1488 /*-
1489 * Implements the :[first..last] modifier.
1490 * This is a special case of ModifyWords since we want to be able
1491 * to scan the list backwards if first > last.
1492 */
1493 static char *
1494 VarSelectWords(char sep, Boolean oneBigWord, const char *str, int first,
1495 int last)
1496 {
1497 Words words;
1498 int len, start, end, step;
1499 int i;
1500
1501 SepBuf buf;
1502 SepBuf_Init(&buf, sep);
1503
1504 if (oneBigWord) {
1505 /* fake what Str_Words() would do if there were only one word */
1506 words.len = 1;
1507 words.words = bmake_malloc((words.len + 1) * sizeof(char *));
1508 words.freeIt = bmake_strdup(str);
1509 words.words[0] = words.freeIt;
1510 words.words[1] = NULL;
1511 } else {
1512 words = Str_Words(str, FALSE);
1513 }
1514
1515 /*
1516 * Now sanitize the given range.
1517 * If first or last are negative, convert them to the positive equivalents
1518 * (-1 gets converted to ac, -2 gets converted to (ac - 1), etc.).
1519 */
1520 len = (int)words.len;
1521 if (first < 0)
1522 first += len + 1;
1523 if (last < 0)
1524 last += len + 1;
1525
1526 /*
1527 * We avoid scanning more of the list than we need to.
1528 */
1529 if (first > last) {
1530 start = (first > len ? len : first) - 1;
1531 end = last < 1 ? 0 : last - 1;
1532 step = -1;
1533 } else {
1534 start = first < 1 ? 0 : first - 1;
1535 end = last > len ? len : last;
1536 step = 1;
1537 }
1538
1539 for (i = start; (step < 0) == (i >= end); i += step) {
1540 SepBuf_AddStr(&buf, words.words[i]);
1541 SepBuf_Sep(&buf);
1542 }
1543
1544 Words_Free(words);
1545
1546 return SepBuf_Destroy(&buf, FALSE);
1547 }
1548
1549
1550 /* Callback for ModifyWords to implement the :tA modifier.
1551 * Replace each word with the result of realpath() if successful. */
1552 static void
1553 ModifyWord_Realpath(const char *word, SepBuf *buf, void *data MAKE_ATTR_UNUSED)
1554 {
1555 struct stat st;
1556 char rbuf[MAXPATHLEN];
1557
1558 const char *rp = cached_realpath(word, rbuf);
1559 if (rp != NULL && *rp == '/' && stat(rp, &st) == 0)
1560 word = rp;
1561
1562 SepBuf_AddStr(buf, word);
1563 }
1564
1565 /*-
1566 *-----------------------------------------------------------------------
1567 * Modify each of the words of the passed string using the given function.
1568 *
1569 * Input:
1570 * str String whose words should be modified
1571 * modifyWord Function that modifies a single word
1572 * modifyWord_args Custom arguments for modifyWord
1573 *
1574 * Results:
1575 * A string of all the words modified appropriately.
1576 *-----------------------------------------------------------------------
1577 */
1578 static char *
1579 ModifyWords(GNode *ctx, char sep, Boolean oneBigWord, const char *str,
1580 ModifyWordsCallback modifyWord, void *modifyWord_args)
1581 {
1582 SepBuf result;
1583 Words words;
1584 size_t i;
1585
1586 if (oneBigWord) {
1587 SepBuf_Init(&result, sep);
1588 modifyWord(str, &result, modifyWord_args);
1589 return SepBuf_Destroy(&result, FALSE);
1590 }
1591
1592 SepBuf_Init(&result, sep);
1593
1594 words = Str_Words(str, FALSE);
1595
1596 VAR_DEBUG2("ModifyWords: split \"%s\" into %zu words\n", str, words.len);
1597
1598 for (i = 0; i < words.len; i++) {
1599 modifyWord(words.words[i], &result, modifyWord_args);
1600 if (Buf_Len(&result.buf) > 0)
1601 SepBuf_Sep(&result);
1602 }
1603
1604 Words_Free(words);
1605
1606 return SepBuf_Destroy(&result, FALSE);
1607 }
1608
1609
1610 static char *
1611 Words_JoinFree(Words words)
1612 {
1613 Buffer buf;
1614 size_t i;
1615
1616 Buf_Init(&buf, 0);
1617
1618 for (i = 0; i < words.len; i++) {
1619 if (i != 0)
1620 Buf_AddByte(&buf, ' '); /* XXX: st->sep, for consistency */
1621 Buf_AddStr(&buf, words.words[i]);
1622 }
1623
1624 Words_Free(words);
1625
1626 return Buf_Destroy(&buf, FALSE);
1627 }
1628
1629 /* Remove adjacent duplicate words. */
1630 static char *
1631 VarUniq(const char *str)
1632 {
1633 Words words = Str_Words(str, FALSE);
1634
1635 if (words.len > 1) {
1636 size_t i, j;
1637 for (j = 0, i = 1; i < words.len; i++)
1638 if (strcmp(words.words[i], words.words[j]) != 0 && (++j != i))
1639 words.words[j] = words.words[i];
1640 words.len = j + 1;
1641 }
1642
1643 return Words_JoinFree(words);
1644 }
1645
1646
1647 /* Quote shell meta-characters and space characters in the string.
1648 * If quoteDollar is set, also quote and double any '$' characters. */
1649 static char *
1650 VarQuote(const char *str, Boolean quoteDollar)
1651 {
1652 char *res;
1653 Buffer buf;
1654 Buf_Init(&buf, 0);
1655
1656 for (; *str != '\0'; str++) {
1657 if (*str == '\n') {
1658 const char *newline = Shell_GetNewline();
1659 if (newline == NULL)
1660 newline = "\\\n";
1661 Buf_AddStr(&buf, newline);
1662 continue;
1663 }
1664 if (ch_isspace(*str) || ismeta((unsigned char)*str))
1665 Buf_AddByte(&buf, '\\');
1666 Buf_AddByte(&buf, *str);
1667 if (quoteDollar && *str == '$')
1668 Buf_AddStr(&buf, "\\$");
1669 }
1670
1671 res = Buf_Destroy(&buf, FALSE);
1672 VAR_DEBUG1("QuoteMeta: [%s]\n", res);
1673 return res;
1674 }
1675
1676 /* Compute the 32-bit hash of the given string, using the MurmurHash3
1677 * algorithm. Output is encoded as 8 hex digits, in Little Endian order. */
1678 static char *
1679 VarHash(const char *str)
1680 {
1681 static const char hexdigits[16] = "0123456789abcdef";
1682 const unsigned char *ustr = (const unsigned char *)str;
1683
1684 uint32_t h = 0x971e137bU;
1685 uint32_t c1 = 0x95543787U;
1686 uint32_t c2 = 0x2ad7eb25U;
1687 size_t len2 = strlen(str);
1688
1689 char *buf;
1690 size_t i;
1691
1692 size_t len;
1693 for (len = len2; len; ) {
1694 uint32_t k = 0;
1695 switch (len) {
1696 default:
1697 k = ((uint32_t)ustr[3] << 24) |
1698 ((uint32_t)ustr[2] << 16) |
1699 ((uint32_t)ustr[1] << 8) |
1700 (uint32_t)ustr[0];
1701 len -= 4;
1702 ustr += 4;
1703 break;
1704 case 3:
1705 k |= (uint32_t)ustr[2] << 16;
1706 /* FALLTHROUGH */
1707 case 2:
1708 k |= (uint32_t)ustr[1] << 8;
1709 /* FALLTHROUGH */
1710 case 1:
1711 k |= (uint32_t)ustr[0];
1712 len = 0;
1713 }
1714 c1 = c1 * 5 + 0x7b7d159cU;
1715 c2 = c2 * 5 + 0x6bce6396U;
1716 k *= c1;
1717 k = (k << 11) ^ (k >> 21);
1718 k *= c2;
1719 h = (h << 13) ^ (h >> 19);
1720 h = h * 5 + 0x52dce729U;
1721 h ^= k;
1722 }
1723 h ^= (uint32_t)len2;
1724 h *= 0x85ebca6b;
1725 h ^= h >> 13;
1726 h *= 0xc2b2ae35;
1727 h ^= h >> 16;
1728
1729 buf = bmake_malloc(9);
1730 for (i = 0; i < 8; i++) {
1731 buf[i] = hexdigits[h & 0x0f];
1732 h >>= 4;
1733 }
1734 buf[8] = '\0';
1735 return buf;
1736 }
1737
1738 static char *
1739 VarStrftime(const char *fmt, Boolean zulu, time_t tim)
1740 {
1741 char buf[BUFSIZ];
1742
1743 if (!tim)
1744 time(&tim);
1745 if (!*fmt)
1746 fmt = "%c";
1747 strftime(buf, sizeof(buf), fmt, zulu ? gmtime(&tim) : localtime(&tim));
1748
1749 buf[sizeof(buf) - 1] = '\0';
1750 return bmake_strdup(buf);
1751 }
1752
1753 /* The ApplyModifier functions all work in the same way. They get the
1754 * current parsing position (pp) and parse the modifier from there. The
1755 * modifier typically lasts until the next ':', or a closing '}' or ')'
1756 * (taken from st->endc), or the end of the string (parse error).
1757 *
1758 * The high-level behavior of these functions is:
1759 *
1760 * 1. parse the modifier
1761 * 2. evaluate the modifier
1762 * 3. housekeeping
1763 *
1764 * Parsing the modifier
1765 *
1766 * If parsing succeeds, the parsing position *pp is updated to point to the
1767 * first character following the modifier, which typically is either ':' or
1768 * st->endc.
1769 *
1770 * If parsing fails because of a missing delimiter (as in the :S, :C or :@
1771 * modifiers), return AMR_CLEANUP.
1772 *
1773 * If parsing fails because the modifier is unknown, return AMR_UNKNOWN to
1774 * try the SysV modifier ${VAR:from=to} as fallback. This should only be
1775 * done as long as there have been no side effects from evaluating nested
1776 * variables, to avoid evaluating them more than once. In this case, the
1777 * parsing position must not be updated. (XXX: Why not? The original parsing
1778 * position is well-known in ApplyModifiers.)
1779 *
1780 * If parsing fails and the SysV modifier ${VAR:from=to} should not be used
1781 * as a fallback, either issue an error message using Error or Parse_Error
1782 * and then return AMR_CLEANUP, or return AMR_BAD for the default error
1783 * message. Both of these return values will stop processing the variable
1784 * expression. (XXX: As of 2020-08-23, evaluation of the whole string
1785 * continues nevertheless after skipping a few bytes, which essentially is
1786 * undefined behavior. Not in the sense of C, but still it's impossible to
1787 * predict what happens in the parser.)
1788 *
1789 * Evaluating the modifier
1790 *
1791 * After parsing, the modifier is evaluated. The side effects from evaluating
1792 * nested variable expressions in the modifier text often already happen
1793 * during parsing though.
1794 *
1795 * Evaluating the modifier usually takes the current value of the variable
1796 * expression from st->val, or the variable name from st->v->name and stores
1797 * the result in st->newVal.
1798 *
1799 * If evaluating fails (as of 2020-08-23), an error message is printed using
1800 * Error. This function has no side-effects, it really just prints the error
1801 * message. Processing the expression continues as if everything were ok.
1802 * XXX: This should be fixed by adding proper error handling to Var_Subst,
1803 * Var_Parse, ApplyModifiers and ModifyWords.
1804 *
1805 * Housekeeping
1806 *
1807 * Some modifiers such as :D and :U turn undefined expressions into defined
1808 * expressions (see VEF_UNDEF, VEF_DEF).
1809 *
1810 * Some modifiers need to free some memory.
1811 */
1812
1813 typedef enum VarExprFlags {
1814 /* The variable expression is based on an undefined variable. */
1815 VEF_UNDEF = 0x01,
1816 /* The variable expression started as an undefined expression, but one
1817 * of the modifiers (such as :D or :U) has turned the expression from
1818 * undefined to defined. */
1819 VEF_DEF = 0x02
1820 } VarExprFlags;
1821
1822 ENUM_FLAGS_RTTI_2(VarExprFlags,
1823 VEF_UNDEF, VEF_DEF);
1824
1825
1826 typedef struct ApplyModifiersState {
1827 const char startc; /* '\0' or '{' or '(' */
1828 const char endc; /* '\0' or '}' or ')' */
1829 Var * const v;
1830 GNode * const ctxt;
1831 const VarEvalFlags eflags;
1832
1833 char *val; /* The old value of the expression,
1834 * before applying the modifier, never NULL */
1835 char *newVal; /* The new value of the expression,
1836 * after applying the modifier, never NULL */
1837 char sep; /* Word separator in expansions
1838 * (see the :ts modifier) */
1839 Boolean oneBigWord; /* TRUE if some modifiers that otherwise split
1840 * the variable value into words, like :S and
1841 * :C, treat the variable value as a single big
1842 * word, possibly containing spaces. */
1843 VarExprFlags exprFlags;
1844 } ApplyModifiersState;
1845
1846 static void
1847 ApplyModifiersState_Define(ApplyModifiersState *st)
1848 {
1849 if (st->exprFlags & VEF_UNDEF)
1850 st->exprFlags |= VEF_DEF;
1851 }
1852
1853 typedef enum ApplyModifierResult {
1854 AMR_OK, /* Continue parsing */
1855 AMR_UNKNOWN, /* Not a match, try other modifiers as well */
1856 AMR_BAD, /* Error out with "Bad modifier" message */
1857 AMR_CLEANUP /* Error out without error message */
1858 } ApplyModifierResult;
1859
1860 /*-
1861 * Parse a part of a modifier such as the "from" and "to" in :S/from/to/
1862 * or the "var" or "replacement" in :@var@replacement+${var}@, up to and
1863 * including the next unescaped delimiter. The delimiter, as well as the
1864 * backslash or the dollar, can be escaped with a backslash.
1865 *
1866 * Return the parsed (and possibly expanded) string, or NULL if no delimiter
1867 * was found. On successful return, the parsing position pp points right
1868 * after the delimiter. The delimiter is not included in the returned
1869 * value though.
1870 */
1871 static VarParseResult
1872 ParseModifierPart(
1873 const char **pp, /* The parsing position, updated upon return */
1874 int delim, /* Parsing stops at this delimiter */
1875 VarEvalFlags eflags, /* Flags for evaluating nested variables;
1876 * if VARE_WANTRES is not set, the text is
1877 * only parsed */
1878 ApplyModifiersState *st,
1879 char **out_part,
1880 size_t *out_length, /* Optionally stores the length of the returned
1881 * string, just to save another strlen call. */
1882 VarPatternFlags *out_pflags,/* For the first part of the :S modifier,
1883 * sets the VARP_ANCHOR_END flag if the last
1884 * character of the pattern is a $. */
1885 struct ModifyWord_SubstArgs *subst
1886 /* For the second part of the :S modifier,
1887 * allow ampersands to be escaped and replace
1888 * unescaped ampersands with subst->lhs. */
1889 ) {
1890 Buffer buf;
1891 const char *p;
1892
1893 Buf_Init(&buf, 0);
1894
1895 /*
1896 * Skim through until the matching delimiter is found;
1897 * pick up variable substitutions on the way. Also allow
1898 * backslashes to quote the delimiter, $, and \, but don't
1899 * touch other backslashes.
1900 */
1901 p = *pp;
1902 while (*p != '\0' && *p != delim) {
1903 const char *varstart;
1904
1905 Boolean is_escaped = p[0] == '\\' && (
1906 p[1] == delim || p[1] == '\\' || p[1] == '$' ||
1907 (p[1] == '&' && subst != NULL));
1908 if (is_escaped) {
1909 Buf_AddByte(&buf, p[1]);
1910 p += 2;
1911 continue;
1912 }
1913
1914 if (*p != '$') { /* Unescaped, simple text */
1915 if (subst != NULL && *p == '&')
1916 Buf_AddBytes(&buf, subst->lhs, subst->lhsLen);
1917 else
1918 Buf_AddByte(&buf, *p);
1919 p++;
1920 continue;
1921 }
1922
1923 if (p[1] == delim) { /* Unescaped $ at end of pattern */
1924 if (out_pflags != NULL)
1925 *out_pflags |= VARP_ANCHOR_END;
1926 else
1927 Buf_AddByte(&buf, *p);
1928 p++;
1929 continue;
1930 }
1931
1932 if (eflags & VARE_WANTRES) { /* Nested variable, evaluated */
1933 const char *nested_p = p;
1934 const char *nested_val;
1935 void *nested_val_freeIt;
1936 VarEvalFlags nested_eflags = eflags & ~(unsigned)VARE_ASSIGN;
1937
1938 (void)Var_Parse(&nested_p, st->ctxt, nested_eflags,
1939 &nested_val, &nested_val_freeIt);
1940 /* TODO: handle errors */
1941 Buf_AddStr(&buf, nested_val);
1942 free(nested_val_freeIt);
1943 p += nested_p - p;
1944 continue;
1945 }
1946
1947 /* XXX: This whole block is very similar to Var_Parse without
1948 * VARE_WANTRES. There may be subtle edge cases though that are
1949 * not yet covered in the unit tests and that are parsed differently,
1950 * depending on whether they are evaluated or not.
1951 *
1952 * This subtle difference is not documented in the manual page,
1953 * neither is the difference between parsing :D and :M documented.
1954 * No code should ever depend on these details, but who knows. */
1955
1956 varstart = p; /* Nested variable, only parsed */
1957 if (p[1] == '(' || p[1] == '{') {
1958 /*
1959 * Find the end of this variable reference
1960 * and suck it in without further ado.
1961 * It will be interpreted later.
1962 */
1963 char have = p[1];
1964 int want = have == '(' ? ')' : '}';
1965 int depth = 1;
1966
1967 for (p += 2; *p != '\0' && depth > 0; p++) {
1968 if (p[-1] != '\\') {
1969 if (*p == have)
1970 depth++;
1971 if (*p == want)
1972 depth--;
1973 }
1974 }
1975 Buf_AddBytesBetween(&buf, varstart, p);
1976 } else {
1977 Buf_AddByte(&buf, *varstart);
1978 p++;
1979 }
1980 }
1981
1982 if (*p != delim) {
1983 *pp = p;
1984 Error("Unfinished modifier for %s ('%c' missing)", st->v->name, delim);
1985 *out_part = NULL;
1986 return VPR_PARSE_MSG;
1987 }
1988
1989 *pp = ++p;
1990 if (out_length != NULL)
1991 *out_length = Buf_Len(&buf);
1992
1993 *out_part = Buf_Destroy(&buf, FALSE);
1994 VAR_DEBUG1("Modifier part: \"%s\"\n", *out_part);
1995 return VPR_OK;
1996 }
1997
1998 /* Test whether mod starts with modname, followed by a delimiter. */
1999 static Boolean
2000 ModMatch(const char *mod, const char *modname, char endc)
2001 {
2002 size_t n = strlen(modname);
2003 return strncmp(mod, modname, n) == 0 &&
2004 (mod[n] == endc || mod[n] == ':');
2005 }
2006
2007 /* Test whether mod starts with modname, followed by a delimiter or '='. */
2008 static inline Boolean
2009 ModMatchEq(const char *mod, const char *modname, char endc)
2010 {
2011 size_t n = strlen(modname);
2012 return strncmp(mod, modname, n) == 0 &&
2013 (mod[n] == endc || mod[n] == ':' || mod[n] == '=');
2014 }
2015
2016 /* :@var (at) ...${var}...@ */
2017 static ApplyModifierResult
2018 ApplyModifier_Loop(const char **pp, ApplyModifiersState *st)
2019 {
2020 struct ModifyWord_LoopArgs args;
2021 char prev_sep;
2022 VarEvalFlags eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2023 VarParseResult res;
2024
2025 args.ctx = st->ctxt;
2026
2027 (*pp)++; /* Skip the first '@' */
2028 res = ParseModifierPart(pp, '@', eflags, st,
2029 &args.tvar, NULL, NULL, NULL);
2030 if (res != VPR_OK)
2031 return AMR_CLEANUP;
2032 if (DEBUG(LINT) && strchr(args.tvar, '$') != NULL) {
2033 Parse_Error(PARSE_FATAL,
2034 "In the :@ modifier of \"%s\", the variable name \"%s\" "
2035 "must not contain a dollar.",
2036 st->v->name, args.tvar);
2037 return AMR_CLEANUP;
2038 }
2039
2040 res = ParseModifierPart(pp, '@', eflags, st,
2041 &args.str, NULL, NULL, NULL);
2042 if (res != VPR_OK)
2043 return AMR_CLEANUP;
2044
2045 args.eflags = st->eflags & (VARE_UNDEFERR | VARE_WANTRES);
2046 prev_sep = st->sep;
2047 st->sep = ' '; /* XXX: should be st->sep for consistency */
2048 st->newVal = ModifyWords(st->ctxt, st->sep, st->oneBigWord, st->val,
2049 ModifyWord_Loop, &args);
2050 st->sep = prev_sep;
2051 Var_Delete(args.tvar, st->ctxt);
2052 free(args.tvar);
2053 free(args.str);
2054 return AMR_OK;
2055 }
2056
2057 /* :Ddefined or :Uundefined */
2058 static ApplyModifierResult
2059 ApplyModifier_Defined(const char **pp, ApplyModifiersState *st)
2060 {
2061 Buffer buf;
2062 const char *p;
2063
2064 VarEvalFlags eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2065 if (st->eflags & VARE_WANTRES) {
2066 if ((**pp == 'D') == !(st->exprFlags & VEF_UNDEF))
2067 eflags |= VARE_WANTRES;
2068 }
2069
2070 Buf_Init(&buf, 0);
2071 p = *pp + 1;
2072 while (*p != st->endc && *p != ':' && *p != '\0') {
2073
2074 /* Escaped delimiter or other special character */
2075 if (*p == '\\') {
2076 char c = p[1];
2077 if (c == st->endc || c == ':' || c == '$' || c == '\\') {
2078 Buf_AddByte(&buf, c);
2079 p += 2;
2080 continue;
2081 }
2082 }
2083
2084 /* Nested variable expression */
2085 if (*p == '$') {
2086 const char *nested_val;
2087 void *nested_val_freeIt;
2088
2089 (void)Var_Parse(&p, st->ctxt, eflags,
2090 &nested_val, &nested_val_freeIt);
2091 /* TODO: handle errors */
2092 Buf_AddStr(&buf, nested_val);
2093 free(nested_val_freeIt);
2094 continue;
2095 }
2096
2097 /* Ordinary text */
2098 Buf_AddByte(&buf, *p);
2099 p++;
2100 }
2101 *pp = p;
2102
2103 ApplyModifiersState_Define(st);
2104
2105 if (eflags & VARE_WANTRES) {
2106 st->newVal = Buf_Destroy(&buf, FALSE);
2107 } else {
2108 st->newVal = st->val;
2109 Buf_Destroy(&buf, TRUE);
2110 }
2111 return AMR_OK;
2112 }
2113
2114 /* :L */
2115 static ApplyModifierResult
2116 ApplyModifier_Literal(const char **pp, ApplyModifiersState *st)
2117 {
2118 ApplyModifiersState_Define(st);
2119 st->newVal = bmake_strdup(st->v->name);
2120 (*pp)++;
2121 return AMR_OK;
2122 }
2123
2124 /* :gmtime */
2125 static ApplyModifierResult
2126 ApplyModifier_Gmtime(const char **pp, ApplyModifiersState *st)
2127 {
2128 time_t utc;
2129
2130 const char *mod = *pp;
2131 if (!ModMatchEq(mod, "gmtime", st->endc))
2132 return AMR_UNKNOWN;
2133
2134 if (mod[6] == '=') {
2135 char *ep;
2136 utc = (time_t)strtoul(mod + 7, &ep, 10);
2137 *pp = ep;
2138 } else {
2139 utc = 0;
2140 *pp = mod + 6;
2141 }
2142 st->newVal = VarStrftime(st->val, TRUE, utc);
2143 return AMR_OK;
2144 }
2145
2146 /* :localtime */
2147 static ApplyModifierResult
2148 ApplyModifier_Localtime(const char **pp, ApplyModifiersState *st)
2149 {
2150 time_t utc;
2151
2152 const char *mod = *pp;
2153 if (!ModMatchEq(mod, "localtime", st->endc))
2154 return AMR_UNKNOWN;
2155
2156 if (mod[9] == '=') {
2157 char *ep;
2158 utc = (time_t)strtoul(mod + 10, &ep, 10);
2159 *pp = ep;
2160 } else {
2161 utc = 0;
2162 *pp = mod + 9;
2163 }
2164 st->newVal = VarStrftime(st->val, FALSE, utc);
2165 return AMR_OK;
2166 }
2167
2168 /* :hash */
2169 static ApplyModifierResult
2170 ApplyModifier_Hash(const char **pp, ApplyModifiersState *st)
2171 {
2172 if (!ModMatch(*pp, "hash", st->endc))
2173 return AMR_UNKNOWN;
2174
2175 st->newVal = VarHash(st->val);
2176 *pp += 4;
2177 return AMR_OK;
2178 }
2179
2180 /* :P */
2181 static ApplyModifierResult
2182 ApplyModifier_Path(const char **pp, ApplyModifiersState *st)
2183 {
2184 GNode *gn;
2185 char *path;
2186
2187 ApplyModifiersState_Define(st);
2188
2189 gn = Targ_FindNode(st->v->name);
2190 if (gn == NULL || gn->type & OP_NOPATH) {
2191 path = NULL;
2192 } else if (gn->path != NULL) {
2193 path = bmake_strdup(gn->path);
2194 } else {
2195 SearchPath *searchPath = Suff_FindPath(gn);
2196 path = Dir_FindFile(st->v->name, searchPath);
2197 }
2198 if (path == NULL)
2199 path = bmake_strdup(st->v->name);
2200 st->newVal = path;
2201
2202 (*pp)++;
2203 return AMR_OK;
2204 }
2205
2206 /* :!cmd! */
2207 static ApplyModifierResult
2208 ApplyModifier_ShellCommand(const char **pp, ApplyModifiersState *st)
2209 {
2210 char *cmd;
2211 const char *errfmt;
2212 VarParseResult res;
2213
2214 (*pp)++;
2215 res = ParseModifierPart(pp, '!', st->eflags, st,
2216 &cmd, NULL, NULL, NULL);
2217 if (res != VPR_OK)
2218 return AMR_CLEANUP;
2219
2220 errfmt = NULL;
2221 if (st->eflags & VARE_WANTRES)
2222 st->newVal = Cmd_Exec(cmd, &errfmt);
2223 else
2224 st->newVal = emptyString;
2225 free(cmd);
2226
2227 if (errfmt != NULL)
2228 Error(errfmt, st->val); /* XXX: why still return AMR_OK? */
2229
2230 ApplyModifiersState_Define(st);
2231 return AMR_OK;
2232 }
2233
2234 /* The :range modifier generates an integer sequence as long as the words.
2235 * The :range=7 modifier generates an integer sequence from 1 to 7. */
2236 static ApplyModifierResult
2237 ApplyModifier_Range(const char **pp, ApplyModifiersState *st)
2238 {
2239 size_t n;
2240 Buffer buf;
2241 size_t i;
2242
2243 const char *mod = *pp;
2244 if (!ModMatchEq(mod, "range", st->endc))
2245 return AMR_UNKNOWN;
2246
2247 if (mod[5] == '=') {
2248 char *ep;
2249 n = (size_t)strtoul(mod + 6, &ep, 10);
2250 *pp = ep;
2251 } else {
2252 n = 0;
2253 *pp = mod + 5;
2254 }
2255
2256 if (n == 0) {
2257 Words words = Str_Words(st->val, FALSE);
2258 n = words.len;
2259 Words_Free(words);
2260 }
2261
2262 Buf_Init(&buf, 0);
2263
2264 for (i = 0; i < n; i++) {
2265 if (i != 0)
2266 Buf_AddByte(&buf, ' '); /* XXX: st->sep, for consistency */
2267 Buf_AddInt(&buf, 1 + (int)i);
2268 }
2269
2270 st->newVal = Buf_Destroy(&buf, FALSE);
2271 return AMR_OK;
2272 }
2273
2274 /* :Mpattern or :Npattern */
2275 static ApplyModifierResult
2276 ApplyModifier_Match(const char **pp, ApplyModifiersState *st)
2277 {
2278 const char *mod = *pp;
2279 Boolean copy = FALSE; /* pattern should be, or has been, copied */
2280 Boolean needSubst = FALSE;
2281 const char *endpat;
2282 char *pattern;
2283 ModifyWordsCallback callback;
2284
2285 /*
2286 * In the loop below, ignore ':' unless we are at (or back to) the
2287 * original brace level.
2288 * XXX This will likely not work right if $() and ${} are intermixed.
2289 */
2290 int nest = 0;
2291 const char *p;
2292 for (p = mod + 1; *p != '\0' && !(*p == ':' && nest == 0); p++) {
2293 if (*p == '\\' &&
2294 (p[1] == ':' || p[1] == st->endc || p[1] == st->startc)) {
2295 if (!needSubst)
2296 copy = TRUE;
2297 p++;
2298 continue;
2299 }
2300 if (*p == '$')
2301 needSubst = TRUE;
2302 if (*p == '(' || *p == '{')
2303 nest++;
2304 if (*p == ')' || *p == '}') {
2305 nest--;
2306 if (nest < 0)
2307 break;
2308 }
2309 }
2310 *pp = p;
2311 endpat = p;
2312
2313 if (copy) {
2314 char *dst;
2315 const char *src;
2316
2317 /* Compress the \:'s out of the pattern. */
2318 pattern = bmake_malloc((size_t)(endpat - (mod + 1)) + 1);
2319 dst = pattern;
2320 src = mod + 1;
2321 for (; src < endpat; src++, dst++) {
2322 if (src[0] == '\\' && src + 1 < endpat &&
2323 /* XXX: st->startc is missing here; see above */
2324 (src[1] == ':' || src[1] == st->endc))
2325 src++;
2326 *dst = *src;
2327 }
2328 *dst = '\0';
2329 endpat = dst;
2330 } else {
2331 pattern = bmake_strsedup(mod + 1, endpat);
2332 }
2333
2334 if (needSubst) {
2335 /* pattern contains embedded '$', so use Var_Subst to expand it. */
2336 char *old_pattern = pattern;
2337 (void)Var_Subst(pattern, st->ctxt, st->eflags, &pattern);
2338 /* TODO: handle errors */
2339 free(old_pattern);
2340 }
2341
2342 VAR_DEBUG3("Pattern[%s] for [%s] is [%s]\n", st->v->name, st->val, pattern);
2343
2344 callback = mod[0] == 'M' ? ModifyWord_Match : ModifyWord_NoMatch;
2345 st->newVal = ModifyWords(st->ctxt, st->sep, st->oneBigWord, st->val,
2346 callback, pattern);
2347 free(pattern);
2348 return AMR_OK;
2349 }
2350
2351 /* :S,from,to, */
2352 static ApplyModifierResult
2353 ApplyModifier_Subst(const char **pp, ApplyModifiersState *st)
2354 {
2355 struct ModifyWord_SubstArgs args;
2356 char *lhs, *rhs;
2357 Boolean oneBigWord;
2358 VarParseResult res;
2359
2360 char delim = (*pp)[1];
2361 if (delim == '\0') {
2362 Error("Missing delimiter for :S modifier");
2363 (*pp)++;
2364 return AMR_CLEANUP;
2365 }
2366
2367 *pp += 2;
2368
2369 args.pflags = 0;
2370 args.matched = FALSE;
2371
2372 /*
2373 * If pattern begins with '^', it is anchored to the
2374 * start of the word -- skip over it and flag pattern.
2375 */
2376 if (**pp == '^') {
2377 args.pflags |= VARP_ANCHOR_START;
2378 (*pp)++;
2379 }
2380
2381 res = ParseModifierPart(pp, delim, st->eflags, st,
2382 &lhs, &args.lhsLen, &args.pflags, NULL);
2383 if (res != VPR_OK)
2384 return AMR_CLEANUP;
2385 args.lhs = lhs;
2386
2387 res = ParseModifierPart(pp, delim, st->eflags, st,
2388 &rhs, &args.rhsLen, NULL, &args);
2389 if (res != VPR_OK)
2390 return AMR_CLEANUP;
2391 args.rhs = rhs;
2392
2393 oneBigWord = st->oneBigWord;
2394 for (;; (*pp)++) {
2395 switch (**pp) {
2396 case 'g':
2397 args.pflags |= VARP_SUB_GLOBAL;
2398 continue;
2399 case '1':
2400 args.pflags |= VARP_SUB_ONE;
2401 continue;
2402 case 'W':
2403 oneBigWord = TRUE;
2404 continue;
2405 }
2406 break;
2407 }
2408
2409 st->newVal = ModifyWords(st->ctxt, st->sep, oneBigWord, st->val,
2410 ModifyWord_Subst, &args);
2411
2412 free(lhs);
2413 free(rhs);
2414 return AMR_OK;
2415 }
2416
2417 #ifndef NO_REGEX
2418
2419 /* :C,from,to, */
2420 static ApplyModifierResult
2421 ApplyModifier_Regex(const char **pp, ApplyModifiersState *st)
2422 {
2423 char *re;
2424 struct ModifyWord_SubstRegexArgs args;
2425 Boolean oneBigWord;
2426 int error;
2427 VarParseResult res;
2428
2429 char delim = (*pp)[1];
2430 if (delim == '\0') {
2431 Error("Missing delimiter for :C modifier");
2432 (*pp)++;
2433 return AMR_CLEANUP;
2434 }
2435
2436 *pp += 2;
2437
2438 res = ParseModifierPart(pp, delim, st->eflags, st,
2439 &re, NULL, NULL, NULL);
2440 if (res != VPR_OK)
2441 return AMR_CLEANUP;
2442
2443 res = ParseModifierPart(pp, delim, st->eflags, st,
2444 &args.replace, NULL, NULL, NULL);
2445 if (args.replace == NULL) {
2446 free(re);
2447 return AMR_CLEANUP;
2448 }
2449
2450 args.pflags = 0;
2451 args.matched = FALSE;
2452 oneBigWord = st->oneBigWord;
2453 for (;; (*pp)++) {
2454 switch (**pp) {
2455 case 'g':
2456 args.pflags |= VARP_SUB_GLOBAL;
2457 continue;
2458 case '1':
2459 args.pflags |= VARP_SUB_ONE;
2460 continue;
2461 case 'W':
2462 oneBigWord = TRUE;
2463 continue;
2464 }
2465 break;
2466 }
2467
2468 error = regcomp(&args.re, re, REG_EXTENDED);
2469 free(re);
2470 if (error) {
2471 VarREError(error, &args.re, "Regex compilation error");
2472 free(args.replace);
2473 return AMR_CLEANUP;
2474 }
2475
2476 args.nsub = args.re.re_nsub + 1;
2477 if (args.nsub > 10)
2478 args.nsub = 10;
2479 st->newVal = ModifyWords(st->ctxt, st->sep, oneBigWord, st->val,
2480 ModifyWord_SubstRegex, &args);
2481 regfree(&args.re);
2482 free(args.replace);
2483 return AMR_OK;
2484 }
2485 #endif
2486
2487 /* :Q, :q */
2488 static ApplyModifierResult
2489 ApplyModifier_Quote(const char **pp, ApplyModifiersState *st)
2490 {
2491 if ((*pp)[1] == st->endc || (*pp)[1] == ':') {
2492 st->newVal = VarQuote(st->val, **pp == 'q');
2493 (*pp)++;
2494 return AMR_OK;
2495 } else
2496 return AMR_UNKNOWN;
2497 }
2498
2499 static void
2500 ModifyWord_Copy(const char *word, SepBuf *buf, void *data MAKE_ATTR_UNUSED)
2501 {
2502 SepBuf_AddStr(buf, word);
2503 }
2504
2505 /* :ts<separator> */
2506 static ApplyModifierResult
2507 ApplyModifier_ToSep(const char **pp, ApplyModifiersState *st)
2508 {
2509 const char *sep = *pp + 2;
2510
2511 /* ":ts<any><endc>" or ":ts<any>:" */
2512 if (sep[0] != st->endc && (sep[1] == st->endc || sep[1] == ':')) {
2513 st->sep = sep[0];
2514 *pp = sep + 1;
2515 goto ok;
2516 }
2517
2518 /* ":ts<endc>" or ":ts:" */
2519 if (sep[0] == st->endc || sep[0] == ':') {
2520 st->sep = '\0'; /* no separator */
2521 *pp = sep;
2522 goto ok;
2523 }
2524
2525 /* ":ts<unrecognised><unrecognised>". */
2526 if (sep[0] != '\\') {
2527 (*pp)++; /* just for backwards compatibility */
2528 return AMR_BAD;
2529 }
2530
2531 /* ":ts\n" */
2532 if (sep[1] == 'n') {
2533 st->sep = '\n';
2534 *pp = sep + 2;
2535 goto ok;
2536 }
2537
2538 /* ":ts\t" */
2539 if (sep[1] == 't') {
2540 st->sep = '\t';
2541 *pp = sep + 2;
2542 goto ok;
2543 }
2544
2545 /* ":ts\x40" or ":ts\100" */
2546 {
2547 const char *numStart = sep + 1;
2548 int base = 8; /* assume octal */
2549 char *end;
2550
2551 if (sep[1] == 'x') {
2552 base = 16;
2553 numStart++;
2554 } else if (!ch_isdigit(sep[1])) {
2555 (*pp)++; /* just for backwards compatibility */
2556 return AMR_BAD; /* ":ts<backslash><unrecognised>". */
2557 }
2558
2559 st->sep = (char)strtoul(numStart, &end, base);
2560 if (*end != ':' && *end != st->endc) {
2561 (*pp)++; /* just for backwards compatibility */
2562 return AMR_BAD;
2563 }
2564 *pp = end;
2565 }
2566
2567 ok:
2568 st->newVal = ModifyWords(st->ctxt, st->sep, st->oneBigWord, st->val,
2569 ModifyWord_Copy, NULL);
2570 return AMR_OK;
2571 }
2572
2573 /* :tA, :tu, :tl, :ts<separator>, etc. */
2574 static ApplyModifierResult
2575 ApplyModifier_To(const char **pp, ApplyModifiersState *st)
2576 {
2577 const char *mod = *pp;
2578 assert(mod[0] == 't');
2579
2580 if (mod[1] == st->endc || mod[1] == ':' || mod[1] == '\0') {
2581 *pp = mod + 1;
2582 return AMR_BAD; /* Found ":t<endc>" or ":t:". */
2583 }
2584
2585 if (mod[1] == 's')
2586 return ApplyModifier_ToSep(pp, st);
2587
2588 if (mod[2] != st->endc && mod[2] != ':') {
2589 *pp = mod + 1;
2590 return AMR_BAD; /* Found ":t<unrecognised><unrecognised>". */
2591 }
2592
2593 /* Check for two-character options: ":tu", ":tl" */
2594 if (mod[1] == 'A') { /* absolute path */
2595 st->newVal = ModifyWords(st->ctxt, st->sep, st->oneBigWord, st->val,
2596 ModifyWord_Realpath, NULL);
2597 *pp = mod + 2;
2598 return AMR_OK;
2599 }
2600
2601 if (mod[1] == 'u') { /* :tu */
2602 size_t i;
2603 size_t len = strlen(st->val);
2604 st->newVal = bmake_malloc(len + 1);
2605 for (i = 0; i < len + 1; i++)
2606 st->newVal[i] = ch_toupper(st->val[i]);
2607 *pp = mod + 2;
2608 return AMR_OK;
2609 }
2610
2611 if (mod[1] == 'l') { /* :tl */
2612 size_t i;
2613 size_t len = strlen(st->val);
2614 st->newVal = bmake_malloc(len + 1);
2615 for (i = 0; i < len + 1; i++)
2616 st->newVal[i] = ch_tolower(st->val[i]);
2617 *pp = mod + 2;
2618 return AMR_OK;
2619 }
2620
2621 if (mod[1] == 'W' || mod[1] == 'w') { /* :tW, :tw */
2622 st->oneBigWord = mod[1] == 'W';
2623 st->newVal = st->val;
2624 *pp = mod + 2;
2625 return AMR_OK;
2626 }
2627
2628 /* Found ":t<unrecognised>:" or ":t<unrecognised><endc>". */
2629 *pp = mod + 1;
2630 return AMR_BAD;
2631 }
2632
2633 /* :[#], :[1], etc. */
2634 static ApplyModifierResult
2635 ApplyModifier_Words(const char **pp, ApplyModifiersState *st)
2636 {
2637 char *estr;
2638 char *ep;
2639 int first, last;
2640 VarParseResult res;
2641
2642 (*pp)++; /* skip the '[' */
2643 res = ParseModifierPart(pp, ']', st->eflags, st,
2644 &estr, NULL, NULL, NULL);
2645 if (res != VPR_OK)
2646 return AMR_CLEANUP;
2647
2648 /* now *pp points just after the closing ']' */
2649 if (**pp != ':' && **pp != st->endc)
2650 goto bad_modifier; /* Found junk after ']' */
2651
2652 if (estr[0] == '\0')
2653 goto bad_modifier; /* empty square brackets in ":[]". */
2654
2655 if (estr[0] == '#' && estr[1] == '\0') { /* Found ":[#]" */
2656 if (st->oneBigWord) {
2657 st->newVal = bmake_strdup("1");
2658 } else {
2659 Buffer buf;
2660
2661 Words words = Str_Words(st->val, FALSE);
2662 size_t ac = words.len;
2663 Words_Free(words);
2664
2665 Buf_Init(&buf, 4); /* 3 digits + '\0' is usually enough */
2666 Buf_AddInt(&buf, (int)ac);
2667 st->newVal = Buf_Destroy(&buf, FALSE);
2668 }
2669 goto ok;
2670 }
2671
2672 if (estr[0] == '*' && estr[1] == '\0') {
2673 /* Found ":[*]" */
2674 st->oneBigWord = TRUE;
2675 st->newVal = st->val;
2676 goto ok;
2677 }
2678
2679 if (estr[0] == '@' && estr[1] == '\0') {
2680 /* Found ":[@]" */
2681 st->oneBigWord = FALSE;
2682 st->newVal = st->val;
2683 goto ok;
2684 }
2685
2686 /*
2687 * We expect estr to contain a single integer for :[N], or two integers
2688 * separated by ".." for :[start..end].
2689 */
2690 first = (int)strtol(estr, &ep, 0);
2691 if (ep == estr) /* Found junk instead of a number */
2692 goto bad_modifier;
2693
2694 if (ep[0] == '\0') { /* Found only one integer in :[N] */
2695 last = first;
2696 } else if (ep[0] == '.' && ep[1] == '.' && ep[2] != '\0') {
2697 /* Expecting another integer after ".." */
2698 ep += 2;
2699 last = (int)strtol(ep, &ep, 0);
2700 if (ep[0] != '\0') /* Found junk after ".." */
2701 goto bad_modifier;
2702 } else
2703 goto bad_modifier; /* Found junk instead of ".." */
2704
2705 /*
2706 * Now seldata is properly filled in, but we still have to check for 0 as
2707 * a special case.
2708 */
2709 if (first == 0 && last == 0) {
2710 /* ":[0]" or perhaps ":[0..0]" */
2711 st->oneBigWord = TRUE;
2712 st->newVal = st->val;
2713 goto ok;
2714 }
2715
2716 /* ":[0..N]" or ":[N..0]" */
2717 if (first == 0 || last == 0)
2718 goto bad_modifier;
2719
2720 /* Normal case: select the words described by seldata. */
2721 st->newVal = VarSelectWords(st->sep, st->oneBigWord, st->val, first, last);
2722
2723 ok:
2724 free(estr);
2725 return AMR_OK;
2726
2727 bad_modifier:
2728 free(estr);
2729 return AMR_BAD;
2730 }
2731
2732 static int
2733 str_cmp_asc(const void *a, const void *b)
2734 {
2735 return strcmp(*(const char * const *)a, *(const char * const *)b);
2736 }
2737
2738 static int
2739 str_cmp_desc(const void *a, const void *b)
2740 {
2741 return strcmp(*(const char * const *)b, *(const char * const *)a);
2742 }
2743
2744 /* :O (order ascending) or :Or (order descending) or :Ox (shuffle) */
2745 static ApplyModifierResult
2746 ApplyModifier_Order(const char **pp, ApplyModifiersState *st)
2747 {
2748 const char *mod = (*pp)++; /* skip past the 'O' in any case */
2749
2750 Words words = Str_Words(st->val, FALSE);
2751
2752 if (mod[1] == st->endc || mod[1] == ':') {
2753 /* :O sorts ascending */
2754 qsort(words.words, words.len, sizeof(char *), str_cmp_asc);
2755
2756 } else if ((mod[1] == 'r' || mod[1] == 'x') &&
2757 (mod[2] == st->endc || mod[2] == ':')) {
2758 (*pp)++;
2759
2760 if (mod[1] == 'r') {
2761 /* :Or sorts descending */
2762 qsort(words.words, words.len, sizeof(char *), str_cmp_desc);
2763
2764 } else {
2765 /* :Ox shuffles
2766 *
2767 * We will use [ac..2] range for mod factors. This will produce
2768 * random numbers in [(ac-1)..0] interval, and minimal
2769 * reasonable value for mod factor is 2 (the mod 1 will produce
2770 * 0 with probability 1).
2771 */
2772 size_t i;
2773 for (i = words.len - 1; i > 0; i--) {
2774 size_t rndidx = (size_t)random() % (i + 1);
2775 char *t = words.words[i];
2776 words.words[i] = words.words[rndidx];
2777 words.words[rndidx] = t;
2778 }
2779 }
2780 } else {
2781 Words_Free(words);
2782 return AMR_BAD;
2783 }
2784
2785 st->newVal = Words_JoinFree(words);
2786 return AMR_OK;
2787 }
2788
2789 /* :? then : else */
2790 static ApplyModifierResult
2791 ApplyModifier_IfElse(const char **pp, ApplyModifiersState *st)
2792 {
2793 char *then_expr, *else_expr;
2794 VarParseResult res;
2795
2796 Boolean value = FALSE;
2797 VarEvalFlags then_eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2798 VarEvalFlags else_eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2799
2800 int cond_rc = COND_PARSE; /* anything other than COND_INVALID */
2801 if (st->eflags & VARE_WANTRES) {
2802 cond_rc = Cond_EvalCondition(st->v->name, &value);
2803 if (cond_rc != COND_INVALID && value)
2804 then_eflags |= VARE_WANTRES;
2805 if (cond_rc != COND_INVALID && !value)
2806 else_eflags |= VARE_WANTRES;
2807 }
2808
2809 (*pp)++; /* skip past the '?' */
2810 res = ParseModifierPart(pp, ':', then_eflags, st,
2811 &then_expr, NULL, NULL, NULL);
2812 if (res != VPR_OK)
2813 return AMR_CLEANUP;
2814
2815 res = ParseModifierPart(pp, st->endc, else_eflags, st,
2816 &else_expr, NULL, NULL, NULL);
2817 if (res != VPR_OK)
2818 return AMR_CLEANUP;
2819
2820 (*pp)--;
2821 if (cond_rc == COND_INVALID) {
2822 Error("Bad conditional expression `%s' in %s?%s:%s",
2823 st->v->name, st->v->name, then_expr, else_expr);
2824 return AMR_CLEANUP;
2825 }
2826
2827 if (value) {
2828 st->newVal = then_expr;
2829 free(else_expr);
2830 } else {
2831 st->newVal = else_expr;
2832 free(then_expr);
2833 }
2834 ApplyModifiersState_Define(st);
2835 return AMR_OK;
2836 }
2837
2838 /*
2839 * The ::= modifiers actually assign a value to the variable.
2840 * Their main purpose is in supporting modifiers of .for loop
2841 * iterators and other obscure uses. They always expand to
2842 * nothing. In a target rule that would otherwise expand to an
2843 * empty line they can be preceded with @: to keep make happy.
2844 * Eg.
2845 *
2846 * foo: .USE
2847 * .for i in ${.TARGET} ${.TARGET:R}.gz
2848 * @: ${t::=$i}
2849 * @echo blah ${t:T}
2850 * .endfor
2851 *
2852 * ::=<str> Assigns <str> as the new value of variable.
2853 * ::?=<str> Assigns <str> as value of variable if
2854 * it was not already set.
2855 * ::+=<str> Appends <str> to variable.
2856 * ::!=<cmd> Assigns output of <cmd> as the new value of
2857 * variable.
2858 */
2859 static ApplyModifierResult
2860 ApplyModifier_Assign(const char **pp, ApplyModifiersState *st)
2861 {
2862 GNode *v_ctxt;
2863 char delim;
2864 char *val;
2865 VarParseResult res;
2866
2867 const char *mod = *pp;
2868 const char *op = mod + 1;
2869 if (!(op[0] == '=' ||
2870 (op[1] == '=' &&
2871 (op[0] == '!' || op[0] == '+' || op[0] == '?'))))
2872 return AMR_UNKNOWN; /* "::<unrecognised>" */
2873
2874
2875 if (st->v->name[0] == '\0') {
2876 *pp = mod + 1;
2877 return AMR_BAD;
2878 }
2879
2880 v_ctxt = st->ctxt; /* context where v belongs */
2881 if (!(st->exprFlags & VEF_UNDEF) && st->ctxt != VAR_GLOBAL) {
2882 Var *gv = VarFind(st->v->name, st->ctxt, 0);
2883 if (gv == NULL)
2884 v_ctxt = VAR_GLOBAL;
2885 else
2886 VarFreeEnv(gv, TRUE);
2887 }
2888
2889 switch (op[0]) {
2890 case '+':
2891 case '?':
2892 case '!':
2893 *pp = mod + 3;
2894 break;
2895 default:
2896 *pp = mod + 2;
2897 break;
2898 }
2899
2900 delim = st->startc == '(' ? ')' : '}';
2901 res = ParseModifierPart(pp, delim, st->eflags, st, &val, NULL, NULL, NULL);
2902 if (res != VPR_OK)
2903 return AMR_CLEANUP;
2904
2905 (*pp)--;
2906
2907 if (st->eflags & VARE_WANTRES) {
2908 switch (op[0]) {
2909 case '+':
2910 Var_Append(st->v->name, val, v_ctxt);
2911 break;
2912 case '!': {
2913 const char *errfmt;
2914 char *cmd_output = Cmd_Exec(val, &errfmt);
2915 if (errfmt)
2916 Error(errfmt, val);
2917 else
2918 Var_Set(st->v->name, cmd_output, v_ctxt);
2919 free(cmd_output);
2920 break;
2921 }
2922 case '?':
2923 if (!(st->exprFlags & VEF_UNDEF))
2924 break;
2925 /* FALLTHROUGH */
2926 default:
2927 Var_Set(st->v->name, val, v_ctxt);
2928 break;
2929 }
2930 }
2931 free(val);
2932 st->newVal = emptyString;
2933 return AMR_OK;
2934 }
2935
2936 /* remember current value */
2937 static ApplyModifierResult
2938 ApplyModifier_Remember(const char **pp, ApplyModifiersState *st)
2939 {
2940 const char *mod = *pp;
2941 if (!ModMatchEq(mod, "_", st->endc))
2942 return AMR_UNKNOWN;
2943
2944 if (mod[1] == '=') {
2945 size_t n = strcspn(mod + 2, ":)}");
2946 char *name = bmake_strldup(mod + 2, n);
2947 Var_Set(name, st->val, st->ctxt);
2948 free(name);
2949 *pp = mod + 2 + n;
2950 } else {
2951 Var_Set("_", st->val, st->ctxt);
2952 *pp = mod + 1;
2953 }
2954 st->newVal = st->val;
2955 return AMR_OK;
2956 }
2957
2958 /* Apply the given function to each word of the variable value. */
2959 static ApplyModifierResult
2960 ApplyModifier_WordFunc(const char **pp, ApplyModifiersState *st,
2961 ModifyWordsCallback modifyWord)
2962 {
2963 char delim = (*pp)[1];
2964 if (delim != st->endc && delim != ':')
2965 return AMR_UNKNOWN;
2966
2967 st->newVal = ModifyWords(st->ctxt, st->sep, st->oneBigWord,
2968 st->val, modifyWord, NULL);
2969 (*pp)++;
2970 return AMR_OK;
2971 }
2972
2973 static ApplyModifierResult
2974 ApplyModifier_Unique(const char **pp, ApplyModifiersState *st)
2975 {
2976 if ((*pp)[1] == st->endc || (*pp)[1] == ':') {
2977 st->newVal = VarUniq(st->val);
2978 (*pp)++;
2979 return AMR_OK;
2980 } else
2981 return AMR_UNKNOWN;
2982 }
2983
2984 #ifdef SYSVVARSUB
2985 /* :from=to */
2986 static ApplyModifierResult
2987 ApplyModifier_SysV(const char **pp, ApplyModifiersState *st)
2988 {
2989 char *lhs, *rhs;
2990 VarParseResult res;
2991
2992 const char *mod = *pp;
2993 Boolean eqFound = FALSE;
2994
2995 /*
2996 * First we make a pass through the string trying
2997 * to verify it is a SYSV-make-style translation:
2998 * it must be: <string1>=<string2>)
2999 */
3000 int nest = 1;
3001 const char *next = mod;
3002 while (*next != '\0' && nest > 0) {
3003 if (*next == '=') {
3004 eqFound = TRUE;
3005 /* continue looking for st->endc */
3006 } else if (*next == st->endc)
3007 nest--;
3008 else if (*next == st->startc)
3009 nest++;
3010 if (nest > 0)
3011 next++;
3012 }
3013 if (*next != st->endc || !eqFound)
3014 return AMR_UNKNOWN;
3015
3016 *pp = mod;
3017 res = ParseModifierPart(pp, '=', st->eflags, st,
3018 &lhs, NULL, NULL, NULL);
3019 if (res != VPR_OK)
3020 return AMR_CLEANUP;
3021
3022 res = ParseModifierPart(pp, st->endc, st->eflags, st,
3023 &rhs, NULL, NULL, NULL);
3024 if (res != VPR_OK)
3025 return AMR_CLEANUP;
3026
3027 /*
3028 * SYSV modifications happen through the whole
3029 * string. Note the pattern is anchored at the end.
3030 */
3031 (*pp)--;
3032 if (lhs[0] == '\0' && st->val[0] == '\0') {
3033 st->newVal = st->val; /* special case */
3034 } else {
3035 struct ModifyWord_SYSVSubstArgs args = {st->ctxt, lhs, rhs};
3036 st->newVal = ModifyWords(st->ctxt, st->sep, st->oneBigWord, st->val,
3037 ModifyWord_SYSVSubst, &args);
3038 }
3039 free(lhs);
3040 free(rhs);
3041 return AMR_OK;
3042 }
3043 #endif
3044
3045 #ifdef SUNSHCMD
3046 /* :sh */
3047 static ApplyModifierResult
3048 ApplyModifier_SunShell(const char **pp, ApplyModifiersState *st)
3049 {
3050 const char *p = *pp;
3051 if (p[1] == 'h' && (p[2] == st->endc || p[2] == ':')) {
3052 if (st->eflags & VARE_WANTRES) {
3053 const char *errfmt;
3054 st->newVal = Cmd_Exec(st->val, &errfmt);
3055 if (errfmt)
3056 Error(errfmt, st->val);
3057 } else
3058 st->newVal = emptyString;
3059 *pp = p + 2;
3060 return AMR_OK;
3061 } else
3062 return AMR_UNKNOWN;
3063 }
3064 #endif
3065
3066 static void
3067 LogBeforeApply(const ApplyModifiersState *st, const char *mod, const char endc)
3068 {
3069 char eflags_str[VarEvalFlags_ToStringSize];
3070 char vflags_str[VarFlags_ToStringSize];
3071 char exprflags_str[VarExprFlags_ToStringSize];
3072 Boolean is_single_char = mod[0] != '\0' &&
3073 (mod[1] == endc || mod[1] == ':');
3074
3075 /* At this point, only the first character of the modifier can
3076 * be used since the end of the modifier is not yet known. */
3077 debug_printf("Applying ${%s:%c%s} to \"%s\" (%s, %s, %s)\n",
3078 st->v->name, mod[0], is_single_char ? "" : "...", st->val,
3079 Enum_FlagsToString(eflags_str, sizeof eflags_str,
3080 st->eflags, VarEvalFlags_ToStringSpecs),
3081 Enum_FlagsToString(vflags_str, sizeof vflags_str,
3082 st->v->flags, VarFlags_ToStringSpecs),
3083 Enum_FlagsToString(exprflags_str, sizeof exprflags_str,
3084 st->exprFlags,
3085 VarExprFlags_ToStringSpecs));
3086 }
3087
3088 static void
3089 LogAfterApply(ApplyModifiersState *st, const char *p, const char *mod)
3090 {
3091 char eflags_str[VarEvalFlags_ToStringSize];
3092 char vflags_str[VarFlags_ToStringSize];
3093 char exprflags_str[VarExprFlags_ToStringSize];
3094 const char *quot = st->newVal == var_Error ? "" : "\"";
3095 const char *newVal = st->newVal == var_Error ? "error" : st->newVal;
3096
3097 debug_printf("Result of ${%s:%.*s} is %s%s%s (%s, %s, %s)\n",
3098 st->v->name, (int)(p - mod), mod, quot, newVal, quot,
3099 Enum_FlagsToString(eflags_str, sizeof eflags_str,
3100 st->eflags, VarEvalFlags_ToStringSpecs),
3101 Enum_FlagsToString(vflags_str, sizeof vflags_str,
3102 st->v->flags, VarFlags_ToStringSpecs),
3103 Enum_FlagsToString(exprflags_str, sizeof exprflags_str,
3104 st->exprFlags,
3105 VarExprFlags_ToStringSpecs));
3106 }
3107
3108 static ApplyModifierResult
3109 ApplyModifier(const char **pp, ApplyModifiersState *st)
3110 {
3111 switch (**pp) {
3112 case ':':
3113 return ApplyModifier_Assign(pp, st);
3114 case '@':
3115 return ApplyModifier_Loop(pp, st);
3116 case '_':
3117 return ApplyModifier_Remember(pp, st);
3118 case 'D':
3119 case 'U':
3120 return ApplyModifier_Defined(pp, st);
3121 case 'L':
3122 return ApplyModifier_Literal(pp, st);
3123 case 'P':
3124 return ApplyModifier_Path(pp, st);
3125 case '!':
3126 return ApplyModifier_ShellCommand(pp, st);
3127 case '[':
3128 return ApplyModifier_Words(pp, st);
3129 case 'g':
3130 return ApplyModifier_Gmtime(pp, st);
3131 case 'h':
3132 return ApplyModifier_Hash(pp, st);
3133 case 'l':
3134 return ApplyModifier_Localtime(pp, st);
3135 case 't':
3136 return ApplyModifier_To(pp, st);
3137 case 'N':
3138 case 'M':
3139 return ApplyModifier_Match(pp, st);
3140 case 'S':
3141 return ApplyModifier_Subst(pp, st);
3142 case '?':
3143 return ApplyModifier_IfElse(pp, st);
3144 #ifndef NO_REGEX
3145 case 'C':
3146 return ApplyModifier_Regex(pp, st);
3147 #endif
3148 case 'q':
3149 case 'Q':
3150 return ApplyModifier_Quote(pp, st);
3151 case 'T':
3152 return ApplyModifier_WordFunc(pp, st, ModifyWord_Tail);
3153 case 'H':
3154 return ApplyModifier_WordFunc(pp, st, ModifyWord_Head);
3155 case 'E':
3156 return ApplyModifier_WordFunc(pp, st, ModifyWord_Suffix);
3157 case 'R':
3158 return ApplyModifier_WordFunc(pp, st, ModifyWord_Root);
3159 case 'r':
3160 return ApplyModifier_Range(pp, st);
3161 case 'O':
3162 return ApplyModifier_Order(pp, st);
3163 case 'u':
3164 return ApplyModifier_Unique(pp, st);
3165 #ifdef SUNSHCMD
3166 case 's':
3167 return ApplyModifier_SunShell(pp, st);
3168 #endif
3169 default:
3170 return AMR_UNKNOWN;
3171 }
3172 }
3173
3174 /* Apply any modifiers (such as :Mpattern or :@var@loop@ or :Q or ::=value). */
3175 static char *
3176 ApplyModifiers(
3177 const char **pp, /* the parsing position, updated upon return */
3178 char *val, /* the current value of the variable */
3179 char const startc, /* '(' or '{', or '\0' for indirect modifiers */
3180 char const endc, /* ')' or '}', or '\0' for indirect modifiers */
3181 Var * const v,
3182 VarExprFlags *exprFlags,
3183 GNode * const ctxt, /* for looking up and modifying variables */
3184 VarEvalFlags const eflags,
3185 void ** const freePtr /* free this after using the return value */
3186 ) {
3187 ApplyModifiersState st = {
3188 startc, endc, v, ctxt, eflags, val,
3189 var_Error, /* .newVal */
3190 ' ', /* .sep */
3191 FALSE, /* .oneBigWord */
3192 *exprFlags /* .exprFlags */
3193 };
3194 const char *p;
3195 const char *mod;
3196 ApplyModifierResult res;
3197
3198 assert(startc == '(' || startc == '{' || startc == '\0');
3199 assert(endc == ')' || endc == '}' || endc == '\0');
3200 assert(val != NULL);
3201
3202 p = *pp;
3203 while (*p != '\0' && *p != endc) {
3204
3205 if (*p == '$') {
3206 /*
3207 * We may have some complex modifiers in a variable.
3208 */
3209 const char *nested_p = p;
3210 void *freeIt;
3211 const char *rval;
3212 char c;
3213
3214 (void)Var_Parse(&nested_p, st.ctxt, st.eflags, &rval, &freeIt);
3215 /* TODO: handle errors */
3216
3217 /*
3218 * If we have not parsed up to st.endc or ':', we are not
3219 * interested. This means the expression ${VAR:${M_1}${M_2}}
3220 * is not accepted, but ${VAR:${M_1}:${M_2}} is.
3221 */
3222 if (rval[0] != '\0' &&
3223 (c = *nested_p) != '\0' && c != ':' && c != st.endc) {
3224 if (DEBUG(LINT))
3225 Parse_Error(PARSE_FATAL,
3226 "Missing delimiter ':' after indirect modifier \"%.*s\"",
3227 (int)(nested_p - p), p);
3228
3229 free(freeIt);
3230 /* XXX: apply_mods doesn't sound like "not interested". */
3231 /* XXX: Why is the indirect modifier parsed again by
3232 * apply_mods? If any, p should be advanced to nested_p. */
3233 goto apply_mods;
3234 }
3235
3236 VAR_DEBUG3("Indirect modifier \"%s\" from \"%.*s\"\n",
3237 rval, (int)(size_t)(nested_p - p), p);
3238
3239 p = nested_p;
3240
3241 if (rval[0] != '\0') {
3242 const char *rval_pp = rval;
3243 st.val = ApplyModifiers(&rval_pp, st.val, '\0', '\0', v,
3244 &st.exprFlags, ctxt, eflags, freePtr);
3245 if (st.val == var_Error
3246 || (st.val == varUndefined && !(st.eflags & VARE_UNDEFERR))
3247 || *rval_pp != '\0') {
3248 free(freeIt);
3249 goto out; /* error already reported */
3250 }
3251 }
3252 free(freeIt);
3253
3254 if (*p == ':')
3255 p++;
3256 else if (*p == '\0' && endc != '\0') {
3257 Error("Unclosed variable specification after complex "
3258 "modifier (expecting '%c') for %s", st.endc, st.v->name);
3259 goto out;
3260 }
3261 continue;
3262 }
3263 apply_mods:
3264 st.newVal = var_Error; /* default value, in case of errors */
3265 mod = p;
3266
3267 if (DEBUG(VAR))
3268 LogBeforeApply(&st, mod, endc);
3269
3270 res = ApplyModifier(&p, &st);
3271
3272 #ifdef SYSVVARSUB
3273 if (res == AMR_UNKNOWN) {
3274 assert(p == mod);
3275 res = ApplyModifier_SysV(&p, &st);
3276 }
3277 #endif
3278
3279 if (res == AMR_UNKNOWN) {
3280 Error("Unknown modifier '%c'", *mod);
3281 for (p++; *p != ':' && *p != st.endc && *p != '\0'; p++)
3282 continue;
3283 st.newVal = var_Error;
3284 }
3285 if (res == AMR_CLEANUP)
3286 goto cleanup;
3287 if (res == AMR_BAD)
3288 goto bad_modifier;
3289
3290 if (DEBUG(VAR))
3291 LogAfterApply(&st, p, mod);
3292
3293 if (st.newVal != st.val) {
3294 if (*freePtr) {
3295 free(st.val);
3296 *freePtr = NULL;
3297 }
3298 st.val = st.newVal;
3299 if (st.val != var_Error && st.val != varUndefined &&
3300 st.val != emptyString) {
3301 *freePtr = st.val;
3302 }
3303 }
3304 if (*p == '\0' && st.endc != '\0') {
3305 Error("Unclosed variable specification (expecting '%c') "
3306 "for \"%s\" (value \"%s\") modifier %c",
3307 st.endc, st.v->name, st.val, *mod);
3308 } else if (*p == ':') {
3309 p++;
3310 } else if (DEBUG(LINT) && *p != '\0' && *p != endc) {
3311 Parse_Error(PARSE_FATAL,
3312 "Missing delimiter ':' after modifier \"%.*s\"",
3313 (int)(p - mod), mod);
3314 }
3315 }
3316 out:
3317 *pp = p;
3318 assert(st.val != NULL); /* Use var_Error or varUndefined instead. */
3319 *exprFlags = st.exprFlags;
3320 return st.val;
3321
3322 bad_modifier:
3323 Error("Bad modifier `:%.*s' for %s",
3324 (int)strcspn(mod, ":)}"), mod, st.v->name);
3325
3326 cleanup:
3327 *pp = p;
3328 free(*freePtr);
3329 *freePtr = NULL;
3330 *exprFlags = st.exprFlags;
3331 return var_Error;
3332 }
3333
3334 static Boolean
3335 VarIsDynamic(GNode *ctxt, const char *varname, size_t namelen)
3336 {
3337 if ((namelen == 1 ||
3338 (namelen == 2 && (varname[1] == 'F' || varname[1] == 'D'))) &&
3339 (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL))
3340 {
3341 /*
3342 * If substituting a local variable in a non-local context,
3343 * assume it's for dynamic source stuff. We have to handle
3344 * this specially and return the longhand for the variable
3345 * with the dollar sign escaped so it makes it back to the
3346 * caller. Only four of the local variables are treated
3347 * specially as they are the only four that will be set
3348 * when dynamic sources are expanded.
3349 */
3350 switch (varname[0]) {
3351 case '@':
3352 case '%':
3353 case '*':
3354 case '!':
3355 return TRUE;
3356 }
3357 return FALSE;
3358 }
3359
3360 if ((namelen == 7 || namelen == 8) && varname[0] == '.' &&
3361 ch_isupper(varname[1]) && (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL))
3362 {
3363 return strcmp(varname, ".TARGET") == 0 ||
3364 strcmp(varname, ".ARCHIVE") == 0 ||
3365 strcmp(varname, ".PREFIX") == 0 ||
3366 strcmp(varname, ".MEMBER") == 0;
3367 }
3368
3369 return FALSE;
3370 }
3371
3372 static const char *
3373 UndefinedShortVarValue(char varname, const GNode *ctxt, VarEvalFlags eflags)
3374 {
3375 if (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL) {
3376 /*
3377 * If substituting a local variable in a non-local context,
3378 * assume it's for dynamic source stuff. We have to handle
3379 * this specially and return the longhand for the variable
3380 * with the dollar sign escaped so it makes it back to the
3381 * caller. Only four of the local variables are treated
3382 * specially as they are the only four that will be set
3383 * when dynamic sources are expanded.
3384 */
3385 switch (varname) {
3386 case '@':
3387 return "$(.TARGET)";
3388 case '%':
3389 return "$(.MEMBER)";
3390 case '*':
3391 return "$(.PREFIX)";
3392 case '!':
3393 return "$(.ARCHIVE)";
3394 }
3395 }
3396 return eflags & VARE_UNDEFERR ? var_Error : varUndefined;
3397 }
3398
3399 /* Parse a variable name, until the end character or a colon, whichever
3400 * comes first. */
3401 static char *
3402 ParseVarname(const char **pp, char startc, char endc,
3403 GNode *ctxt, VarEvalFlags eflags,
3404 size_t *out_varname_len)
3405 {
3406 Buffer buf;
3407 const char *p = *pp;
3408 int depth = 1;
3409
3410 Buf_Init(&buf, 0);
3411
3412 while (*p != '\0') {
3413 /* Track depth so we can spot parse errors. */
3414 if (*p == startc)
3415 depth++;
3416 if (*p == endc) {
3417 if (--depth == 0)
3418 break;
3419 }
3420 if (*p == ':' && depth == 1)
3421 break;
3422
3423 /* A variable inside a variable, expand. */
3424 if (*p == '$') {
3425 void *freeIt;
3426 const char *rval;
3427 (void)Var_Parse(&p, ctxt, eflags, &rval, &freeIt);
3428 /* TODO: handle errors */
3429 Buf_AddStr(&buf, rval);
3430 free(freeIt);
3431 } else {
3432 Buf_AddByte(&buf, *p);
3433 p++;
3434 }
3435 }
3436 *pp = p;
3437 *out_varname_len = Buf_Len(&buf);
3438 return Buf_Destroy(&buf, FALSE);
3439 }
3440
3441 static Boolean
3442 ValidShortVarname(char varname, const char *start)
3443 {
3444 switch (varname) {
3445 case '\0':
3446 case ')':
3447 case '}':
3448 case ':':
3449 case '$':
3450 break; /* and continue below */
3451 default:
3452 return TRUE;
3453 }
3454
3455 if (!DEBUG(LINT))
3456 return FALSE;
3457
3458 if (varname == '$')
3459 Parse_Error(PARSE_FATAL,
3460 "To escape a dollar, use \\$, not $$, at \"%s\"", start);
3461 else if (varname == '\0')
3462 Parse_Error(PARSE_FATAL, "Dollar followed by nothing");
3463 else
3464 Parse_Error(PARSE_FATAL,
3465 "Invalid variable name '%c', at \"%s\"", varname, start);
3466
3467 return FALSE;
3468 }
3469
3470 /*-
3471 *-----------------------------------------------------------------------
3472 * Var_Parse --
3473 * Given the start of a variable expression (such as $v, $(VAR),
3474 * ${VAR:Mpattern}), extract the variable name, possibly some
3475 * modifiers and find its value by applying the modifiers to the
3476 * original value.
3477 *
3478 * When parsing a condition in ParseEmptyArg, pp may also point to
3479 * the "y" of "empty(VARNAME:Modifiers)", which is syntactically
3480 * identical.
3481 *
3482 * Input:
3483 * str The string to parse
3484 * ctxt The context for the variable
3485 * flags VARE_UNDEFERR if undefineds are an error
3486 * VARE_WANTRES if we actually want the result
3487 * VARE_ASSIGN if we are in a := assignment
3488 * lengthPtr OUT: The length of the specification
3489 * freePtr OUT: Non-NULL if caller should free *freePtr
3490 *
3491 * Results:
3492 * Returns the value of the variable expression, never NULL.
3493 * Returns var_Error if there was a parse error and VARE_UNDEFERR was
3494 * set.
3495 * Returns varUndefined if there was an undefined variable and
3496 * VARE_UNDEFERR was not set.
3497 *
3498 * Parsing should continue at *pp.
3499 * TODO: Document the value of *pp on parse errors. It might be advanced
3500 * by 0, or +1, or the index of the parse error, or the guessed end of the
3501 * variable expression.
3502 *
3503 * If var_Error is returned, a diagnostic may or may not have been
3504 * printed. XXX: This is inconsistent.
3505 *
3506 * If varUndefined is returned, a diagnostic may or may not have been
3507 * printed. XXX: This is inconsistent.
3508 *
3509 * After using the returned value, *freePtr must be freed, preferably
3510 * using bmake_free since it is NULL in most cases.
3511 *
3512 * Side Effects:
3513 * Any effects from the modifiers, such as :!cmd! or ::=value.
3514 *-----------------------------------------------------------------------
3515 */
3516 /* coverity[+alloc : arg-*4] */
3517 VarParseResult
3518 Var_Parse(const char **pp, GNode *ctxt, VarEvalFlags eflags,
3519 const char **out_val, void **freePtr)
3520 {
3521 const char *const start = *pp;
3522 const char *p;
3523 Boolean haveModifier; /* TRUE if have modifiers for the variable */
3524 char startc; /* Starting character if variable in parens
3525 * or braces */
3526 char endc; /* Ending character if variable in parens
3527 * or braces */
3528 Boolean dynamic; /* TRUE if the variable is local and we're
3529 * expanding it in a non-local context. This
3530 * is done to support dynamic sources. The
3531 * result is just the expression, unaltered */
3532 const char *extramodifiers;
3533 Var *v;
3534 char *nstr;
3535 char eflags_str[VarEvalFlags_ToStringSize];
3536 VarExprFlags exprFlags = 0;
3537
3538 VAR_DEBUG3("%s: %s with %s\n", __func__, start,
3539 Enum_FlagsToString(eflags_str, sizeof eflags_str, eflags,
3540 VarEvalFlags_ToStringSpecs));
3541
3542 *freePtr = NULL;
3543 extramodifiers = NULL; /* extra modifiers to apply first */
3544 dynamic = FALSE;
3545
3546 /* Appease GCC, which thinks that the variable might not be
3547 * initialized. */
3548 endc = '\0';
3549
3550 startc = start[1];
3551 if (startc != '(' && startc != '{') {
3552 char name[2];
3553
3554 /*
3555 * If it's not bounded by braces of some sort, life is much simpler.
3556 * We just need to check for the first character and return the
3557 * value if it exists.
3558 */
3559
3560 if (!ValidShortVarname(startc, start)) {
3561 (*pp)++;
3562 *out_val = var_Error;
3563 return VPR_PARSE_MSG;
3564 }
3565
3566 name[0] = startc;
3567 name[1] = '\0';
3568 v = VarFind(name, ctxt, TRUE);
3569 if (v == NULL) {
3570 *pp += 2;
3571
3572 *out_val = UndefinedShortVarValue(startc, ctxt, eflags);
3573 if (DEBUG(LINT) && *out_val == var_Error) {
3574 Parse_Error(PARSE_FATAL, "Variable \"%s\" is undefined", name);
3575 return VPR_UNDEF_MSG;
3576 }
3577 return eflags & VARE_UNDEFERR ? VPR_UNDEF_SILENT : VPR_OK;
3578 } else {
3579 haveModifier = FALSE;
3580 p = start + 1;
3581 }
3582 } else {
3583 size_t namelen;
3584 char *varname;
3585
3586 endc = startc == '(' ? ')' : '}';
3587
3588 p = start + 2;
3589 varname = ParseVarname(&p, startc, endc, ctxt, eflags, &namelen);
3590
3591 if (*p == ':') {
3592 haveModifier = TRUE;
3593 } else if (*p == endc) {
3594 haveModifier = FALSE;
3595 } else {
3596 Parse_Error(PARSE_FATAL, "Unclosed variable \"%s\"", varname);
3597 *pp = p;
3598 free(varname);
3599 *out_val = var_Error;
3600 return VPR_PARSE_MSG;
3601 }
3602
3603 v = VarFind(varname, ctxt, TRUE);
3604
3605 /* At this point, p points just after the variable name,
3606 * either at ':' or at endc. */
3607
3608 /*
3609 * Check also for bogus D and F forms of local variables since we're
3610 * in a local context and the name is the right length.
3611 */
3612 if (v == NULL && ctxt != VAR_CMDLINE && ctxt != VAR_GLOBAL &&
3613 namelen == 2 && (varname[1] == 'F' || varname[1] == 'D') &&
3614 strchr("@%?*!<>", varname[0]) != NULL)
3615 {
3616 /*
3617 * Well, it's local -- go look for it.
3618 */
3619 char name[] = { varname[0], '\0' };
3620 v = VarFind(name, ctxt, 0);
3621
3622 if (v != NULL) {
3623 if (varname[1] == 'D') {
3624 extramodifiers = "H:";
3625 } else { /* F */
3626 extramodifiers = "T:";
3627 }
3628 }
3629 }
3630
3631 if (v == NULL) {
3632 dynamic = VarIsDynamic(ctxt, varname, namelen);
3633
3634 if (!haveModifier) {
3635 p++; /* skip endc */
3636 *pp = p;
3637 if (dynamic) {
3638 char *pstr = bmake_strsedup(start, p);
3639 *freePtr = pstr;
3640 free(varname);
3641 *out_val = pstr;
3642 return VPR_OK;
3643 }
3644
3645 if ((eflags & VARE_UNDEFERR) && (eflags & VARE_WANTRES) &&
3646 DEBUG(LINT))
3647 {
3648 Parse_Error(PARSE_FATAL, "Variable \"%s\" is undefined",
3649 varname);
3650 free(varname);
3651 *out_val = var_Error;
3652 return VPR_UNDEF_MSG;
3653 }
3654
3655 if (eflags & VARE_UNDEFERR) {
3656 free(varname);
3657 *out_val = var_Error;
3658 return VPR_UNDEF_SILENT;
3659 }
3660
3661 free(varname);
3662 *out_val = varUndefined;
3663 return VPR_OK;
3664 }
3665
3666 /* The variable expression is based on an undefined variable.
3667 * Nevertheless it needs a Var, for modifiers that access the
3668 * variable name, such as :L or :?.
3669 *
3670 * Most modifiers leave this expression in the "undefined" state
3671 * (VEF_UNDEF), only a few modifiers like :D, :U, :L, :P turn this
3672 * undefined expression into a defined expression (VEF_DEF).
3673 *
3674 * At the end, after applying all modifiers, if the expression
3675 * is still undefined, Var_Parse will return an empty string
3676 * instead of the actually computed value. */
3677 v = VarNew(varname, varname, "", 0);
3678 exprFlags = VEF_UNDEF;
3679 } else
3680 free(varname);
3681 }
3682
3683 if (v->flags & VAR_IN_USE) {
3684 Fatal("Variable %s is recursive.", v->name);
3685 /*NOTREACHED*/
3686 } else {
3687 v->flags |= VAR_IN_USE;
3688 }
3689
3690 /*
3691 * Before doing any modification, we have to make sure the value
3692 * has been fully expanded. If it looks like recursion might be
3693 * necessary (there's a dollar sign somewhere in the variable's value)
3694 * we just call Var_Subst to do any other substitutions that are
3695 * necessary. Note that the value returned by Var_Subst will have
3696 * been dynamically-allocated, so it will need freeing when we
3697 * return.
3698 */
3699 nstr = Buf_GetAll(&v->val, NULL);
3700 if (strchr(nstr, '$') != NULL && (eflags & VARE_WANTRES)) {
3701 VarEvalFlags nested_eflags = eflags;
3702 if (DEBUG(LINT))
3703 nested_eflags &= ~(unsigned)VARE_UNDEFERR;
3704 (void)Var_Subst(nstr, ctxt, nested_eflags, &nstr);
3705 /* TODO: handle errors */
3706 *freePtr = nstr;
3707 }
3708
3709 v->flags &= ~(unsigned)VAR_IN_USE;
3710
3711 if (haveModifier || extramodifiers != NULL) {
3712 void *extraFree;
3713
3714 extraFree = NULL;
3715 if (extramodifiers != NULL) {
3716 const char *em = extramodifiers;
3717 nstr = ApplyModifiers(&em, nstr, '(', ')',
3718 v, &exprFlags, ctxt, eflags, &extraFree);
3719 }
3720
3721 if (haveModifier) {
3722 /* Skip initial colon. */
3723 p++;
3724
3725 nstr = ApplyModifiers(&p, nstr, startc, endc,
3726 v, &exprFlags, ctxt, eflags, freePtr);
3727 free(extraFree);
3728 } else {
3729 *freePtr = extraFree;
3730 }
3731 }
3732
3733 if (*p != '\0') /* Skip past endc if possible. */
3734 p++;
3735
3736 *pp = p;
3737
3738 if (v->flags & VAR_FROM_ENV) {
3739 /* Free the environment variable now since we own it,
3740 * but don't free the variable value if it will be returned. */
3741 Boolean keepValue = nstr == Buf_GetAll(&v->val, NULL);
3742 if (keepValue)
3743 *freePtr = nstr;
3744 (void)VarFreeEnv(v, !keepValue);
3745
3746 } else if (exprFlags & VEF_UNDEF) {
3747 if (!(exprFlags & VEF_DEF)) {
3748 if (*freePtr != NULL) {
3749 free(*freePtr);
3750 *freePtr = NULL;
3751 }
3752 if (dynamic) {
3753 nstr = bmake_strsedup(start, p);
3754 *freePtr = nstr;
3755 } else {
3756 /* The expression is still undefined, therefore discard the
3757 * actual value and return an error marker instead. */
3758 nstr = (eflags & VARE_UNDEFERR) ? var_Error : varUndefined;
3759 }
3760 }
3761 if (nstr != Buf_GetAll(&v->val, NULL))
3762 Buf_Destroy(&v->val, TRUE);
3763 free(v->name_freeIt);
3764 free(v);
3765 }
3766 *out_val = nstr;
3767 return VPR_UNKNOWN;
3768 }
3769
3770 /* Substitute for all variables in the given string in the given context.
3771 *
3772 * If eflags & VARE_UNDEFERR, Parse_Error will be called when an undefined
3773 * variable is encountered.
3774 *
3775 * If eflags & VARE_WANTRES, any effects from the modifiers, such as ::=,
3776 * :sh or !cmd! take place.
3777 *
3778 * Input:
3779 * str the string which to substitute
3780 * ctxt the context wherein to find variables
3781 * eflags VARE_UNDEFERR if undefineds are an error
3782 * VARE_WANTRES if we actually want the result
3783 * VARE_ASSIGN if we are in a := assignment
3784 *
3785 * Results:
3786 * The resulting string.
3787 */
3788 VarParseResult
3789 Var_Subst(const char *str, GNode *ctxt, VarEvalFlags eflags, char **out_res)
3790 {
3791 const char *p = str;
3792 Buffer buf; /* Buffer for forming things */
3793
3794 /* Set true if an error has already been reported,
3795 * to prevent a plethora of messages when recursing */
3796 static Boolean errorReported;
3797
3798 Buf_Init(&buf, 0);
3799 errorReported = FALSE;
3800
3801 while (*p != '\0') {
3802 if (p[0] == '$' && p[1] == '$') {
3803 /* A dollar sign may be escaped with another dollar sign. */
3804 if (save_dollars && (eflags & VARE_ASSIGN))
3805 Buf_AddByte(&buf, '$');
3806 Buf_AddByte(&buf, '$');
3807 p += 2;
3808 } else if (*p != '$') {
3809 /*
3810 * Skip as many characters as possible -- either to the end of
3811 * the string or to the next dollar sign (variable expression).
3812 */
3813 const char *plainStart = p;
3814
3815 for (p++; *p != '$' && *p != '\0'; p++)
3816 continue;
3817 Buf_AddBytesBetween(&buf, plainStart, p);
3818 } else {
3819 const char *nested_p = p;
3820 void *freeIt;
3821 const char *val;
3822 (void)Var_Parse(&nested_p, ctxt, eflags, &val, &freeIt);
3823 /* TODO: handle errors */
3824
3825 if (val == var_Error || val == varUndefined) {
3826 /*
3827 * If performing old-time variable substitution, skip over
3828 * the variable and continue with the substitution. Otherwise,
3829 * store the dollar sign and advance str so we continue with
3830 * the string...
3831 */
3832 if (oldVars) {
3833 p = nested_p;
3834 } else if ((eflags & VARE_UNDEFERR) || val == var_Error) {
3835 /*
3836 * If variable is undefined, complain and skip the
3837 * variable. The complaint will stop us from doing anything
3838 * when the file is parsed.
3839 */
3840 if (!errorReported) {
3841 Parse_Error(PARSE_FATAL, "Undefined variable \"%.*s\"",
3842 (int)(size_t)(nested_p - p), p);
3843 }
3844 p = nested_p;
3845 errorReported = TRUE;
3846 } else {
3847 /* Copy the initial '$' of the undefined expression,
3848 * thereby deferring expansion of the expression, but
3849 * expand nested expressions if already possible.
3850 * See unit-tests/varparse-undef-partial.mk. */
3851 Buf_AddByte(&buf, *p);
3852 p++;
3853 }
3854 } else {
3855 p = nested_p;
3856 Buf_AddStr(&buf, val);
3857 }
3858 free(freeIt);
3859 freeIt = NULL;
3860 }
3861 }
3862
3863 *out_res = Buf_DestroyCompact(&buf);
3864 return VPR_OK;
3865 }
3866
3867 /* Initialize the variables module. */
3868 void
3869 Var_Init(void)
3870 {
3871 VAR_INTERNAL = Targ_NewGN("Internal");
3872 VAR_GLOBAL = Targ_NewGN("Global");
3873 VAR_CMDLINE = Targ_NewGN("Command");
3874 }
3875
3876 /* Clean up the variables module. */
3877 void
3878 Var_End(void)
3879 {
3880 Var_Stats();
3881 }
3882
3883 void
3884 Var_Stats(void)
3885 {
3886 HashTable_DebugStats(&VAR_GLOBAL->context, "VAR_GLOBAL");
3887 }
3888
3889 /* Print all variables in a context, sorted by name. */
3890 void
3891 Var_Dump(GNode *ctxt)
3892 {
3893 Vector /* of const char * */ vec;
3894 HashIter hi;
3895 size_t i;
3896 const char **varnames;
3897
3898 Vector_Init(&vec, sizeof(const char *));
3899
3900 HashIter_Init(&hi, &ctxt->context);
3901 while (HashIter_Next(&hi) != NULL)
3902 *(const char **)Vector_Push(&vec) = hi.entry->key;
3903 varnames = vec.items;
3904
3905 qsort(varnames, vec.len, sizeof varnames[0], str_cmp_asc);
3906
3907 for (i = 0; i < vec.len; i++) {
3908 const char *varname = varnames[i];
3909 Var *var = HashTable_FindValue(&ctxt->context, varname);
3910 debug_printf("%-16s = %s\n", varname, Buf_GetAll(&var->val, NULL));
3911 }
3912
3913 Vector_Done(&vec);
3914 }
3915