var.c revision 1.681 1 /* $NetBSD: var.c,v 1.681 2020/11/08 19:24:19 rillig Exp $ */
2
3 /*
4 * Copyright (c) 1988, 1989, 1990, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * Adam de Boor.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 */
34
35 /*
36 * Copyright (c) 1989 by Berkeley Softworks
37 * All rights reserved.
38 *
39 * This code is derived from software contributed to Berkeley by
40 * Adam de Boor.
41 *
42 * Redistribution and use in source and binary forms, with or without
43 * modification, are permitted provided that the following conditions
44 * are met:
45 * 1. Redistributions of source code must retain the above copyright
46 * notice, this list of conditions and the following disclaimer.
47 * 2. Redistributions in binary form must reproduce the above copyright
48 * notice, this list of conditions and the following disclaimer in the
49 * documentation and/or other materials provided with the distribution.
50 * 3. All advertising materials mentioning features or use of this software
51 * must display the following acknowledgement:
52 * This product includes software developed by the University of
53 * California, Berkeley and its contributors.
54 * 4. Neither the name of the University nor the names of its contributors
55 * may be used to endorse or promote products derived from this software
56 * without specific prior written permission.
57 *
58 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
59 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
60 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
61 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
62 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
63 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
64 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
65 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
66 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
67 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
68 * SUCH DAMAGE.
69 */
70
71 /*
72 * Handling of variables and the expressions formed from them.
73 *
74 * Variables are set using lines of the form VAR=value. Both the variable
75 * name and the value can contain references to other variables, by using
76 * expressions like ${VAR}, ${VAR:Modifiers}, ${${VARNAME}} or ${VAR:${MODS}}.
77 *
78 * Interface:
79 * Var_Init Initialize this module.
80 *
81 * Var_End Clean up the module.
82 *
83 * Var_Set Set the value of the variable, creating it if
84 * necessary.
85 *
86 * Var_Append Append more characters to the variable, creating it if
87 * necessary. A space is placed between the old value and
88 * the new one.
89 *
90 * Var_Exists See if a variable exists.
91 *
92 * Var_Value Return the unexpanded value of a variable, or NULL if
93 * the variable is undefined.
94 *
95 * Var_Subst Substitute all variable expressions in a string.
96 *
97 * Var_Parse Parse a variable expression such as ${VAR:Mpattern}.
98 *
99 * Var_Delete Delete a variable.
100 *
101 * Var_ExportVars Export some or even all variables to the environment
102 * of this process and its child processes.
103 *
104 * Var_Export Export the variable to the environment of this process
105 * and its child processes.
106 *
107 * Var_UnExport Don't export the variable anymore.
108 *
109 * Debugging:
110 * Var_Stats Print out hashing statistics if in -dh mode.
111 *
112 * Var_Dump Print out all variables defined in the given context.
113 *
114 * XXX: There's a lot of duplication in these functions.
115 */
116
117 #include <sys/stat.h>
118 #ifndef NO_REGEX
119 #include <sys/types.h>
120 #include <regex.h>
121 #endif
122 #include <errno.h>
123 #include <inttypes.h>
124 #include <limits.h>
125 #include <time.h>
126
127 #include "make.h"
128 #include "dir.h"
129 #include "job.h"
130 #include "metachar.h"
131
132 /* "@(#)var.c 8.3 (Berkeley) 3/19/94" */
133 MAKE_RCSID("$NetBSD: var.c,v 1.681 2020/11/08 19:24:19 rillig Exp $");
134
135 #define VAR_DEBUG1(fmt, arg1) DEBUG1(VAR, fmt, arg1)
136 #define VAR_DEBUG2(fmt, arg1, arg2) DEBUG2(VAR, fmt, arg1, arg2)
137 #define VAR_DEBUG3(fmt, arg1, arg2, arg3) DEBUG3(VAR, fmt, arg1, arg2, arg3)
138 #define VAR_DEBUG4(fmt, arg1, arg2, arg3, arg4) DEBUG4(VAR, fmt, arg1, arg2, arg3, arg4)
139
140 ENUM_FLAGS_RTTI_3(VarEvalFlags,
141 VARE_UNDEFERR, VARE_WANTRES, VARE_KEEP_DOLLAR);
142
143 /*
144 * This lets us tell if we have replaced the original environ
145 * (which we cannot free).
146 */
147 char **savedEnv = NULL;
148
149 /* Special return value for Var_Parse, indicating a parse error. It may be
150 * caused by an undefined variable, a syntax error in a modifier or
151 * something entirely different. */
152 char var_Error[] = "";
153
154 /* Special return value for Var_Parse, indicating an undefined variable in
155 * a case where VARE_UNDEFERR is not set. This undefined variable is
156 * typically a dynamic variable such as ${.TARGET}, whose expansion needs to
157 * be deferred until it is defined in an actual target. */
158 static char varUndefined[] = "";
159
160 /*
161 * Traditionally this make consumed $$ during := like any other expansion.
162 * Other make's do not, and this make follows straight since 2016-01-09.
163 *
164 * This knob allows controlling the behavior.
165 * FALSE to consume $$ during := assignment.
166 * TRUE to preserve $$ during := assignment.
167 */
168 #define MAKE_SAVE_DOLLARS ".MAKE.SAVE_DOLLARS"
169 static Boolean save_dollars = TRUE;
170
171 /*
172 * Internally, variables are contained in four different contexts.
173 * 1) the environment. They cannot be changed. If an environment
174 * variable is appended to, the result is placed in the global
175 * context.
176 * 2) the global context. Variables set in the makefiles are located
177 * here.
178 * 3) the command-line context. All variables set on the command line
179 * are placed in this context.
180 * 4) the local context. Each target has associated with it a context
181 * list. On this list are located the structures describing such
182 * local variables as $(@) and $(*)
183 * The four contexts are searched in the reverse order from which they are
184 * listed (but see opts.checkEnvFirst).
185 */
186 GNode *VAR_INTERNAL; /* variables from make itself */
187 GNode *VAR_GLOBAL; /* variables from the makefile */
188 GNode *VAR_CMDLINE; /* variables defined on the command-line */
189
190 typedef enum VarFlags {
191
192 /* The variable's value is currently being used by Var_Parse or Var_Subst.
193 * This marker is used to avoid endless recursion. */
194 VAR_IN_USE = 0x01,
195
196 /* The variable comes from the environment.
197 * These variables are not registered in any GNode, therefore they must
198 * be freed as soon as they are not used anymore. */
199 VAR_FROM_ENV = 0x02,
200
201 /* The variable is exported to the environment, to be used by child
202 * processes. */
203 VAR_EXPORTED = 0x10,
204
205 /* At the point where this variable was exported, it contained an
206 * unresolved reference to another variable. Before any child process is
207 * started, it needs to be exported again, in the hope that the referenced
208 * variable can then be resolved. */
209 VAR_REEXPORT = 0x20,
210
211 /* The variable came from the command line. */
212 VAR_FROM_CMD = 0x40,
213
214 /* The variable value cannot be changed anymore, and the variable cannot
215 * be deleted. Any attempts to do so are ignored. */
216 VAR_READONLY = 0x80
217 } VarFlags;
218
219 ENUM_FLAGS_RTTI_6(VarFlags,
220 VAR_IN_USE, VAR_FROM_ENV,
221 VAR_EXPORTED, VAR_REEXPORT, VAR_FROM_CMD, VAR_READONLY);
222
223 /* Variables are defined using one of the VAR=value assignments. Their
224 * value can be queried by expressions such as $V, ${VAR}, or with modifiers
225 * such as ${VAR:S,from,to,g:Q}.
226 *
227 * There are 3 kinds of variables: context variables, environment variables,
228 * undefined variables.
229 *
230 * Context variables are stored in a GNode.context. The only way to undefine
231 * a context variable is using the .undef directive. In particular, it must
232 * not be possible to undefine a variable during the evaluation of an
233 * expression, or Var.name might point nowhere.
234 *
235 * Environment variables are temporary. They are returned by VarFind, and
236 * after using them, they must be freed using VarFreeEnv.
237 *
238 * Undefined variables occur during evaluation of variable expressions such
239 * as ${UNDEF:Ufallback} in Var_Parse and ApplyModifiers.
240 */
241 typedef struct Var {
242 /* The name of the variable, once set, doesn't change anymore.
243 * For context variables, it aliases the corresponding HashEntry name.
244 * For environment and undefined variables, it is allocated. */
245 const char *name;
246 void *name_freeIt;
247
248 Buffer val; /* its value */
249 VarFlags flags; /* miscellaneous status flags */
250 } Var;
251
252 /*
253 * Exporting vars is expensive so skip it if we can
254 */
255 typedef enum VarExportedMode {
256 VAR_EXPORTED_NONE,
257 VAR_EXPORTED_SOME,
258 VAR_EXPORTED_ALL
259 } VarExportedMode;
260
261 static VarExportedMode var_exportedVars = VAR_EXPORTED_NONE;
262
263 typedef enum VarExportFlags {
264 VAR_EXPORT_NORMAL = 0,
265 /*
266 * We pass this to Var_Export when doing the initial export
267 * or after updating an exported var.
268 */
269 VAR_EXPORT_PARENT = 0x01,
270 /*
271 * We pass this to Var_Export1 to tell it to leave the value alone.
272 */
273 VAR_EXPORT_LITERAL = 0x02
274 } VarExportFlags;
275
276 /* Flags for pattern matching in the :S and :C modifiers */
277 typedef enum VarPatternFlags {
278 VARP_SUB_GLOBAL = 0x01, /* Replace as often as possible ('g') */
279 VARP_SUB_ONE = 0x02, /* Replace only once ('1') */
280 VARP_ANCHOR_START = 0x04, /* Match at start of word ('^') */
281 VARP_ANCHOR_END = 0x08 /* Match at end of word ('$') */
282 } VarPatternFlags;
283
284 static Var *
285 VarNew(const char *name, void *name_freeIt, const char *value, VarFlags flags)
286 {
287 size_t value_len = strlen(value);
288 Var *var = bmake_malloc(sizeof *var);
289 var->name = name;
290 var->name_freeIt = name_freeIt;
291 Buf_InitSize(&var->val, value_len + 1);
292 Buf_AddBytes(&var->val, value, value_len);
293 var->flags = flags;
294 return var;
295 }
296
297 static const char *
298 CanonicalVarname(const char *name)
299 {
300 if (*name == '.' && ch_isupper(name[1])) {
301 switch (name[1]) {
302 case 'A':
303 if (strcmp(name, ".ALLSRC") == 0)
304 name = ALLSRC;
305 if (strcmp(name, ".ARCHIVE") == 0)
306 name = ARCHIVE;
307 break;
308 case 'I':
309 if (strcmp(name, ".IMPSRC") == 0)
310 name = IMPSRC;
311 break;
312 case 'M':
313 if (strcmp(name, ".MEMBER") == 0)
314 name = MEMBER;
315 break;
316 case 'O':
317 if (strcmp(name, ".OODATE") == 0)
318 name = OODATE;
319 break;
320 case 'P':
321 if (strcmp(name, ".PREFIX") == 0)
322 name = PREFIX;
323 break;
324 case 'S':
325 if (strcmp(name, ".SHELL") == 0) {
326 if (!shellPath)
327 Shell_Init();
328 }
329 break;
330 case 'T':
331 if (strcmp(name, ".TARGET") == 0)
332 name = TARGET;
333 break;
334 }
335 }
336
337 /* GNU make has an additional alias $^ == ${.ALLSRC}. */
338
339 return name;
340 }
341
342 static Var *
343 GNode_FindVar(GNode *ctxt, const char *varname, unsigned int hash)
344 {
345 return HashTable_FindValueHash(&ctxt->context, varname, hash);
346 }
347
348 /* Find the variable in the context, and maybe in other contexts as well.
349 *
350 * Input:
351 * name name to find, is not expanded any further
352 * ctxt context in which to look first
353 * elsewhere TRUE to look in other contexts as well
354 *
355 * Results:
356 * The found variable, or NULL if the variable does not exist.
357 * If the variable is an environment variable, it must be freed using
358 * VarFreeEnv after use.
359 */
360 static Var *
361 VarFind(const char *name, GNode *ctxt, Boolean elsewhere)
362 {
363 Var *var;
364 unsigned int nameHash;
365
366 /*
367 * If the variable name begins with a '.', it could very well be one of
368 * the local ones. We check the name against all the local variables
369 * and substitute the short version in for 'name' if it matches one of
370 * them.
371 */
372 name = CanonicalVarname(name);
373 nameHash = Hash_Hash(name);
374
375 /* First look for the variable in the given context. */
376 var = GNode_FindVar(ctxt, name, nameHash);
377 if (!elsewhere)
378 return var;
379
380 /* The variable was not found in the given context. Now look for it in
381 * the other contexts as well. */
382 if (var == NULL && ctxt != VAR_CMDLINE)
383 var = GNode_FindVar(VAR_CMDLINE, name, nameHash);
384
385 if (!opts.checkEnvFirst && var == NULL && ctxt != VAR_GLOBAL) {
386 var = GNode_FindVar(VAR_GLOBAL, name, nameHash);
387 if (var == NULL && ctxt != VAR_INTERNAL) {
388 /* VAR_INTERNAL is subordinate to VAR_GLOBAL */
389 var = GNode_FindVar(VAR_INTERNAL, name, nameHash);
390 }
391 }
392
393 if (var == NULL) {
394 char *env;
395
396 if ((env = getenv(name)) != NULL) {
397 char *varname = bmake_strdup(name);
398 return VarNew(varname, varname, env, VAR_FROM_ENV);
399 }
400
401 if (opts.checkEnvFirst && ctxt != VAR_GLOBAL) {
402 var = GNode_FindVar(VAR_GLOBAL, name, nameHash);
403 if (var == NULL && ctxt != VAR_INTERNAL)
404 var = GNode_FindVar(VAR_INTERNAL, name, nameHash);
405 return var;
406 }
407
408 return NULL;
409 }
410
411 return var;
412 }
413
414 /* If the variable is an environment variable, free it.
415 *
416 * Input:
417 * v the variable
418 * freeValue true if the variable value should be freed as well
419 *
420 * Results:
421 * TRUE if it is an environment variable, FALSE otherwise.
422 */
423 static Boolean
424 VarFreeEnv(Var *v, Boolean freeValue)
425 {
426 if (!(v->flags & VAR_FROM_ENV))
427 return FALSE;
428
429 free(v->name_freeIt);
430 Buf_Destroy(&v->val, freeValue);
431 free(v);
432 return TRUE;
433 }
434
435 /* Add a new variable of the given name and value to the given context.
436 * The name and val arguments are duplicated so they may safely be freed. */
437 static void
438 VarAdd(const char *name, const char *val, GNode *ctxt, VarSetFlags flags)
439 {
440 HashEntry *he = HashTable_CreateEntry(&ctxt->context, name, NULL);
441 Var *v = VarNew(he->key /* aliased */, NULL, val,
442 flags & VAR_SET_READONLY ? VAR_READONLY : 0);
443 HashEntry_Set(he, v);
444 if (!(ctxt->flags & INTERNAL)) {
445 VAR_DEBUG3("%s:%s = %s\n", ctxt->name, name, val);
446 }
447 }
448
449 /* Remove a variable from a context, freeing all related memory as well.
450 * The variable name is expanded once. */
451 void
452 Var_Delete(const char *name, GNode *ctxt)
453 {
454 char *name_freeIt = NULL;
455 HashEntry *he;
456
457 if (strchr(name, '$') != NULL) {
458 (void)Var_Subst(name, VAR_GLOBAL, VARE_WANTRES, &name_freeIt);
459 /* TODO: handle errors */
460 name = name_freeIt;
461 }
462 he = HashTable_FindEntry(&ctxt->context, name);
463 VAR_DEBUG3("%s:delete %s%s\n",
464 ctxt->name, name, he != NULL ? "" : " (not found)");
465 free(name_freeIt);
466
467 if (he != NULL) {
468 Var *v = HashEntry_Get(he);
469 if (v->flags & VAR_EXPORTED)
470 unsetenv(v->name);
471 if (strcmp(v->name, MAKE_EXPORTED) == 0)
472 var_exportedVars = VAR_EXPORTED_NONE;
473 assert(v->name_freeIt == NULL);
474 HashTable_DeleteEntry(&ctxt->context, he);
475 Buf_Destroy(&v->val, TRUE);
476 free(v);
477 }
478 }
479
480 static Boolean
481 MayExport(const char *name)
482 {
483 if (name[0] == '.')
484 return FALSE; /* skip internals */
485 if (name[0] == '-')
486 return FALSE; /* skip misnamed variables */
487 if (name[1] == '\0') {
488 /*
489 * A single char.
490 * If it is one of the vars that should only appear in
491 * local context, skip it, else we can get Var_Subst
492 * into a loop.
493 */
494 switch (name[0]) {
495 case '@':
496 case '%':
497 case '*':
498 case '!':
499 return FALSE;
500 }
501 }
502 return TRUE;
503 }
504
505 /*
506 * Export a single variable.
507 * We ignore make internal variables (those which start with '.').
508 * Also we jump through some hoops to avoid calling setenv
509 * more than necessary since it can leak.
510 * We only manipulate flags of vars if 'parent' is set.
511 */
512 static Boolean
513 Var_Export1(const char *name, VarExportFlags flags)
514 {
515 VarExportFlags parent = flags & VAR_EXPORT_PARENT;
516 Var *v;
517 char *val;
518
519 if (!MayExport(name))
520 return FALSE;
521
522 v = VarFind(name, VAR_GLOBAL, FALSE);
523 if (v == NULL)
524 return FALSE;
525
526 if (!parent && (v->flags & VAR_EXPORTED) && !(v->flags & VAR_REEXPORT))
527 return FALSE; /* nothing to do */
528
529 val = Buf_GetAll(&v->val, NULL);
530 if (!(flags & VAR_EXPORT_LITERAL) && strchr(val, '$') != NULL) {
531 char *expr;
532
533 if (parent) {
534 /*
535 * Flag the variable as something we need to re-export.
536 * No point actually exporting it now though,
537 * the child process can do it at the last minute.
538 */
539 v->flags |= VAR_EXPORTED | VAR_REEXPORT;
540 return TRUE;
541 }
542 if (v->flags & VAR_IN_USE) {
543 /*
544 * We recursed while exporting in a child.
545 * This isn't going to end well, just skip it.
546 */
547 return FALSE;
548 }
549
550 /* XXX: name is injected without escaping it */
551 expr = str_concat3("${", name, "}");
552 (void)Var_Subst(expr, VAR_GLOBAL, VARE_WANTRES, &val);
553 /* TODO: handle errors */
554 setenv(name, val, 1);
555 free(val);
556 free(expr);
557 } else {
558 if (parent)
559 v->flags &= ~(unsigned)VAR_REEXPORT; /* once will do */
560 if (parent || !(v->flags & VAR_EXPORTED))
561 setenv(name, val, 1);
562 }
563
564 /*
565 * This is so Var_Set knows to call Var_Export again...
566 */
567 if (parent) {
568 v->flags |= VAR_EXPORTED;
569 }
570 return TRUE;
571 }
572
573 /*
574 * This gets called from our child processes.
575 */
576 void
577 Var_ExportVars(void)
578 {
579 char *val;
580
581 /*
582 * Several make's support this sort of mechanism for tracking
583 * recursion - but each uses a different name.
584 * We allow the makefiles to update MAKELEVEL and ensure
585 * children see a correctly incremented value.
586 */
587 char tmp[BUFSIZ];
588 snprintf(tmp, sizeof tmp, "%d", makelevel + 1);
589 setenv(MAKE_LEVEL_ENV, tmp, 1);
590
591 if (var_exportedVars == VAR_EXPORTED_NONE)
592 return;
593
594 if (var_exportedVars == VAR_EXPORTED_ALL) {
595 HashIter hi;
596
597 /* Ouch! Exporting all variables at once is crazy... */
598 HashIter_Init(&hi, &VAR_GLOBAL->context);
599 while (HashIter_Next(&hi) != NULL) {
600 Var *var = hi.entry->value;
601 Var_Export1(var->name, VAR_EXPORT_NORMAL);
602 }
603 return;
604 }
605
606 (void)Var_Subst("${" MAKE_EXPORTED ":O:u}", VAR_GLOBAL, VARE_WANTRES, &val);
607 /* TODO: handle errors */
608 if (*val) {
609 Words words = Str_Words(val, FALSE);
610 size_t i;
611
612 for (i = 0; i < words.len; i++)
613 Var_Export1(words.words[i], VAR_EXPORT_NORMAL);
614 Words_Free(words);
615 }
616 free(val);
617 }
618
619 /*
620 * This is called when .export is seen or .MAKE.EXPORTED is modified.
621 *
622 * It is also called when any exported variable is modified.
623 * XXX: Is it really?
624 *
625 * str has the format "[-env|-literal] varname...".
626 */
627 void
628 Var_Export(const char *str, Boolean isExport)
629 {
630 VarExportFlags flags;
631 char *val;
632
633 if (isExport && str[0] == '\0') {
634 var_exportedVars = VAR_EXPORTED_ALL; /* use with caution! */
635 return;
636 }
637
638 if (isExport && strncmp(str, "-env", 4) == 0) {
639 str += 4;
640 flags = 0;
641 } else if (isExport && strncmp(str, "-literal", 8) == 0) {
642 str += 8;
643 flags = VAR_EXPORT_LITERAL;
644 } else {
645 flags = VAR_EXPORT_PARENT;
646 }
647
648 (void)Var_Subst(str, VAR_GLOBAL, VARE_WANTRES, &val);
649 /* TODO: handle errors */
650 if (val[0] != '\0') {
651 Words words = Str_Words(val, FALSE);
652
653 size_t i;
654 for (i = 0; i < words.len; i++) {
655 const char *name = words.words[i];
656 if (Var_Export1(name, flags)) {
657 if (var_exportedVars == VAR_EXPORTED_NONE)
658 var_exportedVars = VAR_EXPORTED_SOME;
659 if (isExport && (flags & VAR_EXPORT_PARENT)) {
660 Var_Append(MAKE_EXPORTED, name, VAR_GLOBAL);
661 }
662 }
663 }
664 Words_Free(words);
665 }
666 free(val);
667 }
668
669
670 extern char **environ;
671
672 /*
673 * This is called when .unexport[-env] is seen.
674 *
675 * str must have the form "unexport[-env] varname...".
676 */
677 void
678 Var_UnExport(const char *str)
679 {
680 const char *varnames;
681 char *varnames_freeIt;
682 Boolean unexport_env;
683
684 varnames = NULL;
685 varnames_freeIt = NULL;
686
687 str += strlen("unexport");
688 unexport_env = strncmp(str, "-env", 4) == 0;
689 if (unexport_env) {
690 const char *cp;
691 char **newenv;
692
693 cp = getenv(MAKE_LEVEL_ENV); /* we should preserve this */
694 if (environ == savedEnv) {
695 /* we have been here before! */
696 newenv = bmake_realloc(environ, 2 * sizeof(char *));
697 } else {
698 if (savedEnv) {
699 free(savedEnv);
700 savedEnv = NULL;
701 }
702 newenv = bmake_malloc(2 * sizeof(char *));
703 }
704
705 /* Note: we cannot safely free() the original environ. */
706 environ = savedEnv = newenv;
707 newenv[0] = NULL;
708 newenv[1] = NULL;
709 if (cp && *cp)
710 setenv(MAKE_LEVEL_ENV, cp, 1);
711 } else {
712 cpp_skip_whitespace(&str);
713 if (str[0] != '\0')
714 varnames = str;
715 }
716
717 if (varnames == NULL) {
718 /* Using .MAKE.EXPORTED */
719 (void)Var_Subst("${" MAKE_EXPORTED ":O:u}", VAR_GLOBAL, VARE_WANTRES,
720 &varnames_freeIt);
721 /* TODO: handle errors */
722 varnames = varnames_freeIt;
723 }
724
725 {
726 Var *v;
727 size_t i;
728
729 Words words = Str_Words(varnames, FALSE);
730 for (i = 0; i < words.len; i++) {
731 const char *varname = words.words[i];
732 v = VarFind(varname, VAR_GLOBAL, FALSE);
733 if (v == NULL) {
734 VAR_DEBUG1("Not unexporting \"%s\" (not found)\n", varname);
735 continue;
736 }
737
738 VAR_DEBUG1("Unexporting \"%s\"\n", varname);
739 if (!unexport_env && (v->flags & VAR_EXPORTED) &&
740 !(v->flags & VAR_REEXPORT))
741 unsetenv(v->name);
742 v->flags &= ~(unsigned)(VAR_EXPORTED | VAR_REEXPORT);
743
744 /*
745 * If we are unexporting a list,
746 * remove each one from .MAKE.EXPORTED.
747 * If we are removing them all,
748 * just delete .MAKE.EXPORTED below.
749 */
750 if (varnames == str) {
751 /* XXX: v->name is injected without escaping it */
752 char *expr = str_concat3("${" MAKE_EXPORTED ":N", v->name, "}");
753 char *cp;
754 (void)Var_Subst(expr, VAR_GLOBAL, VARE_WANTRES, &cp);
755 /* TODO: handle errors */
756 Var_Set(MAKE_EXPORTED, cp, VAR_GLOBAL);
757 free(cp);
758 free(expr);
759 }
760 }
761 Words_Free(words);
762 if (varnames != str) {
763 Var_Delete(MAKE_EXPORTED, VAR_GLOBAL);
764 free(varnames_freeIt);
765 }
766 }
767 }
768
769 /* See Var_Set for documentation. */
770 void
771 Var_SetWithFlags(const char *name, const char *val, GNode *ctxt,
772 VarSetFlags flags)
773 {
774 const char *unexpanded_name = name;
775 char *name_freeIt = NULL;
776 Var *v;
777
778 assert(val != NULL);
779
780 if (strchr(name, '$') != NULL) {
781 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
782 /* TODO: handle errors */
783 name = name_freeIt;
784 }
785
786 if (name[0] == '\0') {
787 VAR_DEBUG2("Var_Set(\"%s\", \"%s\", ...) "
788 "name expands to empty string - ignored\n",
789 unexpanded_name, val);
790 free(name_freeIt);
791 return;
792 }
793
794 if (ctxt == VAR_GLOBAL) {
795 v = VarFind(name, VAR_CMDLINE, FALSE);
796 if (v != NULL) {
797 if (v->flags & VAR_FROM_CMD) {
798 VAR_DEBUG3("%s:%s = %s ignored!\n", ctxt->name, name, val);
799 goto out;
800 }
801 VarFreeEnv(v, TRUE);
802 }
803 }
804
805 /*
806 * We only look for a variable in the given context since anything set
807 * here will override anything in a lower context, so there's not much
808 * point in searching them all just to save a bit of memory...
809 */
810 v = VarFind(name, ctxt, FALSE);
811 if (v == NULL) {
812 if (ctxt == VAR_CMDLINE && !(flags & VAR_SET_NO_EXPORT)) {
813 /*
814 * This var would normally prevent the same name being added
815 * to VAR_GLOBAL, so delete it from there if needed.
816 * Otherwise -V name may show the wrong value.
817 */
818 /* XXX: name is expanded for the second time */
819 Var_Delete(name, VAR_GLOBAL);
820 }
821 VarAdd(name, val, ctxt, flags);
822 } else {
823 if ((v->flags & VAR_READONLY) && !(flags & VAR_SET_READONLY)) {
824 VAR_DEBUG3("%s:%s = %s ignored (read-only)\n",
825 ctxt->name, name, val);
826 goto out;
827 }
828 Buf_Empty(&v->val);
829 Buf_AddStr(&v->val, val);
830
831 VAR_DEBUG3("%s:%s = %s\n", ctxt->name, name, val);
832 if (v->flags & VAR_EXPORTED) {
833 Var_Export1(name, VAR_EXPORT_PARENT);
834 }
835 }
836 /*
837 * Any variables given on the command line are automatically exported
838 * to the environment (as per POSIX standard)
839 * Other than internals.
840 */
841 if (ctxt == VAR_CMDLINE && !(flags & VAR_SET_NO_EXPORT) && name[0] != '.') {
842 if (v == NULL)
843 v = VarFind(name, ctxt, FALSE); /* we just added it */
844 v->flags |= VAR_FROM_CMD;
845
846 /*
847 * If requested, don't export these in the environment
848 * individually. We still put them in MAKEOVERRIDES so
849 * that the command-line settings continue to override
850 * Makefile settings.
851 */
852 if (!opts.varNoExportEnv)
853 setenv(name, val, 1);
854
855 Var_Append(MAKEOVERRIDES, name, VAR_GLOBAL);
856 }
857 if (name[0] == '.' && strcmp(name, MAKE_SAVE_DOLLARS) == 0)
858 save_dollars = ParseBoolean(val, save_dollars);
859
860 out:
861 free(name_freeIt);
862 if (v != NULL)
863 VarFreeEnv(v, TRUE);
864 }
865
866 /*-
867 *-----------------------------------------------------------------------
868 * Var_Set --
869 * Set the variable name to the value val in the given context.
870 *
871 * If the variable doesn't yet exist, it is created.
872 * Otherwise the new value overwrites and replaces the old value.
873 *
874 * Input:
875 * name name of the variable to set, is expanded once
876 * val value to give to the variable
877 * ctxt context in which to set it
878 *
879 * Notes:
880 * The variable is searched for only in its context before being
881 * created in that context. I.e. if the context is VAR_GLOBAL,
882 * only VAR_GLOBAL->context is searched. Likewise if it is VAR_CMDLINE,
883 * only VAR_CMDLINE->context is searched. This is done to avoid the
884 * literally thousands of unnecessary strcmp's that used to be done to
885 * set, say, $(@) or $(<).
886 * If the context is VAR_GLOBAL though, we check if the variable
887 * was set in VAR_CMDLINE from the command line and skip it if so.
888 *-----------------------------------------------------------------------
889 */
890 void
891 Var_Set(const char *name, const char *val, GNode *ctxt)
892 {
893 Var_SetWithFlags(name, val, ctxt, VAR_SET_NONE);
894 }
895
896 /*-
897 *-----------------------------------------------------------------------
898 * Var_Append --
899 * The variable of the given name has the given value appended to it in
900 * the given context.
901 *
902 * If the variable doesn't exist, it is created. Otherwise the strings
903 * are concatenated, with a space in between.
904 *
905 * Input:
906 * name name of the variable to modify, is expanded once
907 * val string to append to it
908 * ctxt context in which this should occur
909 *
910 * Notes:
911 * Only if the variable is being sought in the global context is the
912 * environment searched.
913 * XXX: Knows its calling circumstances in that if called with ctxt
914 * an actual target, it will only search that context since only
915 * a local variable could be being appended to. This is actually
916 * a big win and must be tolerated.
917 *-----------------------------------------------------------------------
918 */
919 void
920 Var_Append(const char *name, const char *val, GNode *ctxt)
921 {
922 char *name_freeIt = NULL;
923 Var *v;
924
925 assert(val != NULL);
926
927 if (strchr(name, '$') != NULL) {
928 const char *unexpanded_name = name;
929 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
930 /* TODO: handle errors */
931 name = name_freeIt;
932 if (name[0] == '\0') {
933 VAR_DEBUG2("Var_Append(\"%s\", \"%s\", ...) "
934 "name expands to empty string - ignored\n",
935 unexpanded_name, val);
936 free(name_freeIt);
937 return;
938 }
939 }
940
941 v = VarFind(name, ctxt, ctxt == VAR_GLOBAL);
942
943 if (v == NULL) {
944 /* XXX: name is expanded for the second time */
945 Var_Set(name, val, ctxt);
946 } else if (v->flags & VAR_READONLY) {
947 VAR_DEBUG1("Ignoring append to %s since it is read-only\n", name);
948 } else if (ctxt == VAR_CMDLINE || !(v->flags & VAR_FROM_CMD)) {
949 Buf_AddByte(&v->val, ' ');
950 Buf_AddStr(&v->val, val);
951
952 VAR_DEBUG3("%s:%s = %s\n",
953 ctxt->name, name, Buf_GetAll(&v->val, NULL));
954
955 if (v->flags & VAR_FROM_ENV) {
956 HashEntry *h;
957
958 /*
959 * If the original variable came from the environment, we
960 * have to install it in the global context (we could place
961 * it in the environment, but then we should provide a way to
962 * export other variables...)
963 */
964 v->flags &= ~(unsigned)VAR_FROM_ENV;
965 h = HashTable_CreateEntry(&ctxt->context, name, NULL);
966 HashEntry_Set(h, v);
967 }
968 }
969 free(name_freeIt);
970 }
971
972 /* See if the given variable exists, in the given context or in other
973 * fallback contexts.
974 *
975 * Input:
976 * name Variable to find, is expanded once
977 * ctxt Context in which to start search
978 */
979 Boolean
980 Var_Exists(const char *name, GNode *ctxt)
981 {
982 char *name_freeIt = NULL;
983 Var *v;
984
985 if (strchr(name, '$') != NULL) {
986 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
987 /* TODO: handle errors */
988 name = name_freeIt;
989 }
990
991 v = VarFind(name, ctxt, TRUE);
992 free(name_freeIt);
993 if (v == NULL)
994 return FALSE;
995
996 (void)VarFreeEnv(v, TRUE);
997 return TRUE;
998 }
999
1000 /*-
1001 *-----------------------------------------------------------------------
1002 * Var_Value --
1003 * Return the unexpanded value of the given variable in the given
1004 * context, or the usual contexts.
1005 *
1006 * Input:
1007 * name name to find, is not expanded any further
1008 * ctxt context in which to search for it
1009 *
1010 * Results:
1011 * The value if the variable exists, NULL if it doesn't.
1012 * If the returned value is not NULL, the caller must free
1013 * out_freeIt when the returned value is no longer needed.
1014 *-----------------------------------------------------------------------
1015 */
1016 const char *
1017 Var_Value(const char *name, GNode *ctxt, void **out_freeIt)
1018 {
1019 Var *v = VarFind(name, ctxt, TRUE);
1020 char *value;
1021
1022 *out_freeIt = NULL;
1023 if (v == NULL)
1024 return NULL;
1025
1026 value = Buf_GetAll(&v->val, NULL);
1027 if (VarFreeEnv(v, FALSE))
1028 *out_freeIt = value;
1029 return value;
1030 }
1031
1032 /* Return the unexpanded variable value from this node, without trying to look
1033 * up the variable in any other context. */
1034 const char *
1035 Var_ValueDirect(const char *name, GNode *ctxt)
1036 {
1037 Var *v = VarFind(name, ctxt, FALSE);
1038 return v != NULL ? Buf_GetAll(&v->val, NULL) : NULL;
1039 }
1040
1041
1042 /* SepBuf is a string being built from words, interleaved with separators. */
1043 typedef struct SepBuf {
1044 Buffer buf;
1045 Boolean needSep;
1046 char sep; /* usually ' ', but see the :ts modifier */
1047 } SepBuf;
1048
1049 static void
1050 SepBuf_Init(SepBuf *buf, char sep)
1051 {
1052 Buf_InitSize(&buf->buf, 32);
1053 buf->needSep = FALSE;
1054 buf->sep = sep;
1055 }
1056
1057 static void
1058 SepBuf_Sep(SepBuf *buf)
1059 {
1060 buf->needSep = TRUE;
1061 }
1062
1063 static void
1064 SepBuf_AddBytes(SepBuf *buf, const char *mem, size_t mem_size)
1065 {
1066 if (mem_size == 0)
1067 return;
1068 if (buf->needSep && buf->sep != '\0') {
1069 Buf_AddByte(&buf->buf, buf->sep);
1070 buf->needSep = FALSE;
1071 }
1072 Buf_AddBytes(&buf->buf, mem, mem_size);
1073 }
1074
1075 static void
1076 SepBuf_AddBytesBetween(SepBuf *buf, const char *start, const char *end)
1077 {
1078 SepBuf_AddBytes(buf, start, (size_t)(end - start));
1079 }
1080
1081 static void
1082 SepBuf_AddStr(SepBuf *buf, const char *str)
1083 {
1084 SepBuf_AddBytes(buf, str, strlen(str));
1085 }
1086
1087 static char *
1088 SepBuf_Destroy(SepBuf *buf, Boolean free_buf)
1089 {
1090 return Buf_Destroy(&buf->buf, free_buf);
1091 }
1092
1093
1094 /* This callback for ModifyWords gets a single word from a variable expression
1095 * and typically adds a modification of this word to the buffer. It may also
1096 * do nothing or add several words.
1097 *
1098 * For example, in ${:Ua b c:M*2}, the callback is called 3 times, once for
1099 * each word of "a b c". */
1100 typedef void (*ModifyWordsCallback)(const char *word, SepBuf *buf, void *data);
1101
1102
1103 /* Callback for ModifyWords to implement the :H modifier.
1104 * Add the dirname of the given word to the buffer. */
1105 static void
1106 ModifyWord_Head(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1107 {
1108 const char *slash = strrchr(word, '/');
1109 if (slash != NULL)
1110 SepBuf_AddBytesBetween(buf, word, slash);
1111 else
1112 SepBuf_AddStr(buf, ".");
1113 }
1114
1115 /* Callback for ModifyWords to implement the :T modifier.
1116 * Add the basename of the given word to the buffer. */
1117 static void
1118 ModifyWord_Tail(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1119 {
1120 const char *slash = strrchr(word, '/');
1121 const char *base = slash != NULL ? slash + 1 : word;
1122 SepBuf_AddStr(buf, base);
1123 }
1124
1125 /* Callback for ModifyWords to implement the :E modifier.
1126 * Add the filename suffix of the given word to the buffer, if it exists. */
1127 static void
1128 ModifyWord_Suffix(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1129 {
1130 const char *dot = strrchr(word, '.');
1131 if (dot != NULL)
1132 SepBuf_AddStr(buf, dot + 1);
1133 }
1134
1135 /* Callback for ModifyWords to implement the :R modifier.
1136 * Add the basename of the given word to the buffer. */
1137 static void
1138 ModifyWord_Root(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1139 {
1140 const char *dot = strrchr(word, '.');
1141 size_t len = dot != NULL ? (size_t)(dot - word) : strlen(word);
1142 SepBuf_AddBytes(buf, word, len);
1143 }
1144
1145 /* Callback for ModifyWords to implement the :M modifier.
1146 * Place the word in the buffer if it matches the given pattern. */
1147 static void
1148 ModifyWord_Match(const char *word, SepBuf *buf, void *data)
1149 {
1150 const char *pattern = data;
1151 VAR_DEBUG2("VarMatch [%s] [%s]\n", word, pattern);
1152 if (Str_Match(word, pattern))
1153 SepBuf_AddStr(buf, word);
1154 }
1155
1156 /* Callback for ModifyWords to implement the :N modifier.
1157 * Place the word in the buffer if it doesn't match the given pattern. */
1158 static void
1159 ModifyWord_NoMatch(const char *word, SepBuf *buf, void *data)
1160 {
1161 const char *pattern = data;
1162 if (!Str_Match(word, pattern))
1163 SepBuf_AddStr(buf, word);
1164 }
1165
1166 #ifdef SYSVVARSUB
1167 /* Check word against pattern for a match (% is a wildcard).
1168 *
1169 * Input:
1170 * word Word to examine
1171 * pattern Pattern to examine against
1172 *
1173 * Results:
1174 * Returns the start of the match, or NULL.
1175 * out_match_len returns the length of the match, if any.
1176 * out_hasPercent returns whether the pattern contains a percent.
1177 */
1178 static const char *
1179 SysVMatch(const char *word, const char *pattern,
1180 size_t *out_match_len, Boolean *out_hasPercent)
1181 {
1182 const char *p = pattern;
1183 const char *w = word;
1184 const char *percent;
1185 size_t w_len;
1186 size_t p_len;
1187 const char *w_tail;
1188
1189 *out_hasPercent = FALSE;
1190 percent = strchr(p, '%');
1191 if (percent != NULL) { /* ${VAR:...%...=...} */
1192 *out_hasPercent = TRUE;
1193 if (*w == '\0')
1194 return NULL; /* empty word does not match pattern */
1195
1196 /* check that the prefix matches */
1197 for (; p != percent && *w != '\0' && *w == *p; w++, p++)
1198 continue;
1199 if (p != percent)
1200 return NULL; /* No match */
1201
1202 p++; /* Skip the percent */
1203 if (*p == '\0') {
1204 /* No more pattern, return the rest of the string */
1205 *out_match_len = strlen(w);
1206 return w;
1207 }
1208 }
1209
1210 /* Test whether the tail matches */
1211 w_len = strlen(w);
1212 p_len = strlen(p);
1213 if (w_len < p_len)
1214 return NULL;
1215
1216 w_tail = w + w_len - p_len;
1217 if (memcmp(p, w_tail, p_len) != 0)
1218 return NULL;
1219
1220 *out_match_len = (size_t)(w_tail - w);
1221 return w;
1222 }
1223
1224 struct ModifyWord_SYSVSubstArgs {
1225 GNode *ctx;
1226 const char *lhs;
1227 const char *rhs;
1228 };
1229
1230 /* Callback for ModifyWords to implement the :%.from=%.to modifier. */
1231 static void
1232 ModifyWord_SYSVSubst(const char *word, SepBuf *buf, void *data)
1233 {
1234 const struct ModifyWord_SYSVSubstArgs *args = data;
1235 char *rhs_expanded;
1236 const char *rhs;
1237 const char *percent;
1238
1239 size_t match_len;
1240 Boolean lhsPercent;
1241 const char *match = SysVMatch(word, args->lhs, &match_len, &lhsPercent);
1242 if (match == NULL) {
1243 SepBuf_AddStr(buf, word);
1244 return;
1245 }
1246
1247 /* Append rhs to the buffer, substituting the first '%' with the
1248 * match, but only if the lhs had a '%' as well. */
1249
1250 (void)Var_Subst(args->rhs, args->ctx, VARE_WANTRES, &rhs_expanded);
1251 /* TODO: handle errors */
1252
1253 rhs = rhs_expanded;
1254 percent = strchr(rhs, '%');
1255
1256 if (percent != NULL && lhsPercent) {
1257 /* Copy the prefix of the replacement pattern */
1258 SepBuf_AddBytesBetween(buf, rhs, percent);
1259 rhs = percent + 1;
1260 }
1261 if (percent != NULL || !lhsPercent)
1262 SepBuf_AddBytes(buf, match, match_len);
1263
1264 /* Append the suffix of the replacement pattern */
1265 SepBuf_AddStr(buf, rhs);
1266
1267 free(rhs_expanded);
1268 }
1269 #endif
1270
1271
1272 struct ModifyWord_SubstArgs {
1273 const char *lhs;
1274 size_t lhsLen;
1275 const char *rhs;
1276 size_t rhsLen;
1277 VarPatternFlags pflags;
1278 Boolean matched;
1279 };
1280
1281 /* Callback for ModifyWords to implement the :S,from,to, modifier.
1282 * Perform a string substitution on the given word. */
1283 static void
1284 ModifyWord_Subst(const char *word, SepBuf *buf, void *data)
1285 {
1286 size_t wordLen = strlen(word);
1287 struct ModifyWord_SubstArgs *args = data;
1288 const char *match;
1289
1290 if ((args->pflags & VARP_SUB_ONE) && args->matched)
1291 goto nosub;
1292
1293 if (args->pflags & VARP_ANCHOR_START) {
1294 if (wordLen < args->lhsLen ||
1295 memcmp(word, args->lhs, args->lhsLen) != 0)
1296 goto nosub;
1297
1298 if ((args->pflags & VARP_ANCHOR_END) && wordLen != args->lhsLen)
1299 goto nosub;
1300
1301 /* :S,^prefix,replacement, or :S,^whole$,replacement, */
1302 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1303 SepBuf_AddBytes(buf, word + args->lhsLen, wordLen - args->lhsLen);
1304 args->matched = TRUE;
1305 return;
1306 }
1307
1308 if (args->pflags & VARP_ANCHOR_END) {
1309 const char *start;
1310
1311 if (wordLen < args->lhsLen)
1312 goto nosub;
1313
1314 start = word + (wordLen - args->lhsLen);
1315 if (memcmp(start, args->lhs, args->lhsLen) != 0)
1316 goto nosub;
1317
1318 /* :S,suffix$,replacement, */
1319 SepBuf_AddBytesBetween(buf, word, start);
1320 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1321 args->matched = TRUE;
1322 return;
1323 }
1324
1325 if (args->lhs[0] == '\0')
1326 goto nosub;
1327
1328 /* unanchored case, may match more than once */
1329 while ((match = strstr(word, args->lhs)) != NULL) {
1330 SepBuf_AddBytesBetween(buf, word, match);
1331 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1332 args->matched = TRUE;
1333 wordLen -= (size_t)(match - word) + args->lhsLen;
1334 word += (size_t)(match - word) + args->lhsLen;
1335 if (wordLen == 0 || !(args->pflags & VARP_SUB_GLOBAL))
1336 break;
1337 }
1338 nosub:
1339 SepBuf_AddBytes(buf, word, wordLen);
1340 }
1341
1342 #ifndef NO_REGEX
1343 /* Print the error caused by a regcomp or regexec call. */
1344 static void
1345 VarREError(int reerr, const regex_t *pat, const char *str)
1346 {
1347 size_t errlen = regerror(reerr, pat, NULL, 0);
1348 char *errbuf = bmake_malloc(errlen);
1349 regerror(reerr, pat, errbuf, errlen);
1350 Error("%s: %s", str, errbuf);
1351 free(errbuf);
1352 }
1353
1354 struct ModifyWord_SubstRegexArgs {
1355 regex_t re;
1356 size_t nsub;
1357 char *replace;
1358 VarPatternFlags pflags;
1359 Boolean matched;
1360 };
1361
1362 /* Callback for ModifyWords to implement the :C/from/to/ modifier.
1363 * Perform a regex substitution on the given word. */
1364 static void
1365 ModifyWord_SubstRegex(const char *word, SepBuf *buf, void *data)
1366 {
1367 struct ModifyWord_SubstRegexArgs *args = data;
1368 int xrv;
1369 const char *wp = word;
1370 char *rp;
1371 int flags = 0;
1372 regmatch_t m[10];
1373
1374 if ((args->pflags & VARP_SUB_ONE) && args->matched)
1375 goto nosub;
1376
1377 tryagain:
1378 xrv = regexec(&args->re, wp, args->nsub, m, flags);
1379
1380 switch (xrv) {
1381 case 0:
1382 args->matched = TRUE;
1383 SepBuf_AddBytes(buf, wp, (size_t)m[0].rm_so);
1384
1385 for (rp = args->replace; *rp; rp++) {
1386 if (*rp == '\\' && (rp[1] == '&' || rp[1] == '\\')) {
1387 SepBuf_AddBytes(buf, rp + 1, 1);
1388 rp++;
1389 continue;
1390 }
1391
1392 if (*rp == '&') {
1393 SepBuf_AddBytesBetween(buf, wp + m[0].rm_so, wp + m[0].rm_eo);
1394 continue;
1395 }
1396
1397 if (*rp != '\\' || !ch_isdigit(rp[1])) {
1398 SepBuf_AddBytes(buf, rp, 1);
1399 continue;
1400 }
1401
1402 { /* \0 to \9 backreference */
1403 size_t n = (size_t)(rp[1] - '0');
1404 rp++;
1405
1406 if (n >= args->nsub) {
1407 Error("No subexpression \\%zu", n);
1408 } else if (m[n].rm_so == -1) {
1409 Error("No match for subexpression \\%zu", n);
1410 } else {
1411 SepBuf_AddBytesBetween(buf, wp + m[n].rm_so,
1412 wp + m[n].rm_eo);
1413 }
1414 }
1415 }
1416
1417 wp += m[0].rm_eo;
1418 if (args->pflags & VARP_SUB_GLOBAL) {
1419 flags |= REG_NOTBOL;
1420 if (m[0].rm_so == 0 && m[0].rm_eo == 0) {
1421 SepBuf_AddBytes(buf, wp, 1);
1422 wp++;
1423 }
1424 if (*wp)
1425 goto tryagain;
1426 }
1427 if (*wp) {
1428 SepBuf_AddStr(buf, wp);
1429 }
1430 break;
1431 default:
1432 VarREError(xrv, &args->re, "Unexpected regex error");
1433 /* FALLTHROUGH */
1434 case REG_NOMATCH:
1435 nosub:
1436 SepBuf_AddStr(buf, wp);
1437 break;
1438 }
1439 }
1440 #endif
1441
1442
1443 struct ModifyWord_LoopArgs {
1444 GNode *ctx;
1445 char *tvar; /* name of temporary variable */
1446 char *str; /* string to expand */
1447 VarEvalFlags eflags;
1448 };
1449
1450 /* Callback for ModifyWords to implement the :@var (at) ...@ modifier of ODE make. */
1451 static void
1452 ModifyWord_Loop(const char *word, SepBuf *buf, void *data)
1453 {
1454 const struct ModifyWord_LoopArgs *args;
1455 char *s;
1456
1457 if (word[0] == '\0')
1458 return;
1459
1460 args = data;
1461 Var_SetWithFlags(args->tvar, word, args->ctx, VAR_SET_NO_EXPORT);
1462 (void)Var_Subst(args->str, args->ctx, args->eflags, &s);
1463 /* TODO: handle errors */
1464
1465 VAR_DEBUG4("ModifyWord_Loop: "
1466 "in \"%s\", replace \"%s\" with \"%s\" to \"%s\"\n",
1467 word, args->tvar, args->str, s);
1468
1469 if (s[0] == '\n' || Buf_EndsWith(&buf->buf, '\n'))
1470 buf->needSep = FALSE;
1471 SepBuf_AddStr(buf, s);
1472 free(s);
1473 }
1474
1475
1476 /* The :[first..last] modifier selects words from the expression.
1477 * It can also reverse the words. */
1478 static char *
1479 VarSelectWords(char sep, Boolean oneBigWord, const char *str, int first,
1480 int last)
1481 {
1482 Words words;
1483 int len, start, end, step;
1484 int i;
1485
1486 SepBuf buf;
1487 SepBuf_Init(&buf, sep);
1488
1489 if (oneBigWord) {
1490 /* fake what Str_Words() would do if there were only one word */
1491 words.len = 1;
1492 words.words = bmake_malloc((words.len + 1) * sizeof(words.words[0]));
1493 words.freeIt = bmake_strdup(str);
1494 words.words[0] = words.freeIt;
1495 words.words[1] = NULL;
1496 } else {
1497 words = Str_Words(str, FALSE);
1498 }
1499
1500 /*
1501 * Now sanitize the given range.
1502 * If first or last are negative, convert them to the positive equivalents
1503 * (-1 gets converted to ac, -2 gets converted to (ac - 1), etc.).
1504 */
1505 len = (int)words.len;
1506 if (first < 0)
1507 first += len + 1;
1508 if (last < 0)
1509 last += len + 1;
1510
1511 /*
1512 * We avoid scanning more of the list than we need to.
1513 */
1514 if (first > last) {
1515 start = (first > len ? len : first) - 1;
1516 end = last < 1 ? 0 : last - 1;
1517 step = -1;
1518 } else {
1519 start = first < 1 ? 0 : first - 1;
1520 end = last > len ? len : last;
1521 step = 1;
1522 }
1523
1524 for (i = start; (step < 0) == (i >= end); i += step) {
1525 SepBuf_AddStr(&buf, words.words[i]);
1526 SepBuf_Sep(&buf);
1527 }
1528
1529 Words_Free(words);
1530
1531 return SepBuf_Destroy(&buf, FALSE);
1532 }
1533
1534
1535 /* Callback for ModifyWords to implement the :tA modifier.
1536 * Replace each word with the result of realpath() if successful. */
1537 static void
1538 ModifyWord_Realpath(const char *word, SepBuf *buf, void *data MAKE_ATTR_UNUSED)
1539 {
1540 struct stat st;
1541 char rbuf[MAXPATHLEN];
1542
1543 const char *rp = cached_realpath(word, rbuf);
1544 if (rp != NULL && *rp == '/' && stat(rp, &st) == 0)
1545 word = rp;
1546
1547 SepBuf_AddStr(buf, word);
1548 }
1549
1550 /* Modify each of the words of the passed string using the given function.
1551 *
1552 * Input:
1553 * str String whose words should be modified
1554 * modifyWord Function that modifies a single word
1555 * modifyWord_args Custom arguments for modifyWord
1556 *
1557 * Results:
1558 * A string of all the words modified appropriately.
1559 *-----------------------------------------------------------------------
1560 */
1561 static char *
1562 ModifyWords(const char *str,
1563 ModifyWordsCallback modifyWord, void *modifyWord_args,
1564 Boolean oneBigWord, char sep)
1565 {
1566 SepBuf result;
1567 Words words;
1568 size_t i;
1569
1570 if (oneBigWord) {
1571 SepBuf_Init(&result, sep);
1572 modifyWord(str, &result, modifyWord_args);
1573 return SepBuf_Destroy(&result, FALSE);
1574 }
1575
1576 SepBuf_Init(&result, sep);
1577
1578 words = Str_Words(str, FALSE);
1579
1580 VAR_DEBUG2("ModifyWords: split \"%s\" into %zu words\n", str, words.len);
1581
1582 for (i = 0; i < words.len; i++) {
1583 modifyWord(words.words[i], &result, modifyWord_args);
1584 if (Buf_Len(&result.buf) > 0)
1585 SepBuf_Sep(&result);
1586 }
1587
1588 Words_Free(words);
1589
1590 return SepBuf_Destroy(&result, FALSE);
1591 }
1592
1593
1594 static char *
1595 Words_JoinFree(Words words)
1596 {
1597 Buffer buf;
1598 size_t i;
1599
1600 Buf_Init(&buf);
1601
1602 for (i = 0; i < words.len; i++) {
1603 if (i != 0)
1604 Buf_AddByte(&buf, ' '); /* XXX: st->sep, for consistency */
1605 Buf_AddStr(&buf, words.words[i]);
1606 }
1607
1608 Words_Free(words);
1609
1610 return Buf_Destroy(&buf, FALSE);
1611 }
1612
1613 /* Remove adjacent duplicate words. */
1614 static char *
1615 VarUniq(const char *str)
1616 {
1617 Words words = Str_Words(str, FALSE);
1618
1619 if (words.len > 1) {
1620 size_t i, j;
1621 for (j = 0, i = 1; i < words.len; i++)
1622 if (strcmp(words.words[i], words.words[j]) != 0 && (++j != i))
1623 words.words[j] = words.words[i];
1624 words.len = j + 1;
1625 }
1626
1627 return Words_JoinFree(words);
1628 }
1629
1630
1631 /* Quote shell meta-characters and space characters in the string.
1632 * If quoteDollar is set, also quote and double any '$' characters. */
1633 static char *
1634 VarQuote(const char *str, Boolean quoteDollar)
1635 {
1636 Buffer buf;
1637 Buf_Init(&buf);
1638
1639 for (; *str != '\0'; str++) {
1640 if (*str == '\n') {
1641 const char *newline = Shell_GetNewline();
1642 if (newline == NULL)
1643 newline = "\\\n";
1644 Buf_AddStr(&buf, newline);
1645 continue;
1646 }
1647 if (ch_isspace(*str) || is_shell_metachar((unsigned char)*str))
1648 Buf_AddByte(&buf, '\\');
1649 Buf_AddByte(&buf, *str);
1650 if (quoteDollar && *str == '$')
1651 Buf_AddStr(&buf, "\\$");
1652 }
1653
1654 return Buf_Destroy(&buf, FALSE);
1655 }
1656
1657 /* Compute the 32-bit hash of the given string, using the MurmurHash3
1658 * algorithm. Output is encoded as 8 hex digits, in Little Endian order. */
1659 static char *
1660 VarHash(const char *str)
1661 {
1662 static const char hexdigits[16] = "0123456789abcdef";
1663 const unsigned char *ustr = (const unsigned char *)str;
1664
1665 uint32_t h = 0x971e137bU;
1666 uint32_t c1 = 0x95543787U;
1667 uint32_t c2 = 0x2ad7eb25U;
1668 size_t len2 = strlen(str);
1669
1670 char *buf;
1671 size_t i;
1672
1673 size_t len;
1674 for (len = len2; len; ) {
1675 uint32_t k = 0;
1676 switch (len) {
1677 default:
1678 k = ((uint32_t)ustr[3] << 24) |
1679 ((uint32_t)ustr[2] << 16) |
1680 ((uint32_t)ustr[1] << 8) |
1681 (uint32_t)ustr[0];
1682 len -= 4;
1683 ustr += 4;
1684 break;
1685 case 3:
1686 k |= (uint32_t)ustr[2] << 16;
1687 /* FALLTHROUGH */
1688 case 2:
1689 k |= (uint32_t)ustr[1] << 8;
1690 /* FALLTHROUGH */
1691 case 1:
1692 k |= (uint32_t)ustr[0];
1693 len = 0;
1694 }
1695 c1 = c1 * 5 + 0x7b7d159cU;
1696 c2 = c2 * 5 + 0x6bce6396U;
1697 k *= c1;
1698 k = (k << 11) ^ (k >> 21);
1699 k *= c2;
1700 h = (h << 13) ^ (h >> 19);
1701 h = h * 5 + 0x52dce729U;
1702 h ^= k;
1703 }
1704 h ^= (uint32_t)len2;
1705 h *= 0x85ebca6b;
1706 h ^= h >> 13;
1707 h *= 0xc2b2ae35;
1708 h ^= h >> 16;
1709
1710 buf = bmake_malloc(9);
1711 for (i = 0; i < 8; i++) {
1712 buf[i] = hexdigits[h & 0x0f];
1713 h >>= 4;
1714 }
1715 buf[8] = '\0';
1716 return buf;
1717 }
1718
1719 static char *
1720 VarStrftime(const char *fmt, Boolean zulu, time_t tim)
1721 {
1722 char buf[BUFSIZ];
1723
1724 if (tim == 0)
1725 time(&tim);
1726 if (*fmt == '\0')
1727 fmt = "%c";
1728 strftime(buf, sizeof buf, fmt, zulu ? gmtime(&tim) : localtime(&tim));
1729
1730 buf[sizeof buf - 1] = '\0';
1731 return bmake_strdup(buf);
1732 }
1733
1734 /* The ApplyModifier functions all work in the same way. They get the
1735 * current parsing position (pp) and parse the modifier from there. The
1736 * modifier typically lasts until the next ':', or a closing '}' or ')'
1737 * (taken from st->endc), or the end of the string (parse error).
1738 *
1739 * The high-level behavior of these functions is:
1740 *
1741 * 1. parse the modifier
1742 * 2. evaluate the modifier
1743 * 3. housekeeping
1744 *
1745 * Parsing the modifier
1746 *
1747 * If parsing succeeds, the parsing position *pp is updated to point to the
1748 * first character following the modifier, which typically is either ':' or
1749 * st->endc.
1750 *
1751 * If parsing fails because of a missing delimiter (as in the :S, :C or :@
1752 * modifiers), return AMR_CLEANUP.
1753 *
1754 * If parsing fails because the modifier is unknown, return AMR_UNKNOWN to
1755 * try the SysV modifier ${VAR:from=to} as fallback. This should only be
1756 * done as long as there have been no side effects from evaluating nested
1757 * variables, to avoid evaluating them more than once. In this case, the
1758 * parsing position must not be updated. (XXX: Why not? The original parsing
1759 * position is well-known in ApplyModifiers.)
1760 *
1761 * If parsing fails and the SysV modifier ${VAR:from=to} should not be used
1762 * as a fallback, either issue an error message using Error or Parse_Error
1763 * and then return AMR_CLEANUP, or return AMR_BAD for the default error
1764 * message. Both of these return values will stop processing the variable
1765 * expression. (XXX: As of 2020-08-23, evaluation of the whole string
1766 * continues nevertheless after skipping a few bytes, which essentially is
1767 * undefined behavior. Not in the sense of C, but still it's impossible to
1768 * predict what happens in the parser.)
1769 *
1770 * Evaluating the modifier
1771 *
1772 * After parsing, the modifier is evaluated. The side effects from evaluating
1773 * nested variable expressions in the modifier text often already happen
1774 * during parsing though.
1775 *
1776 * Evaluating the modifier usually takes the current value of the variable
1777 * expression from st->val, or the variable name from st->v->name and stores
1778 * the result in st->newVal.
1779 *
1780 * If evaluating fails (as of 2020-08-23), an error message is printed using
1781 * Error. This function has no side-effects, it really just prints the error
1782 * message. Processing the expression continues as if everything were ok.
1783 * XXX: This should be fixed by adding proper error handling to Var_Subst,
1784 * Var_Parse, ApplyModifiers and ModifyWords.
1785 *
1786 * Housekeeping
1787 *
1788 * Some modifiers such as :D and :U turn undefined expressions into defined
1789 * expressions (see VEF_UNDEF, VEF_DEF).
1790 *
1791 * Some modifiers need to free some memory.
1792 */
1793
1794 typedef enum VarExprFlags {
1795 /* The variable expression is based on an undefined variable. */
1796 VEF_UNDEF = 0x01,
1797 /* The variable expression started as an undefined expression, but one
1798 * of the modifiers (such as :D or :U) has turned the expression from
1799 * undefined to defined. */
1800 VEF_DEF = 0x02
1801 } VarExprFlags;
1802
1803 ENUM_FLAGS_RTTI_2(VarExprFlags,
1804 VEF_UNDEF, VEF_DEF);
1805
1806
1807 typedef struct ApplyModifiersState {
1808 const char startc; /* '\0' or '{' or '(' */
1809 const char endc; /* '\0' or '}' or ')' */
1810 Var * const v;
1811 GNode * const ctxt;
1812 const VarEvalFlags eflags;
1813
1814 char *val; /* The old value of the expression,
1815 * before applying the modifier, never NULL */
1816 char *newVal; /* The new value of the expression,
1817 * after applying the modifier, never NULL */
1818 char sep; /* Word separator in expansions
1819 * (see the :ts modifier) */
1820 Boolean oneBigWord; /* TRUE if some modifiers that otherwise split
1821 * the variable value into words, like :S and
1822 * :C, treat the variable value as a single big
1823 * word, possibly containing spaces. */
1824 VarExprFlags exprFlags;
1825 } ApplyModifiersState;
1826
1827 static void
1828 ApplyModifiersState_Define(ApplyModifiersState *st)
1829 {
1830 if (st->exprFlags & VEF_UNDEF)
1831 st->exprFlags |= VEF_DEF;
1832 }
1833
1834 typedef enum ApplyModifierResult {
1835 AMR_OK, /* Continue parsing */
1836 AMR_UNKNOWN, /* Not a match, try other modifiers as well */
1837 AMR_BAD, /* Error out with "Bad modifier" message */
1838 AMR_CLEANUP /* Error out without error message */
1839 } ApplyModifierResult;
1840
1841 /* Allow backslashes to escape the delimiter, $, and \, but don't touch other
1842 * backslashes. */
1843 static Boolean
1844 IsEscapedModifierPart(const char *p, char delim,
1845 struct ModifyWord_SubstArgs *subst)
1846 {
1847 if (p[0] != '\\')
1848 return FALSE;
1849 if (p[1] == delim || p[1] == '\\' || p[1] == '$')
1850 return TRUE;
1851 return p[1] == '&' && subst != NULL;
1852 }
1853
1854 /*
1855 * Parse a part of a modifier such as the "from" and "to" in :S/from/to/ or
1856 * the "var" or "replacement ${var}" in :@var@replacement ${var}@, up to and
1857 * including the next unescaped delimiter. The delimiter, as well as the
1858 * backslash or the dollar, can be escaped with a backslash.
1859 *
1860 * Return the parsed (and possibly expanded) string, or NULL if no delimiter
1861 * was found. On successful return, the parsing position pp points right
1862 * after the delimiter. The delimiter is not included in the returned
1863 * value though.
1864 */
1865 static VarParseResult
1866 ParseModifierPart(
1867 const char **pp, /* The parsing position, updated upon return */
1868 char delim, /* Parsing stops at this delimiter */
1869 VarEvalFlags eflags, /* Flags for evaluating nested variables;
1870 * if VARE_WANTRES is not set, the text is
1871 * only parsed */
1872 ApplyModifiersState *st,
1873 char **out_part,
1874 size_t *out_length, /* Optionally stores the length of the returned
1875 * string, just to save another strlen call. */
1876 VarPatternFlags *out_pflags,/* For the first part of the :S modifier,
1877 * sets the VARP_ANCHOR_END flag if the last
1878 * character of the pattern is a $. */
1879 struct ModifyWord_SubstArgs *subst
1880 /* For the second part of the :S modifier,
1881 * allow ampersands to be escaped and replace
1882 * unescaped ampersands with subst->lhs. */
1883 ) {
1884 Buffer buf;
1885 const char *p;
1886
1887 Buf_Init(&buf);
1888
1889 /*
1890 * Skim through until the matching delimiter is found; pick up variable
1891 * expressions on the way.
1892 */
1893 p = *pp;
1894 while (*p != '\0' && *p != delim) {
1895 const char *varstart;
1896
1897 if (IsEscapedModifierPart(p, delim, subst)) {
1898 Buf_AddByte(&buf, p[1]);
1899 p += 2;
1900 continue;
1901 }
1902
1903 if (*p != '$') { /* Unescaped, simple text */
1904 if (subst != NULL && *p == '&')
1905 Buf_AddBytes(&buf, subst->lhs, subst->lhsLen);
1906 else
1907 Buf_AddByte(&buf, *p);
1908 p++;
1909 continue;
1910 }
1911
1912 if (p[1] == delim) { /* Unescaped $ at end of pattern */
1913 if (out_pflags != NULL)
1914 *out_pflags |= VARP_ANCHOR_END;
1915 else
1916 Buf_AddByte(&buf, *p);
1917 p++;
1918 continue;
1919 }
1920
1921 if (eflags & VARE_WANTRES) { /* Nested variable, evaluated */
1922 const char *nested_p = p;
1923 const char *nested_val;
1924 void *nested_val_freeIt;
1925 VarEvalFlags nested_eflags = eflags & ~(unsigned)VARE_KEEP_DOLLAR;
1926
1927 (void)Var_Parse(&nested_p, st->ctxt, nested_eflags,
1928 &nested_val, &nested_val_freeIt);
1929 /* TODO: handle errors */
1930 Buf_AddStr(&buf, nested_val);
1931 free(nested_val_freeIt);
1932 p += nested_p - p;
1933 continue;
1934 }
1935
1936 /* XXX: This whole block is very similar to Var_Parse without
1937 * VARE_WANTRES. There may be subtle edge cases though that are
1938 * not yet covered in the unit tests and that are parsed differently,
1939 * depending on whether they are evaluated or not.
1940 *
1941 * This subtle difference is not documented in the manual page,
1942 * neither is the difference between parsing :D and :M documented.
1943 * No code should ever depend on these details, but who knows. */
1944
1945 varstart = p; /* Nested variable, only parsed */
1946 if (p[1] == '(' || p[1] == '{') {
1947 /*
1948 * Find the end of this variable reference
1949 * and suck it in without further ado.
1950 * It will be interpreted later.
1951 */
1952 char startc = p[1];
1953 int endc = startc == '(' ? ')' : '}';
1954 int depth = 1;
1955
1956 for (p += 2; *p != '\0' && depth > 0; p++) {
1957 if (p[-1] != '\\') {
1958 if (*p == startc)
1959 depth++;
1960 if (*p == endc)
1961 depth--;
1962 }
1963 }
1964 Buf_AddBytesBetween(&buf, varstart, p);
1965 } else {
1966 Buf_AddByte(&buf, *varstart);
1967 p++;
1968 }
1969 }
1970
1971 if (*p != delim) {
1972 *pp = p;
1973 Error("Unfinished modifier for %s ('%c' missing)", st->v->name, delim);
1974 *out_part = NULL;
1975 return VPR_PARSE_MSG;
1976 }
1977
1978 *pp = ++p;
1979 if (out_length != NULL)
1980 *out_length = Buf_Len(&buf);
1981
1982 *out_part = Buf_Destroy(&buf, FALSE);
1983 VAR_DEBUG1("Modifier part: \"%s\"\n", *out_part);
1984 return VPR_OK;
1985 }
1986
1987 /* Test whether mod starts with modname, followed by a delimiter. */
1988 static Boolean
1989 ModMatch(const char *mod, const char *modname, char endc)
1990 {
1991 size_t n = strlen(modname);
1992 return strncmp(mod, modname, n) == 0 &&
1993 (mod[n] == endc || mod[n] == ':');
1994 }
1995
1996 /* Test whether mod starts with modname, followed by a delimiter or '='. */
1997 static inline Boolean
1998 ModMatchEq(const char *mod, const char *modname, char endc)
1999 {
2000 size_t n = strlen(modname);
2001 return strncmp(mod, modname, n) == 0 &&
2002 (mod[n] == endc || mod[n] == ':' || mod[n] == '=');
2003 }
2004
2005 static Boolean
2006 TryParseIntBase0(const char **pp, int *out_num)
2007 {
2008 char *end;
2009 long n;
2010
2011 errno = 0;
2012 n = strtol(*pp, &end, 0);
2013 if ((n == LONG_MIN || n == LONG_MAX) && errno == ERANGE)
2014 return FALSE;
2015 if (n < INT_MIN || n > INT_MAX)
2016 return FALSE;
2017
2018 *pp = end;
2019 *out_num = (int)n;
2020 return TRUE;
2021 }
2022
2023 static Boolean
2024 TryParseSize(const char **pp, size_t *out_num)
2025 {
2026 char *end;
2027 unsigned long n;
2028
2029 if (!ch_isdigit(**pp))
2030 return FALSE;
2031
2032 errno = 0;
2033 n = strtoul(*pp, &end, 10);
2034 if (n == ULONG_MAX && errno == ERANGE)
2035 return FALSE;
2036 if (n > SIZE_MAX)
2037 return FALSE;
2038
2039 *pp = end;
2040 *out_num = (size_t)n;
2041 return TRUE;
2042 }
2043
2044 static Boolean
2045 TryParseChar(const char **pp, int base, char *out_ch)
2046 {
2047 char *end;
2048 unsigned long n;
2049
2050 if (!ch_isalnum(**pp))
2051 return FALSE;
2052
2053 errno = 0;
2054 n = strtoul(*pp, &end, base);
2055 if (n == ULONG_MAX && errno == ERANGE)
2056 return FALSE;
2057 if (n > UCHAR_MAX)
2058 return FALSE;
2059
2060 *pp = end;
2061 *out_ch = (char)n;
2062 return TRUE;
2063 }
2064
2065 /* :@var (at) ...${var}...@ */
2066 static ApplyModifierResult
2067 ApplyModifier_Loop(const char **pp, ApplyModifiersState *st)
2068 {
2069 struct ModifyWord_LoopArgs args;
2070 char prev_sep;
2071 VarParseResult res;
2072
2073 args.ctx = st->ctxt;
2074
2075 (*pp)++; /* Skip the first '@' */
2076 res = ParseModifierPart(pp, '@', VARE_NONE, st,
2077 &args.tvar, NULL, NULL, NULL);
2078 if (res != VPR_OK)
2079 return AMR_CLEANUP;
2080 if (DEBUG(LINT) && strchr(args.tvar, '$') != NULL) {
2081 Parse_Error(PARSE_FATAL,
2082 "In the :@ modifier of \"%s\", the variable name \"%s\" "
2083 "must not contain a dollar.",
2084 st->v->name, args.tvar);
2085 return AMR_CLEANUP;
2086 }
2087
2088 res = ParseModifierPart(pp, '@', VARE_NONE, st,
2089 &args.str, NULL, NULL, NULL);
2090 if (res != VPR_OK)
2091 return AMR_CLEANUP;
2092
2093 args.eflags = st->eflags & ~(unsigned)VARE_KEEP_DOLLAR;
2094 prev_sep = st->sep;
2095 st->sep = ' '; /* XXX: should be st->sep for consistency */
2096 st->newVal = ModifyWords(st->val, ModifyWord_Loop, &args,
2097 st->oneBigWord, st->sep);
2098 st->sep = prev_sep;
2099 Var_Delete(args.tvar, st->ctxt);
2100 free(args.tvar);
2101 free(args.str);
2102 return AMR_OK;
2103 }
2104
2105 /* :Ddefined or :Uundefined */
2106 static ApplyModifierResult
2107 ApplyModifier_Defined(const char **pp, ApplyModifiersState *st)
2108 {
2109 Buffer buf;
2110 const char *p;
2111
2112 VarEvalFlags eflags = VARE_NONE;
2113 if (st->eflags & VARE_WANTRES)
2114 if ((**pp == 'D') == !(st->exprFlags & VEF_UNDEF))
2115 eflags = st->eflags;
2116
2117 Buf_Init(&buf);
2118 p = *pp + 1;
2119 while (*p != st->endc && *p != ':' && *p != '\0') {
2120
2121 /* Escaped delimiter or other special character */
2122 if (*p == '\\') {
2123 char c = p[1];
2124 if (c == st->endc || c == ':' || c == '$' || c == '\\') {
2125 Buf_AddByte(&buf, c);
2126 p += 2;
2127 continue;
2128 }
2129 }
2130
2131 /* Nested variable expression */
2132 if (*p == '$') {
2133 const char *nested_val;
2134 void *nested_val_freeIt;
2135
2136 (void)Var_Parse(&p, st->ctxt, eflags,
2137 &nested_val, &nested_val_freeIt);
2138 /* TODO: handle errors */
2139 Buf_AddStr(&buf, nested_val);
2140 free(nested_val_freeIt);
2141 continue;
2142 }
2143
2144 /* Ordinary text */
2145 Buf_AddByte(&buf, *p);
2146 p++;
2147 }
2148 *pp = p;
2149
2150 ApplyModifiersState_Define(st);
2151
2152 if (eflags & VARE_WANTRES) {
2153 st->newVal = Buf_Destroy(&buf, FALSE);
2154 } else {
2155 st->newVal = st->val;
2156 Buf_Destroy(&buf, TRUE);
2157 }
2158 return AMR_OK;
2159 }
2160
2161 /* :L */
2162 static ApplyModifierResult
2163 ApplyModifier_Literal(const char **pp, ApplyModifiersState *st)
2164 {
2165 ApplyModifiersState_Define(st);
2166 st->newVal = bmake_strdup(st->v->name);
2167 (*pp)++;
2168 return AMR_OK;
2169 }
2170
2171 static Boolean
2172 TryParseTime(const char **pp, time_t *out_time)
2173 {
2174 char *end;
2175 unsigned long n;
2176
2177 if (!ch_isdigit(**pp))
2178 return FALSE;
2179
2180 errno = 0;
2181 n = strtoul(*pp, &end, 10);
2182 if (n == ULONG_MAX && errno == ERANGE)
2183 return FALSE;
2184
2185 *pp = end;
2186 *out_time = (time_t)n; /* ignore possible truncation for now */
2187 return TRUE;
2188 }
2189
2190 /* :gmtime */
2191 static ApplyModifierResult
2192 ApplyModifier_Gmtime(const char **pp, ApplyModifiersState *st)
2193 {
2194 time_t utc;
2195
2196 const char *mod = *pp;
2197 if (!ModMatchEq(mod, "gmtime", st->endc))
2198 return AMR_UNKNOWN;
2199
2200 if (mod[6] == '=') {
2201 const char *arg = mod + 7;
2202 if (!TryParseTime(&arg, &utc)) {
2203 Parse_Error(PARSE_FATAL, "Invalid time value: %s\n", mod + 7);
2204 return AMR_CLEANUP;
2205 }
2206 *pp = arg;
2207 } else {
2208 utc = 0;
2209 *pp = mod + 6;
2210 }
2211 st->newVal = VarStrftime(st->val, TRUE, utc);
2212 return AMR_OK;
2213 }
2214
2215 /* :localtime */
2216 static ApplyModifierResult
2217 ApplyModifier_Localtime(const char **pp, ApplyModifiersState *st)
2218 {
2219 time_t utc;
2220
2221 const char *mod = *pp;
2222 if (!ModMatchEq(mod, "localtime", st->endc))
2223 return AMR_UNKNOWN;
2224
2225 if (mod[9] == '=') {
2226 const char *arg = mod + 10;
2227 if (!TryParseTime(&arg, &utc)) {
2228 Parse_Error(PARSE_FATAL, "Invalid time value: %s\n", mod + 10);
2229 return AMR_CLEANUP;
2230 }
2231 *pp = arg;
2232 } else {
2233 utc = 0;
2234 *pp = mod + 9;
2235 }
2236 st->newVal = VarStrftime(st->val, FALSE, utc);
2237 return AMR_OK;
2238 }
2239
2240 /* :hash */
2241 static ApplyModifierResult
2242 ApplyModifier_Hash(const char **pp, ApplyModifiersState *st)
2243 {
2244 if (!ModMatch(*pp, "hash", st->endc))
2245 return AMR_UNKNOWN;
2246
2247 st->newVal = VarHash(st->val);
2248 *pp += 4;
2249 return AMR_OK;
2250 }
2251
2252 /* :P */
2253 static ApplyModifierResult
2254 ApplyModifier_Path(const char **pp, ApplyModifiersState *st)
2255 {
2256 GNode *gn;
2257 char *path;
2258
2259 ApplyModifiersState_Define(st);
2260
2261 gn = Targ_FindNode(st->v->name);
2262 if (gn == NULL || gn->type & OP_NOPATH) {
2263 path = NULL;
2264 } else if (gn->path != NULL) {
2265 path = bmake_strdup(gn->path);
2266 } else {
2267 SearchPath *searchPath = Suff_FindPath(gn);
2268 path = Dir_FindFile(st->v->name, searchPath);
2269 }
2270 if (path == NULL)
2271 path = bmake_strdup(st->v->name);
2272 st->newVal = path;
2273
2274 (*pp)++;
2275 return AMR_OK;
2276 }
2277
2278 /* :!cmd! */
2279 static ApplyModifierResult
2280 ApplyModifier_ShellCommand(const char **pp, ApplyModifiersState *st)
2281 {
2282 char *cmd;
2283 const char *errfmt;
2284 VarParseResult res;
2285
2286 (*pp)++;
2287 res = ParseModifierPart(pp, '!', st->eflags, st,
2288 &cmd, NULL, NULL, NULL);
2289 if (res != VPR_OK)
2290 return AMR_CLEANUP;
2291
2292 errfmt = NULL;
2293 if (st->eflags & VARE_WANTRES)
2294 st->newVal = Cmd_Exec(cmd, &errfmt);
2295 else
2296 st->newVal = bmake_strdup("");
2297 free(cmd);
2298
2299 if (errfmt != NULL)
2300 Error(errfmt, st->val); /* XXX: why still return AMR_OK? */
2301
2302 ApplyModifiersState_Define(st);
2303 return AMR_OK;
2304 }
2305
2306 /* The :range modifier generates an integer sequence as long as the words.
2307 * The :range=7 modifier generates an integer sequence from 1 to 7. */
2308 static ApplyModifierResult
2309 ApplyModifier_Range(const char **pp, ApplyModifiersState *st)
2310 {
2311 size_t n;
2312 Buffer buf;
2313 size_t i;
2314
2315 const char *mod = *pp;
2316 if (!ModMatchEq(mod, "range", st->endc))
2317 return AMR_UNKNOWN;
2318
2319 if (mod[5] == '=') {
2320 const char *p = mod + 6;
2321 if (!TryParseSize(&p, &n)) {
2322 Parse_Error(PARSE_FATAL, "Invalid number: %s\n", mod + 6);
2323 return AMR_CLEANUP;
2324 }
2325 *pp = p;
2326 } else {
2327 n = 0;
2328 *pp = mod + 5;
2329 }
2330
2331 if (n == 0) {
2332 Words words = Str_Words(st->val, FALSE);
2333 n = words.len;
2334 Words_Free(words);
2335 }
2336
2337 Buf_Init(&buf);
2338
2339 for (i = 0; i < n; i++) {
2340 if (i != 0)
2341 Buf_AddByte(&buf, ' '); /* XXX: st->sep, for consistency */
2342 Buf_AddInt(&buf, 1 + (int)i);
2343 }
2344
2345 st->newVal = Buf_Destroy(&buf, FALSE);
2346 return AMR_OK;
2347 }
2348
2349 /* :Mpattern or :Npattern */
2350 static ApplyModifierResult
2351 ApplyModifier_Match(const char **pp, ApplyModifiersState *st)
2352 {
2353 const char *mod = *pp;
2354 Boolean copy = FALSE; /* pattern should be, or has been, copied */
2355 Boolean needSubst = FALSE;
2356 const char *endpat;
2357 char *pattern;
2358 ModifyWordsCallback callback;
2359
2360 /*
2361 * In the loop below, ignore ':' unless we are at (or back to) the
2362 * original brace level.
2363 * XXX This will likely not work right if $() and ${} are intermixed.
2364 */
2365 int nest = 0;
2366 const char *p;
2367 for (p = mod + 1; *p != '\0' && !(*p == ':' && nest == 0); p++) {
2368 if (*p == '\\' &&
2369 (p[1] == ':' || p[1] == st->endc || p[1] == st->startc)) {
2370 if (!needSubst)
2371 copy = TRUE;
2372 p++;
2373 continue;
2374 }
2375 if (*p == '$')
2376 needSubst = TRUE;
2377 if (*p == '(' || *p == '{')
2378 nest++;
2379 if (*p == ')' || *p == '}') {
2380 nest--;
2381 if (nest < 0)
2382 break;
2383 }
2384 }
2385 *pp = p;
2386 endpat = p;
2387
2388 if (copy) {
2389 char *dst;
2390 const char *src;
2391
2392 /* Compress the \:'s out of the pattern. */
2393 pattern = bmake_malloc((size_t)(endpat - (mod + 1)) + 1);
2394 dst = pattern;
2395 src = mod + 1;
2396 for (; src < endpat; src++, dst++) {
2397 if (src[0] == '\\' && src + 1 < endpat &&
2398 /* XXX: st->startc is missing here; see above */
2399 (src[1] == ':' || src[1] == st->endc))
2400 src++;
2401 *dst = *src;
2402 }
2403 *dst = '\0';
2404 endpat = dst;
2405 } else {
2406 pattern = bmake_strsedup(mod + 1, endpat);
2407 }
2408
2409 if (needSubst) {
2410 /* pattern contains embedded '$', so use Var_Subst to expand it. */
2411 char *old_pattern = pattern;
2412 (void)Var_Subst(pattern, st->ctxt, st->eflags, &pattern);
2413 /* TODO: handle errors */
2414 free(old_pattern);
2415 }
2416
2417 VAR_DEBUG3("Pattern[%s] for [%s] is [%s]\n", st->v->name, st->val, pattern);
2418
2419 callback = mod[0] == 'M' ? ModifyWord_Match : ModifyWord_NoMatch;
2420 st->newVal = ModifyWords(st->val, callback, pattern,
2421 st->oneBigWord, st->sep);
2422 free(pattern);
2423 return AMR_OK;
2424 }
2425
2426 /* :S,from,to, */
2427 static ApplyModifierResult
2428 ApplyModifier_Subst(const char **pp, ApplyModifiersState *st)
2429 {
2430 struct ModifyWord_SubstArgs args;
2431 char *lhs, *rhs;
2432 Boolean oneBigWord;
2433 VarParseResult res;
2434
2435 char delim = (*pp)[1];
2436 if (delim == '\0') {
2437 Error("Missing delimiter for :S modifier");
2438 (*pp)++;
2439 return AMR_CLEANUP;
2440 }
2441
2442 *pp += 2;
2443
2444 args.pflags = 0;
2445 args.matched = FALSE;
2446
2447 /*
2448 * If pattern begins with '^', it is anchored to the
2449 * start of the word -- skip over it and flag pattern.
2450 */
2451 if (**pp == '^') {
2452 args.pflags |= VARP_ANCHOR_START;
2453 (*pp)++;
2454 }
2455
2456 res = ParseModifierPart(pp, delim, st->eflags, st,
2457 &lhs, &args.lhsLen, &args.pflags, NULL);
2458 if (res != VPR_OK)
2459 return AMR_CLEANUP;
2460 args.lhs = lhs;
2461
2462 res = ParseModifierPart(pp, delim, st->eflags, st,
2463 &rhs, &args.rhsLen, NULL, &args);
2464 if (res != VPR_OK)
2465 return AMR_CLEANUP;
2466 args.rhs = rhs;
2467
2468 oneBigWord = st->oneBigWord;
2469 for (;; (*pp)++) {
2470 switch (**pp) {
2471 case 'g':
2472 args.pflags |= VARP_SUB_GLOBAL;
2473 continue;
2474 case '1':
2475 args.pflags |= VARP_SUB_ONE;
2476 continue;
2477 case 'W':
2478 oneBigWord = TRUE;
2479 continue;
2480 }
2481 break;
2482 }
2483
2484 st->newVal = ModifyWords(st->val, ModifyWord_Subst, &args,
2485 oneBigWord, st->sep);
2486
2487 free(lhs);
2488 free(rhs);
2489 return AMR_OK;
2490 }
2491
2492 #ifndef NO_REGEX
2493
2494 /* :C,from,to, */
2495 static ApplyModifierResult
2496 ApplyModifier_Regex(const char **pp, ApplyModifiersState *st)
2497 {
2498 char *re;
2499 struct ModifyWord_SubstRegexArgs args;
2500 Boolean oneBigWord;
2501 int error;
2502 VarParseResult res;
2503
2504 char delim = (*pp)[1];
2505 if (delim == '\0') {
2506 Error("Missing delimiter for :C modifier");
2507 (*pp)++;
2508 return AMR_CLEANUP;
2509 }
2510
2511 *pp += 2;
2512
2513 res = ParseModifierPart(pp, delim, st->eflags, st,
2514 &re, NULL, NULL, NULL);
2515 if (res != VPR_OK)
2516 return AMR_CLEANUP;
2517
2518 res = ParseModifierPart(pp, delim, st->eflags, st,
2519 &args.replace, NULL, NULL, NULL);
2520 if (args.replace == NULL) {
2521 free(re);
2522 return AMR_CLEANUP;
2523 }
2524
2525 args.pflags = 0;
2526 args.matched = FALSE;
2527 oneBigWord = st->oneBigWord;
2528 for (;; (*pp)++) {
2529 switch (**pp) {
2530 case 'g':
2531 args.pflags |= VARP_SUB_GLOBAL;
2532 continue;
2533 case '1':
2534 args.pflags |= VARP_SUB_ONE;
2535 continue;
2536 case 'W':
2537 oneBigWord = TRUE;
2538 continue;
2539 }
2540 break;
2541 }
2542
2543 error = regcomp(&args.re, re, REG_EXTENDED);
2544 free(re);
2545 if (error) {
2546 VarREError(error, &args.re, "Regex compilation error");
2547 free(args.replace);
2548 return AMR_CLEANUP;
2549 }
2550
2551 args.nsub = args.re.re_nsub + 1;
2552 if (args.nsub > 10)
2553 args.nsub = 10;
2554 st->newVal = ModifyWords(st->val, ModifyWord_SubstRegex, &args,
2555 oneBigWord, st->sep);
2556 regfree(&args.re);
2557 free(args.replace);
2558 return AMR_OK;
2559 }
2560 #endif
2561
2562 /* :Q, :q */
2563 static ApplyModifierResult
2564 ApplyModifier_Quote(const char **pp, ApplyModifiersState *st)
2565 {
2566 if ((*pp)[1] == st->endc || (*pp)[1] == ':') {
2567 st->newVal = VarQuote(st->val, **pp == 'q');
2568 (*pp)++;
2569 return AMR_OK;
2570 } else
2571 return AMR_UNKNOWN;
2572 }
2573
2574 static void
2575 ModifyWord_Copy(const char *word, SepBuf *buf, void *data MAKE_ATTR_UNUSED)
2576 {
2577 SepBuf_AddStr(buf, word);
2578 }
2579
2580 /* :ts<separator> */
2581 static ApplyModifierResult
2582 ApplyModifier_ToSep(const char **pp, ApplyModifiersState *st)
2583 {
2584 const char *sep = *pp + 2;
2585
2586 /* ":ts<any><endc>" or ":ts<any>:" */
2587 if (sep[0] != st->endc && (sep[1] == st->endc || sep[1] == ':')) {
2588 st->sep = sep[0];
2589 *pp = sep + 1;
2590 goto ok;
2591 }
2592
2593 /* ":ts<endc>" or ":ts:" */
2594 if (sep[0] == st->endc || sep[0] == ':') {
2595 st->sep = '\0'; /* no separator */
2596 *pp = sep;
2597 goto ok;
2598 }
2599
2600 /* ":ts<unrecognised><unrecognised>". */
2601 if (sep[0] != '\\') {
2602 (*pp)++; /* just for backwards compatibility */
2603 return AMR_BAD;
2604 }
2605
2606 /* ":ts\n" */
2607 if (sep[1] == 'n') {
2608 st->sep = '\n';
2609 *pp = sep + 2;
2610 goto ok;
2611 }
2612
2613 /* ":ts\t" */
2614 if (sep[1] == 't') {
2615 st->sep = '\t';
2616 *pp = sep + 2;
2617 goto ok;
2618 }
2619
2620 /* ":ts\x40" or ":ts\100" */
2621 {
2622 const char *p = sep + 1;
2623 int base = 8; /* assume octal */
2624
2625 if (sep[1] == 'x') {
2626 base = 16;
2627 p++;
2628 } else if (!ch_isdigit(sep[1])) {
2629 (*pp)++; /* just for backwards compatibility */
2630 return AMR_BAD; /* ":ts<backslash><unrecognised>". */
2631 }
2632
2633 if (!TryParseChar(&p, base, &st->sep)) {
2634 Parse_Error(PARSE_FATAL, "Invalid character number: %s\n", p);
2635 return AMR_CLEANUP;
2636 }
2637 if (*p != ':' && *p != st->endc) {
2638 (*pp)++; /* just for backwards compatibility */
2639 return AMR_BAD;
2640 }
2641
2642 *pp = p;
2643 }
2644
2645 ok:
2646 st->newVal = ModifyWords(st->val, ModifyWord_Copy, NULL,
2647 st->oneBigWord, st->sep);
2648 return AMR_OK;
2649 }
2650
2651 /* :tA, :tu, :tl, :ts<separator>, etc. */
2652 static ApplyModifierResult
2653 ApplyModifier_To(const char **pp, ApplyModifiersState *st)
2654 {
2655 const char *mod = *pp;
2656 assert(mod[0] == 't');
2657
2658 if (mod[1] == st->endc || mod[1] == ':' || mod[1] == '\0') {
2659 *pp = mod + 1;
2660 return AMR_BAD; /* Found ":t<endc>" or ":t:". */
2661 }
2662
2663 if (mod[1] == 's')
2664 return ApplyModifier_ToSep(pp, st);
2665
2666 if (mod[2] != st->endc && mod[2] != ':') {
2667 *pp = mod + 1;
2668 return AMR_BAD; /* Found ":t<unrecognised><unrecognised>". */
2669 }
2670
2671 /* Check for two-character options: ":tu", ":tl" */
2672 if (mod[1] == 'A') { /* absolute path */
2673 st->newVal = ModifyWords(st->val, ModifyWord_Realpath, NULL,
2674 st->oneBigWord, st->sep);
2675 *pp = mod + 2;
2676 return AMR_OK;
2677 }
2678
2679 if (mod[1] == 'u') { /* :tu */
2680 size_t i;
2681 size_t len = strlen(st->val);
2682 st->newVal = bmake_malloc(len + 1);
2683 for (i = 0; i < len + 1; i++)
2684 st->newVal[i] = ch_toupper(st->val[i]);
2685 *pp = mod + 2;
2686 return AMR_OK;
2687 }
2688
2689 if (mod[1] == 'l') { /* :tl */
2690 size_t i;
2691 size_t len = strlen(st->val);
2692 st->newVal = bmake_malloc(len + 1);
2693 for (i = 0; i < len + 1; i++)
2694 st->newVal[i] = ch_tolower(st->val[i]);
2695 *pp = mod + 2;
2696 return AMR_OK;
2697 }
2698
2699 if (mod[1] == 'W' || mod[1] == 'w') { /* :tW, :tw */
2700 st->oneBigWord = mod[1] == 'W';
2701 st->newVal = st->val;
2702 *pp = mod + 2;
2703 return AMR_OK;
2704 }
2705
2706 /* Found ":t<unrecognised>:" or ":t<unrecognised><endc>". */
2707 *pp = mod + 1;
2708 return AMR_BAD;
2709 }
2710
2711 /* :[#], :[1], :[-1..1], etc. */
2712 static ApplyModifierResult
2713 ApplyModifier_Words(const char **pp, ApplyModifiersState *st)
2714 {
2715 char *estr;
2716 int first, last;
2717 VarParseResult res;
2718 const char *p;
2719
2720 (*pp)++; /* skip the '[' */
2721 res = ParseModifierPart(pp, ']', st->eflags, st,
2722 &estr, NULL, NULL, NULL);
2723 if (res != VPR_OK)
2724 return AMR_CLEANUP;
2725
2726 /* now *pp points just after the closing ']' */
2727 if (**pp != ':' && **pp != st->endc)
2728 goto bad_modifier; /* Found junk after ']' */
2729
2730 if (estr[0] == '\0')
2731 goto bad_modifier; /* empty square brackets in ":[]". */
2732
2733 if (estr[0] == '#' && estr[1] == '\0') { /* Found ":[#]" */
2734 if (st->oneBigWord) {
2735 st->newVal = bmake_strdup("1");
2736 } else {
2737 Buffer buf;
2738
2739 Words words = Str_Words(st->val, FALSE);
2740 size_t ac = words.len;
2741 Words_Free(words);
2742
2743 Buf_InitSize(&buf, 4); /* 3 digits + '\0' is usually enough */
2744 Buf_AddInt(&buf, (int)ac);
2745 st->newVal = Buf_Destroy(&buf, FALSE);
2746 }
2747 goto ok;
2748 }
2749
2750 if (estr[0] == '*' && estr[1] == '\0') {
2751 /* Found ":[*]" */
2752 st->oneBigWord = TRUE;
2753 st->newVal = st->val;
2754 goto ok;
2755 }
2756
2757 if (estr[0] == '@' && estr[1] == '\0') {
2758 /* Found ":[@]" */
2759 st->oneBigWord = FALSE;
2760 st->newVal = st->val;
2761 goto ok;
2762 }
2763
2764 /*
2765 * We expect estr to contain a single integer for :[N], or two integers
2766 * separated by ".." for :[start..end].
2767 */
2768 p = estr;
2769 if (!TryParseIntBase0(&p, &first))
2770 goto bad_modifier; /* Found junk instead of a number */
2771
2772 if (p[0] == '\0') { /* Found only one integer in :[N] */
2773 last = first;
2774 } else if (p[0] == '.' && p[1] == '.' && p[2] != '\0') {
2775 /* Expecting another integer after ".." */
2776 p += 2;
2777 if (!TryParseIntBase0(&p, &last) || *p != '\0')
2778 goto bad_modifier; /* Found junk after ".." */
2779 } else
2780 goto bad_modifier; /* Found junk instead of ".." */
2781
2782 /*
2783 * Now first and last are properly filled in, but we still have to check
2784 * for 0 as a special case.
2785 */
2786 if (first == 0 && last == 0) {
2787 /* ":[0]" or perhaps ":[0..0]" */
2788 st->oneBigWord = TRUE;
2789 st->newVal = st->val;
2790 goto ok;
2791 }
2792
2793 /* ":[0..N]" or ":[N..0]" */
2794 if (first == 0 || last == 0)
2795 goto bad_modifier;
2796
2797 /* Normal case: select the words described by first and last. */
2798 st->newVal = VarSelectWords(st->sep, st->oneBigWord, st->val, first, last);
2799
2800 ok:
2801 free(estr);
2802 return AMR_OK;
2803
2804 bad_modifier:
2805 free(estr);
2806 return AMR_BAD;
2807 }
2808
2809 static int
2810 str_cmp_asc(const void *a, const void *b)
2811 {
2812 return strcmp(*(const char * const *)a, *(const char * const *)b);
2813 }
2814
2815 static int
2816 str_cmp_desc(const void *a, const void *b)
2817 {
2818 return strcmp(*(const char * const *)b, *(const char * const *)a);
2819 }
2820
2821 /* :O (order ascending) or :Or (order descending) or :Ox (shuffle) */
2822 static ApplyModifierResult
2823 ApplyModifier_Order(const char **pp, ApplyModifiersState *st)
2824 {
2825 const char *mod = (*pp)++; /* skip past the 'O' in any case */
2826
2827 Words words = Str_Words(st->val, FALSE);
2828
2829 if (mod[1] == st->endc || mod[1] == ':') {
2830 /* :O sorts ascending */
2831 qsort(words.words, words.len, sizeof words.words[0], str_cmp_asc);
2832
2833 } else if ((mod[1] == 'r' || mod[1] == 'x') &&
2834 (mod[2] == st->endc || mod[2] == ':')) {
2835 (*pp)++;
2836
2837 if (mod[1] == 'r') {
2838 /* :Or sorts descending */
2839 qsort(words.words, words.len, sizeof words.words[0], str_cmp_desc);
2840
2841 } else {
2842 /* :Ox shuffles
2843 *
2844 * We will use [ac..2] range for mod factors. This will produce
2845 * random numbers in [(ac-1)..0] interval, and minimal
2846 * reasonable value for mod factor is 2 (the mod 1 will produce
2847 * 0 with probability 1).
2848 */
2849 size_t i;
2850 for (i = words.len - 1; i > 0; i--) {
2851 size_t rndidx = (size_t)random() % (i + 1);
2852 char *t = words.words[i];
2853 words.words[i] = words.words[rndidx];
2854 words.words[rndidx] = t;
2855 }
2856 }
2857 } else {
2858 Words_Free(words);
2859 return AMR_BAD;
2860 }
2861
2862 st->newVal = Words_JoinFree(words);
2863 return AMR_OK;
2864 }
2865
2866 /* :? then : else */
2867 static ApplyModifierResult
2868 ApplyModifier_IfElse(const char **pp, ApplyModifiersState *st)
2869 {
2870 char *then_expr, *else_expr;
2871 VarParseResult res;
2872
2873 Boolean value = FALSE;
2874 VarEvalFlags then_eflags = VARE_NONE;
2875 VarEvalFlags else_eflags = VARE_NONE;
2876
2877 int cond_rc = COND_PARSE; /* anything other than COND_INVALID */
2878 if (st->eflags & VARE_WANTRES) {
2879 cond_rc = Cond_EvalCondition(st->v->name, &value);
2880 if (cond_rc != COND_INVALID && value)
2881 then_eflags = st->eflags;
2882 if (cond_rc != COND_INVALID && !value)
2883 else_eflags = st->eflags;
2884 }
2885
2886 (*pp)++; /* skip past the '?' */
2887 res = ParseModifierPart(pp, ':', then_eflags, st,
2888 &then_expr, NULL, NULL, NULL);
2889 if (res != VPR_OK)
2890 return AMR_CLEANUP;
2891
2892 res = ParseModifierPart(pp, st->endc, else_eflags, st,
2893 &else_expr, NULL, NULL, NULL);
2894 if (res != VPR_OK)
2895 return AMR_CLEANUP;
2896
2897 (*pp)--;
2898 if (cond_rc == COND_INVALID) {
2899 Error("Bad conditional expression `%s' in %s?%s:%s",
2900 st->v->name, st->v->name, then_expr, else_expr);
2901 return AMR_CLEANUP;
2902 }
2903
2904 if (value) {
2905 st->newVal = then_expr;
2906 free(else_expr);
2907 } else {
2908 st->newVal = else_expr;
2909 free(then_expr);
2910 }
2911 ApplyModifiersState_Define(st);
2912 return AMR_OK;
2913 }
2914
2915 /*
2916 * The ::= modifiers actually assign a value to the variable.
2917 * Their main purpose is in supporting modifiers of .for loop
2918 * iterators and other obscure uses. They always expand to
2919 * nothing. In a target rule that would otherwise expand to an
2920 * empty line they can be preceded with @: to keep make happy.
2921 * Eg.
2922 *
2923 * foo: .USE
2924 * .for i in ${.TARGET} ${.TARGET:R}.gz
2925 * @: ${t::=$i}
2926 * @echo blah ${t:T}
2927 * .endfor
2928 *
2929 * ::=<str> Assigns <str> as the new value of variable.
2930 * ::?=<str> Assigns <str> as value of variable if
2931 * it was not already set.
2932 * ::+=<str> Appends <str> to variable.
2933 * ::!=<cmd> Assigns output of <cmd> as the new value of
2934 * variable.
2935 */
2936 static ApplyModifierResult
2937 ApplyModifier_Assign(const char **pp, ApplyModifiersState *st)
2938 {
2939 GNode *v_ctxt;
2940 char delim;
2941 char *val;
2942 VarParseResult res;
2943
2944 const char *mod = *pp;
2945 const char *op = mod + 1;
2946
2947 if (op[0] == '=')
2948 goto ok;
2949 if ((op[0] == '!' || op[0] == '+' || op[0] == '?') && op[1] == '=')
2950 goto ok;
2951 return AMR_UNKNOWN; /* "::<unrecognised>" */
2952 ok:
2953
2954 if (st->v->name[0] == '\0') {
2955 *pp = mod + 1;
2956 return AMR_BAD;
2957 }
2958
2959 v_ctxt = st->ctxt; /* context where v belongs */
2960 if (!(st->exprFlags & VEF_UNDEF) && st->ctxt != VAR_GLOBAL) {
2961 Var *gv = VarFind(st->v->name, st->ctxt, FALSE);
2962 if (gv == NULL)
2963 v_ctxt = VAR_GLOBAL;
2964 else
2965 VarFreeEnv(gv, TRUE);
2966 }
2967
2968 switch (op[0]) {
2969 case '+':
2970 case '?':
2971 case '!':
2972 *pp = mod + 3;
2973 break;
2974 default:
2975 *pp = mod + 2;
2976 break;
2977 }
2978
2979 delim = st->startc == '(' ? ')' : '}';
2980 /* TODO: Add test for using the ::= modifier in a := assignment line.
2981 * Probably st->eflags should be passed down without VARE_KEEP_DOLLAR
2982 * here. */
2983 res = ParseModifierPart(pp, delim, st->eflags, st, &val, NULL, NULL, NULL);
2984 if (res != VPR_OK)
2985 return AMR_CLEANUP;
2986
2987 (*pp)--;
2988
2989 if (st->eflags & VARE_WANTRES) {
2990 switch (op[0]) {
2991 case '+':
2992 Var_Append(st->v->name, val, v_ctxt);
2993 break;
2994 case '!': {
2995 const char *errfmt;
2996 char *cmd_output = Cmd_Exec(val, &errfmt);
2997 if (errfmt)
2998 Error(errfmt, val);
2999 else
3000 Var_Set(st->v->name, cmd_output, v_ctxt);
3001 free(cmd_output);
3002 break;
3003 }
3004 case '?':
3005 if (!(st->exprFlags & VEF_UNDEF))
3006 break;
3007 /* FALLTHROUGH */
3008 default:
3009 Var_Set(st->v->name, val, v_ctxt);
3010 break;
3011 }
3012 }
3013 free(val);
3014 st->newVal = bmake_strdup("");
3015 return AMR_OK;
3016 }
3017
3018 /* :_=...
3019 * remember current value */
3020 static ApplyModifierResult
3021 ApplyModifier_Remember(const char **pp, ApplyModifiersState *st)
3022 {
3023 const char *mod = *pp;
3024 if (!ModMatchEq(mod, "_", st->endc))
3025 return AMR_UNKNOWN;
3026
3027 if (mod[1] == '=') {
3028 size_t n = strcspn(mod + 2, ":)}");
3029 char *name = bmake_strldup(mod + 2, n);
3030 Var_Set(name, st->val, st->ctxt);
3031 free(name);
3032 *pp = mod + 2 + n;
3033 } else {
3034 Var_Set("_", st->val, st->ctxt);
3035 *pp = mod + 1;
3036 }
3037 st->newVal = st->val;
3038 return AMR_OK;
3039 }
3040
3041 /* Apply the given function to each word of the variable value,
3042 * for a single-letter modifier such as :H, :T. */
3043 static ApplyModifierResult
3044 ApplyModifier_WordFunc(const char **pp, ApplyModifiersState *st,
3045 ModifyWordsCallback modifyWord)
3046 {
3047 char delim = (*pp)[1];
3048 if (delim != st->endc && delim != ':')
3049 return AMR_UNKNOWN;
3050
3051 st->newVal = ModifyWords(st->val, modifyWord, NULL,
3052 st->oneBigWord, st->sep);
3053 (*pp)++;
3054 return AMR_OK;
3055 }
3056
3057 static ApplyModifierResult
3058 ApplyModifier_Unique(const char **pp, ApplyModifiersState *st)
3059 {
3060 if ((*pp)[1] == st->endc || (*pp)[1] == ':') {
3061 st->newVal = VarUniq(st->val);
3062 (*pp)++;
3063 return AMR_OK;
3064 } else
3065 return AMR_UNKNOWN;
3066 }
3067
3068 #ifdef SYSVVARSUB
3069 /* :from=to */
3070 static ApplyModifierResult
3071 ApplyModifier_SysV(const char **pp, ApplyModifiersState *st)
3072 {
3073 char *lhs, *rhs;
3074 VarParseResult res;
3075
3076 const char *mod = *pp;
3077 Boolean eqFound = FALSE;
3078
3079 /*
3080 * First we make a pass through the string trying to verify it is a
3081 * SysV-make-style translation. It must be: <lhs>=<rhs>
3082 */
3083 int depth = 1;
3084 const char *p = mod;
3085 while (*p != '\0' && depth > 0) {
3086 if (*p == '=') { /* XXX: should also test depth == 1 */
3087 eqFound = TRUE;
3088 /* continue looking for st->endc */
3089 } else if (*p == st->endc)
3090 depth--;
3091 else if (*p == st->startc)
3092 depth++;
3093 if (depth > 0)
3094 p++;
3095 }
3096 if (*p != st->endc || !eqFound)
3097 return AMR_UNKNOWN;
3098
3099 *pp = mod;
3100 res = ParseModifierPart(pp, '=', st->eflags, st,
3101 &lhs, NULL, NULL, NULL);
3102 if (res != VPR_OK)
3103 return AMR_CLEANUP;
3104
3105 /* The SysV modifier lasts until the end of the variable expression. */
3106 res = ParseModifierPart(pp, st->endc, st->eflags, st,
3107 &rhs, NULL, NULL, NULL);
3108 if (res != VPR_OK)
3109 return AMR_CLEANUP;
3110
3111 (*pp)--;
3112 if (lhs[0] == '\0' && st->val[0] == '\0') {
3113 st->newVal = st->val; /* special case */
3114 } else {
3115 struct ModifyWord_SYSVSubstArgs args = {st->ctxt, lhs, rhs};
3116 st->newVal = ModifyWords(st->val, ModifyWord_SYSVSubst, &args,
3117 st->oneBigWord, st->sep);
3118 }
3119 free(lhs);
3120 free(rhs);
3121 return AMR_OK;
3122 }
3123 #endif
3124
3125 #ifdef SUNSHCMD
3126 /* :sh */
3127 static ApplyModifierResult
3128 ApplyModifier_SunShell(const char **pp, ApplyModifiersState *st)
3129 {
3130 const char *p = *pp;
3131 if (p[1] == 'h' && (p[2] == st->endc || p[2] == ':')) {
3132 if (st->eflags & VARE_WANTRES) {
3133 const char *errfmt;
3134 st->newVal = Cmd_Exec(st->val, &errfmt);
3135 if (errfmt)
3136 Error(errfmt, st->val);
3137 } else
3138 st->newVal = bmake_strdup("");
3139 *pp = p + 2;
3140 return AMR_OK;
3141 } else
3142 return AMR_UNKNOWN;
3143 }
3144 #endif
3145
3146 static void
3147 LogBeforeApply(const ApplyModifiersState *st, const char *mod, const char endc)
3148 {
3149 char eflags_str[VarEvalFlags_ToStringSize];
3150 char vflags_str[VarFlags_ToStringSize];
3151 char exprflags_str[VarExprFlags_ToStringSize];
3152 Boolean is_single_char = mod[0] != '\0' &&
3153 (mod[1] == endc || mod[1] == ':');
3154
3155 /* At this point, only the first character of the modifier can
3156 * be used since the end of the modifier is not yet known. */
3157 debug_printf("Applying ${%s:%c%s} to \"%s\" (%s, %s, %s)\n",
3158 st->v->name, mod[0], is_single_char ? "" : "...", st->val,
3159 Enum_FlagsToString(eflags_str, sizeof eflags_str,
3160 st->eflags, VarEvalFlags_ToStringSpecs),
3161 Enum_FlagsToString(vflags_str, sizeof vflags_str,
3162 st->v->flags, VarFlags_ToStringSpecs),
3163 Enum_FlagsToString(exprflags_str, sizeof exprflags_str,
3164 st->exprFlags,
3165 VarExprFlags_ToStringSpecs));
3166 }
3167
3168 static void
3169 LogAfterApply(ApplyModifiersState *st, const char *p, const char *mod)
3170 {
3171 char eflags_str[VarEvalFlags_ToStringSize];
3172 char vflags_str[VarFlags_ToStringSize];
3173 char exprflags_str[VarExprFlags_ToStringSize];
3174 const char *quot = st->newVal == var_Error ? "" : "\"";
3175 const char *newVal = st->newVal == var_Error ? "error" : st->newVal;
3176
3177 debug_printf("Result of ${%s:%.*s} is %s%s%s (%s, %s, %s)\n",
3178 st->v->name, (int)(p - mod), mod, quot, newVal, quot,
3179 Enum_FlagsToString(eflags_str, sizeof eflags_str,
3180 st->eflags, VarEvalFlags_ToStringSpecs),
3181 Enum_FlagsToString(vflags_str, sizeof vflags_str,
3182 st->v->flags, VarFlags_ToStringSpecs),
3183 Enum_FlagsToString(exprflags_str, sizeof exprflags_str,
3184 st->exprFlags,
3185 VarExprFlags_ToStringSpecs));
3186 }
3187
3188 static ApplyModifierResult
3189 ApplyModifier(const char **pp, ApplyModifiersState *st)
3190 {
3191 switch (**pp) {
3192 case ':':
3193 return ApplyModifier_Assign(pp, st);
3194 case '@':
3195 return ApplyModifier_Loop(pp, st);
3196 case '_':
3197 return ApplyModifier_Remember(pp, st);
3198 case 'D':
3199 case 'U':
3200 return ApplyModifier_Defined(pp, st);
3201 case 'L':
3202 return ApplyModifier_Literal(pp, st);
3203 case 'P':
3204 return ApplyModifier_Path(pp, st);
3205 case '!':
3206 return ApplyModifier_ShellCommand(pp, st);
3207 case '[':
3208 return ApplyModifier_Words(pp, st);
3209 case 'g':
3210 return ApplyModifier_Gmtime(pp, st);
3211 case 'h':
3212 return ApplyModifier_Hash(pp, st);
3213 case 'l':
3214 return ApplyModifier_Localtime(pp, st);
3215 case 't':
3216 return ApplyModifier_To(pp, st);
3217 case 'N':
3218 case 'M':
3219 return ApplyModifier_Match(pp, st);
3220 case 'S':
3221 return ApplyModifier_Subst(pp, st);
3222 case '?':
3223 return ApplyModifier_IfElse(pp, st);
3224 #ifndef NO_REGEX
3225 case 'C':
3226 return ApplyModifier_Regex(pp, st);
3227 #endif
3228 case 'q':
3229 case 'Q':
3230 return ApplyModifier_Quote(pp, st);
3231 case 'T':
3232 return ApplyModifier_WordFunc(pp, st, ModifyWord_Tail);
3233 case 'H':
3234 return ApplyModifier_WordFunc(pp, st, ModifyWord_Head);
3235 case 'E':
3236 return ApplyModifier_WordFunc(pp, st, ModifyWord_Suffix);
3237 case 'R':
3238 return ApplyModifier_WordFunc(pp, st, ModifyWord_Root);
3239 case 'r':
3240 return ApplyModifier_Range(pp, st);
3241 case 'O':
3242 return ApplyModifier_Order(pp, st);
3243 case 'u':
3244 return ApplyModifier_Unique(pp, st);
3245 #ifdef SUNSHCMD
3246 case 's':
3247 return ApplyModifier_SunShell(pp, st);
3248 #endif
3249 default:
3250 return AMR_UNKNOWN;
3251 }
3252 }
3253
3254 static char *ApplyModifiers(const char **, char *, char, char, Var *,
3255 VarExprFlags *, GNode *, VarEvalFlags, void **);
3256
3257 typedef enum ApplyModifiersIndirectResult {
3258 AMIR_CONTINUE,
3259 AMIR_APPLY_MODS,
3260 AMIR_OUT
3261 } ApplyModifiersIndirectResult;
3262
3263 /* While expanding a variable expression, expand and apply indirect
3264 * modifiers such as in ${VAR:${M_indirect}}. */
3265 static ApplyModifiersIndirectResult
3266 ApplyModifiersIndirect(
3267 ApplyModifiersState *const st,
3268 const char **const inout_p,
3269 void **const inout_freeIt
3270 ) {
3271 const char *p = *inout_p;
3272 const char *mods;
3273 void *mods_freeIt;
3274
3275 (void)Var_Parse(&p, st->ctxt, st->eflags, &mods, &mods_freeIt);
3276 /* TODO: handle errors */
3277
3278 /*
3279 * If we have not parsed up to st->endc or ':', we are not
3280 * interested. This means the expression ${VAR:${M_1}${M_2}}
3281 * is not accepted, but ${VAR:${M_1}:${M_2}} is.
3282 */
3283 if (mods[0] != '\0' && *p != '\0' && *p != ':' && *p != st->endc) {
3284 if (DEBUG(LINT))
3285 Parse_Error(PARSE_FATAL,
3286 "Missing delimiter ':' after indirect modifier \"%.*s\"",
3287 (int)(p - *inout_p), *inout_p);
3288
3289 free(mods_freeIt);
3290 /* XXX: apply_mods doesn't sound like "not interested". */
3291 /* XXX: Why is the indirect modifier parsed once more by
3292 * apply_mods? If any, p should be advanced to nested_p. */
3293 return AMIR_APPLY_MODS;
3294 }
3295
3296 VAR_DEBUG3("Indirect modifier \"%s\" from \"%.*s\"\n",
3297 mods, (int)(p - *inout_p), *inout_p);
3298
3299 if (mods[0] != '\0') {
3300 const char *rval_pp = mods;
3301 st->val = ApplyModifiers(&rval_pp, st->val, '\0', '\0', st->v,
3302 &st->exprFlags, st->ctxt, st->eflags,
3303 inout_freeIt);
3304 if (st->val == var_Error || st->val == varUndefined ||
3305 *rval_pp != '\0') {
3306 free(mods_freeIt);
3307 *inout_p = p;
3308 return AMIR_OUT; /* error already reported */
3309 }
3310 }
3311 free(mods_freeIt);
3312
3313 if (*p == ':')
3314 p++;
3315 else if (*p == '\0' && st->endc != '\0') {
3316 Error("Unclosed variable specification after complex "
3317 "modifier (expecting '%c') for %s", st->endc, st->v->name);
3318 *inout_p = p;
3319 return AMIR_OUT;
3320 }
3321
3322 *inout_p = p;
3323 return AMIR_CONTINUE;
3324 }
3325
3326 /* Apply any modifiers (such as :Mpattern or :@var@loop@ or :Q or ::=value). */
3327 static char *
3328 ApplyModifiers(
3329 const char **const pp, /* the parsing position, updated upon return */
3330 char *const val, /* the current value of the expression */
3331 char const startc, /* '(' or '{', or '\0' for indirect modifiers */
3332 char const endc, /* ')' or '}', or '\0' for indirect modifiers */
3333 Var *const v,
3334 VarExprFlags *const exprFlags,
3335 GNode *const ctxt, /* for looking up and modifying variables */
3336 VarEvalFlags const eflags,
3337 void **const inout_freeIt /* free this after using the return value */
3338 ) {
3339 ApplyModifiersState st = {
3340 startc, endc, v, ctxt, eflags,
3341 val, /* .val */
3342 var_Error, /* .newVal */
3343 ' ', /* .sep */
3344 FALSE, /* .oneBigWord */
3345 *exprFlags /* .exprFlags */
3346 };
3347 const char *p;
3348 const char *mod;
3349 ApplyModifierResult res;
3350
3351 assert(startc == '(' || startc == '{' || startc == '\0');
3352 assert(endc == ')' || endc == '}' || endc == '\0');
3353 assert(val != NULL);
3354
3355 p = *pp;
3356
3357 if (*p == '\0' && endc != '\0') {
3358 Error("Unclosed variable expression (expecting '%c') for \"%s\"",
3359 st.endc, st.v->name);
3360 goto cleanup;
3361 }
3362
3363 while (*p != '\0' && *p != endc) {
3364
3365 if (*p == '$') {
3366 ApplyModifiersIndirectResult amir;
3367 amir = ApplyModifiersIndirect(&st, &p, inout_freeIt);
3368 if (amir == AMIR_CONTINUE)
3369 continue;
3370 if (amir == AMIR_OUT)
3371 goto out;
3372 }
3373 st.newVal = var_Error; /* default value, in case of errors */
3374 mod = p;
3375
3376 if (DEBUG(VAR))
3377 LogBeforeApply(&st, mod, endc);
3378
3379 res = ApplyModifier(&p, &st);
3380
3381 #ifdef SYSVVARSUB
3382 if (res == AMR_UNKNOWN) {
3383 assert(p == mod);
3384 res = ApplyModifier_SysV(&p, &st);
3385 }
3386 #endif
3387
3388 if (res == AMR_UNKNOWN) {
3389 Error("Unknown modifier '%c'", *mod);
3390 /* Guess the end of the current modifier.
3391 * XXX: Skipping the rest of the modifier hides errors and leads
3392 * to wrong results. Parsing should rather stop here. */
3393 for (p++; *p != ':' && *p != st.endc && *p != '\0'; p++)
3394 continue;
3395 st.newVal = var_Error;
3396 }
3397 if (res == AMR_CLEANUP)
3398 goto cleanup;
3399 if (res == AMR_BAD)
3400 goto bad_modifier;
3401
3402 if (DEBUG(VAR))
3403 LogAfterApply(&st, p, mod);
3404
3405 if (st.newVal != st.val) {
3406 if (*inout_freeIt != NULL) {
3407 free(st.val);
3408 *inout_freeIt = NULL;
3409 }
3410 st.val = st.newVal;
3411 if (st.val != var_Error && st.val != varUndefined)
3412 *inout_freeIt = st.val;
3413 }
3414 if (*p == '\0' && st.endc != '\0') {
3415 Error("Unclosed variable specification (expecting '%c') "
3416 "for \"%s\" (value \"%s\") modifier %c",
3417 st.endc, st.v->name, st.val, *mod);
3418 } else if (*p == ':') {
3419 p++;
3420 } else if (DEBUG(LINT) && *p != '\0' && *p != endc) {
3421 Parse_Error(PARSE_FATAL,
3422 "Missing delimiter ':' after modifier \"%.*s\"",
3423 (int)(p - mod), mod);
3424 /* TODO: propagate parse error to the enclosing expression */
3425 }
3426 }
3427 out:
3428 *pp = p;
3429 assert(st.val != NULL); /* Use var_Error or varUndefined instead. */
3430 *exprFlags = st.exprFlags;
3431 return st.val;
3432
3433 bad_modifier:
3434 /* XXX: The modifier end is only guessed. */
3435 Error("Bad modifier `:%.*s' for %s",
3436 (int)strcspn(mod, ":)}"), mod, st.v->name);
3437
3438 cleanup:
3439 *pp = p;
3440 free(*inout_freeIt);
3441 *inout_freeIt = NULL;
3442 *exprFlags = st.exprFlags;
3443 return var_Error;
3444 }
3445
3446 /* Only four of the local variables are treated specially as they are the
3447 * only four that will be set when dynamic sources are expanded. */
3448 static Boolean
3449 VarnameIsDynamic(const char *name, size_t len)
3450 {
3451 if (len == 1 || (len == 2 && (name[1] == 'F' || name[1] == 'D'))) {
3452 switch (name[0]) {
3453 case '@':
3454 case '%':
3455 case '*':
3456 case '!':
3457 return TRUE;
3458 }
3459 return FALSE;
3460 }
3461
3462 if ((len == 7 || len == 8) && name[0] == '.' && ch_isupper(name[1])) {
3463 return strcmp(name, ".TARGET") == 0 ||
3464 strcmp(name, ".ARCHIVE") == 0 ||
3465 strcmp(name, ".PREFIX") == 0 ||
3466 strcmp(name, ".MEMBER") == 0;
3467 }
3468
3469 return FALSE;
3470 }
3471
3472 static const char *
3473 UndefinedShortVarValue(char varname, const GNode *ctxt, VarEvalFlags eflags)
3474 {
3475 if (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL) {
3476 /*
3477 * If substituting a local variable in a non-local context,
3478 * assume it's for dynamic source stuff. We have to handle
3479 * this specially and return the longhand for the variable
3480 * with the dollar sign escaped so it makes it back to the
3481 * caller. Only four of the local variables are treated
3482 * specially as they are the only four that will be set
3483 * when dynamic sources are expanded.
3484 */
3485 switch (varname) {
3486 case '@':
3487 return "$(.TARGET)";
3488 case '%':
3489 return "$(.MEMBER)";
3490 case '*':
3491 return "$(.PREFIX)";
3492 case '!':
3493 return "$(.ARCHIVE)";
3494 }
3495 }
3496 return eflags & VARE_UNDEFERR ? var_Error : varUndefined;
3497 }
3498
3499 /* Parse a variable name, until the end character or a colon, whichever
3500 * comes first. */
3501 static char *
3502 ParseVarname(const char **pp, char startc, char endc,
3503 GNode *ctxt, VarEvalFlags eflags,
3504 size_t *out_varname_len)
3505 {
3506 Buffer buf;
3507 const char *p = *pp;
3508 int depth = 1;
3509
3510 Buf_Init(&buf);
3511
3512 while (*p != '\0') {
3513 /* Track depth so we can spot parse errors. */
3514 if (*p == startc)
3515 depth++;
3516 if (*p == endc) {
3517 if (--depth == 0)
3518 break;
3519 }
3520 if (*p == ':' && depth == 1)
3521 break;
3522
3523 /* A variable inside a variable, expand. */
3524 if (*p == '$') {
3525 const char *nested_val;
3526 void *nested_val_freeIt;
3527 (void)Var_Parse(&p, ctxt, eflags, &nested_val, &nested_val_freeIt);
3528 /* TODO: handle errors */
3529 Buf_AddStr(&buf, nested_val);
3530 free(nested_val_freeIt);
3531 } else {
3532 Buf_AddByte(&buf, *p);
3533 p++;
3534 }
3535 }
3536 *pp = p;
3537 *out_varname_len = Buf_Len(&buf);
3538 return Buf_Destroy(&buf, FALSE);
3539 }
3540
3541 static VarParseResult
3542 ValidShortVarname(char varname, const char *start)
3543 {
3544 switch (varname) {
3545 case '\0':
3546 case ')':
3547 case '}':
3548 case ':':
3549 case '$':
3550 break; /* and continue below */
3551 default:
3552 return VPR_OK;
3553 }
3554
3555 if (!DEBUG(LINT))
3556 return VPR_PARSE_SILENT;
3557
3558 if (varname == '$')
3559 Parse_Error(PARSE_FATAL,
3560 "To escape a dollar, use \\$, not $$, at \"%s\"", start);
3561 else if (varname == '\0')
3562 Parse_Error(PARSE_FATAL, "Dollar followed by nothing");
3563 else
3564 Parse_Error(PARSE_FATAL,
3565 "Invalid variable name '%c', at \"%s\"", varname, start);
3566
3567 return VPR_PARSE_MSG;
3568 }
3569
3570 /* Parse a single-character variable name such as $V or $@.
3571 * Return whether to continue parsing. */
3572 static Boolean
3573 ParseVarnameShort(char startc, const char **pp, GNode *ctxt,
3574 VarEvalFlags eflags,
3575 VarParseResult *out_FALSE_res, const char **out_FALSE_val,
3576 Var **out_TRUE_var)
3577 {
3578 char name[2];
3579 Var *v;
3580 VarParseResult vpr;
3581
3582 /*
3583 * If it's not bounded by braces of some sort, life is much simpler.
3584 * We just need to check for the first character and return the
3585 * value if it exists.
3586 */
3587
3588 vpr = ValidShortVarname(startc, *pp);
3589 if (vpr != VPR_OK) {
3590 (*pp)++;
3591 *out_FALSE_val = var_Error;
3592 *out_FALSE_res = vpr;
3593 return FALSE;
3594 }
3595
3596 name[0] = startc;
3597 name[1] = '\0';
3598 v = VarFind(name, ctxt, TRUE);
3599 if (v == NULL) {
3600 *pp += 2;
3601
3602 *out_FALSE_val = UndefinedShortVarValue(startc, ctxt, eflags);
3603 if (DEBUG(LINT) && *out_FALSE_val == var_Error) {
3604 Parse_Error(PARSE_FATAL, "Variable \"%s\" is undefined", name);
3605 *out_FALSE_res = VPR_UNDEF_MSG;
3606 return FALSE;
3607 }
3608 *out_FALSE_res = eflags & VARE_UNDEFERR ? VPR_UNDEF_SILENT : VPR_OK;
3609 return FALSE;
3610 }
3611
3612 *out_TRUE_var = v;
3613 return TRUE;
3614 }
3615
3616 /* Find variables like @F or <D. */
3617 static Var *
3618 FindLocalLegacyVar(const char *varname, size_t namelen, GNode *ctxt,
3619 const char **out_extraModifiers)
3620 {
3621 /* Only resolve these variables if ctxt is a "real" target. */
3622 if (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL)
3623 return NULL;
3624
3625 if (namelen != 2)
3626 return NULL;
3627 if (varname[1] != 'F' && varname[1] != 'D')
3628 return NULL;
3629 if (strchr("@%?*!<>", varname[0]) == NULL)
3630 return NULL;
3631
3632 {
3633 char name[] = { varname[0], '\0' };
3634 Var *v = VarFind(name, ctxt, FALSE);
3635
3636 if (v != NULL) {
3637 if (varname[1] == 'D') {
3638 *out_extraModifiers = "H:";
3639 } else { /* F */
3640 *out_extraModifiers = "T:";
3641 }
3642 }
3643 return v;
3644 }
3645 }
3646
3647 static VarParseResult
3648 EvalUndefined(Boolean dynamic, const char *start, const char *p, char *varname,
3649 VarEvalFlags eflags,
3650 void **out_freeIt, const char **out_val)
3651 {
3652 if (dynamic) {
3653 char *pstr = bmake_strsedup(start, p);
3654 free(varname);
3655 *out_freeIt = pstr;
3656 *out_val = pstr;
3657 return VPR_OK;
3658 }
3659
3660 if ((eflags & VARE_UNDEFERR) && (eflags & VARE_WANTRES) &&
3661 DEBUG(LINT))
3662 {
3663 Parse_Error(PARSE_FATAL, "Variable \"%s\" is undefined", varname);
3664 free(varname);
3665 *out_val = var_Error;
3666 return VPR_UNDEF_MSG;
3667 }
3668
3669 if (eflags & VARE_UNDEFERR) {
3670 free(varname);
3671 *out_val = var_Error;
3672 return VPR_UNDEF_SILENT;
3673 }
3674
3675 free(varname);
3676 *out_val = varUndefined;
3677 return VPR_OK;
3678 }
3679
3680 /* Parse a long variable name enclosed in braces or parentheses such as $(VAR)
3681 * or ${VAR}, up to the closing brace or parenthesis, or in the case of
3682 * ${VAR:Modifiers}, up to the ':' that starts the modifiers.
3683 * Return whether to continue parsing. */
3684 static Boolean
3685 ParseVarnameLong(
3686 const char *p,
3687 char startc,
3688 GNode *ctxt,
3689 VarEvalFlags eflags,
3690
3691 const char **out_FALSE_pp,
3692 VarParseResult *out_FALSE_res,
3693 const char **out_FALSE_val,
3694 void **out_FALSE_freeIt,
3695
3696 char *out_TRUE_endc,
3697 const char **out_TRUE_p,
3698 Var **out_TRUE_v,
3699 Boolean *out_TRUE_haveModifier,
3700 const char **out_TRUE_extraModifiers,
3701 Boolean *out_TRUE_dynamic,
3702 VarExprFlags *out_TRUE_exprFlags
3703 ) {
3704 size_t namelen;
3705 char *varname;
3706 Var *v;
3707 Boolean haveModifier;
3708 Boolean dynamic = FALSE;
3709
3710 const char *const start = p;
3711 char endc = startc == '(' ? ')' : '}';
3712
3713 p += 2; /* skip "${" or "$(" or "y(" */
3714 varname = ParseVarname(&p, startc, endc, ctxt, eflags, &namelen);
3715
3716 if (*p == ':') {
3717 haveModifier = TRUE;
3718 } else if (*p == endc) {
3719 haveModifier = FALSE;
3720 } else {
3721 Parse_Error(PARSE_FATAL, "Unclosed variable \"%s\"", varname);
3722 free(varname);
3723 *out_FALSE_pp = p;
3724 *out_FALSE_val = var_Error;
3725 *out_FALSE_res = VPR_PARSE_MSG;
3726 return FALSE;
3727 }
3728
3729 v = VarFind(varname, ctxt, TRUE);
3730
3731 /* At this point, p points just after the variable name,
3732 * either at ':' or at endc. */
3733
3734 if (v == NULL)
3735 v = FindLocalLegacyVar(varname, namelen, ctxt, out_TRUE_extraModifiers);
3736
3737 if (v == NULL) {
3738 /* Defer expansion of dynamic variables if they appear in non-local
3739 * context since they are not defined there. */
3740 dynamic = VarnameIsDynamic(varname, namelen) &&
3741 (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL);
3742
3743 if (!haveModifier) {
3744 p++; /* skip endc */
3745 *out_FALSE_pp = p;
3746 *out_FALSE_res = EvalUndefined(dynamic, start, p, varname, eflags,
3747 out_FALSE_freeIt, out_FALSE_val);
3748 return FALSE;
3749 }
3750
3751 /* The variable expression is based on an undefined variable.
3752 * Nevertheless it needs a Var, for modifiers that access the
3753 * variable name, such as :L or :?.
3754 *
3755 * Most modifiers leave this expression in the "undefined" state
3756 * (VEF_UNDEF), only a few modifiers like :D, :U, :L, :P turn this
3757 * undefined expression into a defined expression (VEF_DEF).
3758 *
3759 * At the end, after applying all modifiers, if the expression
3760 * is still undefined, Var_Parse will return an empty string
3761 * instead of the actually computed value. */
3762 v = VarNew(varname, varname, "", 0);
3763 *out_TRUE_exprFlags = VEF_UNDEF;
3764 } else
3765 free(varname);
3766
3767 *out_TRUE_endc = endc;
3768 *out_TRUE_p = p;
3769 *out_TRUE_v = v;
3770 *out_TRUE_haveModifier = haveModifier;
3771 *out_TRUE_dynamic = dynamic;
3772 return TRUE;
3773 }
3774
3775 /*
3776 * Given the start of a variable expression (such as $v, $(VAR),
3777 * ${VAR:Mpattern}), extract the variable name and value, and the modifiers,
3778 * if any. While doing that, apply the modifiers to the value of the
3779 * expression, forming its final value. A few of the modifiers such as :!cmd!
3780 * or ::= have side effects.
3781 *
3782 * Input:
3783 * *pp The string to parse.
3784 * When parsing a condition in ParseEmptyArg, it may also
3785 * point to the "y" of "empty(VARNAME:Modifiers)", which
3786 * is syntactically the same.
3787 * ctxt The context for finding variables
3788 * eflags Control the exact details of parsing
3789 *
3790 * Output:
3791 * *pp The position where to continue parsing.
3792 * TODO: After a parse error, the value of *pp is
3793 * unspecified. It may not have been updated at all,
3794 * point to some random character in the string, to the
3795 * location of the parse error, or at the end of the
3796 * string.
3797 * *out_val The value of the variable expression, never NULL.
3798 * *out_val var_Error if there was a parse error.
3799 * *out_val var_Error if the base variable of the expression was
3800 * undefined, eflags contains VARE_UNDEFERR, and none of
3801 * the modifiers turned the undefined expression into a
3802 * defined expression.
3803 * XXX: It is not guaranteed that an error message has
3804 * been printed.
3805 * *out_val varUndefined if the base variable of the expression
3806 * was undefined, eflags did not contain VARE_UNDEFERR,
3807 * and none of the modifiers turned the undefined
3808 * expression into a defined expression.
3809 * XXX: It is not guaranteed that an error message has
3810 * been printed.
3811 * *out_val_freeIt Must be freed by the caller after using *out_val.
3812 */
3813 /* coverity[+alloc : arg-*4] */
3814 VarParseResult
3815 Var_Parse(const char **pp, GNode *ctxt, VarEvalFlags eflags,
3816 const char **out_val, void **out_val_freeIt)
3817 {
3818 const char *p = *pp;
3819 const char *const start = p;
3820 Boolean haveModifier; /* TRUE if have modifiers for the variable */
3821 char startc; /* Starting character if variable in parens
3822 * or braces */
3823 char endc; /* Ending character if variable in parens
3824 * or braces */
3825 Boolean dynamic; /* TRUE if the variable is local and we're
3826 * expanding it in a non-local context. This
3827 * is done to support dynamic sources. The
3828 * result is just the expression, unaltered */
3829 const char *extramodifiers;
3830 Var *v;
3831 char *value;
3832 char eflags_str[VarEvalFlags_ToStringSize];
3833 VarExprFlags exprFlags = 0;
3834
3835 VAR_DEBUG2("Var_Parse: %s with %s\n", start,
3836 Enum_FlagsToString(eflags_str, sizeof eflags_str, eflags,
3837 VarEvalFlags_ToStringSpecs));
3838
3839 *out_val_freeIt = NULL;
3840 extramodifiers = NULL; /* extra modifiers to apply first */
3841 dynamic = FALSE;
3842
3843 /* Appease GCC, which thinks that the variable might not be
3844 * initialized. */
3845 endc = '\0';
3846
3847 startc = p[1];
3848 if (startc != '(' && startc != '{') {
3849 VarParseResult res;
3850 if (!ParseVarnameShort(startc, pp, ctxt, eflags, &res, out_val, &v))
3851 return res;
3852 haveModifier = FALSE;
3853 p++;
3854 } else {
3855 VarParseResult res;
3856 if (!ParseVarnameLong(p, startc, ctxt, eflags,
3857 pp, &res, out_val, out_val_freeIt,
3858 &endc, &p, &v, &haveModifier, &extramodifiers,
3859 &dynamic, &exprFlags))
3860 return res;
3861 }
3862
3863 if (v->flags & VAR_IN_USE)
3864 Fatal("Variable %s is recursive.", v->name);
3865
3866 value = Buf_GetAll(&v->val, NULL);
3867
3868 /* Before applying any modifiers, expand any nested expressions from the
3869 * variable value. */
3870 if (strchr(value, '$') != NULL && (eflags & VARE_WANTRES)) {
3871 VarEvalFlags nested_eflags = eflags;
3872 if (DEBUG(LINT))
3873 nested_eflags &= ~(unsigned)VARE_UNDEFERR;
3874 v->flags |= VAR_IN_USE;
3875 (void)Var_Subst(value, ctxt, nested_eflags, &value);
3876 v->flags &= ~(unsigned)VAR_IN_USE;
3877 /* TODO: handle errors */
3878 *out_val_freeIt = value;
3879 }
3880
3881 if (haveModifier || extramodifiers != NULL) {
3882 void *extraFree;
3883
3884 extraFree = NULL;
3885 if (extramodifiers != NULL) {
3886 const char *em = extramodifiers;
3887 value = ApplyModifiers(&em, value, '\0', '\0',
3888 v, &exprFlags, ctxt, eflags, &extraFree);
3889 }
3890
3891 if (haveModifier) {
3892 /* Skip initial colon. */
3893 p++;
3894
3895 value = ApplyModifiers(&p, value, startc, endc,
3896 v, &exprFlags, ctxt, eflags, out_val_freeIt);
3897 free(extraFree);
3898 } else {
3899 *out_val_freeIt = extraFree;
3900 }
3901 }
3902
3903 if (*p != '\0') /* Skip past endc if possible. */
3904 p++;
3905
3906 *pp = p;
3907
3908 if (v->flags & VAR_FROM_ENV) {
3909 /* Free the environment variable now since we own it,
3910 * but don't free the variable value if it will be returned. */
3911 Boolean keepValue = value == Buf_GetAll(&v->val, NULL);
3912 if (keepValue)
3913 *out_val_freeIt = value;
3914 (void)VarFreeEnv(v, !keepValue);
3915
3916 } else if (exprFlags & VEF_UNDEF) {
3917 if (!(exprFlags & VEF_DEF)) {
3918 if (*out_val_freeIt != NULL) {
3919 free(*out_val_freeIt);
3920 *out_val_freeIt = NULL;
3921 }
3922 if (dynamic) {
3923 value = bmake_strsedup(start, p);
3924 *out_val_freeIt = value;
3925 } else {
3926 /* The expression is still undefined, therefore discard the
3927 * actual value and return an error marker instead. */
3928 value = (eflags & VARE_UNDEFERR) ? var_Error : varUndefined;
3929 }
3930 }
3931 if (value != Buf_GetAll(&v->val, NULL))
3932 Buf_Destroy(&v->val, TRUE);
3933 free(v->name_freeIt);
3934 free(v);
3935 }
3936 *out_val = value;
3937 return VPR_UNKNOWN;
3938 }
3939
3940 static void
3941 VarSubstNested(const char **const pp, Buffer *const buf, GNode *const ctxt,
3942 VarEvalFlags const eflags, Boolean *inout_errorReported)
3943 {
3944 const char *p = *pp;
3945 const char *nested_p = p;
3946 const char *val;
3947 void *val_freeIt;
3948
3949 (void)Var_Parse(&nested_p, ctxt, eflags, &val, &val_freeIt);
3950 /* TODO: handle errors */
3951
3952 if (val == var_Error || val == varUndefined) {
3953 if (!preserveUndefined) {
3954 p = nested_p;
3955 } else if ((eflags & VARE_UNDEFERR) || val == var_Error) {
3956 /* XXX: This condition is wrong. If val == var_Error,
3957 * this doesn't necessarily mean there was an undefined
3958 * variable. It could equally well be a parse error; see
3959 * unit-tests/varmod-order.exp. */
3960
3961 /*
3962 * If variable is undefined, complain and skip the
3963 * variable. The complaint will stop us from doing anything
3964 * when the file is parsed.
3965 */
3966 if (!*inout_errorReported) {
3967 Parse_Error(PARSE_FATAL, "Undefined variable \"%.*s\"",
3968 (int)(size_t)(nested_p - p), p);
3969 }
3970 p = nested_p;
3971 *inout_errorReported = TRUE;
3972 } else {
3973 /* Copy the initial '$' of the undefined expression,
3974 * thereby deferring expansion of the expression, but
3975 * expand nested expressions if already possible.
3976 * See unit-tests/varparse-undef-partial.mk. */
3977 Buf_AddByte(buf, *p);
3978 p++;
3979 }
3980 } else {
3981 p = nested_p;
3982 Buf_AddStr(buf, val);
3983 }
3984
3985 free(val_freeIt);
3986
3987 *pp = p;
3988 }
3989
3990 /* Expand all variable expressions like $V, ${VAR}, $(VAR:Modifiers) in the
3991 * given string.
3992 *
3993 * Input:
3994 * str The string in which the variable expressions are
3995 * expanded.
3996 * ctxt The context in which to start searching for
3997 * variables. The other contexts are searched as well.
3998 * eflags Special effects during expansion.
3999 */
4000 VarParseResult
4001 Var_Subst(const char *str, GNode *ctxt, VarEvalFlags eflags, char **out_res)
4002 {
4003 const char *p = str;
4004 Buffer buf; /* Buffer for forming things */
4005
4006 /* Set true if an error has already been reported,
4007 * to prevent a plethora of messages when recursing */
4008 static Boolean errorReported;
4009
4010 Buf_Init(&buf);
4011 errorReported = FALSE;
4012
4013 while (*p != '\0') {
4014 if (p[0] == '$' && p[1] == '$') {
4015 /* A dollar sign may be escaped with another dollar sign. */
4016 if (save_dollars && (eflags & VARE_KEEP_DOLLAR))
4017 Buf_AddByte(&buf, '$');
4018 Buf_AddByte(&buf, '$');
4019 p += 2;
4020
4021 } else if (p[0] == '$') {
4022 VarSubstNested(&p, &buf, ctxt, eflags, &errorReported);
4023
4024 } else {
4025 /*
4026 * Skip as many characters as possible -- either to the end of
4027 * the string or to the next dollar sign (variable expression).
4028 */
4029 const char *plainStart = p;
4030
4031 for (p++; *p != '$' && *p != '\0'; p++)
4032 continue;
4033 Buf_AddBytesBetween(&buf, plainStart, p);
4034 }
4035 }
4036
4037 *out_res = Buf_DestroyCompact(&buf);
4038 return VPR_OK;
4039 }
4040
4041 /* Initialize the variables module. */
4042 void
4043 Var_Init(void)
4044 {
4045 VAR_INTERNAL = Targ_NewGN("Internal");
4046 VAR_GLOBAL = Targ_NewGN("Global");
4047 VAR_CMDLINE = Targ_NewGN("Command");
4048 }
4049
4050 /* Clean up the variables module. */
4051 void
4052 Var_End(void)
4053 {
4054 Var_Stats();
4055 }
4056
4057 void
4058 Var_Stats(void)
4059 {
4060 HashTable_DebugStats(&VAR_GLOBAL->context, "VAR_GLOBAL");
4061 }
4062
4063 /* Print all variables in a context, sorted by name. */
4064 void
4065 Var_Dump(GNode *ctxt)
4066 {
4067 Vector /* of const char * */ vec;
4068 HashIter hi;
4069 size_t i;
4070 const char **varnames;
4071
4072 Vector_Init(&vec, sizeof(const char *));
4073
4074 HashIter_Init(&hi, &ctxt->context);
4075 while (HashIter_Next(&hi) != NULL)
4076 *(const char **)Vector_Push(&vec) = hi.entry->key;
4077 varnames = vec.items;
4078
4079 qsort(varnames, vec.len, sizeof varnames[0], str_cmp_asc);
4080
4081 for (i = 0; i < vec.len; i++) {
4082 const char *varname = varnames[i];
4083 Var *var = HashTable_FindValue(&ctxt->context, varname);
4084 debug_printf("%-16s = %s\n", varname, Buf_GetAll(&var->val, NULL));
4085 }
4086
4087 Vector_Done(&vec);
4088 }
4089