var.c revision 1.623 1 /* $NetBSD: var.c,v 1.623 2020/10/31 14:40:34 rillig Exp $ */
2
3 /*
4 * Copyright (c) 1988, 1989, 1990, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * Adam de Boor.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 */
34
35 /*
36 * Copyright (c) 1989 by Berkeley Softworks
37 * All rights reserved.
38 *
39 * This code is derived from software contributed to Berkeley by
40 * Adam de Boor.
41 *
42 * Redistribution and use in source and binary forms, with or without
43 * modification, are permitted provided that the following conditions
44 * are met:
45 * 1. Redistributions of source code must retain the above copyright
46 * notice, this list of conditions and the following disclaimer.
47 * 2. Redistributions in binary form must reproduce the above copyright
48 * notice, this list of conditions and the following disclaimer in the
49 * documentation and/or other materials provided with the distribution.
50 * 3. All advertising materials mentioning features or use of this software
51 * must display the following acknowledgement:
52 * This product includes software developed by the University of
53 * California, Berkeley and its contributors.
54 * 4. Neither the name of the University nor the names of its contributors
55 * may be used to endorse or promote products derived from this software
56 * without specific prior written permission.
57 *
58 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
59 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
60 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
61 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
62 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
63 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
64 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
65 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
66 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
67 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
68 * SUCH DAMAGE.
69 */
70
71 /*
72 * Handling of variables and the expressions formed from them.
73 *
74 * Variables are set using lines of the form VAR=value. Both the variable
75 * name and the value can contain references to other variables, by using
76 * expressions like ${VAR}, ${VAR:Modifiers}, ${${VARNAME}} or ${VAR:${MODS}}.
77 *
78 * Interface:
79 * Var_Init Initialize this module.
80 *
81 * Var_End Clean up the module.
82 *
83 * Var_Set Set the value of the variable, creating it if
84 * necessary.
85 *
86 * Var_Append Append more characters to the variable, creating it if
87 * necessary. A space is placed between the old value and
88 * the new one.
89 *
90 * Var_Exists See if a variable exists.
91 *
92 * Var_Value Return the unexpanded value of a variable, or NULL if
93 * the variable is undefined.
94 *
95 * Var_Subst Substitute all variable expressions in a string.
96 *
97 * Var_Parse Parse a variable expression such as ${VAR:Mpattern}.
98 *
99 * Var_Delete Delete a variable.
100 *
101 * Var_ExportVars Export some or even all variables to the environment
102 * of this process and its child processes.
103 *
104 * Var_Export Export the variable to the environment of this process
105 * and its child processes.
106 *
107 * Var_UnExport Don't export the variable anymore.
108 *
109 * Debugging:
110 * Var_Stats Print out hashing statistics if in -dh mode.
111 *
112 * Var_Dump Print out all variables defined in the given context.
113 *
114 * XXX: There's a lot of duplication in these functions.
115 */
116
117 #include <sys/stat.h>
118 #ifndef NO_REGEX
119 #include <sys/types.h>
120 #include <regex.h>
121 #endif
122 #include <inttypes.h>
123 #include <limits.h>
124 #include <time.h>
125
126 #include "make.h"
127 #include "dir.h"
128 #include "job.h"
129 #include "metachar.h"
130
131 /* "@(#)var.c 8.3 (Berkeley) 3/19/94" */
132 MAKE_RCSID("$NetBSD: var.c,v 1.623 2020/10/31 14:40:34 rillig Exp $");
133
134 #define VAR_DEBUG1(fmt, arg1) DEBUG1(VAR, fmt, arg1)
135 #define VAR_DEBUG2(fmt, arg1, arg2) DEBUG2(VAR, fmt, arg1, arg2)
136 #define VAR_DEBUG3(fmt, arg1, arg2, arg3) DEBUG3(VAR, fmt, arg1, arg2, arg3)
137 #define VAR_DEBUG4(fmt, arg1, arg2, arg3, arg4) DEBUG4(VAR, fmt, arg1, arg2, arg3, arg4)
138
139 ENUM_FLAGS_RTTI_3(VarEvalFlags,
140 VARE_UNDEFERR, VARE_WANTRES, VARE_ASSIGN);
141
142 /*
143 * This lets us tell if we have replaced the original environ
144 * (which we cannot free).
145 */
146 char **savedEnv = NULL;
147
148 /* Special return value for Var_Parse, indicating a parse error. It may be
149 * caused by an undefined variable, a syntax error in a modifier or
150 * something entirely different. */
151 char var_Error[] = "";
152
153 /* Special return value for Var_Parse, indicating an undefined variable in
154 * a case where VARE_UNDEFERR is not set. This undefined variable is
155 * typically a dynamic variable such as ${.TARGET}, whose expansion needs to
156 * be deferred until it is defined in an actual target. */
157 static char varUndefined[] = "";
158
159 /* Special return value for Var_Parse, just to avoid allocating empty strings.
160 * In contrast to var_Error and varUndefined, this is not an error marker but
161 * just an ordinary successful return value. */
162 static char emptyString[] = "";
163
164 /*
165 * Traditionally this make consumed $$ during := like any other expansion.
166 * Other make's do not, and this make follows straight since 2016-01-09.
167 *
168 * This knob allows controlling the behavior.
169 * FALSE to consume $$ during := assignment.
170 * TRUE to preserve $$ during := assignment.
171 */
172 #define MAKE_SAVE_DOLLARS ".MAKE.SAVE_DOLLARS"
173 static Boolean save_dollars = TRUE;
174
175 /*
176 * Internally, variables are contained in four different contexts.
177 * 1) the environment. They cannot be changed. If an environment
178 * variable is appended to, the result is placed in the global
179 * context.
180 * 2) the global context. Variables set in the makefiles are located
181 * here.
182 * 3) the command-line context. All variables set on the command line
183 * are placed in this context.
184 * 4) the local context. Each target has associated with it a context
185 * list. On this list are located the structures describing such
186 * local variables as $(@) and $(*)
187 * The four contexts are searched in the reverse order from which they are
188 * listed (but see opts.checkEnvFirst).
189 */
190 GNode *VAR_INTERNAL; /* variables from make itself */
191 GNode *VAR_GLOBAL; /* variables from the makefile */
192 GNode *VAR_CMDLINE; /* variables defined on the command-line */
193
194 typedef enum VarFlags {
195
196 /* The variable's value is currently being used by Var_Parse or Var_Subst.
197 * This marker is used to avoid endless recursion. */
198 VAR_IN_USE = 0x01,
199
200 /* The variable comes from the environment.
201 * These variables are not registered in any GNode, therefore they must
202 * be freed as soon as they are not used anymore. */
203 VAR_FROM_ENV = 0x02,
204
205 /* The variable is exported to the environment, to be used by child
206 * processes. */
207 VAR_EXPORTED = 0x10,
208
209 /* At the point where this variable was exported, it contained an
210 * unresolved reference to another variable. Before any child process is
211 * started, it needs to be exported again, in the hope that the referenced
212 * variable can then be resolved. */
213 VAR_REEXPORT = 0x20,
214
215 /* The variable came from the command line. */
216 VAR_FROM_CMD = 0x40,
217
218 /* The variable value cannot be changed anymore, and the variable cannot
219 * be deleted. Any attempts to do so are ignored. */
220 VAR_READONLY = 0x80
221 } VarFlags;
222
223 ENUM_FLAGS_RTTI_6(VarFlags,
224 VAR_IN_USE, VAR_FROM_ENV,
225 VAR_EXPORTED, VAR_REEXPORT, VAR_FROM_CMD, VAR_READONLY);
226
227 /* Variables are defined using one of the VAR=value assignments. Their
228 * value can be queried by expressions such as $V, ${VAR}, or with modifiers
229 * such as ${VAR:S,from,to,g:Q}.
230 *
231 * There are 3 kinds of variables: context variables, environment variables,
232 * undefined variables.
233 *
234 * Context variables are stored in a GNode.context. The only way to undefine
235 * a context variable is using the .undef directive. In particular, it must
236 * not be possible to undefine a variable during the evaluation of an
237 * expression, or Var.name might point nowhere.
238 *
239 * Environment variables are temporary. They are returned by VarFind, and
240 * after using them, they must be freed using VarFreeEnv.
241 *
242 * Undefined variables occur during evaluation of variable expressions such
243 * as ${UNDEF:Ufallback} in Var_Parse and ApplyModifiers.
244 */
245 typedef struct Var {
246 /* The name of the variable, once set, doesn't change anymore.
247 * For context variables, it aliases the corresponding HashEntry name.
248 * For environment and undefined variables, it is allocated. */
249 const char *name;
250 void *name_freeIt;
251
252 Buffer val; /* its value */
253 VarFlags flags; /* miscellaneous status flags */
254 } Var;
255
256 /*
257 * Exporting vars is expensive so skip it if we can
258 */
259 typedef enum VarExportedMode {
260 VAR_EXPORTED_NONE,
261 VAR_EXPORTED_SOME,
262 VAR_EXPORTED_ALL
263 } VarExportedMode;
264
265 static VarExportedMode var_exportedVars = VAR_EXPORTED_NONE;
266
267 typedef enum VarExportFlags {
268 /*
269 * We pass this to Var_Export when doing the initial export
270 * or after updating an exported var.
271 */
272 VAR_EXPORT_PARENT = 0x01,
273 /*
274 * We pass this to Var_Export1 to tell it to leave the value alone.
275 */
276 VAR_EXPORT_LITERAL = 0x02
277 } VarExportFlags;
278
279 /* Flags for pattern matching in the :S and :C modifiers */
280 typedef enum VarPatternFlags {
281 VARP_SUB_GLOBAL = 0x01, /* Replace as often as possible ('g') */
282 VARP_SUB_ONE = 0x02, /* Replace only once ('1') */
283 VARP_ANCHOR_START = 0x04, /* Match at start of word ('^') */
284 VARP_ANCHOR_END = 0x08 /* Match at end of word ('$') */
285 } VarPatternFlags;
286
287 static Var *
288 VarNew(const char *name, void *name_freeIt, const char *value, VarFlags flags)
289 {
290 size_t value_len = strlen(value);
291 Var *var = bmake_malloc(sizeof *var);
292 var->name = name;
293 var->name_freeIt = name_freeIt;
294 Buf_Init(&var->val, value_len + 1);
295 Buf_AddBytes(&var->val, value, value_len);
296 var->flags = flags;
297 return var;
298 }
299
300 static const char *
301 CanonicalVarname(const char *name)
302 {
303 if (*name == '.' && ch_isupper(name[1])) {
304 switch (name[1]) {
305 case 'A':
306 if (strcmp(name, ".ALLSRC") == 0)
307 name = ALLSRC;
308 if (strcmp(name, ".ARCHIVE") == 0)
309 name = ARCHIVE;
310 break;
311 case 'I':
312 if (strcmp(name, ".IMPSRC") == 0)
313 name = IMPSRC;
314 break;
315 case 'M':
316 if (strcmp(name, ".MEMBER") == 0)
317 name = MEMBER;
318 break;
319 case 'O':
320 if (strcmp(name, ".OODATE") == 0)
321 name = OODATE;
322 break;
323 case 'P':
324 if (strcmp(name, ".PREFIX") == 0)
325 name = PREFIX;
326 break;
327 case 'S':
328 if (strcmp(name, ".SHELL") == 0) {
329 if (!shellPath)
330 Shell_Init();
331 }
332 break;
333 case 'T':
334 if (strcmp(name, ".TARGET") == 0)
335 name = TARGET;
336 break;
337 }
338 }
339
340 /* GNU make has an additional alias $^ == ${.ALLSRC}. */
341
342 return name;
343 }
344
345 static Var *
346 GNode_FindVar(GNode *ctxt, const char *varname, unsigned int hash)
347 {
348 return HashTable_FindValueHash(&ctxt->context, varname, hash);
349 }
350
351 /* Find the variable in the context, and maybe in other contexts as well.
352 *
353 * Input:
354 * name name to find, is not expanded any further
355 * ctxt context in which to look first
356 * elsewhere TRUE to look in other contexts as well
357 *
358 * Results:
359 * The found variable, or NULL if the variable does not exist.
360 * If the variable is an environment variable, it must be freed using
361 * VarFreeEnv after use.
362 */
363 static Var *
364 VarFind(const char *name, GNode *ctxt, Boolean elsewhere)
365 {
366 Var *var;
367 unsigned int nameHash;
368
369 /*
370 * If the variable name begins with a '.', it could very well be one of
371 * the local ones. We check the name against all the local variables
372 * and substitute the short version in for 'name' if it matches one of
373 * them.
374 */
375 name = CanonicalVarname(name);
376 nameHash = Hash_Hash(name);
377
378 /* First look for the variable in the given context. */
379 var = GNode_FindVar(ctxt, name, nameHash);
380 if (!elsewhere)
381 return var;
382
383 /* The variable was not found in the given context. Now look for it in
384 * the other contexts as well. */
385 if (var == NULL && ctxt != VAR_CMDLINE)
386 var = GNode_FindVar(VAR_CMDLINE, name, nameHash);
387
388 if (!opts.checkEnvFirst && var == NULL && ctxt != VAR_GLOBAL) {
389 var = GNode_FindVar(VAR_GLOBAL, name, nameHash);
390 if (var == NULL && ctxt != VAR_INTERNAL) {
391 /* VAR_INTERNAL is subordinate to VAR_GLOBAL */
392 var = GNode_FindVar(VAR_INTERNAL, name, nameHash);
393 }
394 }
395
396 if (var == NULL) {
397 char *env;
398
399 if ((env = getenv(name)) != NULL) {
400 char *varname = bmake_strdup(name);
401 return VarNew(varname, varname, env, VAR_FROM_ENV);
402 }
403
404 if (opts.checkEnvFirst && ctxt != VAR_GLOBAL) {
405 var = GNode_FindVar(VAR_GLOBAL, name, nameHash);
406 if (var == NULL && ctxt != VAR_INTERNAL)
407 var = GNode_FindVar(VAR_INTERNAL, name, nameHash);
408 return var;
409 }
410
411 return NULL;
412 }
413
414 return var;
415 }
416
417 /* If the variable is an environment variable, free it.
418 *
419 * Input:
420 * v the variable
421 * freeValue true if the variable value should be freed as well
422 *
423 * Results:
424 * TRUE if it is an environment variable, FALSE otherwise.
425 */
426 static Boolean
427 VarFreeEnv(Var *v, Boolean freeValue)
428 {
429 if (!(v->flags & VAR_FROM_ENV))
430 return FALSE;
431
432 free(v->name_freeIt);
433 Buf_Destroy(&v->val, freeValue);
434 free(v);
435 return TRUE;
436 }
437
438 /* Add a new variable of the given name and value to the given context.
439 * The name and val arguments are duplicated so they may safely be freed. */
440 static void
441 VarAdd(const char *name, const char *val, GNode *ctxt, VarSet_Flags flags)
442 {
443 HashEntry *he = HashTable_CreateEntry(&ctxt->context, name, NULL);
444 Var *v = VarNew(he->key /* aliased */, NULL, val,
445 flags & VAR_SET_READONLY ? VAR_READONLY : 0);
446 HashEntry_Set(he, v);
447 if (!(ctxt->flags & INTERNAL)) {
448 VAR_DEBUG3("%s:%s = %s\n", ctxt->name, name, val);
449 }
450 }
451
452 /* Remove a variable from a context, freeing all related memory as well.
453 * The variable name is expanded once. */
454 void
455 Var_Delete(const char *name, GNode *ctxt)
456 {
457 char *name_freeIt = NULL;
458 HashEntry *he;
459
460 if (strchr(name, '$') != NULL) {
461 (void)Var_Subst(name, VAR_GLOBAL, VARE_WANTRES, &name_freeIt);
462 /* TODO: handle errors */
463 name = name_freeIt;
464 }
465 he = HashTable_FindEntry(&ctxt->context, name);
466 VAR_DEBUG3("%s:delete %s%s\n",
467 ctxt->name, name, he != NULL ? "" : " (not found)");
468 free(name_freeIt);
469
470 if (he != NULL) {
471 Var *v = HashEntry_Get(he);
472 if (v->flags & VAR_EXPORTED)
473 unsetenv(v->name);
474 if (strcmp(v->name, MAKE_EXPORTED) == 0)
475 var_exportedVars = VAR_EXPORTED_NONE;
476 assert(v->name_freeIt == NULL);
477 HashTable_DeleteEntry(&ctxt->context, he);
478 Buf_Destroy(&v->val, TRUE);
479 free(v);
480 }
481 }
482
483 static Boolean
484 MayExport(const char *name)
485 {
486 if (name[0] == '.')
487 return FALSE; /* skip internals */
488 if (name[0] == '-')
489 return FALSE; /* skip misnamed variables */
490 if (name[1] == '\0') {
491 /*
492 * A single char.
493 * If it is one of the vars that should only appear in
494 * local context, skip it, else we can get Var_Subst
495 * into a loop.
496 */
497 switch (name[0]) {
498 case '@':
499 case '%':
500 case '*':
501 case '!':
502 return FALSE;
503 }
504 }
505 return TRUE;
506 }
507
508 /*
509 * Export a single variable.
510 * We ignore make internal variables (those which start with '.').
511 * Also we jump through some hoops to avoid calling setenv
512 * more than necessary since it can leak.
513 * We only manipulate flags of vars if 'parent' is set.
514 */
515 static Boolean
516 Var_Export1(const char *name, VarExportFlags flags)
517 {
518 VarExportFlags parent = flags & VAR_EXPORT_PARENT;
519 Var *v;
520 char *val;
521
522 if (!MayExport(name))
523 return FALSE;
524
525 v = VarFind(name, VAR_GLOBAL, 0);
526 if (v == NULL)
527 return FALSE;
528
529 if (!parent && (v->flags & VAR_EXPORTED) && !(v->flags & VAR_REEXPORT))
530 return FALSE; /* nothing to do */
531
532 val = Buf_GetAll(&v->val, NULL);
533 if (!(flags & VAR_EXPORT_LITERAL) && strchr(val, '$') != NULL) {
534 char *expr;
535
536 if (parent) {
537 /*
538 * Flag the variable as something we need to re-export.
539 * No point actually exporting it now though,
540 * the child process can do it at the last minute.
541 */
542 v->flags |= VAR_EXPORTED | VAR_REEXPORT;
543 return TRUE;
544 }
545 if (v->flags & VAR_IN_USE) {
546 /*
547 * We recursed while exporting in a child.
548 * This isn't going to end well, just skip it.
549 */
550 return FALSE;
551 }
552
553 /* XXX: name is injected without escaping it */
554 expr = str_concat3("${", name, "}");
555 (void)Var_Subst(expr, VAR_GLOBAL, VARE_WANTRES, &val);
556 /* TODO: handle errors */
557 setenv(name, val, 1);
558 free(val);
559 free(expr);
560 } else {
561 if (parent)
562 v->flags &= ~(unsigned)VAR_REEXPORT; /* once will do */
563 if (parent || !(v->flags & VAR_EXPORTED))
564 setenv(name, val, 1);
565 }
566
567 /*
568 * This is so Var_Set knows to call Var_Export again...
569 */
570 if (parent) {
571 v->flags |= VAR_EXPORTED;
572 }
573 return TRUE;
574 }
575
576 /*
577 * This gets called from our child processes.
578 */
579 void
580 Var_ExportVars(void)
581 {
582 char *val;
583
584 /*
585 * Several make's support this sort of mechanism for tracking
586 * recursion - but each uses a different name.
587 * We allow the makefiles to update MAKELEVEL and ensure
588 * children see a correctly incremented value.
589 */
590 char tmp[BUFSIZ];
591 snprintf(tmp, sizeof(tmp), "%d", makelevel + 1);
592 setenv(MAKE_LEVEL_ENV, tmp, 1);
593
594 if (var_exportedVars == VAR_EXPORTED_NONE)
595 return;
596
597 if (var_exportedVars == VAR_EXPORTED_ALL) {
598 HashIter hi;
599
600 /* Ouch! Exporting all variables at once is crazy... */
601 HashIter_Init(&hi, &VAR_GLOBAL->context);
602 while (HashIter_Next(&hi) != NULL) {
603 Var *var = hi.entry->value;
604 Var_Export1(var->name, 0);
605 }
606 return;
607 }
608
609 (void)Var_Subst("${" MAKE_EXPORTED ":O:u}", VAR_GLOBAL, VARE_WANTRES, &val);
610 /* TODO: handle errors */
611 if (*val) {
612 Words words = Str_Words(val, FALSE);
613 size_t i;
614
615 for (i = 0; i < words.len; i++)
616 Var_Export1(words.words[i], 0);
617 Words_Free(words);
618 }
619 free(val);
620 }
621
622 /*
623 * This is called when .export is seen or .MAKE.EXPORTED is modified.
624 *
625 * It is also called when any exported variable is modified.
626 * XXX: Is it really?
627 *
628 * str has the format "[-env|-literal] varname...".
629 */
630 void
631 Var_Export(const char *str, Boolean isExport)
632 {
633 VarExportFlags flags;
634 char *val;
635
636 if (isExport && str[0] == '\0') {
637 var_exportedVars = VAR_EXPORTED_ALL; /* use with caution! */
638 return;
639 }
640
641 if (isExport && strncmp(str, "-env", 4) == 0) {
642 str += 4;
643 flags = 0;
644 } else if (isExport && strncmp(str, "-literal", 8) == 0) {
645 str += 8;
646 flags = VAR_EXPORT_LITERAL;
647 } else {
648 flags = VAR_EXPORT_PARENT;
649 }
650
651 (void)Var_Subst(str, VAR_GLOBAL, VARE_WANTRES, &val);
652 /* TODO: handle errors */
653 if (val[0] != '\0') {
654 Words words = Str_Words(val, FALSE);
655
656 size_t i;
657 for (i = 0; i < words.len; i++) {
658 const char *name = words.words[i];
659 if (Var_Export1(name, flags)) {
660 if (var_exportedVars == VAR_EXPORTED_NONE)
661 var_exportedVars = VAR_EXPORTED_SOME;
662 if (isExport && (flags & VAR_EXPORT_PARENT)) {
663 Var_Append(MAKE_EXPORTED, name, VAR_GLOBAL);
664 }
665 }
666 }
667 Words_Free(words);
668 }
669 free(val);
670 }
671
672
673 extern char **environ;
674
675 /*
676 * This is called when .unexport[-env] is seen.
677 *
678 * str must have the form "unexport[-env] varname...".
679 */
680 void
681 Var_UnExport(const char *str)
682 {
683 const char *varnames;
684 char *varnames_freeIt;
685 Boolean unexport_env;
686
687 varnames = NULL;
688 varnames_freeIt = NULL;
689
690 str += strlen("unexport");
691 unexport_env = strncmp(str, "-env", 4) == 0;
692 if (unexport_env) {
693 const char *cp;
694 char **newenv;
695
696 cp = getenv(MAKE_LEVEL_ENV); /* we should preserve this */
697 if (environ == savedEnv) {
698 /* we have been here before! */
699 newenv = bmake_realloc(environ, 2 * sizeof(char *));
700 } else {
701 if (savedEnv) {
702 free(savedEnv);
703 savedEnv = NULL;
704 }
705 newenv = bmake_malloc(2 * sizeof(char *));
706 }
707
708 /* Note: we cannot safely free() the original environ. */
709 environ = savedEnv = newenv;
710 newenv[0] = NULL;
711 newenv[1] = NULL;
712 if (cp && *cp)
713 setenv(MAKE_LEVEL_ENV, cp, 1);
714 } else {
715 cpp_skip_whitespace(&str);
716 if (str[0] != '\0')
717 varnames = str;
718 }
719
720 if (varnames == NULL) {
721 /* Using .MAKE.EXPORTED */
722 (void)Var_Subst("${" MAKE_EXPORTED ":O:u}", VAR_GLOBAL, VARE_WANTRES,
723 &varnames_freeIt);
724 /* TODO: handle errors */
725 varnames = varnames_freeIt;
726 }
727
728 {
729 Var *v;
730 size_t i;
731
732 Words words = Str_Words(varnames, FALSE);
733 for (i = 0; i < words.len; i++) {
734 const char *varname = words.words[i];
735 v = VarFind(varname, VAR_GLOBAL, 0);
736 if (v == NULL) {
737 VAR_DEBUG1("Not unexporting \"%s\" (not found)\n", varname);
738 continue;
739 }
740
741 VAR_DEBUG1("Unexporting \"%s\"\n", varname);
742 if (!unexport_env && (v->flags & VAR_EXPORTED) &&
743 !(v->flags & VAR_REEXPORT))
744 unsetenv(v->name);
745 v->flags &= ~(unsigned)(VAR_EXPORTED | VAR_REEXPORT);
746
747 /*
748 * If we are unexporting a list,
749 * remove each one from .MAKE.EXPORTED.
750 * If we are removing them all,
751 * just delete .MAKE.EXPORTED below.
752 */
753 if (varnames == str) {
754 /* XXX: v->name is injected without escaping it */
755 char *expr = str_concat3("${" MAKE_EXPORTED ":N", v->name, "}");
756 char *cp;
757 (void)Var_Subst(expr, VAR_GLOBAL, VARE_WANTRES, &cp);
758 /* TODO: handle errors */
759 Var_Set(MAKE_EXPORTED, cp, VAR_GLOBAL);
760 free(cp);
761 free(expr);
762 }
763 }
764 Words_Free(words);
765 if (varnames != str) {
766 Var_Delete(MAKE_EXPORTED, VAR_GLOBAL);
767 free(varnames_freeIt);
768 }
769 }
770 }
771
772 /* See Var_Set for documentation. */
773 void
774 Var_Set_with_flags(const char *name, const char *val, GNode *ctxt,
775 VarSet_Flags flags)
776 {
777 const char *unexpanded_name = name;
778 char *name_freeIt = NULL;
779 Var *v;
780
781 assert(val != NULL);
782
783 if (strchr(name, '$') != NULL) {
784 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
785 /* TODO: handle errors */
786 name = name_freeIt;
787 }
788
789 if (name[0] == '\0') {
790 VAR_DEBUG2("Var_Set(\"%s\", \"%s\", ...) "
791 "name expands to empty string - ignored\n",
792 unexpanded_name, val);
793 free(name_freeIt);
794 return;
795 }
796
797 if (ctxt == VAR_GLOBAL) {
798 v = VarFind(name, VAR_CMDLINE, 0);
799 if (v != NULL) {
800 if (v->flags & VAR_FROM_CMD) {
801 VAR_DEBUG3("%s:%s = %s ignored!\n", ctxt->name, name, val);
802 goto out;
803 }
804 VarFreeEnv(v, TRUE);
805 }
806 }
807
808 /*
809 * We only look for a variable in the given context since anything set
810 * here will override anything in a lower context, so there's not much
811 * point in searching them all just to save a bit of memory...
812 */
813 v = VarFind(name, ctxt, 0);
814 if (v == NULL) {
815 if (ctxt == VAR_CMDLINE && !(flags & VAR_NO_EXPORT)) {
816 /*
817 * This var would normally prevent the same name being added
818 * to VAR_GLOBAL, so delete it from there if needed.
819 * Otherwise -V name may show the wrong value.
820 */
821 /* XXX: name is expanded for the second time */
822 Var_Delete(name, VAR_GLOBAL);
823 }
824 VarAdd(name, val, ctxt, flags);
825 } else {
826 if ((v->flags & VAR_READONLY) && !(flags & VAR_SET_READONLY)) {
827 VAR_DEBUG3("%s:%s = %s ignored (read-only)\n",
828 ctxt->name, name, val);
829 goto out;
830 }
831 Buf_Empty(&v->val);
832 Buf_AddStr(&v->val, val);
833
834 VAR_DEBUG3("%s:%s = %s\n", ctxt->name, name, val);
835 if (v->flags & VAR_EXPORTED) {
836 Var_Export1(name, VAR_EXPORT_PARENT);
837 }
838 }
839 /*
840 * Any variables given on the command line are automatically exported
841 * to the environment (as per POSIX standard)
842 * Other than internals.
843 */
844 if (ctxt == VAR_CMDLINE && !(flags & VAR_NO_EXPORT) && name[0] != '.') {
845 if (v == NULL)
846 v = VarFind(name, ctxt, 0); /* we just added it */
847 v->flags |= VAR_FROM_CMD;
848
849 /*
850 * If requested, don't export these in the environment
851 * individually. We still put them in MAKEOVERRIDES so
852 * that the command-line settings continue to override
853 * Makefile settings.
854 */
855 if (!opts.varNoExportEnv)
856 setenv(name, val, 1);
857
858 Var_Append(MAKEOVERRIDES, name, VAR_GLOBAL);
859 }
860 if (name[0] == '.' && strcmp(name, MAKE_SAVE_DOLLARS) == 0)
861 save_dollars = s2Boolean(val, save_dollars);
862
863 out:
864 free(name_freeIt);
865 if (v != NULL)
866 VarFreeEnv(v, TRUE);
867 }
868
869 /*-
870 *-----------------------------------------------------------------------
871 * Var_Set --
872 * Set the variable name to the value val in the given context.
873 *
874 * If the variable doesn't yet exist, it is created.
875 * Otherwise the new value overwrites and replaces the old value.
876 *
877 * Input:
878 * name name of the variable to set, is expanded once
879 * val value to give to the variable
880 * ctxt context in which to set it
881 *
882 * Notes:
883 * The variable is searched for only in its context before being
884 * created in that context. I.e. if the context is VAR_GLOBAL,
885 * only VAR_GLOBAL->context is searched. Likewise if it is VAR_CMDLINE,
886 * only VAR_CMDLINE->context is searched. This is done to avoid the
887 * literally thousands of unnecessary strcmp's that used to be done to
888 * set, say, $(@) or $(<).
889 * If the context is VAR_GLOBAL though, we check if the variable
890 * was set in VAR_CMDLINE from the command line and skip it if so.
891 *-----------------------------------------------------------------------
892 */
893 void
894 Var_Set(const char *name, const char *val, GNode *ctxt)
895 {
896 Var_Set_with_flags(name, val, ctxt, 0);
897 }
898
899 /*-
900 *-----------------------------------------------------------------------
901 * Var_Append --
902 * The variable of the given name has the given value appended to it in
903 * the given context.
904 *
905 * If the variable doesn't exist, it is created. Otherwise the strings
906 * are concatenated, with a space in between.
907 *
908 * Input:
909 * name name of the variable to modify, is expanded once
910 * val string to append to it
911 * ctxt context in which this should occur
912 *
913 * Notes:
914 * Only if the variable is being sought in the global context is the
915 * environment searched.
916 * XXX: Knows its calling circumstances in that if called with ctxt
917 * an actual target, it will only search that context since only
918 * a local variable could be being appended to. This is actually
919 * a big win and must be tolerated.
920 *-----------------------------------------------------------------------
921 */
922 void
923 Var_Append(const char *name, const char *val, GNode *ctxt)
924 {
925 char *name_freeIt = NULL;
926 Var *v;
927
928 assert(val != NULL);
929
930 if (strchr(name, '$') != NULL) {
931 const char *unexpanded_name = name;
932 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
933 /* TODO: handle errors */
934 name = name_freeIt;
935 if (name[0] == '\0') {
936 VAR_DEBUG2("Var_Append(\"%s\", \"%s\", ...) "
937 "name expands to empty string - ignored\n",
938 unexpanded_name, val);
939 free(name_freeIt);
940 return;
941 }
942 }
943
944 v = VarFind(name, ctxt, ctxt == VAR_GLOBAL);
945
946 if (v == NULL) {
947 /* XXX: name is expanded for the second time */
948 Var_Set(name, val, ctxt);
949 } else if (v->flags & VAR_READONLY) {
950 VAR_DEBUG1("Ignoring append to %s since it is read-only\n", name);
951 } else if (ctxt == VAR_CMDLINE || !(v->flags & VAR_FROM_CMD)) {
952 Buf_AddByte(&v->val, ' ');
953 Buf_AddStr(&v->val, val);
954
955 VAR_DEBUG3("%s:%s = %s\n",
956 ctxt->name, name, Buf_GetAll(&v->val, NULL));
957
958 if (v->flags & VAR_FROM_ENV) {
959 HashEntry *h;
960
961 /*
962 * If the original variable came from the environment, we
963 * have to install it in the global context (we could place
964 * it in the environment, but then we should provide a way to
965 * export other variables...)
966 */
967 v->flags &= ~(unsigned)VAR_FROM_ENV;
968 h = HashTable_CreateEntry(&ctxt->context, name, NULL);
969 HashEntry_Set(h, v);
970 }
971 }
972 free(name_freeIt);
973 }
974
975 /* See if the given variable exists, in the given context or in other
976 * fallback contexts.
977 *
978 * Input:
979 * name Variable to find, is expanded once
980 * ctxt Context in which to start search
981 */
982 Boolean
983 Var_Exists(const char *name, GNode *ctxt)
984 {
985 char *name_freeIt = NULL;
986 Var *v;
987
988 if (strchr(name, '$') != NULL) {
989 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
990 /* TODO: handle errors */
991 name = name_freeIt;
992 }
993
994 v = VarFind(name, ctxt, TRUE);
995 free(name_freeIt);
996 if (v == NULL)
997 return FALSE;
998
999 (void)VarFreeEnv(v, TRUE);
1000 return TRUE;
1001 }
1002
1003 /*-
1004 *-----------------------------------------------------------------------
1005 * Var_Value --
1006 * Return the unexpanded value of the given variable in the given
1007 * context, or the usual contexts.
1008 *
1009 * Input:
1010 * name name to find, is not expanded any further
1011 * ctxt context in which to search for it
1012 *
1013 * Results:
1014 * The value if the variable exists, NULL if it doesn't.
1015 * If the returned value is not NULL, the caller must free *freeIt
1016 * as soon as the returned value is no longer needed.
1017 *-----------------------------------------------------------------------
1018 */
1019 const char *
1020 Var_Value(const char *name, GNode *ctxt, void **freeIt)
1021 {
1022 Var *v = VarFind(name, ctxt, TRUE);
1023 char *value;
1024
1025 *freeIt = NULL;
1026 if (v == NULL)
1027 return NULL;
1028
1029 value = Buf_GetAll(&v->val, NULL);
1030 if (VarFreeEnv(v, FALSE))
1031 *freeIt = value;
1032 return value;
1033 }
1034
1035 /* Return the unexpanded variable value from this node, without trying to look
1036 * up the variable in any other context. */
1037 const char *
1038 Var_ValueDirect(const char *name, GNode *ctxt)
1039 {
1040 Var *v = VarFind(name, ctxt, FALSE);
1041 return v != NULL ? Buf_GetAll(&v->val, NULL) : NULL;
1042 }
1043
1044
1045 /* SepBuf is a string being built from words, interleaved with separators. */
1046 typedef struct SepBuf {
1047 Buffer buf;
1048 Boolean needSep;
1049 char sep; /* usually ' ', but see the :ts modifier */
1050 } SepBuf;
1051
1052 static void
1053 SepBuf_Init(SepBuf *buf, char sep)
1054 {
1055 Buf_Init(&buf->buf, 32 /* bytes */);
1056 buf->needSep = FALSE;
1057 buf->sep = sep;
1058 }
1059
1060 static void
1061 SepBuf_Sep(SepBuf *buf)
1062 {
1063 buf->needSep = TRUE;
1064 }
1065
1066 static void
1067 SepBuf_AddBytes(SepBuf *buf, const char *mem, size_t mem_size)
1068 {
1069 if (mem_size == 0)
1070 return;
1071 if (buf->needSep && buf->sep != '\0') {
1072 Buf_AddByte(&buf->buf, buf->sep);
1073 buf->needSep = FALSE;
1074 }
1075 Buf_AddBytes(&buf->buf, mem, mem_size);
1076 }
1077
1078 static void
1079 SepBuf_AddBytesBetween(SepBuf *buf, const char *start, const char *end)
1080 {
1081 SepBuf_AddBytes(buf, start, (size_t)(end - start));
1082 }
1083
1084 static void
1085 SepBuf_AddStr(SepBuf *buf, const char *str)
1086 {
1087 SepBuf_AddBytes(buf, str, strlen(str));
1088 }
1089
1090 static char *
1091 SepBuf_Destroy(SepBuf *buf, Boolean free_buf)
1092 {
1093 return Buf_Destroy(&buf->buf, free_buf);
1094 }
1095
1096
1097 /* This callback for ModifyWords gets a single word from a variable expression
1098 * and typically adds a modification of this word to the buffer. It may also
1099 * do nothing or add several words.
1100 *
1101 * For example, in ${:Ua b c:M*2}, the callback is called 3 times, once for
1102 * each word of "a b c". */
1103 typedef void (*ModifyWordsCallback)(const char *word, SepBuf *buf, void *data);
1104
1105
1106 /* Callback for ModifyWords to implement the :H modifier.
1107 * Add the dirname of the given word to the buffer. */
1108 static void
1109 ModifyWord_Head(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1110 {
1111 const char *slash = strrchr(word, '/');
1112 if (slash != NULL)
1113 SepBuf_AddBytesBetween(buf, word, slash);
1114 else
1115 SepBuf_AddStr(buf, ".");
1116 }
1117
1118 /* Callback for ModifyWords to implement the :T modifier.
1119 * Add the basename of the given word to the buffer. */
1120 static void
1121 ModifyWord_Tail(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1122 {
1123 const char *slash = strrchr(word, '/');
1124 const char *base = slash != NULL ? slash + 1 : word;
1125 SepBuf_AddStr(buf, base);
1126 }
1127
1128 /* Callback for ModifyWords to implement the :E modifier.
1129 * Add the filename suffix of the given word to the buffer, if it exists. */
1130 static void
1131 ModifyWord_Suffix(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1132 {
1133 const char *dot = strrchr(word, '.');
1134 if (dot != NULL)
1135 SepBuf_AddStr(buf, dot + 1);
1136 }
1137
1138 /* Callback for ModifyWords to implement the :R modifier.
1139 * Add the basename of the given word to the buffer. */
1140 static void
1141 ModifyWord_Root(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1142 {
1143 const char *dot = strrchr(word, '.');
1144 size_t len = dot != NULL ? (size_t)(dot - word) : strlen(word);
1145 SepBuf_AddBytes(buf, word, len);
1146 }
1147
1148 /* Callback for ModifyWords to implement the :M modifier.
1149 * Place the word in the buffer if it matches the given pattern. */
1150 static void
1151 ModifyWord_Match(const char *word, SepBuf *buf, void *data)
1152 {
1153 const char *pattern = data;
1154 VAR_DEBUG2("VarMatch [%s] [%s]\n", word, pattern);
1155 if (Str_Match(word, pattern))
1156 SepBuf_AddStr(buf, word);
1157 }
1158
1159 /* Callback for ModifyWords to implement the :N modifier.
1160 * Place the word in the buffer if it doesn't match the given pattern. */
1161 static void
1162 ModifyWord_NoMatch(const char *word, SepBuf *buf, void *data)
1163 {
1164 const char *pattern = data;
1165 if (!Str_Match(word, pattern))
1166 SepBuf_AddStr(buf, word);
1167 }
1168
1169 #ifdef SYSVVARSUB
1170 /* Check word against pattern for a match (% is a wildcard).
1171 *
1172 * Input:
1173 * word Word to examine
1174 * pattern Pattern to examine against
1175 *
1176 * Results:
1177 * Returns the start of the match, or NULL.
1178 * out_match_len returns the length of the match, if any.
1179 * out_hasPercent returns whether the pattern contains a percent.
1180 */
1181 static const char *
1182 SysVMatch(const char *word, const char *pattern,
1183 size_t *out_match_len, Boolean *out_hasPercent)
1184 {
1185 const char *p = pattern;
1186 const char *w = word;
1187 const char *percent;
1188 size_t w_len;
1189 size_t p_len;
1190 const char *w_tail;
1191
1192 *out_hasPercent = FALSE;
1193 percent = strchr(p, '%');
1194 if (percent != NULL) { /* ${VAR:...%...=...} */
1195 *out_hasPercent = TRUE;
1196 if (*w == '\0')
1197 return NULL; /* empty word does not match pattern */
1198
1199 /* check that the prefix matches */
1200 for (; p != percent && *w != '\0' && *w == *p; w++, p++)
1201 continue;
1202 if (p != percent)
1203 return NULL; /* No match */
1204
1205 p++; /* Skip the percent */
1206 if (*p == '\0') {
1207 /* No more pattern, return the rest of the string */
1208 *out_match_len = strlen(w);
1209 return w;
1210 }
1211 }
1212
1213 /* Test whether the tail matches */
1214 w_len = strlen(w);
1215 p_len = strlen(p);
1216 if (w_len < p_len)
1217 return NULL;
1218
1219 w_tail = w + w_len - p_len;
1220 if (memcmp(p, w_tail, p_len) != 0)
1221 return NULL;
1222
1223 *out_match_len = (size_t)(w_tail - w);
1224 return w;
1225 }
1226
1227 struct ModifyWord_SYSVSubstArgs {
1228 GNode *ctx;
1229 const char *lhs;
1230 const char *rhs;
1231 };
1232
1233 /* Callback for ModifyWords to implement the :%.from=%.to modifier. */
1234 static void
1235 ModifyWord_SYSVSubst(const char *word, SepBuf *buf, void *data)
1236 {
1237 const struct ModifyWord_SYSVSubstArgs *args = data;
1238 char *rhs_expanded;
1239 const char *rhs;
1240 const char *percent;
1241
1242 size_t match_len;
1243 Boolean lhsPercent;
1244 const char *match = SysVMatch(word, args->lhs, &match_len, &lhsPercent);
1245 if (match == NULL) {
1246 SepBuf_AddStr(buf, word);
1247 return;
1248 }
1249
1250 /* Append rhs to the buffer, substituting the first '%' with the
1251 * match, but only if the lhs had a '%' as well. */
1252
1253 (void)Var_Subst(args->rhs, args->ctx, VARE_WANTRES, &rhs_expanded);
1254 /* TODO: handle errors */
1255
1256 rhs = rhs_expanded;
1257 percent = strchr(rhs, '%');
1258
1259 if (percent != NULL && lhsPercent) {
1260 /* Copy the prefix of the replacement pattern */
1261 SepBuf_AddBytesBetween(buf, rhs, percent);
1262 rhs = percent + 1;
1263 }
1264 if (percent != NULL || !lhsPercent)
1265 SepBuf_AddBytes(buf, match, match_len);
1266
1267 /* Append the suffix of the replacement pattern */
1268 SepBuf_AddStr(buf, rhs);
1269
1270 free(rhs_expanded);
1271 }
1272 #endif
1273
1274
1275 struct ModifyWord_SubstArgs {
1276 const char *lhs;
1277 size_t lhsLen;
1278 const char *rhs;
1279 size_t rhsLen;
1280 VarPatternFlags pflags;
1281 Boolean matched;
1282 };
1283
1284 /* Callback for ModifyWords to implement the :S,from,to, modifier.
1285 * Perform a string substitution on the given word. */
1286 static void
1287 ModifyWord_Subst(const char *word, SepBuf *buf, void *data)
1288 {
1289 size_t wordLen = strlen(word);
1290 struct ModifyWord_SubstArgs *args = data;
1291 const char *match;
1292
1293 if ((args->pflags & VARP_SUB_ONE) && args->matched)
1294 goto nosub;
1295
1296 if (args->pflags & VARP_ANCHOR_START) {
1297 if (wordLen < args->lhsLen ||
1298 memcmp(word, args->lhs, args->lhsLen) != 0)
1299 goto nosub;
1300
1301 if ((args->pflags & VARP_ANCHOR_END) && wordLen != args->lhsLen)
1302 goto nosub;
1303
1304 /* :S,^prefix,replacement, or :S,^whole$,replacement, */
1305 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1306 SepBuf_AddBytes(buf, word + args->lhsLen, wordLen - args->lhsLen);
1307 args->matched = TRUE;
1308 return;
1309 }
1310
1311 if (args->pflags & VARP_ANCHOR_END) {
1312 const char *start;
1313
1314 if (wordLen < args->lhsLen)
1315 goto nosub;
1316
1317 start = word + (wordLen - args->lhsLen);
1318 if (memcmp(start, args->lhs, args->lhsLen) != 0)
1319 goto nosub;
1320
1321 /* :S,suffix$,replacement, */
1322 SepBuf_AddBytesBetween(buf, word, start);
1323 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1324 args->matched = TRUE;
1325 return;
1326 }
1327
1328 if (args->lhs[0] == '\0')
1329 goto nosub;
1330
1331 /* unanchored case, may match more than once */
1332 while ((match = strstr(word, args->lhs)) != NULL) {
1333 SepBuf_AddBytesBetween(buf, word, match);
1334 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1335 args->matched = TRUE;
1336 wordLen -= (size_t)(match - word) + args->lhsLen;
1337 word += (size_t)(match - word) + args->lhsLen;
1338 if (wordLen == 0 || !(args->pflags & VARP_SUB_GLOBAL))
1339 break;
1340 }
1341 nosub:
1342 SepBuf_AddBytes(buf, word, wordLen);
1343 }
1344
1345 #ifndef NO_REGEX
1346 /* Print the error caused by a regcomp or regexec call. */
1347 static void
1348 VarREError(int reerr, regex_t *pat, const char *str)
1349 {
1350 size_t errlen = regerror(reerr, pat, 0, 0);
1351 char *errbuf = bmake_malloc(errlen);
1352 regerror(reerr, pat, errbuf, errlen);
1353 Error("%s: %s", str, errbuf);
1354 free(errbuf);
1355 }
1356
1357 struct ModifyWord_SubstRegexArgs {
1358 regex_t re;
1359 size_t nsub;
1360 char *replace;
1361 VarPatternFlags pflags;
1362 Boolean matched;
1363 };
1364
1365 /* Callback for ModifyWords to implement the :C/from/to/ modifier.
1366 * Perform a regex substitution on the given word. */
1367 static void
1368 ModifyWord_SubstRegex(const char *word, SepBuf *buf, void *data)
1369 {
1370 struct ModifyWord_SubstRegexArgs *args = data;
1371 int xrv;
1372 const char *wp = word;
1373 char *rp;
1374 int flags = 0;
1375 regmatch_t m[10];
1376
1377 if ((args->pflags & VARP_SUB_ONE) && args->matched)
1378 goto nosub;
1379
1380 tryagain:
1381 xrv = regexec(&args->re, wp, args->nsub, m, flags);
1382
1383 switch (xrv) {
1384 case 0:
1385 args->matched = TRUE;
1386 SepBuf_AddBytes(buf, wp, (size_t)m[0].rm_so);
1387
1388 for (rp = args->replace; *rp; rp++) {
1389 if (*rp == '\\' && (rp[1] == '&' || rp[1] == '\\')) {
1390 SepBuf_AddBytes(buf, rp + 1, 1);
1391 rp++;
1392 continue;
1393 }
1394
1395 if (*rp == '&') {
1396 SepBuf_AddBytesBetween(buf, wp + m[0].rm_so, wp + m[0].rm_eo);
1397 continue;
1398 }
1399
1400 if (*rp != '\\' || !ch_isdigit(rp[1])) {
1401 SepBuf_AddBytes(buf, rp, 1);
1402 continue;
1403 }
1404
1405 { /* \0 to \9 backreference */
1406 size_t n = (size_t)(rp[1] - '0');
1407 rp++;
1408
1409 if (n >= args->nsub) {
1410 Error("No subexpression \\%zu", n);
1411 } else if (m[n].rm_so == -1) {
1412 Error("No match for subexpression \\%zu", n);
1413 } else {
1414 SepBuf_AddBytesBetween(buf, wp + m[n].rm_so,
1415 wp + m[n].rm_eo);
1416 }
1417 }
1418 }
1419
1420 wp += m[0].rm_eo;
1421 if (args->pflags & VARP_SUB_GLOBAL) {
1422 flags |= REG_NOTBOL;
1423 if (m[0].rm_so == 0 && m[0].rm_eo == 0) {
1424 SepBuf_AddBytes(buf, wp, 1);
1425 wp++;
1426 }
1427 if (*wp)
1428 goto tryagain;
1429 }
1430 if (*wp) {
1431 SepBuf_AddStr(buf, wp);
1432 }
1433 break;
1434 default:
1435 VarREError(xrv, &args->re, "Unexpected regex error");
1436 /* FALLTHROUGH */
1437 case REG_NOMATCH:
1438 nosub:
1439 SepBuf_AddStr(buf, wp);
1440 break;
1441 }
1442 }
1443 #endif
1444
1445
1446 struct ModifyWord_LoopArgs {
1447 GNode *ctx;
1448 char *tvar; /* name of temporary variable */
1449 char *str; /* string to expand */
1450 VarEvalFlags eflags;
1451 };
1452
1453 /* Callback for ModifyWords to implement the :@var (at) ...@ modifier of ODE make. */
1454 static void
1455 ModifyWord_Loop(const char *word, SepBuf *buf, void *data)
1456 {
1457 const struct ModifyWord_LoopArgs *args;
1458 char *s;
1459
1460 if (word[0] == '\0')
1461 return;
1462
1463 args = data;
1464 Var_Set_with_flags(args->tvar, word, args->ctx, VAR_NO_EXPORT);
1465 (void)Var_Subst(args->str, args->ctx, args->eflags, &s);
1466 /* TODO: handle errors */
1467
1468 VAR_DEBUG4("ModifyWord_Loop: "
1469 "in \"%s\", replace \"%s\" with \"%s\" to \"%s\"\n",
1470 word, args->tvar, args->str, s);
1471
1472 if (s[0] == '\n' || Buf_EndsWith(&buf->buf, '\n'))
1473 buf->needSep = FALSE;
1474 SepBuf_AddStr(buf, s);
1475 free(s);
1476 }
1477
1478
1479 /* The :[first..last] modifier selects words from the expression.
1480 * It can also reverse the words. */
1481 static char *
1482 VarSelectWords(char sep, Boolean oneBigWord, const char *str, int first,
1483 int last)
1484 {
1485 Words words;
1486 int len, start, end, step;
1487 int i;
1488
1489 SepBuf buf;
1490 SepBuf_Init(&buf, sep);
1491
1492 if (oneBigWord) {
1493 /* fake what Str_Words() would do if there were only one word */
1494 words.len = 1;
1495 words.words = bmake_malloc((words.len + 1) * sizeof(char *));
1496 words.freeIt = bmake_strdup(str);
1497 words.words[0] = words.freeIt;
1498 words.words[1] = NULL;
1499 } else {
1500 words = Str_Words(str, FALSE);
1501 }
1502
1503 /*
1504 * Now sanitize the given range.
1505 * If first or last are negative, convert them to the positive equivalents
1506 * (-1 gets converted to ac, -2 gets converted to (ac - 1), etc.).
1507 */
1508 len = (int)words.len;
1509 if (first < 0)
1510 first += len + 1;
1511 if (last < 0)
1512 last += len + 1;
1513
1514 /*
1515 * We avoid scanning more of the list than we need to.
1516 */
1517 if (first > last) {
1518 start = (first > len ? len : first) - 1;
1519 end = last < 1 ? 0 : last - 1;
1520 step = -1;
1521 } else {
1522 start = first < 1 ? 0 : first - 1;
1523 end = last > len ? len : last;
1524 step = 1;
1525 }
1526
1527 for (i = start; (step < 0) == (i >= end); i += step) {
1528 SepBuf_AddStr(&buf, words.words[i]);
1529 SepBuf_Sep(&buf);
1530 }
1531
1532 Words_Free(words);
1533
1534 return SepBuf_Destroy(&buf, FALSE);
1535 }
1536
1537
1538 /* Callback for ModifyWords to implement the :tA modifier.
1539 * Replace each word with the result of realpath() if successful. */
1540 static void
1541 ModifyWord_Realpath(const char *word, SepBuf *buf, void *data MAKE_ATTR_UNUSED)
1542 {
1543 struct stat st;
1544 char rbuf[MAXPATHLEN];
1545
1546 const char *rp = cached_realpath(word, rbuf);
1547 if (rp != NULL && *rp == '/' && stat(rp, &st) == 0)
1548 word = rp;
1549
1550 SepBuf_AddStr(buf, word);
1551 }
1552
1553 /* Modify each of the words of the passed string using the given function.
1554 *
1555 * Input:
1556 * str String whose words should be modified
1557 * modifyWord Function that modifies a single word
1558 * modifyWord_args Custom arguments for modifyWord
1559 *
1560 * Results:
1561 * A string of all the words modified appropriately.
1562 *-----------------------------------------------------------------------
1563 */
1564 static char *
1565 ModifyWords(const char *str,
1566 ModifyWordsCallback modifyWord, void *modifyWord_args,
1567 Boolean oneBigWord, char sep)
1568 {
1569 SepBuf result;
1570 Words words;
1571 size_t i;
1572
1573 if (oneBigWord) {
1574 SepBuf_Init(&result, sep);
1575 modifyWord(str, &result, modifyWord_args);
1576 return SepBuf_Destroy(&result, FALSE);
1577 }
1578
1579 SepBuf_Init(&result, sep);
1580
1581 words = Str_Words(str, FALSE);
1582
1583 VAR_DEBUG2("ModifyWords: split \"%s\" into %zu words\n", str, words.len);
1584
1585 for (i = 0; i < words.len; i++) {
1586 modifyWord(words.words[i], &result, modifyWord_args);
1587 if (Buf_Len(&result.buf) > 0)
1588 SepBuf_Sep(&result);
1589 }
1590
1591 Words_Free(words);
1592
1593 return SepBuf_Destroy(&result, FALSE);
1594 }
1595
1596
1597 static char *
1598 Words_JoinFree(Words words)
1599 {
1600 Buffer buf;
1601 size_t i;
1602
1603 Buf_Init(&buf, 0);
1604
1605 for (i = 0; i < words.len; i++) {
1606 if (i != 0)
1607 Buf_AddByte(&buf, ' '); /* XXX: st->sep, for consistency */
1608 Buf_AddStr(&buf, words.words[i]);
1609 }
1610
1611 Words_Free(words);
1612
1613 return Buf_Destroy(&buf, FALSE);
1614 }
1615
1616 /* Remove adjacent duplicate words. */
1617 static char *
1618 VarUniq(const char *str)
1619 {
1620 Words words = Str_Words(str, FALSE);
1621
1622 if (words.len > 1) {
1623 size_t i, j;
1624 for (j = 0, i = 1; i < words.len; i++)
1625 if (strcmp(words.words[i], words.words[j]) != 0 && (++j != i))
1626 words.words[j] = words.words[i];
1627 words.len = j + 1;
1628 }
1629
1630 return Words_JoinFree(words);
1631 }
1632
1633
1634 /* Quote shell meta-characters and space characters in the string.
1635 * If quoteDollar is set, also quote and double any '$' characters. */
1636 static char *
1637 VarQuote(const char *str, Boolean quoteDollar)
1638 {
1639 char *res;
1640 Buffer buf;
1641 Buf_Init(&buf, 0);
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) || ismeta((unsigned char)*str))
1652 Buf_AddByte(&buf, '\\');
1653 Buf_AddByte(&buf, *str);
1654 if (quoteDollar && *str == '$')
1655 Buf_AddStr(&buf, "\\$");
1656 }
1657
1658 res = Buf_Destroy(&buf, FALSE);
1659 VAR_DEBUG1("QuoteMeta: [%s]\n", res);
1660 return res;
1661 }
1662
1663 /* Compute the 32-bit hash of the given string, using the MurmurHash3
1664 * algorithm. Output is encoded as 8 hex digits, in Little Endian order. */
1665 static char *
1666 VarHash(const char *str)
1667 {
1668 static const char hexdigits[16] = "0123456789abcdef";
1669 const unsigned char *ustr = (const unsigned char *)str;
1670
1671 uint32_t h = 0x971e137bU;
1672 uint32_t c1 = 0x95543787U;
1673 uint32_t c2 = 0x2ad7eb25U;
1674 size_t len2 = strlen(str);
1675
1676 char *buf;
1677 size_t i;
1678
1679 size_t len;
1680 for (len = len2; len; ) {
1681 uint32_t k = 0;
1682 switch (len) {
1683 default:
1684 k = ((uint32_t)ustr[3] << 24) |
1685 ((uint32_t)ustr[2] << 16) |
1686 ((uint32_t)ustr[1] << 8) |
1687 (uint32_t)ustr[0];
1688 len -= 4;
1689 ustr += 4;
1690 break;
1691 case 3:
1692 k |= (uint32_t)ustr[2] << 16;
1693 /* FALLTHROUGH */
1694 case 2:
1695 k |= (uint32_t)ustr[1] << 8;
1696 /* FALLTHROUGH */
1697 case 1:
1698 k |= (uint32_t)ustr[0];
1699 len = 0;
1700 }
1701 c1 = c1 * 5 + 0x7b7d159cU;
1702 c2 = c2 * 5 + 0x6bce6396U;
1703 k *= c1;
1704 k = (k << 11) ^ (k >> 21);
1705 k *= c2;
1706 h = (h << 13) ^ (h >> 19);
1707 h = h * 5 + 0x52dce729U;
1708 h ^= k;
1709 }
1710 h ^= (uint32_t)len2;
1711 h *= 0x85ebca6b;
1712 h ^= h >> 13;
1713 h *= 0xc2b2ae35;
1714 h ^= h >> 16;
1715
1716 buf = bmake_malloc(9);
1717 for (i = 0; i < 8; i++) {
1718 buf[i] = hexdigits[h & 0x0f];
1719 h >>= 4;
1720 }
1721 buf[8] = '\0';
1722 return buf;
1723 }
1724
1725 static char *
1726 VarStrftime(const char *fmt, Boolean zulu, time_t tim)
1727 {
1728 char buf[BUFSIZ];
1729
1730 if (!tim)
1731 time(&tim);
1732 if (!*fmt)
1733 fmt = "%c";
1734 strftime(buf, sizeof(buf), fmt, zulu ? gmtime(&tim) : localtime(&tim));
1735
1736 buf[sizeof(buf) - 1] = '\0';
1737 return bmake_strdup(buf);
1738 }
1739
1740 /* The ApplyModifier functions all work in the same way. They get the
1741 * current parsing position (pp) and parse the modifier from there. The
1742 * modifier typically lasts until the next ':', or a closing '}' or ')'
1743 * (taken from st->endc), or the end of the string (parse error).
1744 *
1745 * The high-level behavior of these functions is:
1746 *
1747 * 1. parse the modifier
1748 * 2. evaluate the modifier
1749 * 3. housekeeping
1750 *
1751 * Parsing the modifier
1752 *
1753 * If parsing succeeds, the parsing position *pp is updated to point to the
1754 * first character following the modifier, which typically is either ':' or
1755 * st->endc.
1756 *
1757 * If parsing fails because of a missing delimiter (as in the :S, :C or :@
1758 * modifiers), return AMR_CLEANUP.
1759 *
1760 * If parsing fails because the modifier is unknown, return AMR_UNKNOWN to
1761 * try the SysV modifier ${VAR:from=to} as fallback. This should only be
1762 * done as long as there have been no side effects from evaluating nested
1763 * variables, to avoid evaluating them more than once. In this case, the
1764 * parsing position must not be updated. (XXX: Why not? The original parsing
1765 * position is well-known in ApplyModifiers.)
1766 *
1767 * If parsing fails and the SysV modifier ${VAR:from=to} should not be used
1768 * as a fallback, either issue an error message using Error or Parse_Error
1769 * and then return AMR_CLEANUP, or return AMR_BAD for the default error
1770 * message. Both of these return values will stop processing the variable
1771 * expression. (XXX: As of 2020-08-23, evaluation of the whole string
1772 * continues nevertheless after skipping a few bytes, which essentially is
1773 * undefined behavior. Not in the sense of C, but still it's impossible to
1774 * predict what happens in the parser.)
1775 *
1776 * Evaluating the modifier
1777 *
1778 * After parsing, the modifier is evaluated. The side effects from evaluating
1779 * nested variable expressions in the modifier text often already happen
1780 * during parsing though.
1781 *
1782 * Evaluating the modifier usually takes the current value of the variable
1783 * expression from st->val, or the variable name from st->v->name and stores
1784 * the result in st->newVal.
1785 *
1786 * If evaluating fails (as of 2020-08-23), an error message is printed using
1787 * Error. This function has no side-effects, it really just prints the error
1788 * message. Processing the expression continues as if everything were ok.
1789 * XXX: This should be fixed by adding proper error handling to Var_Subst,
1790 * Var_Parse, ApplyModifiers and ModifyWords.
1791 *
1792 * Housekeeping
1793 *
1794 * Some modifiers such as :D and :U turn undefined expressions into defined
1795 * expressions (see VEF_UNDEF, VEF_DEF).
1796 *
1797 * Some modifiers need to free some memory.
1798 */
1799
1800 typedef enum VarExprFlags {
1801 /* The variable expression is based on an undefined variable. */
1802 VEF_UNDEF = 0x01,
1803 /* The variable expression started as an undefined expression, but one
1804 * of the modifiers (such as :D or :U) has turned the expression from
1805 * undefined to defined. */
1806 VEF_DEF = 0x02
1807 } VarExprFlags;
1808
1809 ENUM_FLAGS_RTTI_2(VarExprFlags,
1810 VEF_UNDEF, VEF_DEF);
1811
1812
1813 typedef struct ApplyModifiersState {
1814 const char startc; /* '\0' or '{' or '(' */
1815 const char endc; /* '\0' or '}' or ')' */
1816 Var * const v;
1817 GNode * const ctxt;
1818 const VarEvalFlags eflags;
1819
1820 char *val; /* The old value of the expression,
1821 * before applying the modifier, never NULL */
1822 char *newVal; /* The new value of the expression,
1823 * after applying the modifier, never NULL */
1824 char sep; /* Word separator in expansions
1825 * (see the :ts modifier) */
1826 Boolean oneBigWord; /* TRUE if some modifiers that otherwise split
1827 * the variable value into words, like :S and
1828 * :C, treat the variable value as a single big
1829 * word, possibly containing spaces. */
1830 VarExprFlags exprFlags;
1831 } ApplyModifiersState;
1832
1833 static void
1834 ApplyModifiersState_Define(ApplyModifiersState *st)
1835 {
1836 if (st->exprFlags & VEF_UNDEF)
1837 st->exprFlags |= VEF_DEF;
1838 }
1839
1840 typedef enum ApplyModifierResult {
1841 AMR_OK, /* Continue parsing */
1842 AMR_UNKNOWN, /* Not a match, try other modifiers as well */
1843 AMR_BAD, /* Error out with "Bad modifier" message */
1844 AMR_CLEANUP /* Error out without error message */
1845 } ApplyModifierResult;
1846
1847 /* Allow backslashes to escape the delimiter, $, and \, but don't touch other
1848 * backslashes. */
1849 static Boolean
1850 IsEscapedModifierPart(const char *p, char delim,
1851 struct ModifyWord_SubstArgs *subst)
1852 {
1853 if (p[0] != '\\')
1854 return FALSE;
1855 if (p[1] == delim || p[1] == '\\' || p[1] == '$')
1856 return TRUE;
1857 return p[1] == '&' && subst != NULL;
1858 }
1859
1860 /*
1861 * Parse a part of a modifier such as the "from" and "to" in :S/from/to/ or
1862 * the "var" or "replacement ${var}" in :@var@replacement ${var}@, up to and
1863 * including the next unescaped delimiter. The delimiter, as well as the
1864 * backslash or the dollar, can be escaped with a backslash.
1865 *
1866 * Return the parsed (and possibly expanded) string, or NULL if no delimiter
1867 * was found. On successful return, the parsing position pp points right
1868 * after the delimiter. The delimiter is not included in the returned
1869 * value though.
1870 */
1871 static VarParseResult
1872 ParseModifierPart(
1873 const char **pp, /* The parsing position, updated upon return */
1874 char delim, /* Parsing stops at this delimiter */
1875 VarEvalFlags eflags, /* Flags for evaluating nested variables;
1876 * if VARE_WANTRES is not set, the text is
1877 * only parsed */
1878 ApplyModifiersState *st,
1879 char **out_part,
1880 size_t *out_length, /* Optionally stores the length of the returned
1881 * string, just to save another strlen call. */
1882 VarPatternFlags *out_pflags,/* For the first part of the :S modifier,
1883 * sets the VARP_ANCHOR_END flag if the last
1884 * character of the pattern is a $. */
1885 struct ModifyWord_SubstArgs *subst
1886 /* For the second part of the :S modifier,
1887 * allow ampersands to be escaped and replace
1888 * unescaped ampersands with subst->lhs. */
1889 ) {
1890 Buffer buf;
1891 const char *p;
1892
1893 Buf_Init(&buf, 0);
1894
1895 /*
1896 * Skim through until the matching delimiter is found; pick up variable
1897 * expressions on the way.
1898 */
1899 p = *pp;
1900 while (*p != '\0' && *p != delim) {
1901 const char *varstart;
1902
1903 if (IsEscapedModifierPart(p, delim, subst)) {
1904 Buf_AddByte(&buf, p[1]);
1905 p += 2;
1906 continue;
1907 }
1908
1909 if (*p != '$') { /* Unescaped, simple text */
1910 if (subst != NULL && *p == '&')
1911 Buf_AddBytes(&buf, subst->lhs, subst->lhsLen);
1912 else
1913 Buf_AddByte(&buf, *p);
1914 p++;
1915 continue;
1916 }
1917
1918 if (p[1] == delim) { /* Unescaped $ at end of pattern */
1919 if (out_pflags != NULL)
1920 *out_pflags |= VARP_ANCHOR_END;
1921 else
1922 Buf_AddByte(&buf, *p);
1923 p++;
1924 continue;
1925 }
1926
1927 if (eflags & VARE_WANTRES) { /* Nested variable, evaluated */
1928 const char *nested_p = p;
1929 const char *nested_val;
1930 void *nested_val_freeIt;
1931 VarEvalFlags nested_eflags = eflags & ~(unsigned)VARE_ASSIGN;
1932
1933 (void)Var_Parse(&nested_p, st->ctxt, nested_eflags,
1934 &nested_val, &nested_val_freeIt);
1935 /* TODO: handle errors */
1936 Buf_AddStr(&buf, nested_val);
1937 free(nested_val_freeIt);
1938 p += nested_p - p;
1939 continue;
1940 }
1941
1942 /* XXX: This whole block is very similar to Var_Parse without
1943 * VARE_WANTRES. There may be subtle edge cases though that are
1944 * not yet covered in the unit tests and that are parsed differently,
1945 * depending on whether they are evaluated or not.
1946 *
1947 * This subtle difference is not documented in the manual page,
1948 * neither is the difference between parsing :D and :M documented.
1949 * No code should ever depend on these details, but who knows. */
1950
1951 varstart = p; /* Nested variable, only parsed */
1952 if (p[1] == '(' || p[1] == '{') {
1953 /*
1954 * Find the end of this variable reference
1955 * and suck it in without further ado.
1956 * It will be interpreted later.
1957 */
1958 char have = p[1];
1959 int want = have == '(' ? ')' : '}';
1960 int depth = 1;
1961
1962 for (p += 2; *p != '\0' && depth > 0; p++) {
1963 if (p[-1] != '\\') {
1964 if (*p == have)
1965 depth++;
1966 if (*p == want)
1967 depth--;
1968 }
1969 }
1970 Buf_AddBytesBetween(&buf, varstart, p);
1971 } else {
1972 Buf_AddByte(&buf, *varstart);
1973 p++;
1974 }
1975 }
1976
1977 if (*p != delim) {
1978 *pp = p;
1979 Error("Unfinished modifier for %s ('%c' missing)", st->v->name, delim);
1980 *out_part = NULL;
1981 return VPR_PARSE_MSG;
1982 }
1983
1984 *pp = ++p;
1985 if (out_length != NULL)
1986 *out_length = Buf_Len(&buf);
1987
1988 *out_part = Buf_Destroy(&buf, FALSE);
1989 VAR_DEBUG1("Modifier part: \"%s\"\n", *out_part);
1990 return VPR_OK;
1991 }
1992
1993 /* Test whether mod starts with modname, followed by a delimiter. */
1994 static Boolean
1995 ModMatch(const char *mod, const char *modname, char endc)
1996 {
1997 size_t n = strlen(modname);
1998 return strncmp(mod, modname, n) == 0 &&
1999 (mod[n] == endc || mod[n] == ':');
2000 }
2001
2002 /* Test whether mod starts with modname, followed by a delimiter or '='. */
2003 static inline Boolean
2004 ModMatchEq(const char *mod, const char *modname, char endc)
2005 {
2006 size_t n = strlen(modname);
2007 return strncmp(mod, modname, n) == 0 &&
2008 (mod[n] == endc || mod[n] == ':' || mod[n] == '=');
2009 }
2010
2011 /* :@var (at) ...${var}...@ */
2012 static ApplyModifierResult
2013 ApplyModifier_Loop(const char **pp, ApplyModifiersState *st)
2014 {
2015 struct ModifyWord_LoopArgs args;
2016 char prev_sep;
2017 VarEvalFlags eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2018 VarParseResult res;
2019
2020 args.ctx = st->ctxt;
2021
2022 (*pp)++; /* Skip the first '@' */
2023 res = ParseModifierPart(pp, '@', eflags, st,
2024 &args.tvar, NULL, NULL, NULL);
2025 if (res != VPR_OK)
2026 return AMR_CLEANUP;
2027 if (DEBUG(LINT) && strchr(args.tvar, '$') != NULL) {
2028 Parse_Error(PARSE_FATAL,
2029 "In the :@ modifier of \"%s\", the variable name \"%s\" "
2030 "must not contain a dollar.",
2031 st->v->name, args.tvar);
2032 return AMR_CLEANUP;
2033 }
2034
2035 res = ParseModifierPart(pp, '@', eflags, st,
2036 &args.str, NULL, NULL, NULL);
2037 if (res != VPR_OK)
2038 return AMR_CLEANUP;
2039
2040 args.eflags = st->eflags & (VARE_UNDEFERR | VARE_WANTRES);
2041 prev_sep = st->sep;
2042 st->sep = ' '; /* XXX: should be st->sep for consistency */
2043 st->newVal = ModifyWords(st->val, ModifyWord_Loop, &args,
2044 st->oneBigWord, st->sep);
2045 st->sep = prev_sep;
2046 Var_Delete(args.tvar, st->ctxt);
2047 free(args.tvar);
2048 free(args.str);
2049 return AMR_OK;
2050 }
2051
2052 /* :Ddefined or :Uundefined */
2053 static ApplyModifierResult
2054 ApplyModifier_Defined(const char **pp, ApplyModifiersState *st)
2055 {
2056 Buffer buf;
2057 const char *p;
2058
2059 VarEvalFlags eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2060 if (st->eflags & VARE_WANTRES) {
2061 if ((**pp == 'D') == !(st->exprFlags & VEF_UNDEF))
2062 eflags |= VARE_WANTRES;
2063 }
2064
2065 Buf_Init(&buf, 0);
2066 p = *pp + 1;
2067 while (*p != st->endc && *p != ':' && *p != '\0') {
2068
2069 /* Escaped delimiter or other special character */
2070 if (*p == '\\') {
2071 char c = p[1];
2072 if (c == st->endc || c == ':' || c == '$' || c == '\\') {
2073 Buf_AddByte(&buf, c);
2074 p += 2;
2075 continue;
2076 }
2077 }
2078
2079 /* Nested variable expression */
2080 if (*p == '$') {
2081 const char *nested_val;
2082 void *nested_val_freeIt;
2083
2084 (void)Var_Parse(&p, st->ctxt, eflags,
2085 &nested_val, &nested_val_freeIt);
2086 /* TODO: handle errors */
2087 Buf_AddStr(&buf, nested_val);
2088 free(nested_val_freeIt);
2089 continue;
2090 }
2091
2092 /* Ordinary text */
2093 Buf_AddByte(&buf, *p);
2094 p++;
2095 }
2096 *pp = p;
2097
2098 ApplyModifiersState_Define(st);
2099
2100 if (eflags & VARE_WANTRES) {
2101 st->newVal = Buf_Destroy(&buf, FALSE);
2102 } else {
2103 st->newVal = st->val;
2104 Buf_Destroy(&buf, TRUE);
2105 }
2106 return AMR_OK;
2107 }
2108
2109 /* :L */
2110 static ApplyModifierResult
2111 ApplyModifier_Literal(const char **pp, ApplyModifiersState *st)
2112 {
2113 ApplyModifiersState_Define(st);
2114 st->newVal = bmake_strdup(st->v->name);
2115 (*pp)++;
2116 return AMR_OK;
2117 }
2118
2119 /* :gmtime */
2120 static ApplyModifierResult
2121 ApplyModifier_Gmtime(const char **pp, ApplyModifiersState *st)
2122 {
2123 time_t utc;
2124
2125 const char *mod = *pp;
2126 if (!ModMatchEq(mod, "gmtime", st->endc))
2127 return AMR_UNKNOWN;
2128
2129 if (mod[6] == '=') {
2130 char *ep;
2131 utc = (time_t)strtoul(mod + 7, &ep, 10);
2132 *pp = ep;
2133 } else {
2134 utc = 0;
2135 *pp = mod + 6;
2136 }
2137 st->newVal = VarStrftime(st->val, TRUE, utc);
2138 return AMR_OK;
2139 }
2140
2141 /* :localtime */
2142 static ApplyModifierResult
2143 ApplyModifier_Localtime(const char **pp, ApplyModifiersState *st)
2144 {
2145 time_t utc;
2146
2147 const char *mod = *pp;
2148 if (!ModMatchEq(mod, "localtime", st->endc))
2149 return AMR_UNKNOWN;
2150
2151 if (mod[9] == '=') {
2152 char *ep;
2153 utc = (time_t)strtoul(mod + 10, &ep, 10);
2154 *pp = ep;
2155 } else {
2156 utc = 0;
2157 *pp = mod + 9;
2158 }
2159 st->newVal = VarStrftime(st->val, FALSE, utc);
2160 return AMR_OK;
2161 }
2162
2163 /* :hash */
2164 static ApplyModifierResult
2165 ApplyModifier_Hash(const char **pp, ApplyModifiersState *st)
2166 {
2167 if (!ModMatch(*pp, "hash", st->endc))
2168 return AMR_UNKNOWN;
2169
2170 st->newVal = VarHash(st->val);
2171 *pp += 4;
2172 return AMR_OK;
2173 }
2174
2175 /* :P */
2176 static ApplyModifierResult
2177 ApplyModifier_Path(const char **pp, ApplyModifiersState *st)
2178 {
2179 GNode *gn;
2180 char *path;
2181
2182 ApplyModifiersState_Define(st);
2183
2184 gn = Targ_FindNode(st->v->name);
2185 if (gn == NULL || gn->type & OP_NOPATH) {
2186 path = NULL;
2187 } else if (gn->path != NULL) {
2188 path = bmake_strdup(gn->path);
2189 } else {
2190 SearchPath *searchPath = Suff_FindPath(gn);
2191 path = Dir_FindFile(st->v->name, searchPath);
2192 }
2193 if (path == NULL)
2194 path = bmake_strdup(st->v->name);
2195 st->newVal = path;
2196
2197 (*pp)++;
2198 return AMR_OK;
2199 }
2200
2201 /* :!cmd! */
2202 static ApplyModifierResult
2203 ApplyModifier_ShellCommand(const char **pp, ApplyModifiersState *st)
2204 {
2205 char *cmd;
2206 const char *errfmt;
2207 VarParseResult res;
2208
2209 (*pp)++;
2210 res = ParseModifierPart(pp, '!', st->eflags, st,
2211 &cmd, NULL, NULL, NULL);
2212 if (res != VPR_OK)
2213 return AMR_CLEANUP;
2214
2215 errfmt = NULL;
2216 if (st->eflags & VARE_WANTRES)
2217 st->newVal = Cmd_Exec(cmd, &errfmt);
2218 else
2219 st->newVal = emptyString;
2220 free(cmd);
2221
2222 if (errfmt != NULL)
2223 Error(errfmt, st->val); /* XXX: why still return AMR_OK? */
2224
2225 ApplyModifiersState_Define(st);
2226 return AMR_OK;
2227 }
2228
2229 /* The :range modifier generates an integer sequence as long as the words.
2230 * The :range=7 modifier generates an integer sequence from 1 to 7. */
2231 static ApplyModifierResult
2232 ApplyModifier_Range(const char **pp, ApplyModifiersState *st)
2233 {
2234 size_t n;
2235 Buffer buf;
2236 size_t i;
2237
2238 const char *mod = *pp;
2239 if (!ModMatchEq(mod, "range", st->endc))
2240 return AMR_UNKNOWN;
2241
2242 if (mod[5] == '=') {
2243 char *ep;
2244 n = (size_t)strtoul(mod + 6, &ep, 10);
2245 *pp = ep;
2246 } else {
2247 n = 0;
2248 *pp = mod + 5;
2249 }
2250
2251 if (n == 0) {
2252 Words words = Str_Words(st->val, FALSE);
2253 n = words.len;
2254 Words_Free(words);
2255 }
2256
2257 Buf_Init(&buf, 0);
2258
2259 for (i = 0; i < n; i++) {
2260 if (i != 0)
2261 Buf_AddByte(&buf, ' '); /* XXX: st->sep, for consistency */
2262 Buf_AddInt(&buf, 1 + (int)i);
2263 }
2264
2265 st->newVal = Buf_Destroy(&buf, FALSE);
2266 return AMR_OK;
2267 }
2268
2269 /* :Mpattern or :Npattern */
2270 static ApplyModifierResult
2271 ApplyModifier_Match(const char **pp, ApplyModifiersState *st)
2272 {
2273 const char *mod = *pp;
2274 Boolean copy = FALSE; /* pattern should be, or has been, copied */
2275 Boolean needSubst = FALSE;
2276 const char *endpat;
2277 char *pattern;
2278 ModifyWordsCallback callback;
2279
2280 /*
2281 * In the loop below, ignore ':' unless we are at (or back to) the
2282 * original brace level.
2283 * XXX This will likely not work right if $() and ${} are intermixed.
2284 */
2285 int nest = 0;
2286 const char *p;
2287 for (p = mod + 1; *p != '\0' && !(*p == ':' && nest == 0); p++) {
2288 if (*p == '\\' &&
2289 (p[1] == ':' || p[1] == st->endc || p[1] == st->startc)) {
2290 if (!needSubst)
2291 copy = TRUE;
2292 p++;
2293 continue;
2294 }
2295 if (*p == '$')
2296 needSubst = TRUE;
2297 if (*p == '(' || *p == '{')
2298 nest++;
2299 if (*p == ')' || *p == '}') {
2300 nest--;
2301 if (nest < 0)
2302 break;
2303 }
2304 }
2305 *pp = p;
2306 endpat = p;
2307
2308 if (copy) {
2309 char *dst;
2310 const char *src;
2311
2312 /* Compress the \:'s out of the pattern. */
2313 pattern = bmake_malloc((size_t)(endpat - (mod + 1)) + 1);
2314 dst = pattern;
2315 src = mod + 1;
2316 for (; src < endpat; src++, dst++) {
2317 if (src[0] == '\\' && src + 1 < endpat &&
2318 /* XXX: st->startc is missing here; see above */
2319 (src[1] == ':' || src[1] == st->endc))
2320 src++;
2321 *dst = *src;
2322 }
2323 *dst = '\0';
2324 endpat = dst;
2325 } else {
2326 pattern = bmake_strsedup(mod + 1, endpat);
2327 }
2328
2329 if (needSubst) {
2330 /* pattern contains embedded '$', so use Var_Subst to expand it. */
2331 char *old_pattern = pattern;
2332 (void)Var_Subst(pattern, st->ctxt, st->eflags, &pattern);
2333 /* TODO: handle errors */
2334 free(old_pattern);
2335 }
2336
2337 VAR_DEBUG3("Pattern[%s] for [%s] is [%s]\n", st->v->name, st->val, pattern);
2338
2339 callback = mod[0] == 'M' ? ModifyWord_Match : ModifyWord_NoMatch;
2340 st->newVal = ModifyWords(st->val, callback, pattern,
2341 st->oneBigWord, st->sep);
2342 free(pattern);
2343 return AMR_OK;
2344 }
2345
2346 /* :S,from,to, */
2347 static ApplyModifierResult
2348 ApplyModifier_Subst(const char **pp, ApplyModifiersState *st)
2349 {
2350 struct ModifyWord_SubstArgs args;
2351 char *lhs, *rhs;
2352 Boolean oneBigWord;
2353 VarParseResult res;
2354
2355 char delim = (*pp)[1];
2356 if (delim == '\0') {
2357 Error("Missing delimiter for :S modifier");
2358 (*pp)++;
2359 return AMR_CLEANUP;
2360 }
2361
2362 *pp += 2;
2363
2364 args.pflags = 0;
2365 args.matched = FALSE;
2366
2367 /*
2368 * If pattern begins with '^', it is anchored to the
2369 * start of the word -- skip over it and flag pattern.
2370 */
2371 if (**pp == '^') {
2372 args.pflags |= VARP_ANCHOR_START;
2373 (*pp)++;
2374 }
2375
2376 res = ParseModifierPart(pp, delim, st->eflags, st,
2377 &lhs, &args.lhsLen, &args.pflags, NULL);
2378 if (res != VPR_OK)
2379 return AMR_CLEANUP;
2380 args.lhs = lhs;
2381
2382 res = ParseModifierPart(pp, delim, st->eflags, st,
2383 &rhs, &args.rhsLen, NULL, &args);
2384 if (res != VPR_OK)
2385 return AMR_CLEANUP;
2386 args.rhs = rhs;
2387
2388 oneBigWord = st->oneBigWord;
2389 for (;; (*pp)++) {
2390 switch (**pp) {
2391 case 'g':
2392 args.pflags |= VARP_SUB_GLOBAL;
2393 continue;
2394 case '1':
2395 args.pflags |= VARP_SUB_ONE;
2396 continue;
2397 case 'W':
2398 oneBigWord = TRUE;
2399 continue;
2400 }
2401 break;
2402 }
2403
2404 st->newVal = ModifyWords(st->val, ModifyWord_Subst, &args,
2405 oneBigWord, st->sep);
2406
2407 free(lhs);
2408 free(rhs);
2409 return AMR_OK;
2410 }
2411
2412 #ifndef NO_REGEX
2413
2414 /* :C,from,to, */
2415 static ApplyModifierResult
2416 ApplyModifier_Regex(const char **pp, ApplyModifiersState *st)
2417 {
2418 char *re;
2419 struct ModifyWord_SubstRegexArgs args;
2420 Boolean oneBigWord;
2421 int error;
2422 VarParseResult res;
2423
2424 char delim = (*pp)[1];
2425 if (delim == '\0') {
2426 Error("Missing delimiter for :C modifier");
2427 (*pp)++;
2428 return AMR_CLEANUP;
2429 }
2430
2431 *pp += 2;
2432
2433 res = ParseModifierPart(pp, delim, st->eflags, st,
2434 &re, NULL, NULL, NULL);
2435 if (res != VPR_OK)
2436 return AMR_CLEANUP;
2437
2438 res = ParseModifierPart(pp, delim, st->eflags, st,
2439 &args.replace, NULL, NULL, NULL);
2440 if (args.replace == NULL) {
2441 free(re);
2442 return AMR_CLEANUP;
2443 }
2444
2445 args.pflags = 0;
2446 args.matched = FALSE;
2447 oneBigWord = st->oneBigWord;
2448 for (;; (*pp)++) {
2449 switch (**pp) {
2450 case 'g':
2451 args.pflags |= VARP_SUB_GLOBAL;
2452 continue;
2453 case '1':
2454 args.pflags |= VARP_SUB_ONE;
2455 continue;
2456 case 'W':
2457 oneBigWord = TRUE;
2458 continue;
2459 }
2460 break;
2461 }
2462
2463 error = regcomp(&args.re, re, REG_EXTENDED);
2464 free(re);
2465 if (error) {
2466 VarREError(error, &args.re, "Regex compilation error");
2467 free(args.replace);
2468 return AMR_CLEANUP;
2469 }
2470
2471 args.nsub = args.re.re_nsub + 1;
2472 if (args.nsub > 10)
2473 args.nsub = 10;
2474 st->newVal = ModifyWords(st->val, ModifyWord_SubstRegex, &args,
2475 oneBigWord, st->sep);
2476 regfree(&args.re);
2477 free(args.replace);
2478 return AMR_OK;
2479 }
2480 #endif
2481
2482 /* :Q, :q */
2483 static ApplyModifierResult
2484 ApplyModifier_Quote(const char **pp, ApplyModifiersState *st)
2485 {
2486 if ((*pp)[1] == st->endc || (*pp)[1] == ':') {
2487 st->newVal = VarQuote(st->val, **pp == 'q');
2488 (*pp)++;
2489 return AMR_OK;
2490 } else
2491 return AMR_UNKNOWN;
2492 }
2493
2494 static void
2495 ModifyWord_Copy(const char *word, SepBuf *buf, void *data MAKE_ATTR_UNUSED)
2496 {
2497 SepBuf_AddStr(buf, word);
2498 }
2499
2500 /* :ts<separator> */
2501 static ApplyModifierResult
2502 ApplyModifier_ToSep(const char **pp, ApplyModifiersState *st)
2503 {
2504 const char *sep = *pp + 2;
2505
2506 /* ":ts<any><endc>" or ":ts<any>:" */
2507 if (sep[0] != st->endc && (sep[1] == st->endc || sep[1] == ':')) {
2508 st->sep = sep[0];
2509 *pp = sep + 1;
2510 goto ok;
2511 }
2512
2513 /* ":ts<endc>" or ":ts:" */
2514 if (sep[0] == st->endc || sep[0] == ':') {
2515 st->sep = '\0'; /* no separator */
2516 *pp = sep;
2517 goto ok;
2518 }
2519
2520 /* ":ts<unrecognised><unrecognised>". */
2521 if (sep[0] != '\\') {
2522 (*pp)++; /* just for backwards compatibility */
2523 return AMR_BAD;
2524 }
2525
2526 /* ":ts\n" */
2527 if (sep[1] == 'n') {
2528 st->sep = '\n';
2529 *pp = sep + 2;
2530 goto ok;
2531 }
2532
2533 /* ":ts\t" */
2534 if (sep[1] == 't') {
2535 st->sep = '\t';
2536 *pp = sep + 2;
2537 goto ok;
2538 }
2539
2540 /* ":ts\x40" or ":ts\100" */
2541 {
2542 const char *numStart = sep + 1;
2543 int base = 8; /* assume octal */
2544 char *end;
2545
2546 if (sep[1] == 'x') {
2547 base = 16;
2548 numStart++;
2549 } else if (!ch_isdigit(sep[1])) {
2550 (*pp)++; /* just for backwards compatibility */
2551 return AMR_BAD; /* ":ts<backslash><unrecognised>". */
2552 }
2553
2554 st->sep = (char)strtoul(numStart, &end, base);
2555 if (*end != ':' && *end != st->endc) {
2556 (*pp)++; /* just for backwards compatibility */
2557 return AMR_BAD;
2558 }
2559 *pp = end;
2560 }
2561
2562 ok:
2563 st->newVal = ModifyWords(st->val, ModifyWord_Copy, NULL,
2564 st->oneBigWord, st->sep);
2565 return AMR_OK;
2566 }
2567
2568 /* :tA, :tu, :tl, :ts<separator>, etc. */
2569 static ApplyModifierResult
2570 ApplyModifier_To(const char **pp, ApplyModifiersState *st)
2571 {
2572 const char *mod = *pp;
2573 assert(mod[0] == 't');
2574
2575 if (mod[1] == st->endc || mod[1] == ':' || mod[1] == '\0') {
2576 *pp = mod + 1;
2577 return AMR_BAD; /* Found ":t<endc>" or ":t:". */
2578 }
2579
2580 if (mod[1] == 's')
2581 return ApplyModifier_ToSep(pp, st);
2582
2583 if (mod[2] != st->endc && mod[2] != ':') {
2584 *pp = mod + 1;
2585 return AMR_BAD; /* Found ":t<unrecognised><unrecognised>". */
2586 }
2587
2588 /* Check for two-character options: ":tu", ":tl" */
2589 if (mod[1] == 'A') { /* absolute path */
2590 st->newVal = ModifyWords(st->val, ModifyWord_Realpath, NULL,
2591 st->oneBigWord, st->sep);
2592 *pp = mod + 2;
2593 return AMR_OK;
2594 }
2595
2596 if (mod[1] == 'u') { /* :tu */
2597 size_t i;
2598 size_t len = strlen(st->val);
2599 st->newVal = bmake_malloc(len + 1);
2600 for (i = 0; i < len + 1; i++)
2601 st->newVal[i] = ch_toupper(st->val[i]);
2602 *pp = mod + 2;
2603 return AMR_OK;
2604 }
2605
2606 if (mod[1] == 'l') { /* :tl */
2607 size_t i;
2608 size_t len = strlen(st->val);
2609 st->newVal = bmake_malloc(len + 1);
2610 for (i = 0; i < len + 1; i++)
2611 st->newVal[i] = ch_tolower(st->val[i]);
2612 *pp = mod + 2;
2613 return AMR_OK;
2614 }
2615
2616 if (mod[1] == 'W' || mod[1] == 'w') { /* :tW, :tw */
2617 st->oneBigWord = mod[1] == 'W';
2618 st->newVal = st->val;
2619 *pp = mod + 2;
2620 return AMR_OK;
2621 }
2622
2623 /* Found ":t<unrecognised>:" or ":t<unrecognised><endc>". */
2624 *pp = mod + 1;
2625 return AMR_BAD;
2626 }
2627
2628 /* :[#], :[1], etc. */
2629 static ApplyModifierResult
2630 ApplyModifier_Words(const char **pp, ApplyModifiersState *st)
2631 {
2632 char *estr;
2633 char *ep;
2634 int first, last;
2635 VarParseResult res;
2636
2637 (*pp)++; /* skip the '[' */
2638 res = ParseModifierPart(pp, ']', st->eflags, st,
2639 &estr, NULL, NULL, NULL);
2640 if (res != VPR_OK)
2641 return AMR_CLEANUP;
2642
2643 /* now *pp points just after the closing ']' */
2644 if (**pp != ':' && **pp != st->endc)
2645 goto bad_modifier; /* Found junk after ']' */
2646
2647 if (estr[0] == '\0')
2648 goto bad_modifier; /* empty square brackets in ":[]". */
2649
2650 if (estr[0] == '#' && estr[1] == '\0') { /* Found ":[#]" */
2651 if (st->oneBigWord) {
2652 st->newVal = bmake_strdup("1");
2653 } else {
2654 Buffer buf;
2655
2656 Words words = Str_Words(st->val, FALSE);
2657 size_t ac = words.len;
2658 Words_Free(words);
2659
2660 Buf_Init(&buf, 4); /* 3 digits + '\0' is usually enough */
2661 Buf_AddInt(&buf, (int)ac);
2662 st->newVal = Buf_Destroy(&buf, FALSE);
2663 }
2664 goto ok;
2665 }
2666
2667 if (estr[0] == '*' && estr[1] == '\0') {
2668 /* Found ":[*]" */
2669 st->oneBigWord = TRUE;
2670 st->newVal = st->val;
2671 goto ok;
2672 }
2673
2674 if (estr[0] == '@' && estr[1] == '\0') {
2675 /* Found ":[@]" */
2676 st->oneBigWord = FALSE;
2677 st->newVal = st->val;
2678 goto ok;
2679 }
2680
2681 /*
2682 * We expect estr to contain a single integer for :[N], or two integers
2683 * separated by ".." for :[start..end].
2684 */
2685 first = (int)strtol(estr, &ep, 0);
2686 if (ep == estr) /* Found junk instead of a number */
2687 goto bad_modifier;
2688
2689 if (ep[0] == '\0') { /* Found only one integer in :[N] */
2690 last = first;
2691 } else if (ep[0] == '.' && ep[1] == '.' && ep[2] != '\0') {
2692 /* Expecting another integer after ".." */
2693 ep += 2;
2694 last = (int)strtol(ep, &ep, 0);
2695 if (ep[0] != '\0') /* Found junk after ".." */
2696 goto bad_modifier;
2697 } else
2698 goto bad_modifier; /* Found junk instead of ".." */
2699
2700 /*
2701 * Now seldata is properly filled in, but we still have to check for 0 as
2702 * a special case.
2703 */
2704 if (first == 0 && last == 0) {
2705 /* ":[0]" or perhaps ":[0..0]" */
2706 st->oneBigWord = TRUE;
2707 st->newVal = st->val;
2708 goto ok;
2709 }
2710
2711 /* ":[0..N]" or ":[N..0]" */
2712 if (first == 0 || last == 0)
2713 goto bad_modifier;
2714
2715 /* Normal case: select the words described by seldata. */
2716 st->newVal = VarSelectWords(st->sep, st->oneBigWord, st->val, first, last);
2717
2718 ok:
2719 free(estr);
2720 return AMR_OK;
2721
2722 bad_modifier:
2723 free(estr);
2724 return AMR_BAD;
2725 }
2726
2727 static int
2728 str_cmp_asc(const void *a, const void *b)
2729 {
2730 return strcmp(*(const char * const *)a, *(const char * const *)b);
2731 }
2732
2733 static int
2734 str_cmp_desc(const void *a, const void *b)
2735 {
2736 return strcmp(*(const char * const *)b, *(const char * const *)a);
2737 }
2738
2739 /* :O (order ascending) or :Or (order descending) or :Ox (shuffle) */
2740 static ApplyModifierResult
2741 ApplyModifier_Order(const char **pp, ApplyModifiersState *st)
2742 {
2743 const char *mod = (*pp)++; /* skip past the 'O' in any case */
2744
2745 Words words = Str_Words(st->val, FALSE);
2746
2747 if (mod[1] == st->endc || mod[1] == ':') {
2748 /* :O sorts ascending */
2749 qsort(words.words, words.len, sizeof(char *), str_cmp_asc);
2750
2751 } else if ((mod[1] == 'r' || mod[1] == 'x') &&
2752 (mod[2] == st->endc || mod[2] == ':')) {
2753 (*pp)++;
2754
2755 if (mod[1] == 'r') {
2756 /* :Or sorts descending */
2757 qsort(words.words, words.len, sizeof(char *), str_cmp_desc);
2758
2759 } else {
2760 /* :Ox shuffles
2761 *
2762 * We will use [ac..2] range for mod factors. This will produce
2763 * random numbers in [(ac-1)..0] interval, and minimal
2764 * reasonable value for mod factor is 2 (the mod 1 will produce
2765 * 0 with probability 1).
2766 */
2767 size_t i;
2768 for (i = words.len - 1; i > 0; i--) {
2769 size_t rndidx = (size_t)random() % (i + 1);
2770 char *t = words.words[i];
2771 words.words[i] = words.words[rndidx];
2772 words.words[rndidx] = t;
2773 }
2774 }
2775 } else {
2776 Words_Free(words);
2777 return AMR_BAD;
2778 }
2779
2780 st->newVal = Words_JoinFree(words);
2781 return AMR_OK;
2782 }
2783
2784 /* :? then : else */
2785 static ApplyModifierResult
2786 ApplyModifier_IfElse(const char **pp, ApplyModifiersState *st)
2787 {
2788 char *then_expr, *else_expr;
2789 VarParseResult res;
2790
2791 Boolean value = FALSE;
2792 VarEvalFlags then_eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2793 VarEvalFlags else_eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2794
2795 int cond_rc = COND_PARSE; /* anything other than COND_INVALID */
2796 if (st->eflags & VARE_WANTRES) {
2797 cond_rc = Cond_EvalCondition(st->v->name, &value);
2798 if (cond_rc != COND_INVALID && value)
2799 then_eflags |= VARE_WANTRES;
2800 if (cond_rc != COND_INVALID && !value)
2801 else_eflags |= VARE_WANTRES;
2802 }
2803
2804 (*pp)++; /* skip past the '?' */
2805 res = ParseModifierPart(pp, ':', then_eflags, st,
2806 &then_expr, NULL, NULL, NULL);
2807 if (res != VPR_OK)
2808 return AMR_CLEANUP;
2809
2810 res = ParseModifierPart(pp, st->endc, else_eflags, st,
2811 &else_expr, NULL, NULL, NULL);
2812 if (res != VPR_OK)
2813 return AMR_CLEANUP;
2814
2815 (*pp)--;
2816 if (cond_rc == COND_INVALID) {
2817 Error("Bad conditional expression `%s' in %s?%s:%s",
2818 st->v->name, st->v->name, then_expr, else_expr);
2819 return AMR_CLEANUP;
2820 }
2821
2822 if (value) {
2823 st->newVal = then_expr;
2824 free(else_expr);
2825 } else {
2826 st->newVal = else_expr;
2827 free(then_expr);
2828 }
2829 ApplyModifiersState_Define(st);
2830 return AMR_OK;
2831 }
2832
2833 /*
2834 * The ::= modifiers actually assign a value to the variable.
2835 * Their main purpose is in supporting modifiers of .for loop
2836 * iterators and other obscure uses. They always expand to
2837 * nothing. In a target rule that would otherwise expand to an
2838 * empty line they can be preceded with @: to keep make happy.
2839 * Eg.
2840 *
2841 * foo: .USE
2842 * .for i in ${.TARGET} ${.TARGET:R}.gz
2843 * @: ${t::=$i}
2844 * @echo blah ${t:T}
2845 * .endfor
2846 *
2847 * ::=<str> Assigns <str> as the new value of variable.
2848 * ::?=<str> Assigns <str> as value of variable if
2849 * it was not already set.
2850 * ::+=<str> Appends <str> to variable.
2851 * ::!=<cmd> Assigns output of <cmd> as the new value of
2852 * variable.
2853 */
2854 static ApplyModifierResult
2855 ApplyModifier_Assign(const char **pp, ApplyModifiersState *st)
2856 {
2857 GNode *v_ctxt;
2858 char delim;
2859 char *val;
2860 VarParseResult res;
2861
2862 const char *mod = *pp;
2863 const char *op = mod + 1;
2864 if (!(op[0] == '=' ||
2865 (op[1] == '=' &&
2866 (op[0] == '!' || op[0] == '+' || op[0] == '?'))))
2867 return AMR_UNKNOWN; /* "::<unrecognised>" */
2868
2869
2870 if (st->v->name[0] == '\0') {
2871 *pp = mod + 1;
2872 return AMR_BAD;
2873 }
2874
2875 v_ctxt = st->ctxt; /* context where v belongs */
2876 if (!(st->exprFlags & VEF_UNDEF) && st->ctxt != VAR_GLOBAL) {
2877 Var *gv = VarFind(st->v->name, st->ctxt, 0);
2878 if (gv == NULL)
2879 v_ctxt = VAR_GLOBAL;
2880 else
2881 VarFreeEnv(gv, TRUE);
2882 }
2883
2884 switch (op[0]) {
2885 case '+':
2886 case '?':
2887 case '!':
2888 *pp = mod + 3;
2889 break;
2890 default:
2891 *pp = mod + 2;
2892 break;
2893 }
2894
2895 delim = st->startc == '(' ? ')' : '}';
2896 res = ParseModifierPart(pp, delim, st->eflags, st, &val, NULL, NULL, NULL);
2897 if (res != VPR_OK)
2898 return AMR_CLEANUP;
2899
2900 (*pp)--;
2901
2902 if (st->eflags & VARE_WANTRES) {
2903 switch (op[0]) {
2904 case '+':
2905 Var_Append(st->v->name, val, v_ctxt);
2906 break;
2907 case '!': {
2908 const char *errfmt;
2909 char *cmd_output = Cmd_Exec(val, &errfmt);
2910 if (errfmt)
2911 Error(errfmt, val);
2912 else
2913 Var_Set(st->v->name, cmd_output, v_ctxt);
2914 free(cmd_output);
2915 break;
2916 }
2917 case '?':
2918 if (!(st->exprFlags & VEF_UNDEF))
2919 break;
2920 /* FALLTHROUGH */
2921 default:
2922 Var_Set(st->v->name, val, v_ctxt);
2923 break;
2924 }
2925 }
2926 free(val);
2927 st->newVal = emptyString;
2928 return AMR_OK;
2929 }
2930
2931 /* remember current value */
2932 static ApplyModifierResult
2933 ApplyModifier_Remember(const char **pp, ApplyModifiersState *st)
2934 {
2935 const char *mod = *pp;
2936 if (!ModMatchEq(mod, "_", st->endc))
2937 return AMR_UNKNOWN;
2938
2939 if (mod[1] == '=') {
2940 size_t n = strcspn(mod + 2, ":)}");
2941 char *name = bmake_strldup(mod + 2, n);
2942 Var_Set(name, st->val, st->ctxt);
2943 free(name);
2944 *pp = mod + 2 + n;
2945 } else {
2946 Var_Set("_", st->val, st->ctxt);
2947 *pp = mod + 1;
2948 }
2949 st->newVal = st->val;
2950 return AMR_OK;
2951 }
2952
2953 /* Apply the given function to each word of the variable value. */
2954 static ApplyModifierResult
2955 ApplyModifier_WordFunc(const char **pp, ApplyModifiersState *st,
2956 ModifyWordsCallback modifyWord)
2957 {
2958 char delim = (*pp)[1];
2959 if (delim != st->endc && delim != ':')
2960 return AMR_UNKNOWN;
2961
2962 st->newVal = ModifyWords(st->val, modifyWord, NULL,
2963 st->oneBigWord, st->sep);
2964 (*pp)++;
2965 return AMR_OK;
2966 }
2967
2968 static ApplyModifierResult
2969 ApplyModifier_Unique(const char **pp, ApplyModifiersState *st)
2970 {
2971 if ((*pp)[1] == st->endc || (*pp)[1] == ':') {
2972 st->newVal = VarUniq(st->val);
2973 (*pp)++;
2974 return AMR_OK;
2975 } else
2976 return AMR_UNKNOWN;
2977 }
2978
2979 #ifdef SYSVVARSUB
2980 /* :from=to */
2981 static ApplyModifierResult
2982 ApplyModifier_SysV(const char **pp, ApplyModifiersState *st)
2983 {
2984 char *lhs, *rhs;
2985 VarParseResult res;
2986
2987 const char *mod = *pp;
2988 Boolean eqFound = FALSE;
2989
2990 /*
2991 * First we make a pass through the string trying
2992 * to verify it is a SYSV-make-style translation:
2993 * it must be: <string1>=<string2>)
2994 */
2995 int nest = 1;
2996 const char *next = mod;
2997 while (*next != '\0' && nest > 0) {
2998 if (*next == '=') {
2999 eqFound = TRUE;
3000 /* continue looking for st->endc */
3001 } else if (*next == st->endc)
3002 nest--;
3003 else if (*next == st->startc)
3004 nest++;
3005 if (nest > 0)
3006 next++;
3007 }
3008 if (*next != st->endc || !eqFound)
3009 return AMR_UNKNOWN;
3010
3011 *pp = mod;
3012 res = ParseModifierPart(pp, '=', st->eflags, st,
3013 &lhs, NULL, NULL, NULL);
3014 if (res != VPR_OK)
3015 return AMR_CLEANUP;
3016
3017 res = ParseModifierPart(pp, st->endc, st->eflags, st,
3018 &rhs, NULL, NULL, NULL);
3019 if (res != VPR_OK)
3020 return AMR_CLEANUP;
3021
3022 /*
3023 * SYSV modifications happen through the whole
3024 * string. Note the pattern is anchored at the end.
3025 */
3026 (*pp)--;
3027 if (lhs[0] == '\0' && st->val[0] == '\0') {
3028 st->newVal = st->val; /* special case */
3029 } else {
3030 struct ModifyWord_SYSVSubstArgs args = {st->ctxt, lhs, rhs};
3031 st->newVal = ModifyWords(st->val, ModifyWord_SYSVSubst, &args,
3032 st->oneBigWord, st->sep);
3033 }
3034 free(lhs);
3035 free(rhs);
3036 return AMR_OK;
3037 }
3038 #endif
3039
3040 #ifdef SUNSHCMD
3041 /* :sh */
3042 static ApplyModifierResult
3043 ApplyModifier_SunShell(const char **pp, ApplyModifiersState *st)
3044 {
3045 const char *p = *pp;
3046 if (p[1] == 'h' && (p[2] == st->endc || p[2] == ':')) {
3047 if (st->eflags & VARE_WANTRES) {
3048 const char *errfmt;
3049 st->newVal = Cmd_Exec(st->val, &errfmt);
3050 if (errfmt)
3051 Error(errfmt, st->val);
3052 } else
3053 st->newVal = emptyString;
3054 *pp = p + 2;
3055 return AMR_OK;
3056 } else
3057 return AMR_UNKNOWN;
3058 }
3059 #endif
3060
3061 static void
3062 LogBeforeApply(const ApplyModifiersState *st, const char *mod, const char endc)
3063 {
3064 char eflags_str[VarEvalFlags_ToStringSize];
3065 char vflags_str[VarFlags_ToStringSize];
3066 char exprflags_str[VarExprFlags_ToStringSize];
3067 Boolean is_single_char = mod[0] != '\0' &&
3068 (mod[1] == endc || mod[1] == ':');
3069
3070 /* At this point, only the first character of the modifier can
3071 * be used since the end of the modifier is not yet known. */
3072 debug_printf("Applying ${%s:%c%s} to \"%s\" (%s, %s, %s)\n",
3073 st->v->name, mod[0], is_single_char ? "" : "...", st->val,
3074 Enum_FlagsToString(eflags_str, sizeof eflags_str,
3075 st->eflags, VarEvalFlags_ToStringSpecs),
3076 Enum_FlagsToString(vflags_str, sizeof vflags_str,
3077 st->v->flags, VarFlags_ToStringSpecs),
3078 Enum_FlagsToString(exprflags_str, sizeof exprflags_str,
3079 st->exprFlags,
3080 VarExprFlags_ToStringSpecs));
3081 }
3082
3083 static void
3084 LogAfterApply(ApplyModifiersState *st, const char *p, const char *mod)
3085 {
3086 char eflags_str[VarEvalFlags_ToStringSize];
3087 char vflags_str[VarFlags_ToStringSize];
3088 char exprflags_str[VarExprFlags_ToStringSize];
3089 const char *quot = st->newVal == var_Error ? "" : "\"";
3090 const char *newVal = st->newVal == var_Error ? "error" : st->newVal;
3091
3092 debug_printf("Result of ${%s:%.*s} is %s%s%s (%s, %s, %s)\n",
3093 st->v->name, (int)(p - mod), mod, quot, newVal, quot,
3094 Enum_FlagsToString(eflags_str, sizeof eflags_str,
3095 st->eflags, VarEvalFlags_ToStringSpecs),
3096 Enum_FlagsToString(vflags_str, sizeof vflags_str,
3097 st->v->flags, VarFlags_ToStringSpecs),
3098 Enum_FlagsToString(exprflags_str, sizeof exprflags_str,
3099 st->exprFlags,
3100 VarExprFlags_ToStringSpecs));
3101 }
3102
3103 static ApplyModifierResult
3104 ApplyModifier(const char **pp, ApplyModifiersState *st)
3105 {
3106 switch (**pp) {
3107 case ':':
3108 return ApplyModifier_Assign(pp, st);
3109 case '@':
3110 return ApplyModifier_Loop(pp, st);
3111 case '_':
3112 return ApplyModifier_Remember(pp, st);
3113 case 'D':
3114 case 'U':
3115 return ApplyModifier_Defined(pp, st);
3116 case 'L':
3117 return ApplyModifier_Literal(pp, st);
3118 case 'P':
3119 return ApplyModifier_Path(pp, st);
3120 case '!':
3121 return ApplyModifier_ShellCommand(pp, st);
3122 case '[':
3123 return ApplyModifier_Words(pp, st);
3124 case 'g':
3125 return ApplyModifier_Gmtime(pp, st);
3126 case 'h':
3127 return ApplyModifier_Hash(pp, st);
3128 case 'l':
3129 return ApplyModifier_Localtime(pp, st);
3130 case 't':
3131 return ApplyModifier_To(pp, st);
3132 case 'N':
3133 case 'M':
3134 return ApplyModifier_Match(pp, st);
3135 case 'S':
3136 return ApplyModifier_Subst(pp, st);
3137 case '?':
3138 return ApplyModifier_IfElse(pp, st);
3139 #ifndef NO_REGEX
3140 case 'C':
3141 return ApplyModifier_Regex(pp, st);
3142 #endif
3143 case 'q':
3144 case 'Q':
3145 return ApplyModifier_Quote(pp, st);
3146 case 'T':
3147 return ApplyModifier_WordFunc(pp, st, ModifyWord_Tail);
3148 case 'H':
3149 return ApplyModifier_WordFunc(pp, st, ModifyWord_Head);
3150 case 'E':
3151 return ApplyModifier_WordFunc(pp, st, ModifyWord_Suffix);
3152 case 'R':
3153 return ApplyModifier_WordFunc(pp, st, ModifyWord_Root);
3154 case 'r':
3155 return ApplyModifier_Range(pp, st);
3156 case 'O':
3157 return ApplyModifier_Order(pp, st);
3158 case 'u':
3159 return ApplyModifier_Unique(pp, st);
3160 #ifdef SUNSHCMD
3161 case 's':
3162 return ApplyModifier_SunShell(pp, st);
3163 #endif
3164 default:
3165 return AMR_UNKNOWN;
3166 }
3167 }
3168
3169 /* Apply any modifiers (such as :Mpattern or :@var@loop@ or :Q or ::=value). */
3170 static char *
3171 ApplyModifiers(
3172 const char **pp, /* the parsing position, updated upon return */
3173 char *val, /* the current value of the variable */
3174 char const startc, /* '(' or '{', or '\0' for indirect modifiers */
3175 char const endc, /* ')' or '}', or '\0' for indirect modifiers */
3176 Var * const v,
3177 VarExprFlags *exprFlags,
3178 GNode * const ctxt, /* for looking up and modifying variables */
3179 VarEvalFlags const eflags,
3180 void ** const freePtr /* free this after using the return value */
3181 ) {
3182 ApplyModifiersState st = {
3183 startc, endc, v, ctxt, eflags, val,
3184 var_Error, /* .newVal */
3185 ' ', /* .sep */
3186 FALSE, /* .oneBigWord */
3187 *exprFlags /* .exprFlags */
3188 };
3189 const char *p;
3190 const char *mod;
3191 ApplyModifierResult res;
3192
3193 assert(startc == '(' || startc == '{' || startc == '\0');
3194 assert(endc == ')' || endc == '}' || endc == '\0');
3195 assert(val != NULL);
3196
3197 p = *pp;
3198 while (*p != '\0' && *p != endc) {
3199
3200 if (*p == '$') {
3201 /*
3202 * We may have some complex modifiers in a variable.
3203 */
3204 const char *nested_p = p;
3205 void *freeIt;
3206 const char *rval;
3207 char c;
3208
3209 (void)Var_Parse(&nested_p, st.ctxt, st.eflags, &rval, &freeIt);
3210 /* TODO: handle errors */
3211
3212 /*
3213 * If we have not parsed up to st.endc or ':', we are not
3214 * interested. This means the expression ${VAR:${M_1}${M_2}}
3215 * is not accepted, but ${VAR:${M_1}:${M_2}} is.
3216 */
3217 if (rval[0] != '\0' &&
3218 (c = *nested_p) != '\0' && c != ':' && c != st.endc) {
3219 if (DEBUG(LINT))
3220 Parse_Error(PARSE_FATAL,
3221 "Missing delimiter ':' after indirect modifier \"%.*s\"",
3222 (int)(nested_p - p), p);
3223
3224 free(freeIt);
3225 /* XXX: apply_mods doesn't sound like "not interested". */
3226 /* XXX: Why is the indirect modifier parsed again by
3227 * apply_mods? If any, p should be advanced to nested_p. */
3228 goto apply_mods;
3229 }
3230
3231 VAR_DEBUG3("Indirect modifier \"%s\" from \"%.*s\"\n",
3232 rval, (int)(size_t)(nested_p - p), p);
3233
3234 p = nested_p;
3235
3236 if (rval[0] != '\0') {
3237 const char *rval_pp = rval;
3238 st.val = ApplyModifiers(&rval_pp, st.val, '\0', '\0', v,
3239 &st.exprFlags, ctxt, eflags, freePtr);
3240 if (st.val == var_Error
3241 || (st.val == varUndefined && !(st.eflags & VARE_UNDEFERR))
3242 || *rval_pp != '\0') {
3243 free(freeIt);
3244 goto out; /* error already reported */
3245 }
3246 }
3247 free(freeIt);
3248
3249 if (*p == ':')
3250 p++;
3251 else if (*p == '\0' && endc != '\0') {
3252 Error("Unclosed variable specification after complex "
3253 "modifier (expecting '%c') for %s", st.endc, st.v->name);
3254 goto out;
3255 }
3256 continue;
3257 }
3258 apply_mods:
3259 st.newVal = var_Error; /* default value, in case of errors */
3260 mod = p;
3261
3262 if (DEBUG(VAR))
3263 LogBeforeApply(&st, mod, endc);
3264
3265 res = ApplyModifier(&p, &st);
3266
3267 #ifdef SYSVVARSUB
3268 if (res == AMR_UNKNOWN) {
3269 assert(p == mod);
3270 res = ApplyModifier_SysV(&p, &st);
3271 }
3272 #endif
3273
3274 if (res == AMR_UNKNOWN) {
3275 Error("Unknown modifier '%c'", *mod);
3276 for (p++; *p != ':' && *p != st.endc && *p != '\0'; p++)
3277 continue;
3278 st.newVal = var_Error;
3279 }
3280 if (res == AMR_CLEANUP)
3281 goto cleanup;
3282 if (res == AMR_BAD)
3283 goto bad_modifier;
3284
3285 if (DEBUG(VAR))
3286 LogAfterApply(&st, p, mod);
3287
3288 if (st.newVal != st.val) {
3289 if (*freePtr) {
3290 free(st.val);
3291 *freePtr = NULL;
3292 }
3293 st.val = st.newVal;
3294 if (st.val != var_Error && st.val != varUndefined &&
3295 st.val != emptyString) {
3296 *freePtr = st.val;
3297 }
3298 }
3299 if (*p == '\0' && st.endc != '\0') {
3300 Error("Unclosed variable specification (expecting '%c') "
3301 "for \"%s\" (value \"%s\") modifier %c",
3302 st.endc, st.v->name, st.val, *mod);
3303 } else if (*p == ':') {
3304 p++;
3305 } else if (DEBUG(LINT) && *p != '\0' && *p != endc) {
3306 Parse_Error(PARSE_FATAL,
3307 "Missing delimiter ':' after modifier \"%.*s\"",
3308 (int)(p - mod), mod);
3309 }
3310 }
3311 out:
3312 *pp = p;
3313 assert(st.val != NULL); /* Use var_Error or varUndefined instead. */
3314 *exprFlags = st.exprFlags;
3315 return st.val;
3316
3317 bad_modifier:
3318 Error("Bad modifier `:%.*s' for %s",
3319 (int)strcspn(mod, ":)}"), mod, st.v->name);
3320
3321 cleanup:
3322 *pp = p;
3323 free(*freePtr);
3324 *freePtr = NULL;
3325 *exprFlags = st.exprFlags;
3326 return var_Error;
3327 }
3328
3329 static Boolean
3330 VarIsDynamic(GNode *ctxt, const char *varname, size_t namelen)
3331 {
3332 if ((namelen == 1 ||
3333 (namelen == 2 && (varname[1] == 'F' || varname[1] == 'D'))) &&
3334 (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL))
3335 {
3336 /*
3337 * If substituting a local variable in a non-local context,
3338 * assume it's for dynamic source stuff. We have to handle
3339 * this specially and return the longhand for the variable
3340 * with the dollar sign escaped so it makes it back to the
3341 * caller. Only four of the local variables are treated
3342 * specially as they are the only four that will be set
3343 * when dynamic sources are expanded.
3344 */
3345 switch (varname[0]) {
3346 case '@':
3347 case '%':
3348 case '*':
3349 case '!':
3350 return TRUE;
3351 }
3352 return FALSE;
3353 }
3354
3355 if ((namelen == 7 || namelen == 8) && varname[0] == '.' &&
3356 ch_isupper(varname[1]) && (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL))
3357 {
3358 return strcmp(varname, ".TARGET") == 0 ||
3359 strcmp(varname, ".ARCHIVE") == 0 ||
3360 strcmp(varname, ".PREFIX") == 0 ||
3361 strcmp(varname, ".MEMBER") == 0;
3362 }
3363
3364 return FALSE;
3365 }
3366
3367 static const char *
3368 UndefinedShortVarValue(char varname, const GNode *ctxt, VarEvalFlags eflags)
3369 {
3370 if (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL) {
3371 /*
3372 * If substituting a local variable in a non-local context,
3373 * assume it's for dynamic source stuff. We have to handle
3374 * this specially and return the longhand for the variable
3375 * with the dollar sign escaped so it makes it back to the
3376 * caller. Only four of the local variables are treated
3377 * specially as they are the only four that will be set
3378 * when dynamic sources are expanded.
3379 */
3380 switch (varname) {
3381 case '@':
3382 return "$(.TARGET)";
3383 case '%':
3384 return "$(.MEMBER)";
3385 case '*':
3386 return "$(.PREFIX)";
3387 case '!':
3388 return "$(.ARCHIVE)";
3389 }
3390 }
3391 return eflags & VARE_UNDEFERR ? var_Error : varUndefined;
3392 }
3393
3394 /* Parse a variable name, until the end character or a colon, whichever
3395 * comes first. */
3396 static char *
3397 ParseVarname(const char **pp, char startc, char endc,
3398 GNode *ctxt, VarEvalFlags eflags,
3399 size_t *out_varname_len)
3400 {
3401 Buffer buf;
3402 const char *p = *pp;
3403 int depth = 1;
3404
3405 Buf_Init(&buf, 0);
3406
3407 while (*p != '\0') {
3408 /* Track depth so we can spot parse errors. */
3409 if (*p == startc)
3410 depth++;
3411 if (*p == endc) {
3412 if (--depth == 0)
3413 break;
3414 }
3415 if (*p == ':' && depth == 1)
3416 break;
3417
3418 /* A variable inside a variable, expand. */
3419 if (*p == '$') {
3420 void *freeIt;
3421 const char *rval;
3422 (void)Var_Parse(&p, ctxt, eflags, &rval, &freeIt);
3423 /* TODO: handle errors */
3424 Buf_AddStr(&buf, rval);
3425 free(freeIt);
3426 } else {
3427 Buf_AddByte(&buf, *p);
3428 p++;
3429 }
3430 }
3431 *pp = p;
3432 *out_varname_len = Buf_Len(&buf);
3433 return Buf_Destroy(&buf, FALSE);
3434 }
3435
3436 static Boolean
3437 ValidShortVarname(char varname, const char *start)
3438 {
3439 switch (varname) {
3440 case '\0':
3441 case ')':
3442 case '}':
3443 case ':':
3444 case '$':
3445 break; /* and continue below */
3446 default:
3447 return TRUE;
3448 }
3449
3450 if (!DEBUG(LINT))
3451 return FALSE;
3452
3453 if (varname == '$')
3454 Parse_Error(PARSE_FATAL,
3455 "To escape a dollar, use \\$, not $$, at \"%s\"", start);
3456 else if (varname == '\0')
3457 Parse_Error(PARSE_FATAL, "Dollar followed by nothing");
3458 else
3459 Parse_Error(PARSE_FATAL,
3460 "Invalid variable name '%c', at \"%s\"", varname, start);
3461
3462 return FALSE;
3463 }
3464
3465 /* Parse a single-character variable name such as $V or $@.
3466 * Return whether to continue parsing. */
3467 static Boolean
3468 ParseVarnameShort(char const startc, const char **const pp, GNode *const ctxt,
3469 VarEvalFlags const eflags,
3470 const char **const out_FALSE_val,
3471 VarParseResult *const out_FALSE_res,
3472 Var **out_TRUE_var)
3473 {
3474 char name[2];
3475 Var *v;
3476
3477 /*
3478 * If it's not bounded by braces of some sort, life is much simpler.
3479 * We just need to check for the first character and return the
3480 * value if it exists.
3481 */
3482
3483 if (!ValidShortVarname(startc, *pp)) {
3484 (*pp)++;
3485 *out_FALSE_val = var_Error;
3486 *out_FALSE_res = VPR_PARSE_MSG;
3487 return FALSE;
3488 }
3489
3490 name[0] = startc;
3491 name[1] = '\0';
3492 v = VarFind(name, ctxt, TRUE);
3493 if (v == NULL) {
3494 *pp += 2;
3495
3496 *out_FALSE_val = UndefinedShortVarValue(startc, ctxt, eflags);
3497 if (DEBUG(LINT) && *out_FALSE_val == var_Error) {
3498 Parse_Error(PARSE_FATAL, "Variable \"%s\" is undefined", name);
3499 *out_FALSE_res = VPR_UNDEF_MSG;
3500 return FALSE;
3501 }
3502 *out_FALSE_res = eflags & VARE_UNDEFERR ? VPR_UNDEF_SILENT : VPR_OK;
3503 return FALSE;
3504 }
3505
3506 *out_TRUE_var = v;
3507 return TRUE;
3508 }
3509
3510 /* Parse a long variable name enclosed in braces or parentheses such as $(VAR)
3511 * or ${VAR}, up to the closing brace or parenthesis, or in the case of
3512 * ${VAR:Modifiers}, up to the ':' that starts the modifiers.
3513 * Return whether to continue parsing. */
3514 static Boolean
3515 ParseVarnameLong(
3516 const char **const pp,
3517 char const startc,
3518 GNode *const ctxt,
3519 VarEvalFlags const eflags,
3520
3521 VarParseResult *const out_FALSE_res,
3522 const char **const out_FALSE_val,
3523 void **const out_FALSE_freePtr,
3524
3525 char *const out_TRUE_endc,
3526 const char **const out_TRUE_p,
3527 Var **const out_TRUE_v,
3528 Boolean *const out_TRUE_haveModifier,
3529 const char **const out_TRUE_extraModifiers,
3530 Boolean *const out_TRUE_dynamic,
3531 VarExprFlags *const out_TRUE_exprFlags
3532 ) {
3533 size_t namelen;
3534 char *varname;
3535 Var *v;
3536 Boolean haveModifier;
3537 Boolean dynamic = FALSE;
3538
3539 const char *const start = *pp;
3540 char endc = startc == '(' ? ')' : '}';
3541
3542 const char *p = start + 2;
3543 varname = ParseVarname(&p, startc, endc, ctxt, eflags, &namelen);
3544
3545 if (*p == ':') {
3546 haveModifier = TRUE;
3547 } else if (*p == endc) {
3548 haveModifier = FALSE;
3549 } else {
3550 Parse_Error(PARSE_FATAL, "Unclosed variable \"%s\"", varname);
3551 *pp = p;
3552 free(varname);
3553 *out_FALSE_val = var_Error;
3554 *out_FALSE_res = VPR_PARSE_MSG;
3555 return FALSE;
3556 }
3557
3558 v = VarFind(varname, ctxt, TRUE);
3559
3560 /* At this point, p points just after the variable name,
3561 * either at ':' or at endc. */
3562
3563 /*
3564 * Check also for bogus D and F forms of local variables since we're
3565 * in a local context and the name is the right length.
3566 */
3567 if (v == NULL && ctxt != VAR_CMDLINE && ctxt != VAR_GLOBAL &&
3568 namelen == 2 && (varname[1] == 'F' || varname[1] == 'D') &&
3569 strchr("@%?*!<>", varname[0]) != NULL)
3570 {
3571 /*
3572 * Well, it's local -- go look for it.
3573 */
3574 char name[] = { varname[0], '\0' };
3575 v = VarFind(name, ctxt, 0);
3576
3577 if (v != NULL) {
3578 if (varname[1] == 'D') {
3579 *out_TRUE_extraModifiers = "H:";
3580 } else { /* F */
3581 *out_TRUE_extraModifiers = "T:";
3582 }
3583 }
3584 }
3585
3586 if (v == NULL) {
3587 dynamic = VarIsDynamic(ctxt, varname, namelen);
3588
3589 if (!haveModifier) {
3590 p++; /* skip endc */
3591 *pp = p;
3592 if (dynamic) {
3593 char *pstr = bmake_strsedup(start, p);
3594 free(varname);
3595 *out_FALSE_res = VPR_OK;
3596 *out_FALSE_freePtr = pstr;
3597 *out_FALSE_val = pstr;
3598 return FALSE;
3599 }
3600
3601 if ((eflags & VARE_UNDEFERR) && (eflags & VARE_WANTRES) &&
3602 DEBUG(LINT))
3603 {
3604 Parse_Error(PARSE_FATAL, "Variable \"%s\" is undefined",
3605 varname);
3606 free(varname);
3607 *out_FALSE_res = VPR_UNDEF_MSG;
3608 *out_FALSE_val = var_Error;
3609 return FALSE;
3610 }
3611
3612 if (eflags & VARE_UNDEFERR) {
3613 free(varname);
3614 *out_FALSE_res = VPR_UNDEF_SILENT;
3615 *out_FALSE_val = var_Error;
3616 return FALSE;
3617 }
3618
3619 free(varname);
3620 *out_FALSE_res = VPR_OK;
3621 *out_FALSE_val = varUndefined;
3622 return FALSE;
3623 }
3624
3625 /* The variable expression is based on an undefined variable.
3626 * Nevertheless it needs a Var, for modifiers that access the
3627 * variable name, such as :L or :?.
3628 *
3629 * Most modifiers leave this expression in the "undefined" state
3630 * (VEF_UNDEF), only a few modifiers like :D, :U, :L, :P turn this
3631 * undefined expression into a defined expression (VEF_DEF).
3632 *
3633 * At the end, after applying all modifiers, if the expression
3634 * is still undefined, Var_Parse will return an empty string
3635 * instead of the actually computed value. */
3636 v = VarNew(varname, varname, "", 0);
3637 *out_TRUE_exprFlags = VEF_UNDEF;
3638 } else
3639 free(varname);
3640
3641 *out_TRUE_endc = endc;
3642 *out_TRUE_p = p;
3643 *out_TRUE_v = v;
3644 *out_TRUE_haveModifier = haveModifier;
3645 *out_TRUE_dynamic = dynamic;
3646 return TRUE;
3647 }
3648
3649 /*-
3650 *-----------------------------------------------------------------------
3651 * Var_Parse --
3652 * Given the start of a variable expression (such as $v, $(VAR),
3653 * ${VAR:Mpattern}), extract the variable name, possibly some
3654 * modifiers and find its value by applying the modifiers to the
3655 * original value.
3656 *
3657 * When parsing a condition in ParseEmptyArg, pp may also point to
3658 * the "y" of "empty(VARNAME:Modifiers)", which is syntactically
3659 * identical.
3660 *
3661 * Input:
3662 * str The string to parse
3663 * ctxt The context for the variable
3664 * flags VARE_UNDEFERR if undefineds are an error
3665 * VARE_WANTRES if we actually want the result
3666 * VARE_ASSIGN if we are in a := assignment
3667 * lengthPtr OUT: The length of the specification
3668 * freePtr OUT: Non-NULL if caller should free *freePtr
3669 *
3670 * Results:
3671 * Returns the value of the variable expression, never NULL.
3672 * Returns var_Error if there was a parse error and VARE_UNDEFERR was
3673 * set.
3674 * Returns varUndefined if there was an undefined variable and
3675 * VARE_UNDEFERR was not set.
3676 *
3677 * Parsing should continue at *pp.
3678 * TODO: Document the value of *pp on parse errors. It might be advanced
3679 * by 0, or +1, or the index of the parse error, or the guessed end of the
3680 * variable expression.
3681 *
3682 * If var_Error is returned, a diagnostic may or may not have been
3683 * printed. XXX: This is inconsistent.
3684 *
3685 * If varUndefined is returned, a diagnostic may or may not have been
3686 * printed. XXX: This is inconsistent.
3687 *
3688 * After using the returned value, *freePtr must be freed, preferably
3689 * using bmake_free since it is NULL in most cases.
3690 *
3691 * Side Effects:
3692 * Any effects from the modifiers, such as :!cmd! or ::=value.
3693 *-----------------------------------------------------------------------
3694 */
3695 /* coverity[+alloc : arg-*4] */
3696 VarParseResult
3697 Var_Parse(const char **pp, GNode *ctxt, VarEvalFlags eflags,
3698 const char **out_val, void **freePtr)
3699 {
3700 const char *const start = *pp;
3701 const char *p;
3702 Boolean haveModifier; /* TRUE if have modifiers for the variable */
3703 char startc; /* Starting character if variable in parens
3704 * or braces */
3705 char endc; /* Ending character if variable in parens
3706 * or braces */
3707 Boolean dynamic; /* TRUE if the variable is local and we're
3708 * expanding it in a non-local context. This
3709 * is done to support dynamic sources. The
3710 * result is just the expression, unaltered */
3711 const char *extramodifiers;
3712 Var *v;
3713 char *nstr;
3714 char eflags_str[VarEvalFlags_ToStringSize];
3715 VarExprFlags exprFlags = 0;
3716
3717 VAR_DEBUG3("%s: %s with %s\n", __func__, start,
3718 Enum_FlagsToString(eflags_str, sizeof eflags_str, eflags,
3719 VarEvalFlags_ToStringSpecs));
3720
3721 *freePtr = NULL;
3722 extramodifiers = NULL; /* extra modifiers to apply first */
3723 dynamic = FALSE;
3724
3725 /* Appease GCC, which thinks that the variable might not be
3726 * initialized. */
3727 endc = '\0';
3728
3729 startc = start[1];
3730 if (startc != '(' && startc != '{') {
3731 VarParseResult res;
3732 if (!ParseVarnameShort(startc, pp, ctxt, eflags, out_val, &res, &v))
3733 return res;
3734 haveModifier = FALSE;
3735 p = start + 1;
3736 } else {
3737 VarParseResult res;
3738 if (!ParseVarnameLong(pp, startc, ctxt, eflags,
3739 &res, out_val, freePtr,
3740 &endc, &p, &v, &haveModifier, &extramodifiers,
3741 &dynamic, &exprFlags))
3742 return res;
3743 }
3744
3745 if (v->flags & VAR_IN_USE)
3746 Fatal("Variable %s is recursive.", v->name);
3747
3748 /*
3749 * Before doing any modification, we have to make sure the value
3750 * has been fully expanded. If it looks like recursion might be
3751 * necessary (there's a dollar sign somewhere in the variable's value)
3752 * we just call Var_Subst to do any other substitutions that are
3753 * necessary. Note that the value returned by Var_Subst will have
3754 * been dynamically-allocated, so it will need freeing when we
3755 * return.
3756 */
3757 nstr = Buf_GetAll(&v->val, NULL);
3758 if (strchr(nstr, '$') != NULL && (eflags & VARE_WANTRES)) {
3759 VarEvalFlags nested_eflags = eflags;
3760 if (DEBUG(LINT))
3761 nested_eflags &= ~(unsigned)VARE_UNDEFERR;
3762 v->flags |= VAR_IN_USE;
3763 (void)Var_Subst(nstr, ctxt, nested_eflags, &nstr);
3764 v->flags &= ~(unsigned)VAR_IN_USE;
3765 /* TODO: handle errors */
3766 *freePtr = nstr;
3767 }
3768
3769 if (haveModifier || extramodifiers != NULL) {
3770 void *extraFree;
3771
3772 extraFree = NULL;
3773 if (extramodifiers != NULL) {
3774 const char *em = extramodifiers;
3775 nstr = ApplyModifiers(&em, nstr, '(', ')',
3776 v, &exprFlags, ctxt, eflags, &extraFree);
3777 }
3778
3779 if (haveModifier) {
3780 /* Skip initial colon. */
3781 p++;
3782
3783 nstr = ApplyModifiers(&p, nstr, startc, endc,
3784 v, &exprFlags, ctxt, eflags, freePtr);
3785 free(extraFree);
3786 } else {
3787 *freePtr = extraFree;
3788 }
3789 }
3790
3791 if (*p != '\0') /* Skip past endc if possible. */
3792 p++;
3793
3794 *pp = p;
3795
3796 if (v->flags & VAR_FROM_ENV) {
3797 /* Free the environment variable now since we own it,
3798 * but don't free the variable value if it will be returned. */
3799 Boolean keepValue = nstr == Buf_GetAll(&v->val, NULL);
3800 if (keepValue)
3801 *freePtr = nstr;
3802 (void)VarFreeEnv(v, !keepValue);
3803
3804 } else if (exprFlags & VEF_UNDEF) {
3805 if (!(exprFlags & VEF_DEF)) {
3806 if (*freePtr != NULL) {
3807 free(*freePtr);
3808 *freePtr = NULL;
3809 }
3810 if (dynamic) {
3811 nstr = bmake_strsedup(start, p);
3812 *freePtr = nstr;
3813 } else {
3814 /* The expression is still undefined, therefore discard the
3815 * actual value and return an error marker instead. */
3816 nstr = (eflags & VARE_UNDEFERR) ? var_Error : varUndefined;
3817 }
3818 }
3819 if (nstr != Buf_GetAll(&v->val, NULL))
3820 Buf_Destroy(&v->val, TRUE);
3821 free(v->name_freeIt);
3822 free(v);
3823 }
3824 *out_val = nstr;
3825 return VPR_UNKNOWN;
3826 }
3827
3828 /* Substitute for all variables in the given string in the given context.
3829 *
3830 * If eflags & VARE_UNDEFERR, Parse_Error will be called when an undefined
3831 * variable is encountered.
3832 *
3833 * If eflags & VARE_WANTRES, any effects from the modifiers, such as ::=,
3834 * :sh or !cmd! take place.
3835 *
3836 * Input:
3837 * str the string which to substitute
3838 * ctxt the context wherein to find variables
3839 * eflags VARE_UNDEFERR if undefineds are an error
3840 * VARE_WANTRES if we actually want the result
3841 * VARE_ASSIGN if we are in a := assignment
3842 *
3843 * Results:
3844 * The resulting string.
3845 */
3846 VarParseResult
3847 Var_Subst(const char *str, GNode *ctxt, VarEvalFlags eflags, char **out_res)
3848 {
3849 const char *p = str;
3850 Buffer buf; /* Buffer for forming things */
3851
3852 /* Set true if an error has already been reported,
3853 * to prevent a plethora of messages when recursing */
3854 static Boolean errorReported;
3855
3856 Buf_Init(&buf, 0);
3857 errorReported = FALSE;
3858
3859 while (*p != '\0') {
3860 if (p[0] == '$' && p[1] == '$') {
3861 /* A dollar sign may be escaped with another dollar sign. */
3862 if (save_dollars && (eflags & VARE_ASSIGN))
3863 Buf_AddByte(&buf, '$');
3864 Buf_AddByte(&buf, '$');
3865 p += 2;
3866 } else if (*p != '$') {
3867 /*
3868 * Skip as many characters as possible -- either to the end of
3869 * the string or to the next dollar sign (variable expression).
3870 */
3871 const char *plainStart = p;
3872
3873 for (p++; *p != '$' && *p != '\0'; p++)
3874 continue;
3875 Buf_AddBytesBetween(&buf, plainStart, p);
3876 } else {
3877 const char *nested_p = p;
3878 void *freeIt;
3879 const char *val;
3880 (void)Var_Parse(&nested_p, ctxt, eflags, &val, &freeIt);
3881 /* TODO: handle errors */
3882
3883 if (val == var_Error || val == varUndefined) {
3884 /*
3885 * If performing old-time variable substitution, skip over
3886 * the variable and continue with the substitution. Otherwise,
3887 * store the dollar sign and advance str so we continue with
3888 * the string...
3889 */
3890 if (oldVars) {
3891 p = nested_p;
3892 } else if ((eflags & VARE_UNDEFERR) || val == var_Error) {
3893 /*
3894 * If variable is undefined, complain and skip the
3895 * variable. The complaint will stop us from doing anything
3896 * when the file is parsed.
3897 */
3898 if (!errorReported) {
3899 Parse_Error(PARSE_FATAL, "Undefined variable \"%.*s\"",
3900 (int)(size_t)(nested_p - p), p);
3901 }
3902 p = nested_p;
3903 errorReported = TRUE;
3904 } else {
3905 /* Copy the initial '$' of the undefined expression,
3906 * thereby deferring expansion of the expression, but
3907 * expand nested expressions if already possible.
3908 * See unit-tests/varparse-undef-partial.mk. */
3909 Buf_AddByte(&buf, *p);
3910 p++;
3911 }
3912 } else {
3913 p = nested_p;
3914 Buf_AddStr(&buf, val);
3915 }
3916 free(freeIt);
3917 freeIt = NULL;
3918 }
3919 }
3920
3921 *out_res = Buf_DestroyCompact(&buf);
3922 return VPR_OK;
3923 }
3924
3925 /* Initialize the variables module. */
3926 void
3927 Var_Init(void)
3928 {
3929 VAR_INTERNAL = Targ_NewGN("Internal");
3930 VAR_GLOBAL = Targ_NewGN("Global");
3931 VAR_CMDLINE = Targ_NewGN("Command");
3932 }
3933
3934 /* Clean up the variables module. */
3935 void
3936 Var_End(void)
3937 {
3938 Var_Stats();
3939 }
3940
3941 void
3942 Var_Stats(void)
3943 {
3944 HashTable_DebugStats(&VAR_GLOBAL->context, "VAR_GLOBAL");
3945 }
3946
3947 /* Print all variables in a context, sorted by name. */
3948 void
3949 Var_Dump(GNode *ctxt)
3950 {
3951 Vector /* of const char * */ vec;
3952 HashIter hi;
3953 size_t i;
3954 const char **varnames;
3955
3956 Vector_Init(&vec, sizeof(const char *));
3957
3958 HashIter_Init(&hi, &ctxt->context);
3959 while (HashIter_Next(&hi) != NULL)
3960 *(const char **)Vector_Push(&vec) = hi.entry->key;
3961 varnames = vec.items;
3962
3963 qsort(varnames, vec.len, sizeof varnames[0], str_cmp_asc);
3964
3965 for (i = 0; i < vec.len; i++) {
3966 const char *varname = varnames[i];
3967 Var *var = HashTable_FindValue(&ctxt->context, varname);
3968 debug_printf("%-16s = %s\n", varname, Buf_GetAll(&var->val, NULL));
3969 }
3970
3971 Vector_Done(&vec);
3972 }
3973