var.c revision 1.701 1 /* $NetBSD: var.c,v 1.701 2020/12/04 22:47:57 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 ${VAR:Mpattern}.
98 *
99 * Var_Delete Delete a variable.
100 *
101 * Var_ExportVars Export some or even all variables to the environment
102 * of this process and its child processes.
103 *
104 * Var_Export Export the variable to the environment of this process
105 * and its child processes.
106 *
107 * Var_UnExport Don't export the variable anymore.
108 *
109 * Debugging:
110 * Var_Stats Print out hashing statistics if in -dh mode.
111 *
112 * Var_Dump Print out all variables defined in the given context.
113 *
114 * XXX: There's a lot of duplication in these functions.
115 */
116
117 #include <sys/stat.h>
118 #ifndef NO_REGEX
119 #include <sys/types.h>
120 #include <regex.h>
121 #endif
122 #include <errno.h>
123 #include <inttypes.h>
124 #include <limits.h>
125 #include <time.h>
126
127 #include "make.h"
128 #include "dir.h"
129 #include "job.h"
130 #include "metachar.h"
131
132 /* "@(#)var.c 8.3 (Berkeley) 3/19/94" */
133 MAKE_RCSID("$NetBSD: var.c,v 1.701 2020/12/04 22:47:57 rillig Exp $");
134
135 #define VAR_DEBUG1(fmt, arg1) DEBUG1(VAR, fmt, arg1)
136 #define VAR_DEBUG2(fmt, arg1, arg2) DEBUG2(VAR, fmt, arg1, arg2)
137 #define VAR_DEBUG3(fmt, arg1, arg2, arg3) DEBUG3(VAR, fmt, arg1, arg2, arg3)
138 #define VAR_DEBUG4(fmt, arg1, arg2, arg3, arg4) DEBUG4(VAR, fmt, arg1, arg2, arg3, arg4)
139
140 ENUM_FLAGS_RTTI_3(VarEvalFlags,
141 VARE_UNDEFERR, VARE_WANTRES, VARE_KEEP_DOLLAR);
142
143 /*
144 * This lets us tell if we have replaced the original environ
145 * (which we cannot free).
146 */
147 char **savedEnv = NULL;
148
149 /* Special return value for Var_Parse, indicating a parse error. It may be
150 * caused by an undefined variable, a syntax error in a modifier or
151 * something entirely different. */
152 char var_Error[] = "";
153
154 /* Special return value for Var_Parse, indicating an undefined variable in
155 * a case where VARE_UNDEFERR is not set. This undefined variable is
156 * typically a dynamic variable such as ${.TARGET}, whose expansion needs to
157 * be deferred until it is defined in an actual target. */
158 static char varUndefined[] = "";
159
160 /*
161 * Traditionally this make consumed $$ during := like any other expansion.
162 * Other make's do not, and this make follows straight since 2016-01-09.
163 *
164 * This knob allows controlling the behavior.
165 * FALSE to consume $$ during := assignment.
166 * TRUE to preserve $$ during := assignment.
167 */
168 #define MAKE_SAVE_DOLLARS ".MAKE.SAVE_DOLLARS"
169 static Boolean save_dollars = TRUE;
170
171 /*
172 * Internally, variables are contained in four different contexts.
173 * 1) the environment. They cannot be changed. If an environment
174 * variable is appended to, the result is placed in the global
175 * context.
176 * 2) the global context. Variables set in the makefiles are located
177 * here.
178 * 3) the command-line context. All variables set on the command line
179 * are placed in this context.
180 * 4) the local context. Each target has associated with it a context
181 * list. On this list are located the structures describing such
182 * local variables as $(@) and $(*)
183 * The four contexts are searched in the reverse order from which they are
184 * listed (but see opts.checkEnvFirst).
185 */
186 GNode *VAR_INTERNAL; /* variables from make itself */
187 GNode *VAR_GLOBAL; /* variables from the makefile */
188 GNode *VAR_CMDLINE; /* variables defined on the command-line */
189
190 typedef enum VarFlags {
191
192 /* The variable's value is currently being used by Var_Parse or Var_Subst.
193 * This marker is used to avoid endless recursion. */
194 VAR_IN_USE = 0x01,
195
196 /* The variable comes from the environment.
197 * These variables are not registered in any GNode, therefore they must
198 * be freed as soon as they are not used anymore. */
199 VAR_FROM_ENV = 0x02,
200
201 /* The variable is exported to the environment, to be used by child
202 * processes. */
203 VAR_EXPORTED = 0x10,
204
205 /* At the point where this variable was exported, it contained an
206 * unresolved reference to another variable. Before any child process is
207 * started, it needs to be exported again, in the hope that the referenced
208 * variable can then be resolved. */
209 VAR_REEXPORT = 0x20,
210
211 /* The variable came from the command line. */
212 VAR_FROM_CMD = 0x40,
213
214 /* The variable value cannot be changed anymore, and the variable cannot
215 * be deleted. Any attempts to do so are ignored. */
216 VAR_READONLY = 0x80
217 } VarFlags;
218
219 ENUM_FLAGS_RTTI_6(VarFlags,
220 VAR_IN_USE, VAR_FROM_ENV,
221 VAR_EXPORTED, VAR_REEXPORT, VAR_FROM_CMD, VAR_READONLY);
222
223 /* Variables are defined using one of the VAR=value assignments. Their
224 * value can be queried by expressions such as $V, ${VAR}, or with modifiers
225 * such as ${VAR:S,from,to,g:Q}.
226 *
227 * There are 3 kinds of variables: context variables, environment variables,
228 * undefined variables.
229 *
230 * Context variables are stored in a GNode.context. The only way to undefine
231 * a context variable is using the .undef directive. In particular, it must
232 * not be possible to undefine a variable during the evaluation of an
233 * expression, or Var.name might point nowhere.
234 *
235 * Environment variables are temporary. They are returned by VarFind, and
236 * after using them, they must be freed using VarFreeEnv.
237 *
238 * Undefined variables occur during evaluation of variable expressions such
239 * as ${UNDEF:Ufallback} in Var_Parse and ApplyModifiers.
240 */
241 typedef struct Var {
242 /* The name of the variable, once set, doesn't change anymore.
243 * For context variables, it aliases the corresponding HashEntry name.
244 * For environment and undefined variables, it is allocated. */
245 const char *name;
246 void *name_freeIt;
247
248 Buffer val; /* its value */
249 VarFlags flags; /* miscellaneous status flags */
250 } Var;
251
252 /*
253 * Exporting vars is expensive so skip it if we can
254 */
255 typedef enum VarExportedMode {
256 VAR_EXPORTED_NONE,
257 VAR_EXPORTED_SOME,
258 VAR_EXPORTED_ALL
259 } VarExportedMode;
260
261 static VarExportedMode var_exportedVars = VAR_EXPORTED_NONE;
262
263 typedef enum VarExportFlags {
264 VAR_EXPORT_NORMAL = 0,
265 /*
266 * We pass this to Var_Export when doing the initial export
267 * or after updating an exported var.
268 */
269 VAR_EXPORT_PARENT = 0x01,
270 /*
271 * We pass this to Var_Export1 to tell it to leave the value alone.
272 */
273 VAR_EXPORT_LITERAL = 0x02
274 } VarExportFlags;
275
276 /* Flags for pattern matching in the :S and :C modifiers */
277 typedef enum VarPatternFlags {
278 VARP_SUB_GLOBAL = 0x01, /* Replace as often as possible ('g') */
279 VARP_SUB_ONE = 0x02, /* Replace only once ('1') */
280 VARP_ANCHOR_START = 0x04, /* Match at start of word ('^') */
281 VARP_ANCHOR_END = 0x08 /* Match at end of word ('$') */
282 } VarPatternFlags;
283
284 static Var *
285 VarNew(const char *name, void *name_freeIt, const char *value, VarFlags flags)
286 {
287 size_t value_len = strlen(value);
288 Var *var = bmake_malloc(sizeof *var);
289 var->name = name;
290 var->name_freeIt = name_freeIt;
291 Buf_InitSize(&var->val, value_len + 1);
292 Buf_AddBytes(&var->val, value, value_len);
293 var->flags = flags;
294 return var;
295 }
296
297 static const char *
298 CanonicalVarname(const char *name)
299 {
300 if (*name == '.' && ch_isupper(name[1])) {
301 switch (name[1]) {
302 case 'A':
303 if (strcmp(name, ".ALLSRC") == 0)
304 name = ALLSRC;
305 if (strcmp(name, ".ARCHIVE") == 0)
306 name = ARCHIVE;
307 break;
308 case 'I':
309 if (strcmp(name, ".IMPSRC") == 0)
310 name = IMPSRC;
311 break;
312 case 'M':
313 if (strcmp(name, ".MEMBER") == 0)
314 name = MEMBER;
315 break;
316 case 'O':
317 if (strcmp(name, ".OODATE") == 0)
318 name = OODATE;
319 break;
320 case 'P':
321 if (strcmp(name, ".PREFIX") == 0)
322 name = PREFIX;
323 break;
324 case 'S':
325 if (strcmp(name, ".SHELL") == 0) {
326 if (!shellPath)
327 Shell_Init();
328 }
329 break;
330 case 'T':
331 if (strcmp(name, ".TARGET") == 0)
332 name = TARGET;
333 break;
334 }
335 }
336
337 /* GNU make has an additional alias $^ == ${.ALLSRC}. */
338
339 return name;
340 }
341
342 static Var *
343 GNode_FindVar(GNode *ctxt, const char *varname, unsigned int hash)
344 {
345 return HashTable_FindValueHash(&ctxt->vars, varname, hash);
346 }
347
348 /* Find the variable in the context, and maybe in other contexts as well.
349 *
350 * Input:
351 * name name to find, is not expanded any further
352 * ctxt context in which to look first
353 * elsewhere TRUE to look in other contexts as well
354 *
355 * Results:
356 * The found variable, or NULL if the variable does not exist.
357 * If the variable is an environment variable, it must be freed using
358 * VarFreeEnv after use.
359 */
360 static Var *
361 VarFind(const char *name, GNode *ctxt, Boolean elsewhere)
362 {
363 Var *var;
364 unsigned int nameHash;
365
366 /*
367 * If the variable name begins with a '.', it could very well be one of
368 * the local ones. We check the name against all the local variables
369 * and substitute the short version in for 'name' if it matches one of
370 * them.
371 */
372 name = CanonicalVarname(name);
373 nameHash = Hash_Hash(name);
374
375 /* First look for the variable in the given context. */
376 var = GNode_FindVar(ctxt, name, nameHash);
377 if (!elsewhere)
378 return var;
379
380 /* The variable was not found in the given context. Now look for it in
381 * the other contexts as well. */
382 if (var == NULL && ctxt != VAR_CMDLINE)
383 var = GNode_FindVar(VAR_CMDLINE, name, nameHash);
384
385 if (!opts.checkEnvFirst && var == NULL && ctxt != VAR_GLOBAL) {
386 var = GNode_FindVar(VAR_GLOBAL, name, nameHash);
387 if (var == NULL && ctxt != VAR_INTERNAL) {
388 /* VAR_INTERNAL is subordinate to VAR_GLOBAL */
389 var = GNode_FindVar(VAR_INTERNAL, name, nameHash);
390 }
391 }
392
393 if (var == NULL) {
394 char *env;
395
396 if ((env = getenv(name)) != NULL) {
397 char *varname = bmake_strdup(name);
398 return VarNew(varname, varname, env, VAR_FROM_ENV);
399 }
400
401 if (opts.checkEnvFirst && ctxt != VAR_GLOBAL) {
402 var = GNode_FindVar(VAR_GLOBAL, name, nameHash);
403 if (var == NULL && ctxt != VAR_INTERNAL)
404 var = GNode_FindVar(VAR_INTERNAL, name, nameHash);
405 return var;
406 }
407
408 return NULL;
409 }
410
411 return var;
412 }
413
414 /* If the variable is an environment variable, free it.
415 *
416 * Input:
417 * v the variable
418 * freeValue true if the variable value should be freed as well
419 *
420 * Results:
421 * TRUE if it is an environment variable, FALSE otherwise.
422 */
423 static Boolean
424 VarFreeEnv(Var *v, Boolean freeValue)
425 {
426 if (!(v->flags & VAR_FROM_ENV))
427 return FALSE;
428
429 free(v->name_freeIt);
430 Buf_Destroy(&v->val, freeValue);
431 free(v);
432 return TRUE;
433 }
434
435 /* Add a new variable of the given name and value to the given context.
436 * The name and val arguments are duplicated so they may safely be freed. */
437 static void
438 VarAdd(const char *name, const char *val, GNode *ctxt, VarSetFlags flags)
439 {
440 HashEntry *he = HashTable_CreateEntry(&ctxt->vars, name, NULL);
441 Var *v = VarNew(he->key /* aliased */, NULL, val,
442 flags & VAR_SET_READONLY ? VAR_READONLY : 0);
443 HashEntry_Set(he, v);
444 if (!(ctxt->flags & INTERNAL)) {
445 VAR_DEBUG3("%s:%s = %s\n", ctxt->name, name, val);
446 }
447 }
448
449 /* Remove a variable from a context, freeing all related memory as well.
450 * The variable name is expanded once. */
451 void
452 Var_Delete(const char *name, GNode *ctxt)
453 {
454 char *name_freeIt = NULL;
455 HashEntry *he;
456
457 if (strchr(name, '$') != NULL) {
458 (void)Var_Subst(name, VAR_GLOBAL, VARE_WANTRES, &name_freeIt);
459 /* TODO: handle errors */
460 name = name_freeIt;
461 }
462 he = HashTable_FindEntry(&ctxt->vars, name);
463 VAR_DEBUG3("%s:delete %s%s\n",
464 ctxt->name, name, he != NULL ? "" : " (not found)");
465 free(name_freeIt);
466
467 if (he != NULL) {
468 Var *v = HashEntry_Get(he);
469 if (v->flags & VAR_EXPORTED)
470 unsetenv(v->name);
471 if (strcmp(v->name, MAKE_EXPORTED) == 0)
472 var_exportedVars = VAR_EXPORTED_NONE;
473 assert(v->name_freeIt == NULL);
474 HashTable_DeleteEntry(&ctxt->vars, he);
475 Buf_Destroy(&v->val, TRUE);
476 free(v);
477 }
478 }
479
480 static Boolean
481 MayExport(const char *name)
482 {
483 if (name[0] == '.')
484 return FALSE; /* skip internals */
485 if (name[0] == '-')
486 return FALSE; /* skip misnamed variables */
487 if (name[1] == '\0') {
488 /*
489 * A single char.
490 * If it is one of the vars that should only appear in
491 * local context, skip it, else we can get Var_Subst
492 * into a loop.
493 */
494 switch (name[0]) {
495 case '@':
496 case '%':
497 case '*':
498 case '!':
499 return FALSE;
500 }
501 }
502 return TRUE;
503 }
504
505 /*
506 * Export a single variable.
507 * We ignore make internal variables (those which start with '.').
508 * Also we jump through some hoops to avoid calling setenv
509 * more than necessary since it can leak.
510 * We only manipulate flags of vars if 'parent' is set.
511 */
512 static Boolean
513 Var_Export1(const char *name, VarExportFlags flags)
514 {
515 Boolean parent = (flags & VAR_EXPORT_PARENT) != 0;
516 Var *v;
517 char *val;
518
519 if (!MayExport(name))
520 return FALSE;
521
522 v = VarFind(name, VAR_GLOBAL, FALSE);
523 if (v == NULL)
524 return FALSE;
525
526 if (!parent && (v->flags & VAR_EXPORTED) && !(v->flags & VAR_REEXPORT))
527 return FALSE; /* nothing to do */
528
529 val = Buf_GetAll(&v->val, NULL);
530 if (!(flags & VAR_EXPORT_LITERAL) && strchr(val, '$') != NULL) {
531 char *expr;
532
533 if (parent) {
534 /*
535 * Flag the variable as something we need to re-export.
536 * No point actually exporting it now though,
537 * the child process can do it at the last minute.
538 */
539 v->flags |= VAR_EXPORTED | VAR_REEXPORT;
540 return TRUE;
541 }
542 if (v->flags & VAR_IN_USE) {
543 /*
544 * We recursed while exporting in a child.
545 * This isn't going to end well, just skip it.
546 */
547 return FALSE;
548 }
549
550 /* XXX: name is injected without escaping it */
551 expr = str_concat3("${", name, "}");
552 (void)Var_Subst(expr, VAR_GLOBAL, VARE_WANTRES, &val);
553 /* TODO: handle errors */
554 setenv(name, val, 1);
555 free(val);
556 free(expr);
557 } else {
558 if (parent)
559 v->flags &= ~(unsigned)VAR_REEXPORT; /* once will do */
560 if (parent || !(v->flags & VAR_EXPORTED))
561 setenv(name, val, 1);
562 }
563
564 /*
565 * This is so Var_Set knows to call Var_Export again...
566 */
567 if (parent) {
568 v->flags |= VAR_EXPORTED;
569 }
570 return TRUE;
571 }
572
573 /*
574 * This gets called from our child processes.
575 */
576 void
577 Var_ExportVars(void)
578 {
579 char *val;
580
581 /*
582 * Several make's support this sort of mechanism for tracking
583 * recursion - but each uses a different name.
584 * We allow the makefiles to update MAKELEVEL and ensure
585 * children see a correctly incremented value.
586 */
587 char tmp[BUFSIZ];
588 snprintf(tmp, sizeof tmp, "%d", makelevel + 1);
589 setenv(MAKE_LEVEL_ENV, tmp, 1);
590
591 if (var_exportedVars == VAR_EXPORTED_NONE)
592 return;
593
594 if (var_exportedVars == VAR_EXPORTED_ALL) {
595 HashIter hi;
596
597 /* Ouch! Exporting all variables at once is crazy... */
598 HashIter_Init(&hi, &VAR_GLOBAL->vars);
599 while (HashIter_Next(&hi) != NULL) {
600 Var *var = hi.entry->value;
601 Var_Export1(var->name, VAR_EXPORT_NORMAL);
602 }
603 return;
604 }
605
606 (void)Var_Subst("${" MAKE_EXPORTED ":O:u}", VAR_GLOBAL, VARE_WANTRES, &val);
607 /* TODO: handle errors */
608 if (val[0] != '\0') {
609 Words words = Str_Words(val, FALSE);
610 size_t i;
611
612 for (i = 0; i < words.len; i++)
613 Var_Export1(words.words[i], VAR_EXPORT_NORMAL);
614 Words_Free(words);
615 }
616 free(val);
617 }
618
619 /*
620 * This is called when .export is seen or .MAKE.EXPORTED is modified.
621 *
622 * It is also called when any exported variable is modified.
623 * XXX: Is it really?
624 *
625 * str has the format "[-env|-literal] varname...".
626 */
627 void
628 Var_Export(const char *str, Boolean isExport)
629 {
630 VarExportFlags flags;
631 char *val;
632
633 if (isExport && str[0] == '\0') {
634 var_exportedVars = VAR_EXPORTED_ALL; /* use with caution! */
635 return;
636 }
637
638 if (isExport && strncmp(str, "-env", 4) == 0) {
639 str += 4;
640 flags = 0;
641 } else if (isExport && strncmp(str, "-literal", 8) == 0) {
642 str += 8;
643 flags = VAR_EXPORT_LITERAL;
644 } else {
645 flags = VAR_EXPORT_PARENT;
646 }
647
648 (void)Var_Subst(str, VAR_GLOBAL, VARE_WANTRES, &val);
649 /* TODO: handle errors */
650 if (val[0] != '\0') {
651 Words words = Str_Words(val, FALSE);
652
653 size_t i;
654 for (i = 0; i < words.len; i++) {
655 const char *name = words.words[i];
656 if (Var_Export1(name, flags)) {
657 if (var_exportedVars == VAR_EXPORTED_NONE)
658 var_exportedVars = VAR_EXPORTED_SOME;
659 if (isExport && (flags & VAR_EXPORT_PARENT)) {
660 Var_Append(MAKE_EXPORTED, name, VAR_GLOBAL);
661 }
662 }
663 }
664 Words_Free(words);
665 }
666 free(val);
667 }
668
669
670 extern char **environ;
671
672 static void
673 UnexportEnv(void)
674 {
675 const char *cp;
676 char **newenv;
677
678 cp = getenv(MAKE_LEVEL_ENV); /* we should preserve this */
679 if (environ == savedEnv) {
680 /* we have been here before! */
681 newenv = bmake_realloc(environ, 2 * sizeof(char *));
682 } else {
683 if (savedEnv != NULL) {
684 free(savedEnv);
685 savedEnv = NULL;
686 }
687 newenv = bmake_malloc(2 * sizeof(char *));
688 }
689
690 /* Note: we cannot safely free() the original environ. */
691 environ = savedEnv = newenv;
692 newenv[0] = NULL;
693 newenv[1] = NULL;
694 if (cp && *cp)
695 setenv(MAKE_LEVEL_ENV, cp, 1);
696 }
697
698 static void
699 UnexportVar(const char *varname, Boolean unexport_env, Boolean adjust)
700 {
701 Var *v = VarFind(varname, VAR_GLOBAL, FALSE);
702 if (v == NULL) {
703 VAR_DEBUG1("Not unexporting \"%s\" (not found)\n", varname);
704 return;
705 }
706
707 VAR_DEBUG1("Unexporting \"%s\"\n", varname);
708 if (!unexport_env && (v->flags & VAR_EXPORTED) &&
709 !(v->flags & VAR_REEXPORT))
710 unsetenv(v->name);
711 v->flags &= ~(unsigned)(VAR_EXPORTED | VAR_REEXPORT);
712
713 /* If we are unexporting a list, remove each one from .MAKE.EXPORTED. */
714 if (adjust) {
715 /* XXX: v->name is injected without escaping it */
716 char *expr = str_concat3("${" MAKE_EXPORTED ":N", v->name, "}");
717 char *cp;
718 (void)Var_Subst(expr, VAR_GLOBAL, VARE_WANTRES, &cp);
719 /* TODO: handle errors */
720 Var_Set(MAKE_EXPORTED, cp, VAR_GLOBAL);
721 free(cp);
722 free(expr);
723 }
724 }
725
726 /*
727 * This is called when .unexport[-env] is seen.
728 *
729 * str must have the form "unexport[-env] varname...".
730 */
731 void
732 Var_UnExport(const char *str)
733 {
734 const char *varnames;
735 char *varnames_freeIt;
736 Boolean unexport_env;
737
738 varnames = NULL;
739 varnames_freeIt = NULL;
740
741 str += strlen("unexport");
742 unexport_env = strncmp(str, "-env", 4) == 0;
743 if (unexport_env) {
744 UnexportEnv();
745 } else {
746 cpp_skip_whitespace(&str);
747 if (str[0] != '\0')
748 varnames = str;
749 }
750
751 if (varnames == NULL) {
752 /* Using .MAKE.EXPORTED */
753 (void)Var_Subst("${" MAKE_EXPORTED ":O:u}", VAR_GLOBAL, VARE_WANTRES,
754 &varnames_freeIt);
755 /* TODO: handle errors */
756 varnames = varnames_freeIt;
757 }
758
759 {
760 size_t i;
761
762 Words words = Str_Words(varnames, FALSE);
763 for (i = 0; i < words.len; i++) {
764 const char *varname = words.words[i];
765 UnexportVar(varname, unexport_env, varnames == str);
766 }
767 Words_Free(words);
768 if (varnames != str) {
769 Var_Delete(MAKE_EXPORTED, VAR_GLOBAL);
770 free(varnames_freeIt);
771 }
772 }
773 }
774
775 /* See Var_Set for documentation. */
776 void
777 Var_SetWithFlags(const char *name, const char *val, GNode *ctxt,
778 VarSetFlags flags)
779 {
780 const char *unexpanded_name = name;
781 char *name_freeIt = NULL;
782 Var *v;
783
784 assert(val != NULL);
785
786 if (strchr(name, '$') != NULL) {
787 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
788 /* TODO: handle errors */
789 name = name_freeIt;
790 }
791
792 if (name[0] == '\0') {
793 VAR_DEBUG2("Var_Set(\"%s\", \"%s\", ...) "
794 "name expands to empty string - ignored\n",
795 unexpanded_name, val);
796 free(name_freeIt);
797 return;
798 }
799
800 if (ctxt == VAR_GLOBAL) {
801 v = VarFind(name, VAR_CMDLINE, FALSE);
802 if (v != NULL) {
803 if (v->flags & VAR_FROM_CMD) {
804 VAR_DEBUG3("%s:%s = %s ignored!\n", ctxt->name, name, val);
805 goto out;
806 }
807 VarFreeEnv(v, TRUE);
808 }
809 }
810
811 /*
812 * We only look for a variable in the given context since anything set
813 * here will override anything in a lower context, so there's not much
814 * point in searching them all just to save a bit of memory...
815 */
816 v = VarFind(name, ctxt, FALSE);
817 if (v == NULL) {
818 if (ctxt == VAR_CMDLINE && !(flags & VAR_SET_NO_EXPORT)) {
819 /*
820 * This var would normally prevent the same name being added
821 * to VAR_GLOBAL, so delete it from there if needed.
822 * Otherwise -V name may show the wrong value.
823 */
824 /* XXX: name is expanded for the second time */
825 Var_Delete(name, VAR_GLOBAL);
826 }
827 VarAdd(name, val, ctxt, flags);
828 } else {
829 if ((v->flags & VAR_READONLY) && !(flags & VAR_SET_READONLY)) {
830 VAR_DEBUG3("%s:%s = %s ignored (read-only)\n",
831 ctxt->name, name, val);
832 goto out;
833 }
834 Buf_Empty(&v->val);
835 Buf_AddStr(&v->val, val);
836
837 VAR_DEBUG3("%s:%s = %s\n", ctxt->name, name, val);
838 if (v->flags & VAR_EXPORTED) {
839 Var_Export1(name, VAR_EXPORT_PARENT);
840 }
841 }
842 /*
843 * Any variables given on the command line are automatically exported
844 * to the environment (as per POSIX standard)
845 * Other than internals.
846 */
847 if (ctxt == VAR_CMDLINE && !(flags & VAR_SET_NO_EXPORT) && name[0] != '.') {
848 if (v == NULL)
849 v = VarFind(name, ctxt, FALSE); /* we just added it */
850 v->flags |= VAR_FROM_CMD;
851
852 /*
853 * If requested, don't export these in the environment
854 * individually. We still put them in MAKEOVERRIDES so
855 * that the command-line settings continue to override
856 * Makefile settings.
857 */
858 if (!opts.varNoExportEnv)
859 setenv(name, val, 1);
860
861 Var_Append(MAKEOVERRIDES, name, VAR_GLOBAL);
862 }
863 if (name[0] == '.' && strcmp(name, MAKE_SAVE_DOLLARS) == 0)
864 save_dollars = ParseBoolean(val, save_dollars);
865
866 out:
867 free(name_freeIt);
868 if (v != NULL)
869 VarFreeEnv(v, TRUE);
870 }
871
872 /*-
873 *-----------------------------------------------------------------------
874 * Var_Set --
875 * Set the variable name to the value val in the given context.
876 *
877 * If the variable doesn't yet exist, it is created.
878 * Otherwise the new value overwrites and replaces the old value.
879 *
880 * Input:
881 * name name of the variable to set, is expanded once
882 * val value to give to the variable
883 * ctxt context in which to set it
884 *
885 * Notes:
886 * The variable is searched for only in its context before being
887 * created in that context. I.e. if the context is VAR_GLOBAL,
888 * only VAR_GLOBAL->context is searched. Likewise if it is VAR_CMDLINE,
889 * only VAR_CMDLINE->context is searched. This is done to avoid the
890 * literally thousands of unnecessary strcmp's that used to be done to
891 * set, say, $(@) or $(<).
892 * If the context is VAR_GLOBAL though, we check if the variable
893 * was set in VAR_CMDLINE from the command line and skip it if so.
894 *-----------------------------------------------------------------------
895 */
896 void
897 Var_Set(const char *name, const char *val, GNode *ctxt)
898 {
899 Var_SetWithFlags(name, val, ctxt, VAR_SET_NONE);
900 }
901
902 /*-
903 *-----------------------------------------------------------------------
904 * Var_Append --
905 * The variable of the given name has the given value appended to it in
906 * the given context.
907 *
908 * If the variable doesn't exist, it is created. Otherwise the strings
909 * are concatenated, with a space in between.
910 *
911 * Input:
912 * name name of the variable to modify, is expanded once
913 * val string to append to it
914 * ctxt context in which this should occur
915 *
916 * Notes:
917 * Only if the variable is being sought in the global context is the
918 * environment searched.
919 * XXX: Knows its calling circumstances in that if called with ctxt
920 * an actual target, it will only search that context since only
921 * a local variable could be being appended to. This is actually
922 * a big win and must be tolerated.
923 *-----------------------------------------------------------------------
924 */
925 void
926 Var_Append(const char *name, const char *val, GNode *ctxt)
927 {
928 char *name_freeIt = NULL;
929 Var *v;
930
931 assert(val != NULL);
932
933 if (strchr(name, '$') != NULL) {
934 const char *unexpanded_name = name;
935 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
936 /* TODO: handle errors */
937 name = name_freeIt;
938 if (name[0] == '\0') {
939 VAR_DEBUG2("Var_Append(\"%s\", \"%s\", ...) "
940 "name expands to empty string - ignored\n",
941 unexpanded_name, val);
942 free(name_freeIt);
943 return;
944 }
945 }
946
947 v = VarFind(name, ctxt, ctxt == VAR_GLOBAL);
948
949 if (v == NULL) {
950 /* XXX: name is expanded for the second time */
951 Var_Set(name, val, ctxt);
952 } else if (v->flags & VAR_READONLY) {
953 VAR_DEBUG1("Ignoring append to %s since it is read-only\n", name);
954 } else if (ctxt == VAR_CMDLINE || !(v->flags & VAR_FROM_CMD)) {
955 Buf_AddByte(&v->val, ' ');
956 Buf_AddStr(&v->val, val);
957
958 VAR_DEBUG3("%s:%s = %s\n",
959 ctxt->name, name, Buf_GetAll(&v->val, NULL));
960
961 if (v->flags & VAR_FROM_ENV) {
962 /*
963 * If the original variable came from the environment, we
964 * have to install it in the global context (we could place
965 * it in the environment, but then we should provide a way to
966 * export other variables...)
967 */
968 v->flags &= ~(unsigned)VAR_FROM_ENV;
969 /* This is the only place where a variable is created whose
970 * v->name is not the same as ctxt->context->key. */
971 HashTable_Set(&ctxt->vars, name, v);
972 }
973 }
974 free(name_freeIt);
975 }
976
977 /* See if the given variable exists, in the given context or in other
978 * fallback contexts.
979 *
980 * Input:
981 * name Variable to find, is expanded once
982 * ctxt Context in which to start search
983 */
984 Boolean
985 Var_Exists(const char *name, GNode *ctxt)
986 {
987 char *name_freeIt = NULL;
988 Var *v;
989
990 if (strchr(name, '$') != NULL) {
991 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
992 /* TODO: handle errors */
993 name = name_freeIt;
994 }
995
996 v = VarFind(name, ctxt, TRUE);
997 free(name_freeIt);
998 if (v == NULL)
999 return FALSE;
1000
1001 (void)VarFreeEnv(v, TRUE);
1002 return TRUE;
1003 }
1004
1005 /*-
1006 *-----------------------------------------------------------------------
1007 * Var_Value --
1008 * Return the unexpanded value of the given variable in the given
1009 * context, or the usual contexts.
1010 *
1011 * Input:
1012 * name name to find, is not expanded any further
1013 * ctxt context in which to search for it
1014 *
1015 * Results:
1016 * The value if the variable exists, NULL if it doesn't.
1017 * If the returned value is not NULL, the caller must free
1018 * out_freeIt when the returned value is no longer needed.
1019 *-----------------------------------------------------------------------
1020 */
1021 const char *
1022 Var_Value(const char *name, GNode *ctxt, void **out_freeIt)
1023 {
1024 Var *v = VarFind(name, ctxt, TRUE);
1025 char *value;
1026
1027 *out_freeIt = NULL;
1028 if (v == NULL)
1029 return NULL;
1030
1031 value = Buf_GetAll(&v->val, NULL);
1032 if (VarFreeEnv(v, FALSE))
1033 *out_freeIt = value;
1034 return value;
1035 }
1036
1037 /* Return the unexpanded variable value from this node, without trying to look
1038 * up the variable in any other context. */
1039 const char *
1040 Var_ValueDirect(const char *name, GNode *ctxt)
1041 {
1042 Var *v = VarFind(name, ctxt, FALSE);
1043 return v != NULL ? Buf_GetAll(&v->val, NULL) : NULL;
1044 }
1045
1046
1047 /* SepBuf is a string being built from words, interleaved with separators. */
1048 typedef struct SepBuf {
1049 Buffer buf;
1050 Boolean needSep;
1051 char sep; /* usually ' ', but see the :ts modifier */
1052 } SepBuf;
1053
1054 static void
1055 SepBuf_Init(SepBuf *buf, char sep)
1056 {
1057 Buf_InitSize(&buf->buf, 32);
1058 buf->needSep = FALSE;
1059 buf->sep = sep;
1060 }
1061
1062 static void
1063 SepBuf_Sep(SepBuf *buf)
1064 {
1065 buf->needSep = TRUE;
1066 }
1067
1068 static void
1069 SepBuf_AddBytes(SepBuf *buf, const char *mem, size_t mem_size)
1070 {
1071 if (mem_size == 0)
1072 return;
1073 if (buf->needSep && buf->sep != '\0') {
1074 Buf_AddByte(&buf->buf, buf->sep);
1075 buf->needSep = FALSE;
1076 }
1077 Buf_AddBytes(&buf->buf, mem, mem_size);
1078 }
1079
1080 static void
1081 SepBuf_AddBytesBetween(SepBuf *buf, const char *start, const char *end)
1082 {
1083 SepBuf_AddBytes(buf, start, (size_t)(end - start));
1084 }
1085
1086 static void
1087 SepBuf_AddStr(SepBuf *buf, const char *str)
1088 {
1089 SepBuf_AddBytes(buf, str, strlen(str));
1090 }
1091
1092 static char *
1093 SepBuf_Destroy(SepBuf *buf, Boolean free_buf)
1094 {
1095 return Buf_Destroy(&buf->buf, free_buf);
1096 }
1097
1098
1099 /* This callback for ModifyWords gets a single word from a variable expression
1100 * and typically adds a modification of this word to the buffer. It may also
1101 * do nothing or add several words.
1102 *
1103 * For example, in ${:Ua b c:M*2}, the callback is called 3 times, once for
1104 * each word of "a b c". */
1105 typedef void (*ModifyWordsCallback)(const char *word, SepBuf *buf, void *data);
1106
1107
1108 /* Callback for ModifyWords to implement the :H modifier.
1109 * Add the dirname of the given word to the buffer. */
1110 static void
1111 ModifyWord_Head(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1112 {
1113 const char *slash = strrchr(word, '/');
1114 if (slash != NULL)
1115 SepBuf_AddBytesBetween(buf, word, slash);
1116 else
1117 SepBuf_AddStr(buf, ".");
1118 }
1119
1120 /* Callback for ModifyWords to implement the :T modifier.
1121 * Add the basename of the given word to the buffer. */
1122 static void
1123 ModifyWord_Tail(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1124 {
1125 const char *slash = strrchr(word, '/');
1126 const char *base = slash != NULL ? slash + 1 : word;
1127 SepBuf_AddStr(buf, base);
1128 }
1129
1130 /* Callback for ModifyWords to implement the :E modifier.
1131 * Add the filename suffix of the given word to the buffer, if it exists. */
1132 static void
1133 ModifyWord_Suffix(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1134 {
1135 const char *lastDot = strrchr(word, '.');
1136 if (lastDot != NULL)
1137 SepBuf_AddStr(buf, lastDot + 1);
1138 }
1139
1140 /* Callback for ModifyWords to implement the :R modifier.
1141 * Add the basename of the given word to the buffer. */
1142 static void
1143 ModifyWord_Root(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1144 {
1145 const char *lastDot = strrchr(word, '.');
1146 size_t len = lastDot != NULL ? (size_t)(lastDot - word) : strlen(word);
1147 SepBuf_AddBytes(buf, word, len);
1148 }
1149
1150 /* Callback for ModifyWords to implement the :M modifier.
1151 * Place the word in the buffer if it matches the given pattern. */
1152 static void
1153 ModifyWord_Match(const char *word, SepBuf *buf, void *data)
1154 {
1155 const char *pattern = data;
1156 VAR_DEBUG2("VarMatch [%s] [%s]\n", word, pattern);
1157 if (Str_Match(word, pattern))
1158 SepBuf_AddStr(buf, word);
1159 }
1160
1161 /* Callback for ModifyWords to implement the :N modifier.
1162 * Place the word in the buffer if it doesn't match the given pattern. */
1163 static void
1164 ModifyWord_NoMatch(const char *word, SepBuf *buf, void *data)
1165 {
1166 const char *pattern = data;
1167 if (!Str_Match(word, pattern))
1168 SepBuf_AddStr(buf, word);
1169 }
1170
1171 #ifdef SYSVVARSUB
1172 /* Check word against pattern for a match (% is a wildcard).
1173 *
1174 * Input:
1175 * word Word to examine
1176 * pattern Pattern to examine against
1177 *
1178 * Results:
1179 * Returns the start of the match, or NULL.
1180 * out_match_len returns the length of the match, if any.
1181 * out_hasPercent returns whether the pattern contains a percent.
1182 */
1183 static const char *
1184 SysVMatch(const char *word, const char *pattern,
1185 size_t *out_match_len, Boolean *out_hasPercent)
1186 {
1187 const char *p = pattern;
1188 const char *w = word;
1189 const char *percent;
1190 size_t w_len;
1191 size_t p_len;
1192 const char *w_tail;
1193
1194 *out_hasPercent = FALSE;
1195 percent = strchr(p, '%');
1196 if (percent != NULL) { /* ${VAR:...%...=...} */
1197 *out_hasPercent = TRUE;
1198 if (*w == '\0')
1199 return NULL; /* empty word does not match pattern */
1200
1201 /* check that the prefix matches */
1202 for (; p != percent && *w != '\0' && *w == *p; w++, p++)
1203 continue;
1204 if (p != percent)
1205 return NULL; /* No match */
1206
1207 p++; /* Skip the percent */
1208 if (*p == '\0') {
1209 /* No more pattern, return the rest of the string */
1210 *out_match_len = strlen(w);
1211 return w;
1212 }
1213 }
1214
1215 /* Test whether the tail matches */
1216 w_len = strlen(w);
1217 p_len = strlen(p);
1218 if (w_len < p_len)
1219 return NULL;
1220
1221 w_tail = w + w_len - p_len;
1222 if (memcmp(p, w_tail, p_len) != 0)
1223 return NULL;
1224
1225 *out_match_len = (size_t)(w_tail - w);
1226 return w;
1227 }
1228
1229 struct ModifyWord_SYSVSubstArgs {
1230 GNode *ctx;
1231 const char *lhs;
1232 const char *rhs;
1233 };
1234
1235 /* Callback for ModifyWords to implement the :%.from=%.to modifier. */
1236 static void
1237 ModifyWord_SYSVSubst(const char *word, SepBuf *buf, void *data)
1238 {
1239 const struct ModifyWord_SYSVSubstArgs *args = data;
1240 char *rhs_expanded;
1241 const char *rhs;
1242 const char *percent;
1243
1244 size_t match_len;
1245 Boolean lhsPercent;
1246 const char *match = SysVMatch(word, args->lhs, &match_len, &lhsPercent);
1247 if (match == NULL) {
1248 SepBuf_AddStr(buf, word);
1249 return;
1250 }
1251
1252 /* Append rhs to the buffer, substituting the first '%' with the
1253 * match, but only if the lhs had a '%' as well. */
1254
1255 (void)Var_Subst(args->rhs, args->ctx, VARE_WANTRES, &rhs_expanded);
1256 /* TODO: handle errors */
1257
1258 rhs = rhs_expanded;
1259 percent = strchr(rhs, '%');
1260
1261 if (percent != NULL && lhsPercent) {
1262 /* Copy the prefix of the replacement pattern */
1263 SepBuf_AddBytesBetween(buf, rhs, percent);
1264 rhs = percent + 1;
1265 }
1266 if (percent != NULL || !lhsPercent)
1267 SepBuf_AddBytes(buf, match, match_len);
1268
1269 /* Append the suffix of the replacement pattern */
1270 SepBuf_AddStr(buf, rhs);
1271
1272 free(rhs_expanded);
1273 }
1274 #endif
1275
1276
1277 struct ModifyWord_SubstArgs {
1278 const char *lhs;
1279 size_t lhsLen;
1280 const char *rhs;
1281 size_t rhsLen;
1282 VarPatternFlags pflags;
1283 Boolean matched;
1284 };
1285
1286 /* Callback for ModifyWords to implement the :S,from,to, modifier.
1287 * Perform a string substitution on the given word. */
1288 static void
1289 ModifyWord_Subst(const char *word, SepBuf *buf, void *data)
1290 {
1291 size_t wordLen = strlen(word);
1292 struct ModifyWord_SubstArgs *args = data;
1293 const char *match;
1294
1295 if ((args->pflags & VARP_SUB_ONE) && args->matched)
1296 goto nosub;
1297
1298 if (args->pflags & VARP_ANCHOR_START) {
1299 if (wordLen < args->lhsLen ||
1300 memcmp(word, args->lhs, args->lhsLen) != 0)
1301 goto nosub;
1302
1303 if ((args->pflags & VARP_ANCHOR_END) && wordLen != args->lhsLen)
1304 goto nosub;
1305
1306 /* :S,^prefix,replacement, or :S,^whole$,replacement, */
1307 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1308 SepBuf_AddBytes(buf, word + args->lhsLen, wordLen - args->lhsLen);
1309 args->matched = TRUE;
1310 return;
1311 }
1312
1313 if (args->pflags & VARP_ANCHOR_END) {
1314 const char *start;
1315
1316 if (wordLen < args->lhsLen)
1317 goto nosub;
1318
1319 start = word + (wordLen - args->lhsLen);
1320 if (memcmp(start, args->lhs, args->lhsLen) != 0)
1321 goto nosub;
1322
1323 /* :S,suffix$,replacement, */
1324 SepBuf_AddBytesBetween(buf, word, start);
1325 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1326 args->matched = TRUE;
1327 return;
1328 }
1329
1330 if (args->lhs[0] == '\0')
1331 goto nosub;
1332
1333 /* unanchored case, may match more than once */
1334 while ((match = strstr(word, args->lhs)) != NULL) {
1335 SepBuf_AddBytesBetween(buf, word, match);
1336 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1337 args->matched = TRUE;
1338 wordLen -= (size_t)(match - word) + args->lhsLen;
1339 word += (size_t)(match - word) + args->lhsLen;
1340 if (wordLen == 0 || !(args->pflags & VARP_SUB_GLOBAL))
1341 break;
1342 }
1343 nosub:
1344 SepBuf_AddBytes(buf, word, wordLen);
1345 }
1346
1347 #ifndef NO_REGEX
1348 /* Print the error caused by a regcomp or regexec call. */
1349 static void
1350 VarREError(int reerr, const regex_t *pat, const char *str)
1351 {
1352 size_t errlen = regerror(reerr, pat, NULL, 0);
1353 char *errbuf = bmake_malloc(errlen);
1354 regerror(reerr, pat, errbuf, errlen);
1355 Error("%s: %s", str, errbuf);
1356 free(errbuf);
1357 }
1358
1359 struct ModifyWord_SubstRegexArgs {
1360 regex_t re;
1361 size_t nsub;
1362 char *replace;
1363 VarPatternFlags pflags;
1364 Boolean matched;
1365 };
1366
1367 /* Callback for ModifyWords to implement the :C/from/to/ modifier.
1368 * Perform a regex substitution on the given word. */
1369 static void
1370 ModifyWord_SubstRegex(const char *word, SepBuf *buf, void *data)
1371 {
1372 struct ModifyWord_SubstRegexArgs *args = data;
1373 int xrv;
1374 const char *wp = word;
1375 char *rp;
1376 int flags = 0;
1377 regmatch_t m[10];
1378
1379 if ((args->pflags & VARP_SUB_ONE) && args->matched)
1380 goto nosub;
1381
1382 tryagain:
1383 xrv = regexec(&args->re, wp, args->nsub, m, flags);
1384
1385 switch (xrv) {
1386 case 0:
1387 args->matched = TRUE;
1388 SepBuf_AddBytes(buf, wp, (size_t)m[0].rm_so);
1389
1390 for (rp = args->replace; *rp; rp++) {
1391 if (*rp == '\\' && (rp[1] == '&' || rp[1] == '\\')) {
1392 SepBuf_AddBytes(buf, rp + 1, 1);
1393 rp++;
1394 continue;
1395 }
1396
1397 if (*rp == '&') {
1398 SepBuf_AddBytesBetween(buf, wp + m[0].rm_so, wp + m[0].rm_eo);
1399 continue;
1400 }
1401
1402 if (*rp != '\\' || !ch_isdigit(rp[1])) {
1403 SepBuf_AddBytes(buf, rp, 1);
1404 continue;
1405 }
1406
1407 { /* \0 to \9 backreference */
1408 size_t n = (size_t)(rp[1] - '0');
1409 rp++;
1410
1411 if (n >= args->nsub) {
1412 Error("No subexpression \\%zu", n);
1413 } else if (m[n].rm_so == -1) {
1414 Error("No match for subexpression \\%zu", n);
1415 } else {
1416 SepBuf_AddBytesBetween(buf, wp + m[n].rm_so,
1417 wp + m[n].rm_eo);
1418 }
1419 }
1420 }
1421
1422 wp += m[0].rm_eo;
1423 if (args->pflags & VARP_SUB_GLOBAL) {
1424 flags |= REG_NOTBOL;
1425 if (m[0].rm_so == 0 && m[0].rm_eo == 0) {
1426 SepBuf_AddBytes(buf, wp, 1);
1427 wp++;
1428 }
1429 if (*wp != '\0')
1430 goto tryagain;
1431 }
1432 if (*wp != '\0')
1433 SepBuf_AddStr(buf, wp);
1434 break;
1435 default:
1436 VarREError(xrv, &args->re, "Unexpected regex error");
1437 /* FALLTHROUGH */
1438 case REG_NOMATCH:
1439 nosub:
1440 SepBuf_AddStr(buf, wp);
1441 break;
1442 }
1443 }
1444 #endif
1445
1446
1447 struct ModifyWord_LoopArgs {
1448 GNode *ctx;
1449 char *tvar; /* name of temporary variable */
1450 char *str; /* string to expand */
1451 VarEvalFlags eflags;
1452 };
1453
1454 /* Callback for ModifyWords to implement the :@var (at) ...@ modifier of ODE make. */
1455 static void
1456 ModifyWord_Loop(const char *word, SepBuf *buf, void *data)
1457 {
1458 const struct ModifyWord_LoopArgs *args;
1459 char *s;
1460
1461 if (word[0] == '\0')
1462 return;
1463
1464 args = data;
1465 Var_SetWithFlags(args->tvar, word, args->ctx, VAR_SET_NO_EXPORT);
1466 (void)Var_Subst(args->str, args->ctx, args->eflags, &s);
1467 /* TODO: handle errors */
1468
1469 VAR_DEBUG4("ModifyWord_Loop: "
1470 "in \"%s\", replace \"%s\" with \"%s\" to \"%s\"\n",
1471 word, args->tvar, args->str, s);
1472
1473 if (s[0] == '\n' || Buf_EndsWith(&buf->buf, '\n'))
1474 buf->needSep = FALSE;
1475 SepBuf_AddStr(buf, s);
1476 free(s);
1477 }
1478
1479
1480 /* The :[first..last] modifier selects words from the expression.
1481 * It can also reverse the words. */
1482 static char *
1483 VarSelectWords(char sep, Boolean oneBigWord, const char *str, int first,
1484 int last)
1485 {
1486 Words words;
1487 int len, start, end, step;
1488 int i;
1489
1490 SepBuf buf;
1491 SepBuf_Init(&buf, sep);
1492
1493 if (oneBigWord) {
1494 /* fake what Str_Words() would do if there were only one word */
1495 words.len = 1;
1496 words.words = bmake_malloc((words.len + 1) * sizeof(words.words[0]));
1497 words.freeIt = bmake_strdup(str);
1498 words.words[0] = words.freeIt;
1499 words.words[1] = NULL;
1500 } else {
1501 words = Str_Words(str, FALSE);
1502 }
1503
1504 /*
1505 * Now sanitize the given range.
1506 * If first or last are negative, convert them to the positive equivalents
1507 * (-1 gets converted to ac, -2 gets converted to (ac - 1), etc.).
1508 */
1509 len = (int)words.len;
1510 if (first < 0)
1511 first += len + 1;
1512 if (last < 0)
1513 last += len + 1;
1514
1515 /*
1516 * We avoid scanning more of the list than we need to.
1517 */
1518 if (first > last) {
1519 start = (first > len ? len : first) - 1;
1520 end = last < 1 ? 0 : last - 1;
1521 step = -1;
1522 } else {
1523 start = first < 1 ? 0 : first - 1;
1524 end = last > len ? len : last;
1525 step = 1;
1526 }
1527
1528 for (i = start; (step < 0) == (i >= end); i += step) {
1529 SepBuf_AddStr(&buf, words.words[i]);
1530 SepBuf_Sep(&buf);
1531 }
1532
1533 Words_Free(words);
1534
1535 return SepBuf_Destroy(&buf, FALSE);
1536 }
1537
1538
1539 /* Callback for ModifyWords to implement the :tA modifier.
1540 * Replace each word with the result of realpath() if successful. */
1541 static void
1542 ModifyWord_Realpath(const char *word, SepBuf *buf, void *data MAKE_ATTR_UNUSED)
1543 {
1544 struct stat st;
1545 char rbuf[MAXPATHLEN];
1546
1547 const char *rp = cached_realpath(word, rbuf);
1548 if (rp != NULL && *rp == '/' && stat(rp, &st) == 0)
1549 word = rp;
1550
1551 SepBuf_AddStr(buf, word);
1552 }
1553
1554 /* Modify each of the words of the passed string using the given function.
1555 *
1556 * Input:
1557 * str String whose words should be modified
1558 * modifyWord Function that modifies a single word
1559 * modifyWord_args Custom arguments for modifyWord
1560 *
1561 * Results:
1562 * A string of all the words modified appropriately.
1563 *-----------------------------------------------------------------------
1564 */
1565 static char *
1566 ModifyWords(const char *str,
1567 ModifyWordsCallback modifyWord, void *modifyWord_args,
1568 Boolean oneBigWord, char sep)
1569 {
1570 SepBuf result;
1571 Words words;
1572 size_t i;
1573
1574 if (oneBigWord) {
1575 SepBuf_Init(&result, sep);
1576 modifyWord(str, &result, modifyWord_args);
1577 return SepBuf_Destroy(&result, FALSE);
1578 }
1579
1580 SepBuf_Init(&result, sep);
1581
1582 words = Str_Words(str, FALSE);
1583
1584 VAR_DEBUG2("ModifyWords: split \"%s\" into %zu words\n", str, words.len);
1585
1586 for (i = 0; i < words.len; i++) {
1587 modifyWord(words.words[i], &result, modifyWord_args);
1588 if (Buf_Len(&result.buf) > 0)
1589 SepBuf_Sep(&result);
1590 }
1591
1592 Words_Free(words);
1593
1594 return SepBuf_Destroy(&result, FALSE);
1595 }
1596
1597
1598 static char *
1599 Words_JoinFree(Words words)
1600 {
1601 Buffer buf;
1602 size_t i;
1603
1604 Buf_Init(&buf);
1605
1606 for (i = 0; i < words.len; i++) {
1607 if (i != 0)
1608 Buf_AddByte(&buf, ' '); /* XXX: st->sep, for consistency */
1609 Buf_AddStr(&buf, words.words[i]);
1610 }
1611
1612 Words_Free(words);
1613
1614 return Buf_Destroy(&buf, FALSE);
1615 }
1616
1617 /* Remove adjacent duplicate words. */
1618 static char *
1619 VarUniq(const char *str)
1620 {
1621 Words words = Str_Words(str, FALSE);
1622
1623 if (words.len > 1) {
1624 size_t i, j;
1625 for (j = 0, i = 1; i < words.len; i++)
1626 if (strcmp(words.words[i], words.words[j]) != 0 && (++j != i))
1627 words.words[j] = words.words[i];
1628 words.len = j + 1;
1629 }
1630
1631 return Words_JoinFree(words);
1632 }
1633
1634
1635 /* Quote shell meta-characters and space characters in the string.
1636 * If quoteDollar is set, also quote and double any '$' characters. */
1637 static char *
1638 VarQuote(const char *str, Boolean quoteDollar)
1639 {
1640 Buffer buf;
1641 Buf_Init(&buf);
1642
1643 for (; *str != '\0'; str++) {
1644 if (*str == '\n') {
1645 const char *newline = Shell_GetNewline();
1646 if (newline == NULL)
1647 newline = "\\\n";
1648 Buf_AddStr(&buf, newline);
1649 continue;
1650 }
1651 if (ch_isspace(*str) || is_shell_metachar((unsigned char)*str))
1652 Buf_AddByte(&buf, '\\');
1653 Buf_AddByte(&buf, *str);
1654 if (quoteDollar && *str == '$')
1655 Buf_AddStr(&buf, "\\$");
1656 }
1657
1658 return Buf_Destroy(&buf, FALSE);
1659 }
1660
1661 /* Compute the 32-bit hash of the given string, using the MurmurHash3
1662 * algorithm. Output is encoded as 8 hex digits, in Little Endian order. */
1663 static char *
1664 VarHash(const char *str)
1665 {
1666 static const char hexdigits[16] = "0123456789abcdef";
1667 const unsigned char *ustr = (const unsigned char *)str;
1668
1669 uint32_t h = 0x971e137bU;
1670 uint32_t c1 = 0x95543787U;
1671 uint32_t c2 = 0x2ad7eb25U;
1672 size_t len2 = strlen(str);
1673
1674 char *buf;
1675 size_t i;
1676
1677 size_t len;
1678 for (len = len2; len; ) {
1679 uint32_t k = 0;
1680 switch (len) {
1681 default:
1682 k = ((uint32_t)ustr[3] << 24) |
1683 ((uint32_t)ustr[2] << 16) |
1684 ((uint32_t)ustr[1] << 8) |
1685 (uint32_t)ustr[0];
1686 len -= 4;
1687 ustr += 4;
1688 break;
1689 case 3:
1690 k |= (uint32_t)ustr[2] << 16;
1691 /* FALLTHROUGH */
1692 case 2:
1693 k |= (uint32_t)ustr[1] << 8;
1694 /* FALLTHROUGH */
1695 case 1:
1696 k |= (uint32_t)ustr[0];
1697 len = 0;
1698 }
1699 c1 = c1 * 5 + 0x7b7d159cU;
1700 c2 = c2 * 5 + 0x6bce6396U;
1701 k *= c1;
1702 k = (k << 11) ^ (k >> 21);
1703 k *= c2;
1704 h = (h << 13) ^ (h >> 19);
1705 h = h * 5 + 0x52dce729U;
1706 h ^= k;
1707 }
1708 h ^= (uint32_t)len2;
1709 h *= 0x85ebca6b;
1710 h ^= h >> 13;
1711 h *= 0xc2b2ae35;
1712 h ^= h >> 16;
1713
1714 buf = bmake_malloc(9);
1715 for (i = 0; i < 8; i++) {
1716 buf[i] = hexdigits[h & 0x0f];
1717 h >>= 4;
1718 }
1719 buf[8] = '\0';
1720 return buf;
1721 }
1722
1723 static char *
1724 VarStrftime(const char *fmt, Boolean zulu, time_t tim)
1725 {
1726 char buf[BUFSIZ];
1727
1728 if (tim == 0)
1729 time(&tim);
1730 if (*fmt == '\0')
1731 fmt = "%c";
1732 strftime(buf, sizeof buf, fmt, zulu ? gmtime(&tim) : localtime(&tim));
1733
1734 buf[sizeof buf - 1] = '\0';
1735 return bmake_strdup(buf);
1736 }
1737
1738 /*
1739 * The ApplyModifier functions take an expression that is being evaluated.
1740 * Their task is to apply a single modifier to the expression.
1741 * To do this, they parse the modifier and its parameters from pp and apply
1742 * the parsed modifier to the current value of the expression, generating a
1743 * new value from it.
1744 *
1745 * The modifier typically lasts until the next ':', or a closing '}' or ')'
1746 * (taken from st->endc), or the end of the string (parse error).
1747 *
1748 * The high-level behavior of these functions is:
1749 *
1750 * 1. parse the modifier
1751 * 2. evaluate the modifier
1752 * 3. housekeeping
1753 *
1754 * Parsing the modifier
1755 *
1756 * If parsing succeeds, the parsing position *pp is updated to point to the
1757 * first character following the modifier, which typically is either ':' or
1758 * st->endc. The modifier doesn't have to check for this delimiter character,
1759 * this is done by ApplyModifiers.
1760 *
1761 * XXX: As of 2020-11-15, some modifiers such as :S, :C, :P, :L do not
1762 * need to be followed by a ':' or endc; this was an unintended mistake.
1763 *
1764 * If parsing fails because of a missing delimiter (as in the :S, :C or :@
1765 * modifiers), return AMR_CLEANUP.
1766 *
1767 * If parsing fails because the modifier is unknown, return AMR_UNKNOWN to
1768 * try the SysV modifier ${VAR:from=to} as fallback. This should only be
1769 * done as long as there have been no side effects from evaluating nested
1770 * variables, to avoid evaluating them more than once. In this case, the
1771 * parsing position may or may not be updated. (XXX: Why not? The original
1772 * parsing position is well-known in ApplyModifiers.)
1773 *
1774 * If parsing fails and the SysV modifier ${VAR:from=to} should not be used
1775 * as a fallback, either issue an error message using Error or Parse_Error
1776 * and then return AMR_CLEANUP, or return AMR_BAD for the default error
1777 * message. Both of these return values will stop processing the variable
1778 * expression. (XXX: As of 2020-08-23, evaluation of the whole string
1779 * continues nevertheless after skipping a few bytes, which essentially is
1780 * undefined behavior. Not in the sense of C, but still it's impossible to
1781 * predict what happens in the parser.)
1782 *
1783 * Evaluating the modifier
1784 *
1785 * After parsing, the modifier is evaluated. The side effects from evaluating
1786 * nested variable expressions in the modifier text often already happen
1787 * during parsing though.
1788 *
1789 * Evaluating the modifier usually takes the current value of the variable
1790 * expression from st->val, or the variable name from st->var->name and stores
1791 * the result in st->newVal.
1792 *
1793 * If evaluating fails (as of 2020-08-23), an error message is printed using
1794 * Error. This function has no side-effects, it really just prints the error
1795 * message. Processing the expression continues as if everything were ok.
1796 * XXX: This should be fixed by adding proper error handling to Var_Subst,
1797 * Var_Parse, ApplyModifiers and ModifyWords.
1798 *
1799 * Housekeeping
1800 *
1801 * Some modifiers such as :D and :U turn undefined expressions into defined
1802 * expressions (see VEF_UNDEF, VEF_DEF).
1803 *
1804 * Some modifiers need to free some memory.
1805 */
1806
1807 typedef enum VarExprFlags {
1808 /* The variable expression is based on an undefined variable. */
1809 VEF_UNDEF = 0x01,
1810 /* The variable expression started as an undefined expression, but one
1811 * of the modifiers (such as :D or :U) has turned the expression from
1812 * undefined to defined. */
1813 VEF_DEF = 0x02
1814 } VarExprFlags;
1815
1816 ENUM_FLAGS_RTTI_2(VarExprFlags,
1817 VEF_UNDEF, VEF_DEF);
1818
1819
1820 typedef struct ApplyModifiersState {
1821 const char startc; /* '\0' or '{' or '(' */
1822 const char endc; /* '\0' or '}' or ')' */
1823 Var * const var;
1824 GNode * const ctxt;
1825 const VarEvalFlags eflags;
1826
1827 char *val; /* The old value of the expression,
1828 * before applying the modifier, never NULL */
1829 char *newVal; /* The new value of the expression,
1830 * after applying the modifier, never NULL */
1831 char sep; /* Word separator in expansions
1832 * (see the :ts modifier) */
1833 Boolean oneBigWord; /* TRUE if some modifiers that otherwise split
1834 * the variable value into words, like :S and
1835 * :C, treat the variable value as a single big
1836 * word, possibly containing spaces. */
1837 VarExprFlags exprFlags;
1838 } ApplyModifiersState;
1839
1840 static void
1841 ApplyModifiersState_Define(ApplyModifiersState *st)
1842 {
1843 if (st->exprFlags & VEF_UNDEF)
1844 st->exprFlags |= VEF_DEF;
1845 }
1846
1847 typedef enum ApplyModifierResult {
1848 AMR_OK, /* Continue parsing */
1849 AMR_UNKNOWN, /* Not a match, try other modifiers as well */
1850 AMR_BAD, /* Error out with "Bad modifier" message */
1851 AMR_CLEANUP /* Error out without error message */
1852 } ApplyModifierResult;
1853
1854 /* Allow backslashes to escape the delimiter, $, and \, but don't touch other
1855 * backslashes. */
1856 static Boolean
1857 IsEscapedModifierPart(const char *p, char delim,
1858 struct ModifyWord_SubstArgs *subst)
1859 {
1860 if (p[0] != '\\')
1861 return FALSE;
1862 if (p[1] == delim || p[1] == '\\' || p[1] == '$')
1863 return TRUE;
1864 return p[1] == '&' && subst != NULL;
1865 }
1866
1867 /*
1868 * Parse a part of a modifier such as the "from" and "to" in :S/from/to/ or
1869 * the "var" or "replacement ${var}" in :@var@replacement ${var}@, up to and
1870 * including the next unescaped delimiter. The delimiter, as well as the
1871 * backslash or the dollar, can be escaped with a backslash.
1872 *
1873 * Return the parsed (and possibly expanded) string, or NULL if no delimiter
1874 * was found. On successful return, the parsing position pp points right
1875 * after the delimiter. The delimiter is not included in the returned
1876 * value though.
1877 */
1878 static VarParseResult
1879 ParseModifierPart(
1880 const char **pp, /* The parsing position, updated upon return */
1881 char delim, /* Parsing stops at this delimiter */
1882 VarEvalFlags eflags, /* Flags for evaluating nested variables;
1883 * if VARE_WANTRES is not set, the text is
1884 * only parsed */
1885 ApplyModifiersState *st,
1886 char **out_part,
1887 size_t *out_length, /* Optionally stores the length of the returned
1888 * string, just to save another strlen call. */
1889 VarPatternFlags *out_pflags,/* For the first part of the :S modifier,
1890 * sets the VARP_ANCHOR_END flag if the last
1891 * character of the pattern is a $. */
1892 struct ModifyWord_SubstArgs *subst
1893 /* For the second part of the :S modifier,
1894 * allow ampersands to be escaped and replace
1895 * unescaped ampersands with subst->lhs. */
1896 ) {
1897 Buffer buf;
1898 const char *p;
1899
1900 Buf_Init(&buf);
1901
1902 /*
1903 * Skim through until the matching delimiter is found; pick up variable
1904 * expressions on the way.
1905 */
1906 p = *pp;
1907 while (*p != '\0' && *p != delim) {
1908 const char *varstart;
1909
1910 if (IsEscapedModifierPart(p, delim, subst)) {
1911 Buf_AddByte(&buf, p[1]);
1912 p += 2;
1913 continue;
1914 }
1915
1916 if (*p != '$') { /* Unescaped, simple text */
1917 if (subst != NULL && *p == '&')
1918 Buf_AddBytes(&buf, subst->lhs, subst->lhsLen);
1919 else
1920 Buf_AddByte(&buf, *p);
1921 p++;
1922 continue;
1923 }
1924
1925 if (p[1] == delim) { /* Unescaped $ at end of pattern */
1926 if (out_pflags != NULL)
1927 *out_pflags |= VARP_ANCHOR_END;
1928 else
1929 Buf_AddByte(&buf, *p);
1930 p++;
1931 continue;
1932 }
1933
1934 if (eflags & VARE_WANTRES) { /* Nested variable, evaluated */
1935 const char *nested_p = p;
1936 const char *nested_val;
1937 void *nested_val_freeIt;
1938 VarEvalFlags nested_eflags = eflags & ~(unsigned)VARE_KEEP_DOLLAR;
1939
1940 (void)Var_Parse(&nested_p, st->ctxt, nested_eflags,
1941 &nested_val, &nested_val_freeIt);
1942 /* TODO: handle errors */
1943 Buf_AddStr(&buf, nested_val);
1944 free(nested_val_freeIt);
1945 p += nested_p - p;
1946 continue;
1947 }
1948
1949 /* XXX: This whole block is very similar to Var_Parse without
1950 * VARE_WANTRES. There may be subtle edge cases though that are
1951 * not yet covered in the unit tests and that are parsed differently,
1952 * depending on whether they are evaluated or not.
1953 *
1954 * This subtle difference is not documented in the manual page,
1955 * neither is the difference between parsing :D and :M documented.
1956 * No code should ever depend on these details, but who knows. */
1957
1958 varstart = p; /* Nested variable, only parsed */
1959 if (p[1] == '(' || p[1] == '{') {
1960 /*
1961 * Find the end of this variable reference
1962 * and suck it in without further ado.
1963 * It will be interpreted later.
1964 */
1965 char startc = p[1];
1966 int endc = startc == '(' ? ')' : '}';
1967 int depth = 1;
1968
1969 for (p += 2; *p != '\0' && depth > 0; p++) {
1970 if (p[-1] != '\\') {
1971 if (*p == startc)
1972 depth++;
1973 if (*p == endc)
1974 depth--;
1975 }
1976 }
1977 Buf_AddBytesBetween(&buf, varstart, p);
1978 } else {
1979 Buf_AddByte(&buf, *varstart);
1980 p++;
1981 }
1982 }
1983
1984 if (*p != delim) {
1985 *pp = p;
1986 Error("Unfinished modifier for %s ('%c' missing)",
1987 st->var->name, delim);
1988 *out_part = NULL;
1989 return VPR_PARSE_MSG;
1990 }
1991
1992 *pp = ++p;
1993 if (out_length != NULL)
1994 *out_length = Buf_Len(&buf);
1995
1996 *out_part = Buf_Destroy(&buf, FALSE);
1997 VAR_DEBUG1("Modifier part: \"%s\"\n", *out_part);
1998 return VPR_OK;
1999 }
2000
2001 /* Test whether mod starts with modname, followed by a delimiter. */
2002 MAKE_INLINE Boolean
2003 ModMatch(const char *mod, const char *modname, char endc)
2004 {
2005 size_t n = strlen(modname);
2006 return strncmp(mod, modname, n) == 0 &&
2007 (mod[n] == endc || mod[n] == ':');
2008 }
2009
2010 /* Test whether mod starts with modname, followed by a delimiter or '='. */
2011 MAKE_INLINE Boolean
2012 ModMatchEq(const char *mod, const char *modname, char endc)
2013 {
2014 size_t n = strlen(modname);
2015 return strncmp(mod, modname, n) == 0 &&
2016 (mod[n] == endc || mod[n] == ':' || mod[n] == '=');
2017 }
2018
2019 static Boolean
2020 TryParseIntBase0(const char **pp, int *out_num)
2021 {
2022 char *end;
2023 long n;
2024
2025 errno = 0;
2026 n = strtol(*pp, &end, 0);
2027 if ((n == LONG_MIN || n == LONG_MAX) && errno == ERANGE)
2028 return FALSE;
2029 if (n < INT_MIN || n > INT_MAX)
2030 return FALSE;
2031
2032 *pp = end;
2033 *out_num = (int)n;
2034 return TRUE;
2035 }
2036
2037 static Boolean
2038 TryParseSize(const char **pp, size_t *out_num)
2039 {
2040 char *end;
2041 unsigned long n;
2042
2043 if (!ch_isdigit(**pp))
2044 return FALSE;
2045
2046 errno = 0;
2047 n = strtoul(*pp, &end, 10);
2048 if (n == ULONG_MAX && errno == ERANGE)
2049 return FALSE;
2050 if (n > SIZE_MAX)
2051 return FALSE;
2052
2053 *pp = end;
2054 *out_num = (size_t)n;
2055 return TRUE;
2056 }
2057
2058 static Boolean
2059 TryParseChar(const char **pp, int base, char *out_ch)
2060 {
2061 char *end;
2062 unsigned long n;
2063
2064 if (!ch_isalnum(**pp))
2065 return FALSE;
2066
2067 errno = 0;
2068 n = strtoul(*pp, &end, base);
2069 if (n == ULONG_MAX && errno == ERANGE)
2070 return FALSE;
2071 if (n > UCHAR_MAX)
2072 return FALSE;
2073
2074 *pp = end;
2075 *out_ch = (char)n;
2076 return TRUE;
2077 }
2078
2079 /* :@var (at) ...${var}...@ */
2080 static ApplyModifierResult
2081 ApplyModifier_Loop(const char **pp, ApplyModifiersState *st)
2082 {
2083 struct ModifyWord_LoopArgs args;
2084 char prev_sep;
2085 VarParseResult res;
2086
2087 args.ctx = st->ctxt;
2088
2089 (*pp)++; /* Skip the first '@' */
2090 res = ParseModifierPart(pp, '@', VARE_NONE, st,
2091 &args.tvar, NULL, NULL, NULL);
2092 if (res != VPR_OK)
2093 return AMR_CLEANUP;
2094 if (opts.lint && strchr(args.tvar, '$') != NULL) {
2095 Parse_Error(PARSE_FATAL,
2096 "In the :@ modifier of \"%s\", the variable name \"%s\" "
2097 "must not contain a dollar.",
2098 st->var->name, args.tvar);
2099 return AMR_CLEANUP;
2100 }
2101
2102 res = ParseModifierPart(pp, '@', VARE_NONE, st,
2103 &args.str, NULL, NULL, NULL);
2104 if (res != VPR_OK)
2105 return AMR_CLEANUP;
2106
2107 args.eflags = st->eflags & ~(unsigned)VARE_KEEP_DOLLAR;
2108 prev_sep = st->sep;
2109 st->sep = ' '; /* XXX: should be st->sep for consistency */
2110 st->newVal = ModifyWords(st->val, ModifyWord_Loop, &args,
2111 st->oneBigWord, st->sep);
2112 st->sep = prev_sep;
2113 /* XXX: Consider restoring the previous variable instead of deleting. */
2114 Var_Delete(args.tvar, st->ctxt);
2115 free(args.tvar);
2116 free(args.str);
2117 return AMR_OK;
2118 }
2119
2120 /* :Ddefined or :Uundefined */
2121 static ApplyModifierResult
2122 ApplyModifier_Defined(const char **pp, ApplyModifiersState *st)
2123 {
2124 Buffer buf;
2125 const char *p;
2126
2127 VarEvalFlags eflags = VARE_NONE;
2128 if (st->eflags & VARE_WANTRES)
2129 if ((**pp == 'D') == !(st->exprFlags & VEF_UNDEF))
2130 eflags = st->eflags;
2131
2132 Buf_Init(&buf);
2133 p = *pp + 1;
2134 while (*p != st->endc && *p != ':' && *p != '\0') {
2135
2136 /* XXX: This code is similar to the one in Var_Parse.
2137 * See if the code can be merged.
2138 * See also ApplyModifier_Match. */
2139
2140 /* Escaped delimiter or other special character */
2141 if (*p == '\\') {
2142 char c = p[1];
2143 if (c == st->endc || c == ':' || c == '$' || c == '\\') {
2144 Buf_AddByte(&buf, c);
2145 p += 2;
2146 continue;
2147 }
2148 }
2149
2150 /* Nested variable expression */
2151 if (*p == '$') {
2152 const char *nested_val;
2153 void *nested_val_freeIt;
2154
2155 (void)Var_Parse(&p, st->ctxt, eflags,
2156 &nested_val, &nested_val_freeIt);
2157 /* TODO: handle errors */
2158 Buf_AddStr(&buf, nested_val);
2159 free(nested_val_freeIt);
2160 continue;
2161 }
2162
2163 /* Ordinary text */
2164 Buf_AddByte(&buf, *p);
2165 p++;
2166 }
2167 *pp = p;
2168
2169 ApplyModifiersState_Define(st);
2170
2171 if (eflags & VARE_WANTRES) {
2172 st->newVal = Buf_Destroy(&buf, FALSE);
2173 } else {
2174 st->newVal = st->val;
2175 Buf_Destroy(&buf, TRUE);
2176 }
2177 return AMR_OK;
2178 }
2179
2180 /* :L */
2181 static ApplyModifierResult
2182 ApplyModifier_Literal(const char **pp, ApplyModifiersState *st)
2183 {
2184 ApplyModifiersState_Define(st);
2185 st->newVal = bmake_strdup(st->var->name);
2186 (*pp)++;
2187 return AMR_OK;
2188 }
2189
2190 static Boolean
2191 TryParseTime(const char **pp, time_t *out_time)
2192 {
2193 char *end;
2194 unsigned long n;
2195
2196 if (!ch_isdigit(**pp))
2197 return FALSE;
2198
2199 errno = 0;
2200 n = strtoul(*pp, &end, 10);
2201 if (n == ULONG_MAX && errno == ERANGE)
2202 return FALSE;
2203
2204 *pp = end;
2205 *out_time = (time_t)n; /* ignore possible truncation for now */
2206 return TRUE;
2207 }
2208
2209 /* :gmtime */
2210 static ApplyModifierResult
2211 ApplyModifier_Gmtime(const char **pp, ApplyModifiersState *st)
2212 {
2213 time_t utc;
2214
2215 const char *mod = *pp;
2216 if (!ModMatchEq(mod, "gmtime", st->endc))
2217 return AMR_UNKNOWN;
2218
2219 if (mod[6] == '=') {
2220 const char *arg = mod + 7;
2221 if (!TryParseTime(&arg, &utc)) {
2222 Parse_Error(PARSE_FATAL, "Invalid time value: %s\n", mod + 7);
2223 return AMR_CLEANUP;
2224 }
2225 *pp = arg;
2226 } else {
2227 utc = 0;
2228 *pp = mod + 6;
2229 }
2230 st->newVal = VarStrftime(st->val, TRUE, utc);
2231 return AMR_OK;
2232 }
2233
2234 /* :localtime */
2235 static ApplyModifierResult
2236 ApplyModifier_Localtime(const char **pp, ApplyModifiersState *st)
2237 {
2238 time_t utc;
2239
2240 const char *mod = *pp;
2241 if (!ModMatchEq(mod, "localtime", st->endc))
2242 return AMR_UNKNOWN;
2243
2244 if (mod[9] == '=') {
2245 const char *arg = mod + 10;
2246 if (!TryParseTime(&arg, &utc)) {
2247 Parse_Error(PARSE_FATAL, "Invalid time value: %s\n", mod + 10);
2248 return AMR_CLEANUP;
2249 }
2250 *pp = arg;
2251 } else {
2252 utc = 0;
2253 *pp = mod + 9;
2254 }
2255 st->newVal = VarStrftime(st->val, FALSE, utc);
2256 return AMR_OK;
2257 }
2258
2259 /* :hash */
2260 static ApplyModifierResult
2261 ApplyModifier_Hash(const char **pp, ApplyModifiersState *st)
2262 {
2263 if (!ModMatch(*pp, "hash", st->endc))
2264 return AMR_UNKNOWN;
2265
2266 st->newVal = VarHash(st->val);
2267 *pp += 4;
2268 return AMR_OK;
2269 }
2270
2271 /* :P */
2272 static ApplyModifierResult
2273 ApplyModifier_Path(const char **pp, ApplyModifiersState *st)
2274 {
2275 GNode *gn;
2276 char *path;
2277
2278 ApplyModifiersState_Define(st);
2279
2280 gn = Targ_FindNode(st->var->name);
2281 if (gn == NULL || gn->type & OP_NOPATH) {
2282 path = NULL;
2283 } else if (gn->path != NULL) {
2284 path = bmake_strdup(gn->path);
2285 } else {
2286 SearchPath *searchPath = Suff_FindPath(gn);
2287 path = Dir_FindFile(st->var->name, searchPath);
2288 }
2289 if (path == NULL)
2290 path = bmake_strdup(st->var->name);
2291 st->newVal = path;
2292
2293 (*pp)++;
2294 return AMR_OK;
2295 }
2296
2297 /* :!cmd! */
2298 static ApplyModifierResult
2299 ApplyModifier_ShellCommand(const char **pp, ApplyModifiersState *st)
2300 {
2301 char *cmd;
2302 const char *errfmt;
2303 VarParseResult res;
2304
2305 (*pp)++;
2306 res = ParseModifierPart(pp, '!', st->eflags, st,
2307 &cmd, NULL, NULL, NULL);
2308 if (res != VPR_OK)
2309 return AMR_CLEANUP;
2310
2311 errfmt = NULL;
2312 if (st->eflags & VARE_WANTRES)
2313 st->newVal = Cmd_Exec(cmd, &errfmt);
2314 else
2315 st->newVal = bmake_strdup("");
2316 if (errfmt != NULL)
2317 Error(errfmt, cmd); /* XXX: why still return AMR_OK? */
2318 free(cmd);
2319
2320 ApplyModifiersState_Define(st);
2321 return AMR_OK;
2322 }
2323
2324 /* The :range modifier generates an integer sequence as long as the words.
2325 * The :range=7 modifier generates an integer sequence from 1 to 7. */
2326 static ApplyModifierResult
2327 ApplyModifier_Range(const char **pp, ApplyModifiersState *st)
2328 {
2329 size_t n;
2330 Buffer buf;
2331 size_t i;
2332
2333 const char *mod = *pp;
2334 if (!ModMatchEq(mod, "range", st->endc))
2335 return AMR_UNKNOWN;
2336
2337 if (mod[5] == '=') {
2338 const char *p = mod + 6;
2339 if (!TryParseSize(&p, &n)) {
2340 Parse_Error(PARSE_FATAL, "Invalid number: %s\n", mod + 6);
2341 return AMR_CLEANUP;
2342 }
2343 *pp = p;
2344 } else {
2345 n = 0;
2346 *pp = mod + 5;
2347 }
2348
2349 if (n == 0) {
2350 Words words = Str_Words(st->val, FALSE);
2351 n = words.len;
2352 Words_Free(words);
2353 }
2354
2355 Buf_Init(&buf);
2356
2357 for (i = 0; i < n; i++) {
2358 if (i != 0)
2359 Buf_AddByte(&buf, ' '); /* XXX: st->sep, for consistency */
2360 Buf_AddInt(&buf, 1 + (int)i);
2361 }
2362
2363 st->newVal = Buf_Destroy(&buf, FALSE);
2364 return AMR_OK;
2365 }
2366
2367 /* :Mpattern or :Npattern */
2368 static ApplyModifierResult
2369 ApplyModifier_Match(const char **pp, ApplyModifiersState *st)
2370 {
2371 const char *mod = *pp;
2372 Boolean copy = FALSE; /* pattern should be, or has been, copied */
2373 Boolean needSubst = FALSE;
2374 const char *endpat;
2375 char *pattern;
2376 ModifyWordsCallback callback;
2377
2378 /*
2379 * In the loop below, ignore ':' unless we are at (or back to) the
2380 * original brace level.
2381 * XXX: This will likely not work right if $() and ${} are intermixed.
2382 */
2383 /* XXX: This code is similar to the one in Var_Parse.
2384 * See if the code can be merged.
2385 * See also ApplyModifier_Defined. */
2386 int nest = 0;
2387 const char *p;
2388 for (p = mod + 1; *p != '\0' && !(*p == ':' && nest == 0); p++) {
2389 if (*p == '\\' &&
2390 (p[1] == ':' || p[1] == st->endc || p[1] == st->startc)) {
2391 if (!needSubst)
2392 copy = TRUE;
2393 p++;
2394 continue;
2395 }
2396 if (*p == '$')
2397 needSubst = TRUE;
2398 if (*p == '(' || *p == '{')
2399 nest++;
2400 if (*p == ')' || *p == '}') {
2401 nest--;
2402 if (nest < 0)
2403 break;
2404 }
2405 }
2406 *pp = p;
2407 endpat = p;
2408
2409 if (copy) {
2410 char *dst;
2411 const char *src;
2412
2413 /* Compress the \:'s out of the pattern. */
2414 pattern = bmake_malloc((size_t)(endpat - (mod + 1)) + 1);
2415 dst = pattern;
2416 src = mod + 1;
2417 for (; src < endpat; src++, dst++) {
2418 if (src[0] == '\\' && src + 1 < endpat &&
2419 /* XXX: st->startc is missing here; see above */
2420 (src[1] == ':' || src[1] == st->endc))
2421 src++;
2422 *dst = *src;
2423 }
2424 *dst = '\0';
2425 endpat = dst;
2426 } else {
2427 pattern = bmake_strsedup(mod + 1, endpat);
2428 }
2429
2430 if (needSubst) {
2431 /* pattern contains embedded '$', so use Var_Subst to expand it. */
2432 char *old_pattern = pattern;
2433 (void)Var_Subst(pattern, st->ctxt, st->eflags, &pattern);
2434 /* TODO: handle errors */
2435 free(old_pattern);
2436 }
2437
2438 VAR_DEBUG3("Pattern[%s] for [%s] is [%s]\n",
2439 st->var->name, st->val, pattern);
2440
2441 callback = mod[0] == 'M' ? ModifyWord_Match : ModifyWord_NoMatch;
2442 st->newVal = ModifyWords(st->val, callback, pattern,
2443 st->oneBigWord, st->sep);
2444 free(pattern);
2445 return AMR_OK;
2446 }
2447
2448 /* :S,from,to, */
2449 static ApplyModifierResult
2450 ApplyModifier_Subst(const char **pp, ApplyModifiersState *st)
2451 {
2452 struct ModifyWord_SubstArgs args;
2453 char *lhs, *rhs;
2454 Boolean oneBigWord;
2455 VarParseResult res;
2456
2457 char delim = (*pp)[1];
2458 if (delim == '\0') {
2459 Error("Missing delimiter for :S modifier");
2460 (*pp)++;
2461 return AMR_CLEANUP;
2462 }
2463
2464 *pp += 2;
2465
2466 args.pflags = 0;
2467 args.matched = FALSE;
2468
2469 /*
2470 * If pattern begins with '^', it is anchored to the
2471 * start of the word -- skip over it and flag pattern.
2472 */
2473 if (**pp == '^') {
2474 args.pflags |= VARP_ANCHOR_START;
2475 (*pp)++;
2476 }
2477
2478 res = ParseModifierPart(pp, delim, st->eflags, st,
2479 &lhs, &args.lhsLen, &args.pflags, NULL);
2480 if (res != VPR_OK)
2481 return AMR_CLEANUP;
2482 args.lhs = lhs;
2483
2484 res = ParseModifierPart(pp, delim, st->eflags, st,
2485 &rhs, &args.rhsLen, NULL, &args);
2486 if (res != VPR_OK)
2487 return AMR_CLEANUP;
2488 args.rhs = rhs;
2489
2490 oneBigWord = st->oneBigWord;
2491 for (;; (*pp)++) {
2492 switch (**pp) {
2493 case 'g':
2494 args.pflags |= VARP_SUB_GLOBAL;
2495 continue;
2496 case '1':
2497 args.pflags |= VARP_SUB_ONE;
2498 continue;
2499 case 'W':
2500 oneBigWord = TRUE;
2501 continue;
2502 }
2503 break;
2504 }
2505
2506 st->newVal = ModifyWords(st->val, ModifyWord_Subst, &args,
2507 oneBigWord, st->sep);
2508
2509 free(lhs);
2510 free(rhs);
2511 return AMR_OK;
2512 }
2513
2514 #ifndef NO_REGEX
2515
2516 /* :C,from,to, */
2517 static ApplyModifierResult
2518 ApplyModifier_Regex(const char **pp, ApplyModifiersState *st)
2519 {
2520 char *re;
2521 struct ModifyWord_SubstRegexArgs args;
2522 Boolean oneBigWord;
2523 int error;
2524 VarParseResult res;
2525
2526 char delim = (*pp)[1];
2527 if (delim == '\0') {
2528 Error("Missing delimiter for :C modifier");
2529 (*pp)++;
2530 return AMR_CLEANUP;
2531 }
2532
2533 *pp += 2;
2534
2535 res = ParseModifierPart(pp, delim, st->eflags, st,
2536 &re, NULL, NULL, NULL);
2537 if (res != VPR_OK)
2538 return AMR_CLEANUP;
2539
2540 res = ParseModifierPart(pp, delim, st->eflags, st,
2541 &args.replace, NULL, NULL, NULL);
2542 if (args.replace == NULL) {
2543 free(re);
2544 return AMR_CLEANUP;
2545 }
2546
2547 args.pflags = 0;
2548 args.matched = FALSE;
2549 oneBigWord = st->oneBigWord;
2550 for (;; (*pp)++) {
2551 switch (**pp) {
2552 case 'g':
2553 args.pflags |= VARP_SUB_GLOBAL;
2554 continue;
2555 case '1':
2556 args.pflags |= VARP_SUB_ONE;
2557 continue;
2558 case 'W':
2559 oneBigWord = TRUE;
2560 continue;
2561 }
2562 break;
2563 }
2564
2565 error = regcomp(&args.re, re, REG_EXTENDED);
2566 free(re);
2567 if (error != 0) {
2568 VarREError(error, &args.re, "Regex compilation error");
2569 free(args.replace);
2570 return AMR_CLEANUP;
2571 }
2572
2573 args.nsub = args.re.re_nsub + 1;
2574 if (args.nsub > 10)
2575 args.nsub = 10;
2576 st->newVal = ModifyWords(st->val, ModifyWord_SubstRegex, &args,
2577 oneBigWord, st->sep);
2578 regfree(&args.re);
2579 free(args.replace);
2580 return AMR_OK;
2581 }
2582 #endif
2583
2584 /* :Q, :q */
2585 static ApplyModifierResult
2586 ApplyModifier_Quote(const char **pp, ApplyModifiersState *st)
2587 {
2588 if ((*pp)[1] == st->endc || (*pp)[1] == ':') {
2589 st->newVal = VarQuote(st->val, **pp == 'q');
2590 (*pp)++;
2591 return AMR_OK;
2592 } else
2593 return AMR_UNKNOWN;
2594 }
2595
2596 static void
2597 ModifyWord_Copy(const char *word, SepBuf *buf, void *data MAKE_ATTR_UNUSED)
2598 {
2599 SepBuf_AddStr(buf, word);
2600 }
2601
2602 /* :ts<separator> */
2603 static ApplyModifierResult
2604 ApplyModifier_ToSep(const char **pp, ApplyModifiersState *st)
2605 {
2606 const char *sep = *pp + 2;
2607
2608 /* ":ts<any><endc>" or ":ts<any>:" */
2609 if (sep[0] != st->endc && (sep[1] == st->endc || sep[1] == ':')) {
2610 st->sep = sep[0];
2611 *pp = sep + 1;
2612 goto ok;
2613 }
2614
2615 /* ":ts<endc>" or ":ts:" */
2616 if (sep[0] == st->endc || sep[0] == ':') {
2617 st->sep = '\0'; /* no separator */
2618 *pp = sep;
2619 goto ok;
2620 }
2621
2622 /* ":ts<unrecognised><unrecognised>". */
2623 if (sep[0] != '\\') {
2624 (*pp)++; /* just for backwards compatibility */
2625 return AMR_BAD;
2626 }
2627
2628 /* ":ts\n" */
2629 if (sep[1] == 'n') {
2630 st->sep = '\n';
2631 *pp = sep + 2;
2632 goto ok;
2633 }
2634
2635 /* ":ts\t" */
2636 if (sep[1] == 't') {
2637 st->sep = '\t';
2638 *pp = sep + 2;
2639 goto ok;
2640 }
2641
2642 /* ":ts\x40" or ":ts\100" */
2643 {
2644 const char *p = sep + 1;
2645 int base = 8; /* assume octal */
2646
2647 if (sep[1] == 'x') {
2648 base = 16;
2649 p++;
2650 } else if (!ch_isdigit(sep[1])) {
2651 (*pp)++; /* just for backwards compatibility */
2652 return AMR_BAD; /* ":ts<backslash><unrecognised>". */
2653 }
2654
2655 if (!TryParseChar(&p, base, &st->sep)) {
2656 Parse_Error(PARSE_FATAL, "Invalid character number: %s\n", p);
2657 return AMR_CLEANUP;
2658 }
2659 if (*p != ':' && *p != st->endc) {
2660 (*pp)++; /* just for backwards compatibility */
2661 return AMR_BAD;
2662 }
2663
2664 *pp = p;
2665 }
2666
2667 ok:
2668 st->newVal = ModifyWords(st->val, ModifyWord_Copy, NULL,
2669 st->oneBigWord, st->sep);
2670 return AMR_OK;
2671 }
2672
2673 /* :tA, :tu, :tl, :ts<separator>, etc. */
2674 static ApplyModifierResult
2675 ApplyModifier_To(const char **pp, ApplyModifiersState *st)
2676 {
2677 const char *mod = *pp;
2678 assert(mod[0] == 't');
2679
2680 if (mod[1] == st->endc || mod[1] == ':' || mod[1] == '\0') {
2681 *pp = mod + 1;
2682 return AMR_BAD; /* Found ":t<endc>" or ":t:". */
2683 }
2684
2685 if (mod[1] == 's')
2686 return ApplyModifier_ToSep(pp, st);
2687
2688 if (mod[2] != st->endc && mod[2] != ':') {
2689 *pp = mod + 1;
2690 return AMR_BAD; /* Found ":t<unrecognised><unrecognised>". */
2691 }
2692
2693 /* Check for two-character options: ":tu", ":tl" */
2694 if (mod[1] == 'A') { /* absolute path */
2695 st->newVal = ModifyWords(st->val, ModifyWord_Realpath, NULL,
2696 st->oneBigWord, st->sep);
2697 *pp = mod + 2;
2698 return AMR_OK;
2699 }
2700
2701 if (mod[1] == 'u') { /* :tu */
2702 size_t i;
2703 size_t len = strlen(st->val);
2704 st->newVal = bmake_malloc(len + 1);
2705 for (i = 0; i < len + 1; i++)
2706 st->newVal[i] = ch_toupper(st->val[i]);
2707 *pp = mod + 2;
2708 return AMR_OK;
2709 }
2710
2711 if (mod[1] == 'l') { /* :tl */
2712 size_t i;
2713 size_t len = strlen(st->val);
2714 st->newVal = bmake_malloc(len + 1);
2715 for (i = 0; i < len + 1; i++)
2716 st->newVal[i] = ch_tolower(st->val[i]);
2717 *pp = mod + 2;
2718 return AMR_OK;
2719 }
2720
2721 if (mod[1] == 'W' || mod[1] == 'w') { /* :tW, :tw */
2722 st->oneBigWord = mod[1] == 'W';
2723 st->newVal = st->val;
2724 *pp = mod + 2;
2725 return AMR_OK;
2726 }
2727
2728 /* Found ":t<unrecognised>:" or ":t<unrecognised><endc>". */
2729 *pp = mod + 1;
2730 return AMR_BAD;
2731 }
2732
2733 /* :[#], :[1], :[-1..1], etc. */
2734 static ApplyModifierResult
2735 ApplyModifier_Words(const char **pp, ApplyModifiersState *st)
2736 {
2737 char *estr;
2738 int first, last;
2739 VarParseResult res;
2740 const char *p;
2741
2742 (*pp)++; /* skip the '[' */
2743 res = ParseModifierPart(pp, ']', st->eflags, st,
2744 &estr, NULL, NULL, NULL);
2745 if (res != VPR_OK)
2746 return AMR_CLEANUP;
2747
2748 /* now *pp points just after the closing ']' */
2749 if (**pp != ':' && **pp != st->endc)
2750 goto bad_modifier; /* Found junk after ']' */
2751
2752 if (estr[0] == '\0')
2753 goto bad_modifier; /* empty square brackets in ":[]". */
2754
2755 if (estr[0] == '#' && estr[1] == '\0') { /* Found ":[#]" */
2756 if (st->oneBigWord) {
2757 st->newVal = bmake_strdup("1");
2758 } else {
2759 Buffer buf;
2760
2761 Words words = Str_Words(st->val, FALSE);
2762 size_t ac = words.len;
2763 Words_Free(words);
2764
2765 Buf_InitSize(&buf, 4); /* 3 digits + '\0' is usually enough */
2766 Buf_AddInt(&buf, (int)ac);
2767 st->newVal = Buf_Destroy(&buf, FALSE);
2768 }
2769 goto ok;
2770 }
2771
2772 if (estr[0] == '*' && estr[1] == '\0') {
2773 /* Found ":[*]" */
2774 st->oneBigWord = TRUE;
2775 st->newVal = st->val;
2776 goto ok;
2777 }
2778
2779 if (estr[0] == '@' && estr[1] == '\0') {
2780 /* Found ":[@]" */
2781 st->oneBigWord = FALSE;
2782 st->newVal = st->val;
2783 goto ok;
2784 }
2785
2786 /*
2787 * We expect estr to contain a single integer for :[N], or two integers
2788 * separated by ".." for :[start..end].
2789 */
2790 p = estr;
2791 if (!TryParseIntBase0(&p, &first))
2792 goto bad_modifier; /* Found junk instead of a number */
2793
2794 if (p[0] == '\0') { /* Found only one integer in :[N] */
2795 last = first;
2796 } else if (p[0] == '.' && p[1] == '.' && p[2] != '\0') {
2797 /* Expecting another integer after ".." */
2798 p += 2;
2799 if (!TryParseIntBase0(&p, &last) || *p != '\0')
2800 goto bad_modifier; /* Found junk after ".." */
2801 } else
2802 goto bad_modifier; /* Found junk instead of ".." */
2803
2804 /*
2805 * Now first and last are properly filled in, but we still have to check
2806 * for 0 as a special case.
2807 */
2808 if (first == 0 && last == 0) {
2809 /* ":[0]" or perhaps ":[0..0]" */
2810 st->oneBigWord = TRUE;
2811 st->newVal = st->val;
2812 goto ok;
2813 }
2814
2815 /* ":[0..N]" or ":[N..0]" */
2816 if (first == 0 || last == 0)
2817 goto bad_modifier;
2818
2819 /* Normal case: select the words described by first and last. */
2820 st->newVal = VarSelectWords(st->sep, st->oneBigWord, st->val, first, last);
2821
2822 ok:
2823 free(estr);
2824 return AMR_OK;
2825
2826 bad_modifier:
2827 free(estr);
2828 return AMR_BAD;
2829 }
2830
2831 static int
2832 str_cmp_asc(const void *a, const void *b)
2833 {
2834 return strcmp(*(const char * const *)a, *(const char * const *)b);
2835 }
2836
2837 static int
2838 str_cmp_desc(const void *a, const void *b)
2839 {
2840 return strcmp(*(const char * const *)b, *(const char * const *)a);
2841 }
2842
2843 /* :O (order ascending) or :Or (order descending) or :Ox (shuffle) */
2844 static ApplyModifierResult
2845 ApplyModifier_Order(const char **pp, ApplyModifiersState *st)
2846 {
2847 const char *mod = (*pp)++; /* skip past the 'O' in any case */
2848
2849 Words words = Str_Words(st->val, FALSE);
2850
2851 if (mod[1] == st->endc || mod[1] == ':') {
2852 /* :O sorts ascending */
2853 qsort(words.words, words.len, sizeof words.words[0], str_cmp_asc);
2854
2855 } else if ((mod[1] == 'r' || mod[1] == 'x') &&
2856 (mod[2] == st->endc || mod[2] == ':')) {
2857 (*pp)++;
2858
2859 if (mod[1] == 'r') {
2860 /* :Or sorts descending */
2861 qsort(words.words, words.len, sizeof words.words[0], str_cmp_desc);
2862
2863 } else {
2864 /* :Ox shuffles
2865 *
2866 * We will use [ac..2] range for mod factors. This will produce
2867 * random numbers in [(ac-1)..0] interval, and minimal
2868 * reasonable value for mod factor is 2 (the mod 1 will produce
2869 * 0 with probability 1).
2870 */
2871 size_t i;
2872 for (i = words.len - 1; i > 0; i--) {
2873 size_t rndidx = (size_t)random() % (i + 1);
2874 char *t = words.words[i];
2875 words.words[i] = words.words[rndidx];
2876 words.words[rndidx] = t;
2877 }
2878 }
2879 } else {
2880 Words_Free(words);
2881 return AMR_BAD;
2882 }
2883
2884 st->newVal = Words_JoinFree(words);
2885 return AMR_OK;
2886 }
2887
2888 /* :? then : else */
2889 static ApplyModifierResult
2890 ApplyModifier_IfElse(const char **pp, ApplyModifiersState *st)
2891 {
2892 char *then_expr, *else_expr;
2893 VarParseResult res;
2894
2895 Boolean value = FALSE;
2896 VarEvalFlags then_eflags = VARE_NONE;
2897 VarEvalFlags else_eflags = VARE_NONE;
2898
2899 int cond_rc = COND_PARSE; /* anything other than COND_INVALID */
2900 if (st->eflags & VARE_WANTRES) {
2901 cond_rc = Cond_EvalCondition(st->var->name, &value);
2902 if (cond_rc != COND_INVALID && value)
2903 then_eflags = st->eflags;
2904 if (cond_rc != COND_INVALID && !value)
2905 else_eflags = st->eflags;
2906 }
2907
2908 (*pp)++; /* skip past the '?' */
2909 res = ParseModifierPart(pp, ':', then_eflags, st,
2910 &then_expr, NULL, NULL, NULL);
2911 if (res != VPR_OK)
2912 return AMR_CLEANUP;
2913
2914 res = ParseModifierPart(pp, st->endc, else_eflags, st,
2915 &else_expr, NULL, NULL, NULL);
2916 if (res != VPR_OK)
2917 return AMR_CLEANUP;
2918
2919 (*pp)--;
2920 if (cond_rc == COND_INVALID) {
2921 Error("Bad conditional expression `%s' in %s?%s:%s",
2922 st->var->name, st->var->name, then_expr, else_expr);
2923 return AMR_CLEANUP;
2924 }
2925
2926 if (value) {
2927 st->newVal = then_expr;
2928 free(else_expr);
2929 } else {
2930 st->newVal = else_expr;
2931 free(then_expr);
2932 }
2933 ApplyModifiersState_Define(st);
2934 return AMR_OK;
2935 }
2936
2937 /*
2938 * The ::= modifiers actually assign a value to the variable.
2939 * Their main purpose is in supporting modifiers of .for loop
2940 * iterators and other obscure uses. They always expand to
2941 * nothing. In a target rule that would otherwise expand to an
2942 * empty line they can be preceded with @: to keep make happy.
2943 * Eg.
2944 *
2945 * foo: .USE
2946 * .for i in ${.TARGET} ${.TARGET:R}.gz
2947 * @: ${t::=$i}
2948 * @echo blah ${t:T}
2949 * .endfor
2950 *
2951 * ::=<str> Assigns <str> as the new value of variable.
2952 * ::?=<str> Assigns <str> as value of variable if
2953 * it was not already set.
2954 * ::+=<str> Appends <str> to variable.
2955 * ::!=<cmd> Assigns output of <cmd> as the new value of
2956 * variable.
2957 */
2958 static ApplyModifierResult
2959 ApplyModifier_Assign(const char **pp, ApplyModifiersState *st)
2960 {
2961 GNode *ctxt;
2962 char delim;
2963 char *val;
2964 VarParseResult res;
2965
2966 const char *mod = *pp;
2967 const char *op = mod + 1;
2968
2969 if (op[0] == '=')
2970 goto ok;
2971 if ((op[0] == '!' || op[0] == '+' || op[0] == '?') && op[1] == '=')
2972 goto ok;
2973 return AMR_UNKNOWN; /* "::<unrecognised>" */
2974 ok:
2975
2976 if (st->var->name[0] == '\0') {
2977 *pp = mod + 1;
2978 return AMR_BAD;
2979 }
2980
2981 ctxt = st->ctxt; /* context where v belongs */
2982 if (!(st->exprFlags & VEF_UNDEF) && st->ctxt != VAR_GLOBAL) {
2983 Var *gv = VarFind(st->var->name, st->ctxt, FALSE);
2984 if (gv == NULL)
2985 ctxt = VAR_GLOBAL;
2986 else
2987 VarFreeEnv(gv, TRUE);
2988 }
2989
2990 switch (op[0]) {
2991 case '+':
2992 case '?':
2993 case '!':
2994 *pp = mod + 3;
2995 break;
2996 default:
2997 *pp = mod + 2;
2998 break;
2999 }
3000
3001 delim = st->startc == '(' ? ')' : '}';
3002 res = ParseModifierPart(pp, delim, st->eflags, st, &val, NULL, NULL, NULL);
3003 if (res != VPR_OK)
3004 return AMR_CLEANUP;
3005
3006 (*pp)--;
3007
3008 if (st->eflags & VARE_WANTRES) {
3009 switch (op[0]) {
3010 case '+':
3011 Var_Append(st->var->name, val, ctxt);
3012 break;
3013 case '!': {
3014 const char *errfmt;
3015 char *cmd_output = Cmd_Exec(val, &errfmt);
3016 if (errfmt != NULL)
3017 Error(errfmt, val);
3018 else
3019 Var_Set(st->var->name, cmd_output, ctxt);
3020 free(cmd_output);
3021 break;
3022 }
3023 case '?':
3024 if (!(st->exprFlags & VEF_UNDEF))
3025 break;
3026 /* FALLTHROUGH */
3027 default:
3028 Var_Set(st->var->name, val, ctxt);
3029 break;
3030 }
3031 }
3032 free(val);
3033 st->newVal = bmake_strdup("");
3034 return AMR_OK;
3035 }
3036
3037 /* :_=...
3038 * remember current value */
3039 static ApplyModifierResult
3040 ApplyModifier_Remember(const char **pp, ApplyModifiersState *st)
3041 {
3042 const char *mod = *pp;
3043 if (!ModMatchEq(mod, "_", st->endc))
3044 return AMR_UNKNOWN;
3045
3046 if (mod[1] == '=') {
3047 size_t n = strcspn(mod + 2, ":)}");
3048 char *name = bmake_strldup(mod + 2, n);
3049 Var_Set(name, st->val, st->ctxt);
3050 free(name);
3051 *pp = mod + 2 + n;
3052 } else {
3053 Var_Set("_", st->val, st->ctxt);
3054 *pp = mod + 1;
3055 }
3056 st->newVal = st->val;
3057 return AMR_OK;
3058 }
3059
3060 /* Apply the given function to each word of the variable value,
3061 * for a single-letter modifier such as :H, :T. */
3062 static ApplyModifierResult
3063 ApplyModifier_WordFunc(const char **pp, ApplyModifiersState *st,
3064 ModifyWordsCallback modifyWord)
3065 {
3066 char delim = (*pp)[1];
3067 if (delim != st->endc && delim != ':')
3068 return AMR_UNKNOWN;
3069
3070 st->newVal = ModifyWords(st->val, modifyWord, NULL,
3071 st->oneBigWord, st->sep);
3072 (*pp)++;
3073 return AMR_OK;
3074 }
3075
3076 static ApplyModifierResult
3077 ApplyModifier_Unique(const char **pp, ApplyModifiersState *st)
3078 {
3079 if ((*pp)[1] == st->endc || (*pp)[1] == ':') {
3080 st->newVal = VarUniq(st->val);
3081 (*pp)++;
3082 return AMR_OK;
3083 } else
3084 return AMR_UNKNOWN;
3085 }
3086
3087 #ifdef SYSVVARSUB
3088 /* :from=to */
3089 static ApplyModifierResult
3090 ApplyModifier_SysV(const char **pp, ApplyModifiersState *st)
3091 {
3092 char *lhs, *rhs;
3093 VarParseResult res;
3094
3095 const char *mod = *pp;
3096 Boolean eqFound = FALSE;
3097
3098 /*
3099 * First we make a pass through the string trying to verify it is a
3100 * SysV-make-style translation. It must be: <lhs>=<rhs>
3101 */
3102 int depth = 1;
3103 const char *p = mod;
3104 while (*p != '\0' && depth > 0) {
3105 if (*p == '=') { /* XXX: should also test depth == 1 */
3106 eqFound = TRUE;
3107 /* continue looking for st->endc */
3108 } else if (*p == st->endc)
3109 depth--;
3110 else if (*p == st->startc)
3111 depth++;
3112 if (depth > 0)
3113 p++;
3114 }
3115 if (*p != st->endc || !eqFound)
3116 return AMR_UNKNOWN;
3117
3118 *pp = mod;
3119 res = ParseModifierPart(pp, '=', st->eflags, st,
3120 &lhs, NULL, NULL, NULL);
3121 if (res != VPR_OK)
3122 return AMR_CLEANUP;
3123
3124 /* The SysV modifier lasts until the end of the variable expression. */
3125 res = ParseModifierPart(pp, st->endc, st->eflags, st,
3126 &rhs, NULL, NULL, NULL);
3127 if (res != VPR_OK)
3128 return AMR_CLEANUP;
3129
3130 (*pp)--;
3131 if (lhs[0] == '\0' && st->val[0] == '\0') {
3132 st->newVal = st->val; /* special case */
3133 } else {
3134 struct ModifyWord_SYSVSubstArgs args = {st->ctxt, lhs, rhs};
3135 st->newVal = ModifyWords(st->val, ModifyWord_SYSVSubst, &args,
3136 st->oneBigWord, st->sep);
3137 }
3138 free(lhs);
3139 free(rhs);
3140 return AMR_OK;
3141 }
3142 #endif
3143
3144 #ifdef SUNSHCMD
3145 /* :sh */
3146 static ApplyModifierResult
3147 ApplyModifier_SunShell(const char **pp, ApplyModifiersState *st)
3148 {
3149 const char *p = *pp;
3150 if (p[1] == 'h' && (p[2] == st->endc || p[2] == ':')) {
3151 if (st->eflags & VARE_WANTRES) {
3152 const char *errfmt;
3153 st->newVal = Cmd_Exec(st->val, &errfmt);
3154 if (errfmt != NULL)
3155 Error(errfmt, st->val);
3156 } else
3157 st->newVal = bmake_strdup("");
3158 *pp = p + 2;
3159 return AMR_OK;
3160 } else
3161 return AMR_UNKNOWN;
3162 }
3163 #endif
3164
3165 static void
3166 LogBeforeApply(const ApplyModifiersState *st, const char *mod, const char endc)
3167 {
3168 char eflags_str[VarEvalFlags_ToStringSize];
3169 char vflags_str[VarFlags_ToStringSize];
3170 char exprflags_str[VarExprFlags_ToStringSize];
3171 Boolean is_single_char = mod[0] != '\0' &&
3172 (mod[1] == endc || mod[1] == ':');
3173
3174 /* At this point, only the first character of the modifier can
3175 * be used since the end of the modifier is not yet known. */
3176 debug_printf("Applying ${%s:%c%s} to \"%s\" (%s, %s, %s)\n",
3177 st->var->name, mod[0], is_single_char ? "" : "...", st->val,
3178 Enum_FlagsToString(eflags_str, sizeof eflags_str,
3179 st->eflags, VarEvalFlags_ToStringSpecs),
3180 Enum_FlagsToString(vflags_str, sizeof vflags_str,
3181 st->var->flags, VarFlags_ToStringSpecs),
3182 Enum_FlagsToString(exprflags_str, sizeof exprflags_str,
3183 st->exprFlags,
3184 VarExprFlags_ToStringSpecs));
3185 }
3186
3187 static void
3188 LogAfterApply(ApplyModifiersState *st, const char *p, const char *mod)
3189 {
3190 char eflags_str[VarEvalFlags_ToStringSize];
3191 char vflags_str[VarFlags_ToStringSize];
3192 char exprflags_str[VarExprFlags_ToStringSize];
3193 const char *quot = st->newVal == var_Error ? "" : "\"";
3194 const char *newVal = st->newVal == var_Error ? "error" : st->newVal;
3195
3196 debug_printf("Result of ${%s:%.*s} is %s%s%s (%s, %s, %s)\n",
3197 st->var->name, (int)(p - mod), mod, quot, newVal, quot,
3198 Enum_FlagsToString(eflags_str, sizeof eflags_str,
3199 st->eflags, VarEvalFlags_ToStringSpecs),
3200 Enum_FlagsToString(vflags_str, sizeof vflags_str,
3201 st->var->flags, VarFlags_ToStringSpecs),
3202 Enum_FlagsToString(exprflags_str, sizeof exprflags_str,
3203 st->exprFlags,
3204 VarExprFlags_ToStringSpecs));
3205 }
3206
3207 static ApplyModifierResult
3208 ApplyModifier(const char **pp, ApplyModifiersState *st)
3209 {
3210 switch (**pp) {
3211 case ':':
3212 return ApplyModifier_Assign(pp, st);
3213 case '@':
3214 return ApplyModifier_Loop(pp, st);
3215 case '_':
3216 return ApplyModifier_Remember(pp, st);
3217 case 'D':
3218 case 'U':
3219 return ApplyModifier_Defined(pp, st);
3220 case 'L':
3221 return ApplyModifier_Literal(pp, st);
3222 case 'P':
3223 return ApplyModifier_Path(pp, st);
3224 case '!':
3225 return ApplyModifier_ShellCommand(pp, st);
3226 case '[':
3227 return ApplyModifier_Words(pp, st);
3228 case 'g':
3229 return ApplyModifier_Gmtime(pp, st);
3230 case 'h':
3231 return ApplyModifier_Hash(pp, st);
3232 case 'l':
3233 return ApplyModifier_Localtime(pp, st);
3234 case 't':
3235 return ApplyModifier_To(pp, st);
3236 case 'N':
3237 case 'M':
3238 return ApplyModifier_Match(pp, st);
3239 case 'S':
3240 return ApplyModifier_Subst(pp, st);
3241 case '?':
3242 return ApplyModifier_IfElse(pp, st);
3243 #ifndef NO_REGEX
3244 case 'C':
3245 return ApplyModifier_Regex(pp, st);
3246 #endif
3247 case 'q':
3248 case 'Q':
3249 return ApplyModifier_Quote(pp, st);
3250 case 'T':
3251 return ApplyModifier_WordFunc(pp, st, ModifyWord_Tail);
3252 case 'H':
3253 return ApplyModifier_WordFunc(pp, st, ModifyWord_Head);
3254 case 'E':
3255 return ApplyModifier_WordFunc(pp, st, ModifyWord_Suffix);
3256 case 'R':
3257 return ApplyModifier_WordFunc(pp, st, ModifyWord_Root);
3258 case 'r':
3259 return ApplyModifier_Range(pp, st);
3260 case 'O':
3261 return ApplyModifier_Order(pp, st);
3262 case 'u':
3263 return ApplyModifier_Unique(pp, st);
3264 #ifdef SUNSHCMD
3265 case 's':
3266 return ApplyModifier_SunShell(pp, st);
3267 #endif
3268 default:
3269 return AMR_UNKNOWN;
3270 }
3271 }
3272
3273 static char *ApplyModifiers(const char **, char *, char, char, Var *,
3274 VarExprFlags *, GNode *, VarEvalFlags, void **);
3275
3276 typedef enum ApplyModifiersIndirectResult {
3277 AMIR_CONTINUE,
3278 AMIR_APPLY_MODS,
3279 AMIR_OUT
3280 } ApplyModifiersIndirectResult;
3281
3282 /* While expanding a variable expression, expand and apply indirect
3283 * modifiers such as in ${VAR:${M_indirect}}. */
3284 static ApplyModifiersIndirectResult
3285 ApplyModifiersIndirect(
3286 ApplyModifiersState *const st,
3287 const char **const inout_p,
3288 void **const inout_freeIt
3289 ) {
3290 const char *p = *inout_p;
3291 const char *mods;
3292 void *mods_freeIt;
3293
3294 (void)Var_Parse(&p, st->ctxt, st->eflags, &mods, &mods_freeIt);
3295 /* TODO: handle errors */
3296
3297 /*
3298 * If we have not parsed up to st->endc or ':', we are not
3299 * interested. This means the expression ${VAR:${M_1}${M_2}}
3300 * is not accepted, but ${VAR:${M_1}:${M_2}} is.
3301 */
3302 if (mods[0] != '\0' && *p != '\0' && *p != ':' && *p != st->endc) {
3303 if (opts.lint)
3304 Parse_Error(PARSE_FATAL,
3305 "Missing delimiter ':' after indirect modifier \"%.*s\"",
3306 (int)(p - *inout_p), *inout_p);
3307
3308 free(mods_freeIt);
3309 /* XXX: apply_mods doesn't sound like "not interested". */
3310 /* XXX: Why is the indirect modifier parsed once more by
3311 * apply_mods? If any, p should be advanced to nested_p. */
3312 return AMIR_APPLY_MODS;
3313 }
3314
3315 VAR_DEBUG3("Indirect modifier \"%s\" from \"%.*s\"\n",
3316 mods, (int)(p - *inout_p), *inout_p);
3317
3318 if (mods[0] != '\0') {
3319 const char *rval_pp = mods;
3320 st->val = ApplyModifiers(&rval_pp, st->val, '\0', '\0', st->var,
3321 &st->exprFlags, st->ctxt, st->eflags,
3322 inout_freeIt);
3323 if (st->val == var_Error || st->val == varUndefined ||
3324 *rval_pp != '\0') {
3325 free(mods_freeIt);
3326 *inout_p = p;
3327 return AMIR_OUT; /* error already reported */
3328 }
3329 }
3330 free(mods_freeIt);
3331
3332 if (*p == ':')
3333 p++;
3334 else if (*p == '\0' && st->endc != '\0') {
3335 Error("Unclosed variable specification after complex "
3336 "modifier (expecting '%c') for %s", st->endc, st->var->name);
3337 *inout_p = p;
3338 return AMIR_OUT;
3339 }
3340
3341 *inout_p = p;
3342 return AMIR_CONTINUE;
3343 }
3344
3345 /* Apply any modifiers (such as :Mpattern or :@var@loop@ or :Q or ::=value). */
3346 static char *
3347 ApplyModifiers(
3348 const char **const pp, /* the parsing position, updated upon return */
3349 char *const val, /* the current value of the expression */
3350 char const startc, /* '(' or '{', or '\0' for indirect modifiers */
3351 char const endc, /* ')' or '}', or '\0' for indirect modifiers */
3352 Var *const v,
3353 VarExprFlags *const exprFlags,
3354 GNode *const ctxt, /* for looking up and modifying variables */
3355 VarEvalFlags const eflags,
3356 void **const inout_freeIt /* free this after using the return value */
3357 ) {
3358 ApplyModifiersState st = {
3359 startc, endc, v, ctxt, eflags,
3360 val, /* .val */
3361 var_Error, /* .newVal */
3362 ' ', /* .sep */
3363 FALSE, /* .oneBigWord */
3364 *exprFlags /* .exprFlags */
3365 };
3366 const char *p;
3367 const char *mod;
3368 ApplyModifierResult res;
3369
3370 assert(startc == '(' || startc == '{' || startc == '\0');
3371 assert(endc == ')' || endc == '}' || endc == '\0');
3372 assert(val != NULL);
3373
3374 p = *pp;
3375
3376 if (*p == '\0' && endc != '\0') {
3377 Error("Unclosed variable expression (expecting '%c') for \"%s\"",
3378 st.endc, st.var->name);
3379 goto cleanup;
3380 }
3381
3382 while (*p != '\0' && *p != endc) {
3383
3384 if (*p == '$') {
3385 ApplyModifiersIndirectResult amir;
3386 amir = ApplyModifiersIndirect(&st, &p, inout_freeIt);
3387 if (amir == AMIR_CONTINUE)
3388 continue;
3389 if (amir == AMIR_OUT)
3390 goto out;
3391 }
3392 st.newVal = var_Error; /* default value, in case of errors */
3393 mod = p;
3394
3395 if (DEBUG(VAR))
3396 LogBeforeApply(&st, mod, endc);
3397
3398 res = ApplyModifier(&p, &st);
3399
3400 #ifdef SYSVVARSUB
3401 if (res == AMR_UNKNOWN) {
3402 assert(p == mod);
3403 res = ApplyModifier_SysV(&p, &st);
3404 }
3405 #endif
3406
3407 if (res == AMR_UNKNOWN) {
3408 Error("Unknown modifier '%c'", *mod);
3409 /* Guess the end of the current modifier.
3410 * XXX: Skipping the rest of the modifier hides errors and leads
3411 * to wrong results. Parsing should rather stop here. */
3412 for (p++; *p != ':' && *p != st.endc && *p != '\0'; p++)
3413 continue;
3414 st.newVal = var_Error;
3415 }
3416 if (res == AMR_CLEANUP)
3417 goto cleanup;
3418 if (res == AMR_BAD)
3419 goto bad_modifier;
3420
3421 if (DEBUG(VAR))
3422 LogAfterApply(&st, p, mod);
3423
3424 if (st.newVal != st.val) {
3425 if (*inout_freeIt != NULL) {
3426 free(st.val);
3427 *inout_freeIt = NULL;
3428 }
3429 st.val = st.newVal;
3430 if (st.val != var_Error && st.val != varUndefined)
3431 *inout_freeIt = st.val;
3432 }
3433 if (*p == '\0' && st.endc != '\0') {
3434 Error("Unclosed variable specification (expecting '%c') "
3435 "for \"%s\" (value \"%s\") modifier %c",
3436 st.endc, st.var->name, st.val, *mod);
3437 } else if (*p == ':') {
3438 p++;
3439 } else if (opts.lint && *p != '\0' && *p != endc) {
3440 Parse_Error(PARSE_FATAL,
3441 "Missing delimiter ':' after modifier \"%.*s\"",
3442 (int)(p - mod), mod);
3443 /* TODO: propagate parse error to the enclosing expression */
3444 }
3445 }
3446 out:
3447 *pp = p;
3448 assert(st.val != NULL); /* Use var_Error or varUndefined instead. */
3449 *exprFlags = st.exprFlags;
3450 return st.val;
3451
3452 bad_modifier:
3453 /* XXX: The modifier end is only guessed. */
3454 Error("Bad modifier `:%.*s' for %s",
3455 (int)strcspn(mod, ":)}"), mod, st.var->name);
3456
3457 cleanup:
3458 *pp = p;
3459 free(*inout_freeIt);
3460 *inout_freeIt = NULL;
3461 *exprFlags = st.exprFlags;
3462 return var_Error;
3463 }
3464
3465 /* Only four of the local variables are treated specially as they are the
3466 * only four that will be set when dynamic sources are expanded. */
3467 static Boolean
3468 VarnameIsDynamic(const char *name, size_t len)
3469 {
3470 if (len == 1 || (len == 2 && (name[1] == 'F' || name[1] == 'D'))) {
3471 switch (name[0]) {
3472 case '@':
3473 case '%':
3474 case '*':
3475 case '!':
3476 return TRUE;
3477 }
3478 return FALSE;
3479 }
3480
3481 if ((len == 7 || len == 8) && name[0] == '.' && ch_isupper(name[1])) {
3482 return strcmp(name, ".TARGET") == 0 ||
3483 strcmp(name, ".ARCHIVE") == 0 ||
3484 strcmp(name, ".PREFIX") == 0 ||
3485 strcmp(name, ".MEMBER") == 0;
3486 }
3487
3488 return FALSE;
3489 }
3490
3491 static const char *
3492 UndefinedShortVarValue(char varname, const GNode *ctxt, VarEvalFlags eflags)
3493 {
3494 if (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL) {
3495 /*
3496 * If substituting a local variable in a non-local context,
3497 * assume it's for dynamic source stuff. We have to handle
3498 * this specially and return the longhand for the variable
3499 * with the dollar sign escaped so it makes it back to the
3500 * caller. Only four of the local variables are treated
3501 * specially as they are the only four that will be set
3502 * when dynamic sources are expanded.
3503 */
3504 switch (varname) {
3505 case '@':
3506 return "$(.TARGET)";
3507 case '%':
3508 return "$(.MEMBER)";
3509 case '*':
3510 return "$(.PREFIX)";
3511 case '!':
3512 return "$(.ARCHIVE)";
3513 }
3514 }
3515 return eflags & VARE_UNDEFERR ? var_Error : varUndefined;
3516 }
3517
3518 /* Parse a variable name, until the end character or a colon, whichever
3519 * comes first. */
3520 static char *
3521 ParseVarname(const char **pp, char startc, char endc,
3522 GNode *ctxt, VarEvalFlags eflags,
3523 size_t *out_varname_len)
3524 {
3525 Buffer buf;
3526 const char *p = *pp;
3527 int depth = 1;
3528
3529 Buf_Init(&buf);
3530
3531 while (*p != '\0') {
3532 /* Track depth so we can spot parse errors. */
3533 if (*p == startc)
3534 depth++;
3535 if (*p == endc) {
3536 if (--depth == 0)
3537 break;
3538 }
3539 if (*p == ':' && depth == 1)
3540 break;
3541
3542 /* A variable inside a variable, expand. */
3543 if (*p == '$') {
3544 const char *nested_val;
3545 void *nested_val_freeIt;
3546 (void)Var_Parse(&p, ctxt, eflags, &nested_val, &nested_val_freeIt);
3547 /* TODO: handle errors */
3548 Buf_AddStr(&buf, nested_val);
3549 free(nested_val_freeIt);
3550 } else {
3551 Buf_AddByte(&buf, *p);
3552 p++;
3553 }
3554 }
3555 *pp = p;
3556 *out_varname_len = Buf_Len(&buf);
3557 return Buf_Destroy(&buf, FALSE);
3558 }
3559
3560 static VarParseResult
3561 ValidShortVarname(char varname, const char *start)
3562 {
3563 switch (varname) {
3564 case '\0':
3565 case ')':
3566 case '}':
3567 case ':':
3568 case '$':
3569 break; /* and continue below */
3570 default:
3571 return VPR_OK;
3572 }
3573
3574 if (!opts.lint)
3575 return VPR_PARSE_SILENT;
3576
3577 if (varname == '$')
3578 Parse_Error(PARSE_FATAL,
3579 "To escape a dollar, use \\$, not $$, at \"%s\"", start);
3580 else if (varname == '\0')
3581 Parse_Error(PARSE_FATAL, "Dollar followed by nothing");
3582 else
3583 Parse_Error(PARSE_FATAL,
3584 "Invalid variable name '%c', at \"%s\"", varname, start);
3585
3586 return VPR_PARSE_MSG;
3587 }
3588
3589 /* Parse a single-character variable name such as $V or $@.
3590 * Return whether to continue parsing. */
3591 static Boolean
3592 ParseVarnameShort(char startc, const char **pp, GNode *ctxt,
3593 VarEvalFlags eflags,
3594 VarParseResult *out_FALSE_res, const char **out_FALSE_val,
3595 Var **out_TRUE_var)
3596 {
3597 char name[2];
3598 Var *v;
3599 VarParseResult vpr;
3600
3601 /*
3602 * If it's not bounded by braces of some sort, life is much simpler.
3603 * We just need to check for the first character and return the
3604 * value if it exists.
3605 */
3606
3607 vpr = ValidShortVarname(startc, *pp);
3608 if (vpr != VPR_OK) {
3609 (*pp)++;
3610 *out_FALSE_val = var_Error;
3611 *out_FALSE_res = vpr;
3612 return FALSE;
3613 }
3614
3615 name[0] = startc;
3616 name[1] = '\0';
3617 v = VarFind(name, ctxt, TRUE);
3618 if (v == NULL) {
3619 *pp += 2;
3620
3621 *out_FALSE_val = UndefinedShortVarValue(startc, ctxt, eflags);
3622 if (opts.lint && *out_FALSE_val == var_Error) {
3623 Parse_Error(PARSE_FATAL, "Variable \"%s\" is undefined", name);
3624 *out_FALSE_res = VPR_UNDEF_MSG;
3625 return FALSE;
3626 }
3627 *out_FALSE_res = eflags & VARE_UNDEFERR ? VPR_UNDEF_SILENT : VPR_OK;
3628 return FALSE;
3629 }
3630
3631 *out_TRUE_var = v;
3632 return TRUE;
3633 }
3634
3635 /* Find variables like @F or <D. */
3636 static Var *
3637 FindLocalLegacyVar(const char *varname, size_t namelen, GNode *ctxt,
3638 const char **out_extraModifiers)
3639 {
3640 /* Only resolve these variables if ctxt is a "real" target. */
3641 if (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL)
3642 return NULL;
3643
3644 if (namelen != 2)
3645 return NULL;
3646 if (varname[1] != 'F' && varname[1] != 'D')
3647 return NULL;
3648 if (strchr("@%?*!<>", varname[0]) == NULL)
3649 return NULL;
3650
3651 {
3652 char name[] = { varname[0], '\0' };
3653 Var *v = VarFind(name, ctxt, FALSE);
3654
3655 if (v != NULL) {
3656 if (varname[1] == 'D') {
3657 *out_extraModifiers = "H:";
3658 } else { /* F */
3659 *out_extraModifiers = "T:";
3660 }
3661 }
3662 return v;
3663 }
3664 }
3665
3666 static VarParseResult
3667 EvalUndefined(Boolean dynamic, const char *start, const char *p, char *varname,
3668 VarEvalFlags eflags,
3669 const char **out_val, void **out_freeIt)
3670 {
3671 if (dynamic) {
3672 char *pstr = bmake_strsedup(start, p);
3673 free(varname);
3674 *out_val = pstr;
3675 *out_freeIt = pstr;
3676 return VPR_OK;
3677 }
3678
3679 if ((eflags & VARE_UNDEFERR) && opts.lint) {
3680 Parse_Error(PARSE_FATAL, "Variable \"%s\" is undefined", varname);
3681 free(varname);
3682 *out_val = var_Error;
3683 return VPR_UNDEF_MSG;
3684 }
3685
3686 if (eflags & VARE_UNDEFERR) {
3687 free(varname);
3688 *out_val = var_Error;
3689 return VPR_UNDEF_SILENT;
3690 }
3691
3692 free(varname);
3693 *out_val = varUndefined;
3694 return VPR_OK;
3695 }
3696
3697 /* Parse a long variable name enclosed in braces or parentheses such as $(VAR)
3698 * or ${VAR}, up to the closing brace or parenthesis, or in the case of
3699 * ${VAR:Modifiers}, up to the ':' that starts the modifiers.
3700 * Return whether to continue parsing. */
3701 static Boolean
3702 ParseVarnameLong(
3703 const char *p,
3704 char startc,
3705 GNode *ctxt,
3706 VarEvalFlags eflags,
3707
3708 const char **out_FALSE_pp,
3709 VarParseResult *out_FALSE_res,
3710 const char **out_FALSE_val,
3711 void **out_FALSE_freeIt,
3712
3713 char *out_TRUE_endc,
3714 const char **out_TRUE_p,
3715 Var **out_TRUE_v,
3716 Boolean *out_TRUE_haveModifier,
3717 const char **out_TRUE_extraModifiers,
3718 Boolean *out_TRUE_dynamic,
3719 VarExprFlags *out_TRUE_exprFlags
3720 ) {
3721 size_t namelen;
3722 char *varname;
3723 Var *v;
3724 Boolean haveModifier;
3725 Boolean dynamic = FALSE;
3726
3727 const char *const start = p;
3728 char endc = startc == '(' ? ')' : '}';
3729
3730 p += 2; /* skip "${" or "$(" or "y(" */
3731 varname = ParseVarname(&p, startc, endc, ctxt, eflags, &namelen);
3732
3733 if (*p == ':') {
3734 haveModifier = TRUE;
3735 } else if (*p == endc) {
3736 haveModifier = FALSE;
3737 } else {
3738 Parse_Error(PARSE_FATAL, "Unclosed variable \"%s\"", varname);
3739 free(varname);
3740 *out_FALSE_pp = p;
3741 *out_FALSE_val = var_Error;
3742 *out_FALSE_res = VPR_PARSE_MSG;
3743 return FALSE;
3744 }
3745
3746 v = VarFind(varname, ctxt, TRUE);
3747
3748 /* At this point, p points just after the variable name,
3749 * either at ':' or at endc. */
3750
3751 if (v == NULL)
3752 v = FindLocalLegacyVar(varname, namelen, ctxt, out_TRUE_extraModifiers);
3753
3754 if (v == NULL) {
3755 /* Defer expansion of dynamic variables if they appear in non-local
3756 * context since they are not defined there. */
3757 dynamic = VarnameIsDynamic(varname, namelen) &&
3758 (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL);
3759
3760 if (!haveModifier) {
3761 p++; /* skip endc */
3762 *out_FALSE_pp = p;
3763 *out_FALSE_res = EvalUndefined(dynamic, start, p, varname, eflags,
3764 out_FALSE_val, out_FALSE_freeIt);
3765 return FALSE;
3766 }
3767
3768 /* The variable expression is based on an undefined variable.
3769 * Nevertheless it needs a Var, for modifiers that access the
3770 * variable name, such as :L or :?.
3771 *
3772 * Most modifiers leave this expression in the "undefined" state
3773 * (VEF_UNDEF), only a few modifiers like :D, :U, :L, :P turn this
3774 * undefined expression into a defined expression (VEF_DEF).
3775 *
3776 * At the end, after applying all modifiers, if the expression
3777 * is still undefined, Var_Parse will return an empty string
3778 * instead of the actually computed value. */
3779 v = VarNew(varname, varname, "", 0);
3780 *out_TRUE_exprFlags = VEF_UNDEF;
3781 } else
3782 free(varname);
3783
3784 *out_TRUE_endc = endc;
3785 *out_TRUE_p = p;
3786 *out_TRUE_v = v;
3787 *out_TRUE_haveModifier = haveModifier;
3788 *out_TRUE_dynamic = dynamic;
3789 return TRUE;
3790 }
3791
3792 /*
3793 * Given the start of a variable expression (such as $v, $(VAR),
3794 * ${VAR:Mpattern}), extract the variable name and value, and the modifiers,
3795 * if any. While doing that, apply the modifiers to the value of the
3796 * expression, forming its final value. A few of the modifiers such as :!cmd!
3797 * or ::= have side effects.
3798 *
3799 * Input:
3800 * *pp The string to parse.
3801 * When parsing a condition in ParseEmptyArg, it may also
3802 * point to the "y" of "empty(VARNAME:Modifiers)", which
3803 * is syntactically the same.
3804 * ctxt The context for finding variables
3805 * eflags Control the exact details of parsing
3806 *
3807 * Output:
3808 * *pp The position where to continue parsing.
3809 * TODO: After a parse error, the value of *pp is
3810 * unspecified. It may not have been updated at all,
3811 * point to some random character in the string, to the
3812 * location of the parse error, or at the end of the
3813 * string.
3814 * *out_val The value of the variable expression, never NULL.
3815 * *out_val var_Error if there was a parse error.
3816 * *out_val var_Error if the base variable of the expression was
3817 * undefined, eflags contains VARE_UNDEFERR, and none of
3818 * the modifiers turned the undefined expression into a
3819 * defined expression.
3820 * XXX: It is not guaranteed that an error message has
3821 * been printed.
3822 * *out_val varUndefined if the base variable of the expression
3823 * was undefined, eflags did not contain VARE_UNDEFERR,
3824 * and none of the modifiers turned the undefined
3825 * expression into a defined expression.
3826 * XXX: It is not guaranteed that an error message has
3827 * been printed.
3828 * *out_val_freeIt Must be freed by the caller after using *out_val.
3829 */
3830 /* coverity[+alloc : arg-*4] */
3831 VarParseResult
3832 Var_Parse(const char **pp, GNode *ctxt, VarEvalFlags eflags,
3833 const char **out_val, void **out_val_freeIt)
3834 {
3835 const char *p = *pp;
3836 const char *const start = p;
3837 Boolean haveModifier; /* TRUE if have modifiers for the variable */
3838 char startc; /* Starting character if variable in parens
3839 * or braces */
3840 char endc; /* Ending character if variable in parens
3841 * or braces */
3842 Boolean dynamic; /* TRUE if the variable is local and we're
3843 * expanding it in a non-local context. This
3844 * is done to support dynamic sources. The
3845 * result is just the expression, unaltered */
3846 const char *extramodifiers;
3847 Var *v;
3848 char *value;
3849 char eflags_str[VarEvalFlags_ToStringSize];
3850 VarExprFlags exprFlags = 0;
3851
3852 VAR_DEBUG2("Var_Parse: %s with %s\n", start,
3853 Enum_FlagsToString(eflags_str, sizeof eflags_str, eflags,
3854 VarEvalFlags_ToStringSpecs));
3855
3856 *out_val_freeIt = NULL;
3857 extramodifiers = NULL; /* extra modifiers to apply first */
3858 dynamic = FALSE;
3859
3860 /* Appease GCC, which thinks that the variable might not be
3861 * initialized. */
3862 endc = '\0';
3863
3864 startc = p[1];
3865 if (startc != '(' && startc != '{') {
3866 VarParseResult res;
3867 if (!ParseVarnameShort(startc, pp, ctxt, eflags, &res, out_val, &v))
3868 return res;
3869 haveModifier = FALSE;
3870 p++;
3871 } else {
3872 VarParseResult res;
3873 if (!ParseVarnameLong(p, startc, ctxt, eflags,
3874 pp, &res, out_val, out_val_freeIt,
3875 &endc, &p, &v, &haveModifier, &extramodifiers,
3876 &dynamic, &exprFlags))
3877 return res;
3878 }
3879
3880 if (v->flags & VAR_IN_USE)
3881 Fatal("Variable %s is recursive.", v->name);
3882
3883 /* XXX: This assignment creates an alias to the current value of the
3884 * variable. This means that as long as the value of the expression stays
3885 * the same, the value of the variable must not change.
3886 * Using the '::=' modifier, it could be possible to do exactly this.
3887 * At the bottom of this function, the resulting value is compared to the
3888 * then-current value of the variable. This might also invoke undefined
3889 * behavior. */
3890 value = Buf_GetAll(&v->val, NULL);
3891
3892 /* Before applying any modifiers, expand any nested expressions from the
3893 * variable value. */
3894 if (strchr(value, '$') != NULL && (eflags & VARE_WANTRES)) {
3895 VarEvalFlags nested_eflags = eflags;
3896 if (opts.lint)
3897 nested_eflags &= ~(unsigned)VARE_UNDEFERR;
3898 v->flags |= VAR_IN_USE;
3899 (void)Var_Subst(value, ctxt, nested_eflags, &value);
3900 v->flags &= ~(unsigned)VAR_IN_USE;
3901 /* TODO: handle errors */
3902 *out_val_freeIt = value;
3903 }
3904
3905 if (haveModifier || extramodifiers != NULL) {
3906 void *extraFree;
3907
3908 extraFree = NULL;
3909 if (extramodifiers != NULL) {
3910 const char *em = extramodifiers;
3911 value = ApplyModifiers(&em, value, '\0', '\0',
3912 v, &exprFlags, ctxt, eflags, &extraFree);
3913 }
3914
3915 if (haveModifier) {
3916 /* Skip initial colon. */
3917 p++;
3918
3919 value = ApplyModifiers(&p, value, startc, endc,
3920 v, &exprFlags, ctxt, eflags, out_val_freeIt);
3921 free(extraFree);
3922 } else {
3923 *out_val_freeIt = extraFree;
3924 }
3925 }
3926
3927 if (*p != '\0') /* Skip past endc if possible. */
3928 p++;
3929
3930 *pp = p;
3931
3932 if (v->flags & VAR_FROM_ENV) {
3933 /* Free the environment variable now since we own it. */
3934
3935 char *varValue = Buf_Destroy(&v->val, FALSE);
3936 if (value == varValue) {
3937 /* Don't free the variable value since it will be returned. */
3938 *out_val_freeIt = varValue;
3939 } else
3940 free(varValue);
3941
3942 free(v->name_freeIt);
3943 free(v);
3944
3945 } else if (exprFlags & VEF_UNDEF) {
3946 if (!(exprFlags & VEF_DEF)) {
3947 /* TODO: Use a local variable instead of out_val_freeIt.
3948 * Variables named out_* must only be written to. */
3949 if (*out_val_freeIt != NULL) {
3950 free(*out_val_freeIt);
3951 *out_val_freeIt = NULL;
3952 }
3953 if (dynamic) {
3954 value = bmake_strsedup(start, p);
3955 *out_val_freeIt = value;
3956 } else {
3957 /* The expression is still undefined, therefore discard the
3958 * actual value and return an error marker instead. */
3959 value = eflags & VARE_UNDEFERR ? var_Error : varUndefined;
3960 }
3961 }
3962 if (value != Buf_GetAll(&v->val, NULL))
3963 Buf_Destroy(&v->val, TRUE);
3964 free(v->name_freeIt);
3965 free(v);
3966 }
3967 *out_val = value;
3968 return VPR_UNKNOWN;
3969 }
3970
3971 static void
3972 VarSubstNested(const char **const pp, Buffer *const buf, GNode *const ctxt,
3973 VarEvalFlags const eflags, Boolean *inout_errorReported)
3974 {
3975 const char *p = *pp;
3976 const char *nested_p = p;
3977 const char *val;
3978 void *val_freeIt;
3979
3980 (void)Var_Parse(&nested_p, ctxt, eflags, &val, &val_freeIt);
3981 /* TODO: handle errors */
3982
3983 if (val == var_Error || val == varUndefined) {
3984 if (!preserveUndefined) {
3985 p = nested_p;
3986 } else if ((eflags & VARE_UNDEFERR) || val == var_Error) {
3987 /* XXX: This condition is wrong. If val == var_Error,
3988 * this doesn't necessarily mean there was an undefined
3989 * variable. It could equally well be a parse error; see
3990 * unit-tests/varmod-order.exp. */
3991
3992 /*
3993 * If variable is undefined, complain and skip the
3994 * variable. The complaint will stop us from doing anything
3995 * when the file is parsed.
3996 */
3997 if (!*inout_errorReported) {
3998 Parse_Error(PARSE_FATAL, "Undefined variable \"%.*s\"",
3999 (int)(size_t)(nested_p - p), p);
4000 }
4001 p = nested_p;
4002 *inout_errorReported = TRUE;
4003 } else {
4004 /* Copy the initial '$' of the undefined expression,
4005 * thereby deferring expansion of the expression, but
4006 * expand nested expressions if already possible.
4007 * See unit-tests/varparse-undef-partial.mk. */
4008 Buf_AddByte(buf, *p);
4009 p++;
4010 }
4011 } else {
4012 p = nested_p;
4013 Buf_AddStr(buf, val);
4014 }
4015
4016 free(val_freeIt);
4017
4018 *pp = p;
4019 }
4020
4021 /* Expand all variable expressions like $V, ${VAR}, $(VAR:Modifiers) in the
4022 * given string.
4023 *
4024 * Input:
4025 * str The string in which the variable expressions are
4026 * expanded.
4027 * ctxt The context in which to start searching for
4028 * variables. The other contexts are searched as well.
4029 * eflags Special effects during expansion.
4030 */
4031 VarParseResult
4032 Var_Subst(const char *str, GNode *ctxt, VarEvalFlags eflags, char **out_res)
4033 {
4034 const char *p = str;
4035 Buffer res;
4036
4037 /* Set true if an error has already been reported,
4038 * to prevent a plethora of messages when recursing */
4039 /* XXX: Why is the 'static' necessary here? */
4040 static Boolean errorReported;
4041
4042 Buf_Init(&res);
4043 errorReported = FALSE;
4044
4045 while (*p != '\0') {
4046 if (p[0] == '$' && p[1] == '$') {
4047 /* A dollar sign may be escaped with another dollar sign. */
4048 if (save_dollars && (eflags & VARE_KEEP_DOLLAR))
4049 Buf_AddByte(&res, '$');
4050 Buf_AddByte(&res, '$');
4051 p += 2;
4052
4053 } else if (p[0] == '$') {
4054 VarSubstNested(&p, &res, ctxt, eflags, &errorReported);
4055
4056 } else {
4057 /*
4058 * Skip as many characters as possible -- either to the end of
4059 * the string or to the next dollar sign (variable expression).
4060 */
4061 const char *plainStart = p;
4062
4063 for (p++; *p != '$' && *p != '\0'; p++)
4064 continue;
4065 Buf_AddBytesBetween(&res, plainStart, p);
4066 }
4067 }
4068
4069 *out_res = Buf_DestroyCompact(&res);
4070 return VPR_OK;
4071 }
4072
4073 /* Initialize the variables module. */
4074 void
4075 Var_Init(void)
4076 {
4077 VAR_INTERNAL = GNode_New("Internal");
4078 VAR_GLOBAL = GNode_New("Global");
4079 VAR_CMDLINE = GNode_New("Command");
4080 }
4081
4082 /* Clean up the variables module. */
4083 void
4084 Var_End(void)
4085 {
4086 Var_Stats();
4087 }
4088
4089 void
4090 Var_Stats(void)
4091 {
4092 HashTable_DebugStats(&VAR_GLOBAL->vars, "VAR_GLOBAL");
4093 }
4094
4095 /* Print all variables in a context, sorted by name. */
4096 void
4097 Var_Dump(GNode *ctxt)
4098 {
4099 Vector /* of const char * */ vec;
4100 HashIter hi;
4101 size_t i;
4102 const char **varnames;
4103
4104 Vector_Init(&vec, sizeof(const char *));
4105
4106 HashIter_Init(&hi, &ctxt->vars);
4107 while (HashIter_Next(&hi) != NULL)
4108 *(const char **)Vector_Push(&vec) = hi.entry->key;
4109 varnames = vec.items;
4110
4111 qsort(varnames, vec.len, sizeof varnames[0], str_cmp_asc);
4112
4113 for (i = 0; i < vec.len; i++) {
4114 const char *varname = varnames[i];
4115 Var *var = HashTable_FindValue(&ctxt->vars, varname);
4116 debug_printf("%-16s = %s\n", varname, Buf_GetAll(&var->val, NULL));
4117 }
4118
4119 Vector_Done(&vec);
4120 }
4121