var.c revision 1.647 1 /* $NetBSD: var.c,v 1.647 2020/11/02 18:15:12 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 1.647 2020/11/02 18:15:12 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_ASSIGN);
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 /* Special return value for Var_Parse, just to avoid allocating empty strings.
161 * In contrast to var_Error and varUndefined, this is not an error marker but
162 * just an ordinary successful return value. */
163 static char emptyString[] = "";
164
165 /*
166 * Traditionally this make consumed $$ during := like any other expansion.
167 * Other make's do not, and this make follows straight since 2016-01-09.
168 *
169 * This knob allows controlling the behavior.
170 * FALSE to consume $$ during := assignment.
171 * TRUE to preserve $$ during := assignment.
172 */
173 #define MAKE_SAVE_DOLLARS ".MAKE.SAVE_DOLLARS"
174 static Boolean save_dollars = TRUE;
175
176 /*
177 * Internally, variables are contained in four different contexts.
178 * 1) the environment. They cannot be changed. If an environment
179 * variable is appended to, the result is placed in the global
180 * context.
181 * 2) the global context. Variables set in the makefiles are located
182 * here.
183 * 3) the command-line context. All variables set on the command line
184 * are placed in this context.
185 * 4) the local context. Each target has associated with it a context
186 * list. On this list are located the structures describing such
187 * local variables as $(@) and $(*)
188 * The four contexts are searched in the reverse order from which they are
189 * listed (but see opts.checkEnvFirst).
190 */
191 GNode *VAR_INTERNAL; /* variables from make itself */
192 GNode *VAR_GLOBAL; /* variables from the makefile */
193 GNode *VAR_CMDLINE; /* variables defined on the command-line */
194
195 typedef enum VarFlags {
196
197 /* The variable's value is currently being used by Var_Parse or Var_Subst.
198 * This marker is used to avoid endless recursion. */
199 VAR_IN_USE = 0x01,
200
201 /* The variable comes from the environment.
202 * These variables are not registered in any GNode, therefore they must
203 * be freed as soon as they are not used anymore. */
204 VAR_FROM_ENV = 0x02,
205
206 /* The variable is exported to the environment, to be used by child
207 * processes. */
208 VAR_EXPORTED = 0x10,
209
210 /* At the point where this variable was exported, it contained an
211 * unresolved reference to another variable. Before any child process is
212 * started, it needs to be exported again, in the hope that the referenced
213 * variable can then be resolved. */
214 VAR_REEXPORT = 0x20,
215
216 /* The variable came from the command line. */
217 VAR_FROM_CMD = 0x40,
218
219 /* The variable value cannot be changed anymore, and the variable cannot
220 * be deleted. Any attempts to do so are ignored. */
221 VAR_READONLY = 0x80
222 } VarFlags;
223
224 ENUM_FLAGS_RTTI_6(VarFlags,
225 VAR_IN_USE, VAR_FROM_ENV,
226 VAR_EXPORTED, VAR_REEXPORT, VAR_FROM_CMD, VAR_READONLY);
227
228 /* Variables are defined using one of the VAR=value assignments. Their
229 * value can be queried by expressions such as $V, ${VAR}, or with modifiers
230 * such as ${VAR:S,from,to,g:Q}.
231 *
232 * There are 3 kinds of variables: context variables, environment variables,
233 * undefined variables.
234 *
235 * Context variables are stored in a GNode.context. The only way to undefine
236 * a context variable is using the .undef directive. In particular, it must
237 * not be possible to undefine a variable during the evaluation of an
238 * expression, or Var.name might point nowhere.
239 *
240 * Environment variables are temporary. They are returned by VarFind, and
241 * after using them, they must be freed using VarFreeEnv.
242 *
243 * Undefined variables occur during evaluation of variable expressions such
244 * as ${UNDEF:Ufallback} in Var_Parse and ApplyModifiers.
245 */
246 typedef struct Var {
247 /* The name of the variable, once set, doesn't change anymore.
248 * For context variables, it aliases the corresponding HashEntry name.
249 * For environment and undefined variables, it is allocated. */
250 const char *name;
251 void *name_freeIt;
252
253 Buffer val; /* its value */
254 VarFlags flags; /* miscellaneous status flags */
255 } Var;
256
257 /*
258 * Exporting vars is expensive so skip it if we can
259 */
260 typedef enum VarExportedMode {
261 VAR_EXPORTED_NONE,
262 VAR_EXPORTED_SOME,
263 VAR_EXPORTED_ALL
264 } VarExportedMode;
265
266 static VarExportedMode var_exportedVars = VAR_EXPORTED_NONE;
267
268 typedef enum VarExportFlags {
269 /*
270 * We pass this to Var_Export when doing the initial export
271 * or after updating an exported var.
272 */
273 VAR_EXPORT_PARENT = 0x01,
274 /*
275 * We pass this to Var_Export1 to tell it to leave the value alone.
276 */
277 VAR_EXPORT_LITERAL = 0x02
278 } VarExportFlags;
279
280 /* Flags for pattern matching in the :S and :C modifiers */
281 typedef enum VarPatternFlags {
282 VARP_SUB_GLOBAL = 0x01, /* Replace as often as possible ('g') */
283 VARP_SUB_ONE = 0x02, /* Replace only once ('1') */
284 VARP_ANCHOR_START = 0x04, /* Match at start of word ('^') */
285 VARP_ANCHOR_END = 0x08 /* Match at end of word ('$') */
286 } VarPatternFlags;
287
288 static Var *
289 VarNew(const char *name, void *name_freeIt, const char *value, VarFlags flags)
290 {
291 size_t value_len = strlen(value);
292 Var *var = bmake_malloc(sizeof *var);
293 var->name = name;
294 var->name_freeIt = name_freeIt;
295 Buf_Init(&var->val, value_len + 1);
296 Buf_AddBytes(&var->val, value, value_len);
297 var->flags = flags;
298 return var;
299 }
300
301 static const char *
302 CanonicalVarname(const char *name)
303 {
304 if (*name == '.' && ch_isupper(name[1])) {
305 switch (name[1]) {
306 case 'A':
307 if (strcmp(name, ".ALLSRC") == 0)
308 name = ALLSRC;
309 if (strcmp(name, ".ARCHIVE") == 0)
310 name = ARCHIVE;
311 break;
312 case 'I':
313 if (strcmp(name, ".IMPSRC") == 0)
314 name = IMPSRC;
315 break;
316 case 'M':
317 if (strcmp(name, ".MEMBER") == 0)
318 name = MEMBER;
319 break;
320 case 'O':
321 if (strcmp(name, ".OODATE") == 0)
322 name = OODATE;
323 break;
324 case 'P':
325 if (strcmp(name, ".PREFIX") == 0)
326 name = PREFIX;
327 break;
328 case 'S':
329 if (strcmp(name, ".SHELL") == 0) {
330 if (!shellPath)
331 Shell_Init();
332 }
333 break;
334 case 'T':
335 if (strcmp(name, ".TARGET") == 0)
336 name = TARGET;
337 break;
338 }
339 }
340
341 /* GNU make has an additional alias $^ == ${.ALLSRC}. */
342
343 return name;
344 }
345
346 static Var *
347 GNode_FindVar(GNode *ctxt, const char *varname, unsigned int hash)
348 {
349 return HashTable_FindValueHash(&ctxt->context, varname, hash);
350 }
351
352 /* Find the variable in the context, and maybe in other contexts as well.
353 *
354 * Input:
355 * name name to find, is not expanded any further
356 * ctxt context in which to look first
357 * elsewhere TRUE to look in other contexts as well
358 *
359 * Results:
360 * The found variable, or NULL if the variable does not exist.
361 * If the variable is an environment variable, it must be freed using
362 * VarFreeEnv after use.
363 */
364 static Var *
365 VarFind(const char *name, GNode *ctxt, Boolean elsewhere)
366 {
367 Var *var;
368 unsigned int nameHash;
369
370 /*
371 * If the variable name begins with a '.', it could very well be one of
372 * the local ones. We check the name against all the local variables
373 * and substitute the short version in for 'name' if it matches one of
374 * them.
375 */
376 name = CanonicalVarname(name);
377 nameHash = Hash_Hash(name);
378
379 /* First look for the variable in the given context. */
380 var = GNode_FindVar(ctxt, name, nameHash);
381 if (!elsewhere)
382 return var;
383
384 /* The variable was not found in the given context. Now look for it in
385 * the other contexts as well. */
386 if (var == NULL && ctxt != VAR_CMDLINE)
387 var = GNode_FindVar(VAR_CMDLINE, name, nameHash);
388
389 if (!opts.checkEnvFirst && var == NULL && ctxt != VAR_GLOBAL) {
390 var = GNode_FindVar(VAR_GLOBAL, name, nameHash);
391 if (var == NULL && ctxt != VAR_INTERNAL) {
392 /* VAR_INTERNAL is subordinate to VAR_GLOBAL */
393 var = GNode_FindVar(VAR_INTERNAL, name, nameHash);
394 }
395 }
396
397 if (var == NULL) {
398 char *env;
399
400 if ((env = getenv(name)) != NULL) {
401 char *varname = bmake_strdup(name);
402 return VarNew(varname, varname, env, VAR_FROM_ENV);
403 }
404
405 if (opts.checkEnvFirst && ctxt != VAR_GLOBAL) {
406 var = GNode_FindVar(VAR_GLOBAL, name, nameHash);
407 if (var == NULL && ctxt != VAR_INTERNAL)
408 var = GNode_FindVar(VAR_INTERNAL, name, nameHash);
409 return var;
410 }
411
412 return NULL;
413 }
414
415 return var;
416 }
417
418 /* If the variable is an environment variable, free it.
419 *
420 * Input:
421 * v the variable
422 * freeValue true if the variable value should be freed as well
423 *
424 * Results:
425 * TRUE if it is an environment variable, FALSE otherwise.
426 */
427 static Boolean
428 VarFreeEnv(Var *v, Boolean freeValue)
429 {
430 if (!(v->flags & VAR_FROM_ENV))
431 return FALSE;
432
433 free(v->name_freeIt);
434 Buf_Destroy(&v->val, freeValue);
435 free(v);
436 return TRUE;
437 }
438
439 /* Add a new variable of the given name and value to the given context.
440 * The name and val arguments are duplicated so they may safely be freed. */
441 static void
442 VarAdd(const char *name, const char *val, GNode *ctxt, VarSet_Flags flags)
443 {
444 HashEntry *he = HashTable_CreateEntry(&ctxt->context, name, NULL);
445 Var *v = VarNew(he->key /* aliased */, NULL, val,
446 flags & VAR_SET_READONLY ? VAR_READONLY : 0);
447 HashEntry_Set(he, v);
448 if (!(ctxt->flags & INTERNAL)) {
449 VAR_DEBUG3("%s:%s = %s\n", ctxt->name, name, val);
450 }
451 }
452
453 /* Remove a variable from a context, freeing all related memory as well.
454 * The variable name is expanded once. */
455 void
456 Var_Delete(const char *name, GNode *ctxt)
457 {
458 char *name_freeIt = NULL;
459 HashEntry *he;
460
461 if (strchr(name, '$') != NULL) {
462 (void)Var_Subst(name, VAR_GLOBAL, VARE_WANTRES, &name_freeIt);
463 /* TODO: handle errors */
464 name = name_freeIt;
465 }
466 he = HashTable_FindEntry(&ctxt->context, name);
467 VAR_DEBUG3("%s:delete %s%s\n",
468 ctxt->name, name, he != NULL ? "" : " (not found)");
469 free(name_freeIt);
470
471 if (he != NULL) {
472 Var *v = HashEntry_Get(he);
473 if (v->flags & VAR_EXPORTED)
474 unsetenv(v->name);
475 if (strcmp(v->name, MAKE_EXPORTED) == 0)
476 var_exportedVars = VAR_EXPORTED_NONE;
477 assert(v->name_freeIt == NULL);
478 HashTable_DeleteEntry(&ctxt->context, he);
479 Buf_Destroy(&v->val, TRUE);
480 free(v);
481 }
482 }
483
484 static Boolean
485 MayExport(const char *name)
486 {
487 if (name[0] == '.')
488 return FALSE; /* skip internals */
489 if (name[0] == '-')
490 return FALSE; /* skip misnamed variables */
491 if (name[1] == '\0') {
492 /*
493 * A single char.
494 * If it is one of the vars that should only appear in
495 * local context, skip it, else we can get Var_Subst
496 * into a loop.
497 */
498 switch (name[0]) {
499 case '@':
500 case '%':
501 case '*':
502 case '!':
503 return FALSE;
504 }
505 }
506 return TRUE;
507 }
508
509 /*
510 * Export a single variable.
511 * We ignore make internal variables (those which start with '.').
512 * Also we jump through some hoops to avoid calling setenv
513 * more than necessary since it can leak.
514 * We only manipulate flags of vars if 'parent' is set.
515 */
516 static Boolean
517 Var_Export1(const char *name, VarExportFlags flags)
518 {
519 VarExportFlags parent = flags & VAR_EXPORT_PARENT;
520 Var *v;
521 char *val;
522
523 if (!MayExport(name))
524 return FALSE;
525
526 v = VarFind(name, VAR_GLOBAL, 0);
527 if (v == NULL)
528 return FALSE;
529
530 if (!parent && (v->flags & VAR_EXPORTED) && !(v->flags & VAR_REEXPORT))
531 return FALSE; /* nothing to do */
532
533 val = Buf_GetAll(&v->val, NULL);
534 if (!(flags & VAR_EXPORT_LITERAL) && strchr(val, '$') != NULL) {
535 char *expr;
536
537 if (parent) {
538 /*
539 * Flag the variable as something we need to re-export.
540 * No point actually exporting it now though,
541 * the child process can do it at the last minute.
542 */
543 v->flags |= VAR_EXPORTED | VAR_REEXPORT;
544 return TRUE;
545 }
546 if (v->flags & VAR_IN_USE) {
547 /*
548 * We recursed while exporting in a child.
549 * This isn't going to end well, just skip it.
550 */
551 return FALSE;
552 }
553
554 /* XXX: name is injected without escaping it */
555 expr = str_concat3("${", name, "}");
556 (void)Var_Subst(expr, VAR_GLOBAL, VARE_WANTRES, &val);
557 /* TODO: handle errors */
558 setenv(name, val, 1);
559 free(val);
560 free(expr);
561 } else {
562 if (parent)
563 v->flags &= ~(unsigned)VAR_REEXPORT; /* once will do */
564 if (parent || !(v->flags & VAR_EXPORTED))
565 setenv(name, val, 1);
566 }
567
568 /*
569 * This is so Var_Set knows to call Var_Export again...
570 */
571 if (parent) {
572 v->flags |= VAR_EXPORTED;
573 }
574 return TRUE;
575 }
576
577 /*
578 * This gets called from our child processes.
579 */
580 void
581 Var_ExportVars(void)
582 {
583 char *val;
584
585 /*
586 * Several make's support this sort of mechanism for tracking
587 * recursion - but each uses a different name.
588 * We allow the makefiles to update MAKELEVEL and ensure
589 * children see a correctly incremented value.
590 */
591 char tmp[BUFSIZ];
592 snprintf(tmp, sizeof(tmp), "%d", makelevel + 1);
593 setenv(MAKE_LEVEL_ENV, tmp, 1);
594
595 if (var_exportedVars == VAR_EXPORTED_NONE)
596 return;
597
598 if (var_exportedVars == VAR_EXPORTED_ALL) {
599 HashIter hi;
600
601 /* Ouch! Exporting all variables at once is crazy... */
602 HashIter_Init(&hi, &VAR_GLOBAL->context);
603 while (HashIter_Next(&hi) != NULL) {
604 Var *var = hi.entry->value;
605 Var_Export1(var->name, 0);
606 }
607 return;
608 }
609
610 (void)Var_Subst("${" MAKE_EXPORTED ":O:u}", VAR_GLOBAL, VARE_WANTRES, &val);
611 /* TODO: handle errors */
612 if (*val) {
613 Words words = Str_Words(val, FALSE);
614 size_t i;
615
616 for (i = 0; i < words.len; i++)
617 Var_Export1(words.words[i], 0);
618 Words_Free(words);
619 }
620 free(val);
621 }
622
623 /*
624 * This is called when .export is seen or .MAKE.EXPORTED is modified.
625 *
626 * It is also called when any exported variable is modified.
627 * XXX: Is it really?
628 *
629 * str has the format "[-env|-literal] varname...".
630 */
631 void
632 Var_Export(const char *str, Boolean isExport)
633 {
634 VarExportFlags flags;
635 char *val;
636
637 if (isExport && str[0] == '\0') {
638 var_exportedVars = VAR_EXPORTED_ALL; /* use with caution! */
639 return;
640 }
641
642 if (isExport && strncmp(str, "-env", 4) == 0) {
643 str += 4;
644 flags = 0;
645 } else if (isExport && strncmp(str, "-literal", 8) == 0) {
646 str += 8;
647 flags = VAR_EXPORT_LITERAL;
648 } else {
649 flags = VAR_EXPORT_PARENT;
650 }
651
652 (void)Var_Subst(str, VAR_GLOBAL, VARE_WANTRES, &val);
653 /* TODO: handle errors */
654 if (val[0] != '\0') {
655 Words words = Str_Words(val, FALSE);
656
657 size_t i;
658 for (i = 0; i < words.len; i++) {
659 const char *name = words.words[i];
660 if (Var_Export1(name, flags)) {
661 if (var_exportedVars == VAR_EXPORTED_NONE)
662 var_exportedVars = VAR_EXPORTED_SOME;
663 if (isExport && (flags & VAR_EXPORT_PARENT)) {
664 Var_Append(MAKE_EXPORTED, name, VAR_GLOBAL);
665 }
666 }
667 }
668 Words_Free(words);
669 }
670 free(val);
671 }
672
673
674 extern char **environ;
675
676 /*
677 * This is called when .unexport[-env] is seen.
678 *
679 * str must have the form "unexport[-env] varname...".
680 */
681 void
682 Var_UnExport(const char *str)
683 {
684 const char *varnames;
685 char *varnames_freeIt;
686 Boolean unexport_env;
687
688 varnames = NULL;
689 varnames_freeIt = NULL;
690
691 str += strlen("unexport");
692 unexport_env = strncmp(str, "-env", 4) == 0;
693 if (unexport_env) {
694 const char *cp;
695 char **newenv;
696
697 cp = getenv(MAKE_LEVEL_ENV); /* we should preserve this */
698 if (environ == savedEnv) {
699 /* we have been here before! */
700 newenv = bmake_realloc(environ, 2 * sizeof(char *));
701 } else {
702 if (savedEnv) {
703 free(savedEnv);
704 savedEnv = NULL;
705 }
706 newenv = bmake_malloc(2 * sizeof(char *));
707 }
708
709 /* Note: we cannot safely free() the original environ. */
710 environ = savedEnv = newenv;
711 newenv[0] = NULL;
712 newenv[1] = NULL;
713 if (cp && *cp)
714 setenv(MAKE_LEVEL_ENV, cp, 1);
715 } else {
716 cpp_skip_whitespace(&str);
717 if (str[0] != '\0')
718 varnames = str;
719 }
720
721 if (varnames == NULL) {
722 /* Using .MAKE.EXPORTED */
723 (void)Var_Subst("${" MAKE_EXPORTED ":O:u}", VAR_GLOBAL, VARE_WANTRES,
724 &varnames_freeIt);
725 /* TODO: handle errors */
726 varnames = varnames_freeIt;
727 }
728
729 {
730 Var *v;
731 size_t i;
732
733 Words words = Str_Words(varnames, FALSE);
734 for (i = 0; i < words.len; i++) {
735 const char *varname = words.words[i];
736 v = VarFind(varname, VAR_GLOBAL, 0);
737 if (v == NULL) {
738 VAR_DEBUG1("Not unexporting \"%s\" (not found)\n", varname);
739 continue;
740 }
741
742 VAR_DEBUG1("Unexporting \"%s\"\n", varname);
743 if (!unexport_env && (v->flags & VAR_EXPORTED) &&
744 !(v->flags & VAR_REEXPORT))
745 unsetenv(v->name);
746 v->flags &= ~(unsigned)(VAR_EXPORTED | VAR_REEXPORT);
747
748 /*
749 * If we are unexporting a list,
750 * remove each one from .MAKE.EXPORTED.
751 * If we are removing them all,
752 * just delete .MAKE.EXPORTED below.
753 */
754 if (varnames == str) {
755 /* XXX: v->name is injected without escaping it */
756 char *expr = str_concat3("${" MAKE_EXPORTED ":N", v->name, "}");
757 char *cp;
758 (void)Var_Subst(expr, VAR_GLOBAL, VARE_WANTRES, &cp);
759 /* TODO: handle errors */
760 Var_Set(MAKE_EXPORTED, cp, VAR_GLOBAL);
761 free(cp);
762 free(expr);
763 }
764 }
765 Words_Free(words);
766 if (varnames != str) {
767 Var_Delete(MAKE_EXPORTED, VAR_GLOBAL);
768 free(varnames_freeIt);
769 }
770 }
771 }
772
773 /* See Var_Set for documentation. */
774 void
775 Var_Set_with_flags(const char *name, const char *val, GNode *ctxt,
776 VarSet_Flags flags)
777 {
778 const char *unexpanded_name = name;
779 char *name_freeIt = NULL;
780 Var *v;
781
782 assert(val != NULL);
783
784 if (strchr(name, '$') != NULL) {
785 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
786 /* TODO: handle errors */
787 name = name_freeIt;
788 }
789
790 if (name[0] == '\0') {
791 VAR_DEBUG2("Var_Set(\"%s\", \"%s\", ...) "
792 "name expands to empty string - ignored\n",
793 unexpanded_name, val);
794 free(name_freeIt);
795 return;
796 }
797
798 if (ctxt == VAR_GLOBAL) {
799 v = VarFind(name, VAR_CMDLINE, 0);
800 if (v != NULL) {
801 if (v->flags & VAR_FROM_CMD) {
802 VAR_DEBUG3("%s:%s = %s ignored!\n", ctxt->name, name, val);
803 goto out;
804 }
805 VarFreeEnv(v, TRUE);
806 }
807 }
808
809 /*
810 * We only look for a variable in the given context since anything set
811 * here will override anything in a lower context, so there's not much
812 * point in searching them all just to save a bit of memory...
813 */
814 v = VarFind(name, ctxt, 0);
815 if (v == NULL) {
816 if (ctxt == VAR_CMDLINE && !(flags & VAR_NO_EXPORT)) {
817 /*
818 * This var would normally prevent the same name being added
819 * to VAR_GLOBAL, so delete it from there if needed.
820 * Otherwise -V name may show the wrong value.
821 */
822 /* XXX: name is expanded for the second time */
823 Var_Delete(name, VAR_GLOBAL);
824 }
825 VarAdd(name, val, ctxt, flags);
826 } else {
827 if ((v->flags & VAR_READONLY) && !(flags & VAR_SET_READONLY)) {
828 VAR_DEBUG3("%s:%s = %s ignored (read-only)\n",
829 ctxt->name, name, val);
830 goto out;
831 }
832 Buf_Empty(&v->val);
833 Buf_AddStr(&v->val, val);
834
835 VAR_DEBUG3("%s:%s = %s\n", ctxt->name, name, val);
836 if (v->flags & VAR_EXPORTED) {
837 Var_Export1(name, VAR_EXPORT_PARENT);
838 }
839 }
840 /*
841 * Any variables given on the command line are automatically exported
842 * to the environment (as per POSIX standard)
843 * Other than internals.
844 */
845 if (ctxt == VAR_CMDLINE && !(flags & VAR_NO_EXPORT) && name[0] != '.') {
846 if (v == NULL)
847 v = VarFind(name, ctxt, 0); /* we just added it */
848 v->flags |= VAR_FROM_CMD;
849
850 /*
851 * If requested, don't export these in the environment
852 * individually. We still put them in MAKEOVERRIDES so
853 * that the command-line settings continue to override
854 * Makefile settings.
855 */
856 if (!opts.varNoExportEnv)
857 setenv(name, val, 1);
858
859 Var_Append(MAKEOVERRIDES, name, VAR_GLOBAL);
860 }
861 if (name[0] == '.' && strcmp(name, MAKE_SAVE_DOLLARS) == 0)
862 save_dollars = s2Boolean(val, save_dollars);
863
864 out:
865 free(name_freeIt);
866 if (v != NULL)
867 VarFreeEnv(v, TRUE);
868 }
869
870 /*-
871 *-----------------------------------------------------------------------
872 * Var_Set --
873 * Set the variable name to the value val in the given context.
874 *
875 * If the variable doesn't yet exist, it is created.
876 * Otherwise the new value overwrites and replaces the old value.
877 *
878 * Input:
879 * name name of the variable to set, is expanded once
880 * val value to give to the variable
881 * ctxt context in which to set it
882 *
883 * Notes:
884 * The variable is searched for only in its context before being
885 * created in that context. I.e. if the context is VAR_GLOBAL,
886 * only VAR_GLOBAL->context is searched. Likewise if it is VAR_CMDLINE,
887 * only VAR_CMDLINE->context is searched. This is done to avoid the
888 * literally thousands of unnecessary strcmp's that used to be done to
889 * set, say, $(@) or $(<).
890 * If the context is VAR_GLOBAL though, we check if the variable
891 * was set in VAR_CMDLINE from the command line and skip it if so.
892 *-----------------------------------------------------------------------
893 */
894 void
895 Var_Set(const char *name, const char *val, GNode *ctxt)
896 {
897 Var_Set_with_flags(name, val, ctxt, 0);
898 }
899
900 /*-
901 *-----------------------------------------------------------------------
902 * Var_Append --
903 * The variable of the given name has the given value appended to it in
904 * the given context.
905 *
906 * If the variable doesn't exist, it is created. Otherwise the strings
907 * are concatenated, with a space in between.
908 *
909 * Input:
910 * name name of the variable to modify, is expanded once
911 * val string to append to it
912 * ctxt context in which this should occur
913 *
914 * Notes:
915 * Only if the variable is being sought in the global context is the
916 * environment searched.
917 * XXX: Knows its calling circumstances in that if called with ctxt
918 * an actual target, it will only search that context since only
919 * a local variable could be being appended to. This is actually
920 * a big win and must be tolerated.
921 *-----------------------------------------------------------------------
922 */
923 void
924 Var_Append(const char *name, const char *val, GNode *ctxt)
925 {
926 char *name_freeIt = NULL;
927 Var *v;
928
929 assert(val != NULL);
930
931 if (strchr(name, '$') != NULL) {
932 const char *unexpanded_name = name;
933 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
934 /* TODO: handle errors */
935 name = name_freeIt;
936 if (name[0] == '\0') {
937 VAR_DEBUG2("Var_Append(\"%s\", \"%s\", ...) "
938 "name expands to empty string - ignored\n",
939 unexpanded_name, val);
940 free(name_freeIt);
941 return;
942 }
943 }
944
945 v = VarFind(name, ctxt, ctxt == VAR_GLOBAL);
946
947 if (v == NULL) {
948 /* XXX: name is expanded for the second time */
949 Var_Set(name, val, ctxt);
950 } else if (v->flags & VAR_READONLY) {
951 VAR_DEBUG1("Ignoring append to %s since it is read-only\n", name);
952 } else if (ctxt == VAR_CMDLINE || !(v->flags & VAR_FROM_CMD)) {
953 Buf_AddByte(&v->val, ' ');
954 Buf_AddStr(&v->val, val);
955
956 VAR_DEBUG3("%s:%s = %s\n",
957 ctxt->name, name, Buf_GetAll(&v->val, NULL));
958
959 if (v->flags & VAR_FROM_ENV) {
960 HashEntry *h;
961
962 /*
963 * If the original variable came from the environment, we
964 * have to install it in the global context (we could place
965 * it in the environment, but then we should provide a way to
966 * export other variables...)
967 */
968 v->flags &= ~(unsigned)VAR_FROM_ENV;
969 h = HashTable_CreateEntry(&ctxt->context, name, NULL);
970 HashEntry_Set(h, v);
971 }
972 }
973 free(name_freeIt);
974 }
975
976 /* See if the given variable exists, in the given context or in other
977 * fallback contexts.
978 *
979 * Input:
980 * name Variable to find, is expanded once
981 * ctxt Context in which to start search
982 */
983 Boolean
984 Var_Exists(const char *name, GNode *ctxt)
985 {
986 char *name_freeIt = NULL;
987 Var *v;
988
989 if (strchr(name, '$') != NULL) {
990 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
991 /* TODO: handle errors */
992 name = name_freeIt;
993 }
994
995 v = VarFind(name, ctxt, TRUE);
996 free(name_freeIt);
997 if (v == NULL)
998 return FALSE;
999
1000 (void)VarFreeEnv(v, TRUE);
1001 return TRUE;
1002 }
1003
1004 /*-
1005 *-----------------------------------------------------------------------
1006 * Var_Value --
1007 * Return the unexpanded value of the given variable in the given
1008 * context, or the usual contexts.
1009 *
1010 * Input:
1011 * name name to find, is not expanded any further
1012 * ctxt context in which to search for it
1013 *
1014 * Results:
1015 * The value if the variable exists, NULL if it doesn't.
1016 * If the returned value is not NULL, the caller must free
1017 * out_freeIt when the returned value is no longer needed.
1018 *-----------------------------------------------------------------------
1019 */
1020 const char *
1021 Var_Value(const char *name, GNode *ctxt, void **out_freeIt)
1022 {
1023 Var *v = VarFind(name, ctxt, TRUE);
1024 char *value;
1025
1026 *out_freeIt = NULL;
1027 if (v == NULL)
1028 return NULL;
1029
1030 value = Buf_GetAll(&v->val, NULL);
1031 if (VarFreeEnv(v, FALSE))
1032 *out_freeIt = value;
1033 return value;
1034 }
1035
1036 /* Return the unexpanded variable value from this node, without trying to look
1037 * up the variable in any other context. */
1038 const char *
1039 Var_ValueDirect(const char *name, GNode *ctxt)
1040 {
1041 Var *v = VarFind(name, ctxt, FALSE);
1042 return v != NULL ? Buf_GetAll(&v->val, NULL) : NULL;
1043 }
1044
1045
1046 /* SepBuf is a string being built from words, interleaved with separators. */
1047 typedef struct SepBuf {
1048 Buffer buf;
1049 Boolean needSep;
1050 char sep; /* usually ' ', but see the :ts modifier */
1051 } SepBuf;
1052
1053 static void
1054 SepBuf_Init(SepBuf *buf, char sep)
1055 {
1056 Buf_Init(&buf->buf, 32 /* bytes */);
1057 buf->needSep = FALSE;
1058 buf->sep = sep;
1059 }
1060
1061 static void
1062 SepBuf_Sep(SepBuf *buf)
1063 {
1064 buf->needSep = TRUE;
1065 }
1066
1067 static void
1068 SepBuf_AddBytes(SepBuf *buf, const char *mem, size_t mem_size)
1069 {
1070 if (mem_size == 0)
1071 return;
1072 if (buf->needSep && buf->sep != '\0') {
1073 Buf_AddByte(&buf->buf, buf->sep);
1074 buf->needSep = FALSE;
1075 }
1076 Buf_AddBytes(&buf->buf, mem, mem_size);
1077 }
1078
1079 static void
1080 SepBuf_AddBytesBetween(SepBuf *buf, const char *start, const char *end)
1081 {
1082 SepBuf_AddBytes(buf, start, (size_t)(end - start));
1083 }
1084
1085 static void
1086 SepBuf_AddStr(SepBuf *buf, const char *str)
1087 {
1088 SepBuf_AddBytes(buf, str, strlen(str));
1089 }
1090
1091 static char *
1092 SepBuf_Destroy(SepBuf *buf, Boolean free_buf)
1093 {
1094 return Buf_Destroy(&buf->buf, free_buf);
1095 }
1096
1097
1098 /* This callback for ModifyWords gets a single word from a variable expression
1099 * and typically adds a modification of this word to the buffer. It may also
1100 * do nothing or add several words.
1101 *
1102 * For example, in ${:Ua b c:M*2}, the callback is called 3 times, once for
1103 * each word of "a b c". */
1104 typedef void (*ModifyWordsCallback)(const char *word, SepBuf *buf, void *data);
1105
1106
1107 /* Callback for ModifyWords to implement the :H modifier.
1108 * Add the dirname of the given word to the buffer. */
1109 static void
1110 ModifyWord_Head(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1111 {
1112 const char *slash = strrchr(word, '/');
1113 if (slash != NULL)
1114 SepBuf_AddBytesBetween(buf, word, slash);
1115 else
1116 SepBuf_AddStr(buf, ".");
1117 }
1118
1119 /* Callback for ModifyWords to implement the :T modifier.
1120 * Add the basename of the given word to the buffer. */
1121 static void
1122 ModifyWord_Tail(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1123 {
1124 const char *slash = strrchr(word, '/');
1125 const char *base = slash != NULL ? slash + 1 : word;
1126 SepBuf_AddStr(buf, base);
1127 }
1128
1129 /* Callback for ModifyWords to implement the :E modifier.
1130 * Add the filename suffix of the given word to the buffer, if it exists. */
1131 static void
1132 ModifyWord_Suffix(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1133 {
1134 const char *dot = strrchr(word, '.');
1135 if (dot != NULL)
1136 SepBuf_AddStr(buf, dot + 1);
1137 }
1138
1139 /* Callback for ModifyWords to implement the :R modifier.
1140 * Add the basename of the given word to the buffer. */
1141 static void
1142 ModifyWord_Root(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1143 {
1144 const char *dot = strrchr(word, '.');
1145 size_t len = dot != NULL ? (size_t)(dot - word) : strlen(word);
1146 SepBuf_AddBytes(buf, word, len);
1147 }
1148
1149 /* Callback for ModifyWords to implement the :M modifier.
1150 * Place the word in the buffer if it matches the given pattern. */
1151 static void
1152 ModifyWord_Match(const char *word, SepBuf *buf, void *data)
1153 {
1154 const char *pattern = data;
1155 VAR_DEBUG2("VarMatch [%s] [%s]\n", word, pattern);
1156 if (Str_Match(word, pattern))
1157 SepBuf_AddStr(buf, word);
1158 }
1159
1160 /* Callback for ModifyWords to implement the :N modifier.
1161 * Place the word in the buffer if it doesn't match the given pattern. */
1162 static void
1163 ModifyWord_NoMatch(const char *word, SepBuf *buf, void *data)
1164 {
1165 const char *pattern = data;
1166 if (!Str_Match(word, pattern))
1167 SepBuf_AddStr(buf, word);
1168 }
1169
1170 #ifdef SYSVVARSUB
1171 /* Check word against pattern for a match (% is a wildcard).
1172 *
1173 * Input:
1174 * word Word to examine
1175 * pattern Pattern to examine against
1176 *
1177 * Results:
1178 * Returns the start of the match, or NULL.
1179 * out_match_len returns the length of the match, if any.
1180 * out_hasPercent returns whether the pattern contains a percent.
1181 */
1182 static const char *
1183 SysVMatch(const char *word, const char *pattern,
1184 size_t *out_match_len, Boolean *out_hasPercent)
1185 {
1186 const char *p = pattern;
1187 const char *w = word;
1188 const char *percent;
1189 size_t w_len;
1190 size_t p_len;
1191 const char *w_tail;
1192
1193 *out_hasPercent = FALSE;
1194 percent = strchr(p, '%');
1195 if (percent != NULL) { /* ${VAR:...%...=...} */
1196 *out_hasPercent = TRUE;
1197 if (*w == '\0')
1198 return NULL; /* empty word does not match pattern */
1199
1200 /* check that the prefix matches */
1201 for (; p != percent && *w != '\0' && *w == *p; w++, p++)
1202 continue;
1203 if (p != percent)
1204 return NULL; /* No match */
1205
1206 p++; /* Skip the percent */
1207 if (*p == '\0') {
1208 /* No more pattern, return the rest of the string */
1209 *out_match_len = strlen(w);
1210 return w;
1211 }
1212 }
1213
1214 /* Test whether the tail matches */
1215 w_len = strlen(w);
1216 p_len = strlen(p);
1217 if (w_len < p_len)
1218 return NULL;
1219
1220 w_tail = w + w_len - p_len;
1221 if (memcmp(p, w_tail, p_len) != 0)
1222 return NULL;
1223
1224 *out_match_len = (size_t)(w_tail - w);
1225 return w;
1226 }
1227
1228 struct ModifyWord_SYSVSubstArgs {
1229 GNode *ctx;
1230 const char *lhs;
1231 const char *rhs;
1232 };
1233
1234 /* Callback for ModifyWords to implement the :%.from=%.to modifier. */
1235 static void
1236 ModifyWord_SYSVSubst(const char *word, SepBuf *buf, void *data)
1237 {
1238 const struct ModifyWord_SYSVSubstArgs *args = data;
1239 char *rhs_expanded;
1240 const char *rhs;
1241 const char *percent;
1242
1243 size_t match_len;
1244 Boolean lhsPercent;
1245 const char *match = SysVMatch(word, args->lhs, &match_len, &lhsPercent);
1246 if (match == NULL) {
1247 SepBuf_AddStr(buf, word);
1248 return;
1249 }
1250
1251 /* Append rhs to the buffer, substituting the first '%' with the
1252 * match, but only if the lhs had a '%' as well. */
1253
1254 (void)Var_Subst(args->rhs, args->ctx, VARE_WANTRES, &rhs_expanded);
1255 /* TODO: handle errors */
1256
1257 rhs = rhs_expanded;
1258 percent = strchr(rhs, '%');
1259
1260 if (percent != NULL && lhsPercent) {
1261 /* Copy the prefix of the replacement pattern */
1262 SepBuf_AddBytesBetween(buf, rhs, percent);
1263 rhs = percent + 1;
1264 }
1265 if (percent != NULL || !lhsPercent)
1266 SepBuf_AddBytes(buf, match, match_len);
1267
1268 /* Append the suffix of the replacement pattern */
1269 SepBuf_AddStr(buf, rhs);
1270
1271 free(rhs_expanded);
1272 }
1273 #endif
1274
1275
1276 struct ModifyWord_SubstArgs {
1277 const char *lhs;
1278 size_t lhsLen;
1279 const char *rhs;
1280 size_t rhsLen;
1281 VarPatternFlags pflags;
1282 Boolean matched;
1283 };
1284
1285 /* Callback for ModifyWords to implement the :S,from,to, modifier.
1286 * Perform a string substitution on the given word. */
1287 static void
1288 ModifyWord_Subst(const char *word, SepBuf *buf, void *data)
1289 {
1290 size_t wordLen = strlen(word);
1291 struct ModifyWord_SubstArgs *args = data;
1292 const char *match;
1293
1294 if ((args->pflags & VARP_SUB_ONE) && args->matched)
1295 goto nosub;
1296
1297 if (args->pflags & VARP_ANCHOR_START) {
1298 if (wordLen < args->lhsLen ||
1299 memcmp(word, args->lhs, args->lhsLen) != 0)
1300 goto nosub;
1301
1302 if ((args->pflags & VARP_ANCHOR_END) && wordLen != args->lhsLen)
1303 goto nosub;
1304
1305 /* :S,^prefix,replacement, or :S,^whole$,replacement, */
1306 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1307 SepBuf_AddBytes(buf, word + args->lhsLen, wordLen - args->lhsLen);
1308 args->matched = TRUE;
1309 return;
1310 }
1311
1312 if (args->pflags & VARP_ANCHOR_END) {
1313 const char *start;
1314
1315 if (wordLen < args->lhsLen)
1316 goto nosub;
1317
1318 start = word + (wordLen - args->lhsLen);
1319 if (memcmp(start, args->lhs, args->lhsLen) != 0)
1320 goto nosub;
1321
1322 /* :S,suffix$,replacement, */
1323 SepBuf_AddBytesBetween(buf, word, start);
1324 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1325 args->matched = TRUE;
1326 return;
1327 }
1328
1329 if (args->lhs[0] == '\0')
1330 goto nosub;
1331
1332 /* unanchored case, may match more than once */
1333 while ((match = strstr(word, args->lhs)) != NULL) {
1334 SepBuf_AddBytesBetween(buf, word, match);
1335 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1336 args->matched = TRUE;
1337 wordLen -= (size_t)(match - word) + args->lhsLen;
1338 word += (size_t)(match - word) + args->lhsLen;
1339 if (wordLen == 0 || !(args->pflags & VARP_SUB_GLOBAL))
1340 break;
1341 }
1342 nosub:
1343 SepBuf_AddBytes(buf, word, wordLen);
1344 }
1345
1346 #ifndef NO_REGEX
1347 /* Print the error caused by a regcomp or regexec call. */
1348 static void
1349 VarREError(int reerr, regex_t *pat, const char *str)
1350 {
1351 size_t errlen = regerror(reerr, pat, 0, 0);
1352 char *errbuf = bmake_malloc(errlen);
1353 regerror(reerr, pat, errbuf, errlen);
1354 Error("%s: %s", str, errbuf);
1355 free(errbuf);
1356 }
1357
1358 struct ModifyWord_SubstRegexArgs {
1359 regex_t re;
1360 size_t nsub;
1361 char *replace;
1362 VarPatternFlags pflags;
1363 Boolean matched;
1364 };
1365
1366 /* Callback for ModifyWords to implement the :C/from/to/ modifier.
1367 * Perform a regex substitution on the given word. */
1368 static void
1369 ModifyWord_SubstRegex(const char *word, SepBuf *buf, void *data)
1370 {
1371 struct ModifyWord_SubstRegexArgs *args = data;
1372 int xrv;
1373 const char *wp = word;
1374 char *rp;
1375 int flags = 0;
1376 regmatch_t m[10];
1377
1378 if ((args->pflags & VARP_SUB_ONE) && args->matched)
1379 goto nosub;
1380
1381 tryagain:
1382 xrv = regexec(&args->re, wp, args->nsub, m, flags);
1383
1384 switch (xrv) {
1385 case 0:
1386 args->matched = TRUE;
1387 SepBuf_AddBytes(buf, wp, (size_t)m[0].rm_so);
1388
1389 for (rp = args->replace; *rp; rp++) {
1390 if (*rp == '\\' && (rp[1] == '&' || rp[1] == '\\')) {
1391 SepBuf_AddBytes(buf, rp + 1, 1);
1392 rp++;
1393 continue;
1394 }
1395
1396 if (*rp == '&') {
1397 SepBuf_AddBytesBetween(buf, wp + m[0].rm_so, wp + m[0].rm_eo);
1398 continue;
1399 }
1400
1401 if (*rp != '\\' || !ch_isdigit(rp[1])) {
1402 SepBuf_AddBytes(buf, rp, 1);
1403 continue;
1404 }
1405
1406 { /* \0 to \9 backreference */
1407 size_t n = (size_t)(rp[1] - '0');
1408 rp++;
1409
1410 if (n >= args->nsub) {
1411 Error("No subexpression \\%zu", n);
1412 } else if (m[n].rm_so == -1) {
1413 Error("No match for subexpression \\%zu", n);
1414 } else {
1415 SepBuf_AddBytesBetween(buf, wp + m[n].rm_so,
1416 wp + m[n].rm_eo);
1417 }
1418 }
1419 }
1420
1421 wp += m[0].rm_eo;
1422 if (args->pflags & VARP_SUB_GLOBAL) {
1423 flags |= REG_NOTBOL;
1424 if (m[0].rm_so == 0 && m[0].rm_eo == 0) {
1425 SepBuf_AddBytes(buf, wp, 1);
1426 wp++;
1427 }
1428 if (*wp)
1429 goto tryagain;
1430 }
1431 if (*wp) {
1432 SepBuf_AddStr(buf, wp);
1433 }
1434 break;
1435 default:
1436 VarREError(xrv, &args->re, "Unexpected regex error");
1437 /* FALLTHROUGH */
1438 case REG_NOMATCH:
1439 nosub:
1440 SepBuf_AddStr(buf, wp);
1441 break;
1442 }
1443 }
1444 #endif
1445
1446
1447 struct ModifyWord_LoopArgs {
1448 GNode *ctx;
1449 char *tvar; /* name of temporary variable */
1450 char *str; /* string to expand */
1451 VarEvalFlags eflags;
1452 };
1453
1454 /* Callback for ModifyWords to implement the :@var (at) ...@ modifier of ODE make. */
1455 static void
1456 ModifyWord_Loop(const char *word, SepBuf *buf, void *data)
1457 {
1458 const struct ModifyWord_LoopArgs *args;
1459 char *s;
1460
1461 if (word[0] == '\0')
1462 return;
1463
1464 args = data;
1465 Var_Set_with_flags(args->tvar, word, args->ctx, VAR_NO_EXPORT);
1466 (void)Var_Subst(args->str, args->ctx, args->eflags, &s);
1467 /* TODO: handle errors */
1468
1469 VAR_DEBUG4("ModifyWord_Loop: "
1470 "in \"%s\", replace \"%s\" with \"%s\" to \"%s\"\n",
1471 word, args->tvar, args->str, s);
1472
1473 if (s[0] == '\n' || Buf_EndsWith(&buf->buf, '\n'))
1474 buf->needSep = FALSE;
1475 SepBuf_AddStr(buf, s);
1476 free(s);
1477 }
1478
1479
1480 /* The :[first..last] modifier selects words from the expression.
1481 * It can also reverse the words. */
1482 static char *
1483 VarSelectWords(char sep, Boolean oneBigWord, const char *str, int first,
1484 int last)
1485 {
1486 Words words;
1487 int len, start, end, step;
1488 int i;
1489
1490 SepBuf buf;
1491 SepBuf_Init(&buf, sep);
1492
1493 if (oneBigWord) {
1494 /* fake what Str_Words() would do if there were only one word */
1495 words.len = 1;
1496 words.words = bmake_malloc((words.len + 1) * sizeof(char *));
1497 words.freeIt = bmake_strdup(str);
1498 words.words[0] = words.freeIt;
1499 words.words[1] = NULL;
1500 } else {
1501 words = Str_Words(str, FALSE);
1502 }
1503
1504 /*
1505 * Now sanitize the given range.
1506 * If first or last are negative, convert them to the positive equivalents
1507 * (-1 gets converted to ac, -2 gets converted to (ac - 1), etc.).
1508 */
1509 len = (int)words.len;
1510 if (first < 0)
1511 first += len + 1;
1512 if (last < 0)
1513 last += len + 1;
1514
1515 /*
1516 * We avoid scanning more of the list than we need to.
1517 */
1518 if (first > last) {
1519 start = (first > len ? len : first) - 1;
1520 end = last < 1 ? 0 : last - 1;
1521 step = -1;
1522 } else {
1523 start = first < 1 ? 0 : first - 1;
1524 end = last > len ? len : last;
1525 step = 1;
1526 }
1527
1528 for (i = start; (step < 0) == (i >= end); i += step) {
1529 SepBuf_AddStr(&buf, words.words[i]);
1530 SepBuf_Sep(&buf);
1531 }
1532
1533 Words_Free(words);
1534
1535 return SepBuf_Destroy(&buf, FALSE);
1536 }
1537
1538
1539 /* Callback for ModifyWords to implement the :tA modifier.
1540 * Replace each word with the result of realpath() if successful. */
1541 static void
1542 ModifyWord_Realpath(const char *word, SepBuf *buf, void *data MAKE_ATTR_UNUSED)
1543 {
1544 struct stat st;
1545 char rbuf[MAXPATHLEN];
1546
1547 const char *rp = cached_realpath(word, rbuf);
1548 if (rp != NULL && *rp == '/' && stat(rp, &st) == 0)
1549 word = rp;
1550
1551 SepBuf_AddStr(buf, word);
1552 }
1553
1554 /* Modify each of the words of the passed string using the given function.
1555 *
1556 * Input:
1557 * str String whose words should be modified
1558 * modifyWord Function that modifies a single word
1559 * modifyWord_args Custom arguments for modifyWord
1560 *
1561 * Results:
1562 * A string of all the words modified appropriately.
1563 *-----------------------------------------------------------------------
1564 */
1565 static char *
1566 ModifyWords(const char *str,
1567 ModifyWordsCallback modifyWord, void *modifyWord_args,
1568 Boolean oneBigWord, char sep)
1569 {
1570 SepBuf result;
1571 Words words;
1572 size_t i;
1573
1574 if (oneBigWord) {
1575 SepBuf_Init(&result, sep);
1576 modifyWord(str, &result, modifyWord_args);
1577 return SepBuf_Destroy(&result, FALSE);
1578 }
1579
1580 SepBuf_Init(&result, sep);
1581
1582 words = Str_Words(str, FALSE);
1583
1584 VAR_DEBUG2("ModifyWords: split \"%s\" into %zu words\n", str, words.len);
1585
1586 for (i = 0; i < words.len; i++) {
1587 modifyWord(words.words[i], &result, modifyWord_args);
1588 if (Buf_Len(&result.buf) > 0)
1589 SepBuf_Sep(&result);
1590 }
1591
1592 Words_Free(words);
1593
1594 return SepBuf_Destroy(&result, FALSE);
1595 }
1596
1597
1598 static char *
1599 Words_JoinFree(Words words)
1600 {
1601 Buffer buf;
1602 size_t i;
1603
1604 Buf_Init(&buf, 0);
1605
1606 for (i = 0; i < words.len; i++) {
1607 if (i != 0)
1608 Buf_AddByte(&buf, ' '); /* XXX: st->sep, for consistency */
1609 Buf_AddStr(&buf, words.words[i]);
1610 }
1611
1612 Words_Free(words);
1613
1614 return Buf_Destroy(&buf, FALSE);
1615 }
1616
1617 /* Remove adjacent duplicate words. */
1618 static char *
1619 VarUniq(const char *str)
1620 {
1621 Words words = Str_Words(str, FALSE);
1622
1623 if (words.len > 1) {
1624 size_t i, j;
1625 for (j = 0, i = 1; i < words.len; i++)
1626 if (strcmp(words.words[i], words.words[j]) != 0 && (++j != i))
1627 words.words[j] = words.words[i];
1628 words.len = j + 1;
1629 }
1630
1631 return Words_JoinFree(words);
1632 }
1633
1634
1635 /* Quote shell meta-characters and space characters in the string.
1636 * If quoteDollar is set, also quote and double any '$' characters. */
1637 static char *
1638 VarQuote(const char *str, Boolean quoteDollar)
1639 {
1640 Buffer buf;
1641 Buf_Init(&buf, 0);
1642
1643 for (; *str != '\0'; str++) {
1644 if (*str == '\n') {
1645 const char *newline = Shell_GetNewline();
1646 if (newline == NULL)
1647 newline = "\\\n";
1648 Buf_AddStr(&buf, newline);
1649 continue;
1650 }
1651 if (ch_isspace(*str) || is_shell_metachar((unsigned char)*str))
1652 Buf_AddByte(&buf, '\\');
1653 Buf_AddByte(&buf, *str);
1654 if (quoteDollar && *str == '$')
1655 Buf_AddStr(&buf, "\\$");
1656 }
1657
1658 return Buf_Destroy(&buf, FALSE);
1659 }
1660
1661 /* Compute the 32-bit hash of the given string, using the MurmurHash3
1662 * algorithm. Output is encoded as 8 hex digits, in Little Endian order. */
1663 static char *
1664 VarHash(const char *str)
1665 {
1666 static const char hexdigits[16] = "0123456789abcdef";
1667 const unsigned char *ustr = (const unsigned char *)str;
1668
1669 uint32_t h = 0x971e137bU;
1670 uint32_t c1 = 0x95543787U;
1671 uint32_t c2 = 0x2ad7eb25U;
1672 size_t len2 = strlen(str);
1673
1674 char *buf;
1675 size_t i;
1676
1677 size_t len;
1678 for (len = len2; len; ) {
1679 uint32_t k = 0;
1680 switch (len) {
1681 default:
1682 k = ((uint32_t)ustr[3] << 24) |
1683 ((uint32_t)ustr[2] << 16) |
1684 ((uint32_t)ustr[1] << 8) |
1685 (uint32_t)ustr[0];
1686 len -= 4;
1687 ustr += 4;
1688 break;
1689 case 3:
1690 k |= (uint32_t)ustr[2] << 16;
1691 /* FALLTHROUGH */
1692 case 2:
1693 k |= (uint32_t)ustr[1] << 8;
1694 /* FALLTHROUGH */
1695 case 1:
1696 k |= (uint32_t)ustr[0];
1697 len = 0;
1698 }
1699 c1 = c1 * 5 + 0x7b7d159cU;
1700 c2 = c2 * 5 + 0x6bce6396U;
1701 k *= c1;
1702 k = (k << 11) ^ (k >> 21);
1703 k *= c2;
1704 h = (h << 13) ^ (h >> 19);
1705 h = h * 5 + 0x52dce729U;
1706 h ^= k;
1707 }
1708 h ^= (uint32_t)len2;
1709 h *= 0x85ebca6b;
1710 h ^= h >> 13;
1711 h *= 0xc2b2ae35;
1712 h ^= h >> 16;
1713
1714 buf = bmake_malloc(9);
1715 for (i = 0; i < 8; i++) {
1716 buf[i] = hexdigits[h & 0x0f];
1717 h >>= 4;
1718 }
1719 buf[8] = '\0';
1720 return buf;
1721 }
1722
1723 static char *
1724 VarStrftime(const char *fmt, Boolean zulu, time_t tim)
1725 {
1726 char buf[BUFSIZ];
1727
1728 if (!tim)
1729 time(&tim);
1730 if (!*fmt)
1731 fmt = "%c";
1732 strftime(buf, sizeof(buf), fmt, zulu ? gmtime(&tim) : localtime(&tim));
1733
1734 buf[sizeof(buf) - 1] = '\0';
1735 return bmake_strdup(buf);
1736 }
1737
1738 /* The ApplyModifier functions all work in the same way. They get the
1739 * current parsing position (pp) and parse the modifier from there. The
1740 * modifier typically lasts until the next ':', or a closing '}' or ')'
1741 * (taken from st->endc), or the end of the string (parse error).
1742 *
1743 * The high-level behavior of these functions is:
1744 *
1745 * 1. parse the modifier
1746 * 2. evaluate the modifier
1747 * 3. housekeeping
1748 *
1749 * Parsing the modifier
1750 *
1751 * If parsing succeeds, the parsing position *pp is updated to point to the
1752 * first character following the modifier, which typically is either ':' or
1753 * st->endc.
1754 *
1755 * If parsing fails because of a missing delimiter (as in the :S, :C or :@
1756 * modifiers), return AMR_CLEANUP.
1757 *
1758 * If parsing fails because the modifier is unknown, return AMR_UNKNOWN to
1759 * try the SysV modifier ${VAR:from=to} as fallback. This should only be
1760 * done as long as there have been no side effects from evaluating nested
1761 * variables, to avoid evaluating them more than once. In this case, the
1762 * parsing position must not be updated. (XXX: Why not? The original parsing
1763 * position is well-known in ApplyModifiers.)
1764 *
1765 * If parsing fails and the SysV modifier ${VAR:from=to} should not be used
1766 * as a fallback, either issue an error message using Error or Parse_Error
1767 * and then return AMR_CLEANUP, or return AMR_BAD for the default error
1768 * message. Both of these return values will stop processing the variable
1769 * expression. (XXX: As of 2020-08-23, evaluation of the whole string
1770 * continues nevertheless after skipping a few bytes, which essentially is
1771 * undefined behavior. Not in the sense of C, but still it's impossible to
1772 * predict what happens in the parser.)
1773 *
1774 * Evaluating the modifier
1775 *
1776 * After parsing, the modifier is evaluated. The side effects from evaluating
1777 * nested variable expressions in the modifier text often already happen
1778 * during parsing though.
1779 *
1780 * Evaluating the modifier usually takes the current value of the variable
1781 * expression from st->val, or the variable name from st->v->name and stores
1782 * the result in st->newVal.
1783 *
1784 * If evaluating fails (as of 2020-08-23), an error message is printed using
1785 * Error. This function has no side-effects, it really just prints the error
1786 * message. Processing the expression continues as if everything were ok.
1787 * XXX: This should be fixed by adding proper error handling to Var_Subst,
1788 * Var_Parse, ApplyModifiers and ModifyWords.
1789 *
1790 * Housekeeping
1791 *
1792 * Some modifiers such as :D and :U turn undefined expressions into defined
1793 * expressions (see VEF_UNDEF, VEF_DEF).
1794 *
1795 * Some modifiers need to free some memory.
1796 */
1797
1798 typedef enum VarExprFlags {
1799 /* The variable expression is based on an undefined variable. */
1800 VEF_UNDEF = 0x01,
1801 /* The variable expression started as an undefined expression, but one
1802 * of the modifiers (such as :D or :U) has turned the expression from
1803 * undefined to defined. */
1804 VEF_DEF = 0x02
1805 } VarExprFlags;
1806
1807 ENUM_FLAGS_RTTI_2(VarExprFlags,
1808 VEF_UNDEF, VEF_DEF);
1809
1810
1811 typedef struct ApplyModifiersState {
1812 const char startc; /* '\0' or '{' or '(' */
1813 const char endc; /* '\0' or '}' or ')' */
1814 Var * const v;
1815 GNode * const ctxt;
1816 const VarEvalFlags eflags;
1817
1818 char *val; /* The old value of the expression,
1819 * before applying the modifier, never NULL */
1820 char *newVal; /* The new value of the expression,
1821 * after applying the modifier, never NULL */
1822 char sep; /* Word separator in expansions
1823 * (see the :ts modifier) */
1824 Boolean oneBigWord; /* TRUE if some modifiers that otherwise split
1825 * the variable value into words, like :S and
1826 * :C, treat the variable value as a single big
1827 * word, possibly containing spaces. */
1828 VarExprFlags exprFlags;
1829 } ApplyModifiersState;
1830
1831 static void
1832 ApplyModifiersState_Define(ApplyModifiersState *st)
1833 {
1834 if (st->exprFlags & VEF_UNDEF)
1835 st->exprFlags |= VEF_DEF;
1836 }
1837
1838 typedef enum ApplyModifierResult {
1839 AMR_OK, /* Continue parsing */
1840 AMR_UNKNOWN, /* Not a match, try other modifiers as well */
1841 AMR_BAD, /* Error out with "Bad modifier" message */
1842 AMR_CLEANUP /* Error out without error message */
1843 } ApplyModifierResult;
1844
1845 /* Allow backslashes to escape the delimiter, $, and \, but don't touch other
1846 * backslashes. */
1847 static Boolean
1848 IsEscapedModifierPart(const char *p, char delim,
1849 struct ModifyWord_SubstArgs *subst)
1850 {
1851 if (p[0] != '\\')
1852 return FALSE;
1853 if (p[1] == delim || p[1] == '\\' || p[1] == '$')
1854 return TRUE;
1855 return p[1] == '&' && subst != NULL;
1856 }
1857
1858 /*
1859 * Parse a part of a modifier such as the "from" and "to" in :S/from/to/ or
1860 * the "var" or "replacement ${var}" in :@var@replacement ${var}@, up to and
1861 * including the next unescaped delimiter. The delimiter, as well as the
1862 * backslash or the dollar, can be escaped with a backslash.
1863 *
1864 * Return the parsed (and possibly expanded) string, or NULL if no delimiter
1865 * was found. On successful return, the parsing position pp points right
1866 * after the delimiter. The delimiter is not included in the returned
1867 * value though.
1868 */
1869 static VarParseResult
1870 ParseModifierPart(
1871 const char **pp, /* The parsing position, updated upon return */
1872 char delim, /* Parsing stops at this delimiter */
1873 VarEvalFlags eflags, /* Flags for evaluating nested variables;
1874 * if VARE_WANTRES is not set, the text is
1875 * only parsed */
1876 ApplyModifiersState *st,
1877 char **out_part,
1878 size_t *out_length, /* Optionally stores the length of the returned
1879 * string, just to save another strlen call. */
1880 VarPatternFlags *out_pflags,/* For the first part of the :S modifier,
1881 * sets the VARP_ANCHOR_END flag if the last
1882 * character of the pattern is a $. */
1883 struct ModifyWord_SubstArgs *subst
1884 /* For the second part of the :S modifier,
1885 * allow ampersands to be escaped and replace
1886 * unescaped ampersands with subst->lhs. */
1887 ) {
1888 Buffer buf;
1889 const char *p;
1890
1891 Buf_Init(&buf, 0);
1892
1893 /*
1894 * Skim through until the matching delimiter is found; pick up variable
1895 * expressions on the way.
1896 */
1897 p = *pp;
1898 while (*p != '\0' && *p != delim) {
1899 const char *varstart;
1900
1901 if (IsEscapedModifierPart(p, delim, subst)) {
1902 Buf_AddByte(&buf, p[1]);
1903 p += 2;
1904 continue;
1905 }
1906
1907 if (*p != '$') { /* Unescaped, simple text */
1908 if (subst != NULL && *p == '&')
1909 Buf_AddBytes(&buf, subst->lhs, subst->lhsLen);
1910 else
1911 Buf_AddByte(&buf, *p);
1912 p++;
1913 continue;
1914 }
1915
1916 if (p[1] == delim) { /* Unescaped $ at end of pattern */
1917 if (out_pflags != NULL)
1918 *out_pflags |= VARP_ANCHOR_END;
1919 else
1920 Buf_AddByte(&buf, *p);
1921 p++;
1922 continue;
1923 }
1924
1925 if (eflags & VARE_WANTRES) { /* Nested variable, evaluated */
1926 const char *nested_p = p;
1927 const char *nested_val;
1928 void *nested_val_freeIt;
1929 VarEvalFlags nested_eflags = eflags & ~(unsigned)VARE_ASSIGN;
1930
1931 (void)Var_Parse(&nested_p, st->ctxt, nested_eflags,
1932 &nested_val, &nested_val_freeIt);
1933 /* TODO: handle errors */
1934 Buf_AddStr(&buf, nested_val);
1935 free(nested_val_freeIt);
1936 p += nested_p - p;
1937 continue;
1938 }
1939
1940 /* XXX: This whole block is very similar to Var_Parse without
1941 * VARE_WANTRES. There may be subtle edge cases though that are
1942 * not yet covered in the unit tests and that are parsed differently,
1943 * depending on whether they are evaluated or not.
1944 *
1945 * This subtle difference is not documented in the manual page,
1946 * neither is the difference between parsing :D and :M documented.
1947 * No code should ever depend on these details, but who knows. */
1948
1949 varstart = p; /* Nested variable, only parsed */
1950 if (p[1] == '(' || p[1] == '{') {
1951 /*
1952 * Find the end of this variable reference
1953 * and suck it in without further ado.
1954 * It will be interpreted later.
1955 */
1956 char startc = p[1];
1957 int endc = startc == '(' ? ')' : '}';
1958 int depth = 1;
1959
1960 for (p += 2; *p != '\0' && depth > 0; p++) {
1961 if (p[-1] != '\\') {
1962 if (*p == startc)
1963 depth++;
1964 if (*p == endc)
1965 depth--;
1966 }
1967 }
1968 Buf_AddBytesBetween(&buf, varstart, p);
1969 } else {
1970 Buf_AddByte(&buf, *varstart);
1971 p++;
1972 }
1973 }
1974
1975 if (*p != delim) {
1976 *pp = p;
1977 Error("Unfinished modifier for %s ('%c' missing)", st->v->name, delim);
1978 *out_part = NULL;
1979 return VPR_PARSE_MSG;
1980 }
1981
1982 *pp = ++p;
1983 if (out_length != NULL)
1984 *out_length = Buf_Len(&buf);
1985
1986 *out_part = Buf_Destroy(&buf, FALSE);
1987 VAR_DEBUG1("Modifier part: \"%s\"\n", *out_part);
1988 return VPR_OK;
1989 }
1990
1991 /* Test whether mod starts with modname, followed by a delimiter. */
1992 static Boolean
1993 ModMatch(const char *mod, const char *modname, char endc)
1994 {
1995 size_t n = strlen(modname);
1996 return strncmp(mod, modname, n) == 0 &&
1997 (mod[n] == endc || mod[n] == ':');
1998 }
1999
2000 /* Test whether mod starts with modname, followed by a delimiter or '='. */
2001 static inline Boolean
2002 ModMatchEq(const char *mod, const char *modname, char endc)
2003 {
2004 size_t n = strlen(modname);
2005 return strncmp(mod, modname, n) == 0 &&
2006 (mod[n] == endc || mod[n] == ':' || mod[n] == '=');
2007 }
2008
2009 static Boolean
2010 TryParseIntBase0(const char **pp, int *out_num)
2011 {
2012 char *end;
2013 long n;
2014
2015 errno = 0;
2016 n = strtol(*pp, &end, 0);
2017 if ((n == LONG_MIN || n == LONG_MAX) && errno == ERANGE)
2018 return FALSE;
2019 if (n < INT_MIN || n > INT_MAX)
2020 return FALSE;
2021
2022 *pp = end;
2023 *out_num = (int)n;
2024 return TRUE;
2025 }
2026
2027 static Boolean
2028 TryParseSize(const char **pp, size_t *out_num)
2029 {
2030 char *end;
2031 unsigned long n;
2032
2033 if (!ch_isdigit(**pp))
2034 return FALSE;
2035
2036 errno = 0;
2037 n = strtoul(*pp, &end, 10);
2038 if (n == ULONG_MAX && errno == ERANGE)
2039 return FALSE;
2040 if (n > SIZE_MAX)
2041 return FALSE;
2042
2043 *pp = end;
2044 *out_num = (size_t)n;
2045 return TRUE;
2046 }
2047
2048 static Boolean
2049 TryParseChar(const char **pp, int base, char *out_ch)
2050 {
2051 char *end;
2052 unsigned long n;
2053
2054 if (!ch_isalnum(**pp))
2055 return FALSE;
2056
2057 errno = 0;
2058 n = strtoul(*pp, &end, base);
2059 if (n == ULONG_MAX && errno == ERANGE)
2060 return FALSE;
2061 if (n > UCHAR_MAX)
2062 return FALSE;
2063
2064 *pp = end;
2065 *out_ch = (char)n;
2066 return TRUE;
2067 }
2068
2069 /* :@var (at) ...${var}...@ */
2070 static ApplyModifierResult
2071 ApplyModifier_Loop(const char **pp, ApplyModifiersState *st)
2072 {
2073 struct ModifyWord_LoopArgs args;
2074 char prev_sep;
2075 VarEvalFlags eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2076 VarParseResult res;
2077
2078 args.ctx = st->ctxt;
2079
2080 (*pp)++; /* Skip the first '@' */
2081 res = ParseModifierPart(pp, '@', eflags, st,
2082 &args.tvar, NULL, NULL, NULL);
2083 if (res != VPR_OK)
2084 return AMR_CLEANUP;
2085 if (DEBUG(LINT) && strchr(args.tvar, '$') != NULL) {
2086 Parse_Error(PARSE_FATAL,
2087 "In the :@ modifier of \"%s\", the variable name \"%s\" "
2088 "must not contain a dollar.",
2089 st->v->name, args.tvar);
2090 return AMR_CLEANUP;
2091 }
2092
2093 res = ParseModifierPart(pp, '@', eflags, st,
2094 &args.str, NULL, NULL, NULL);
2095 if (res != VPR_OK)
2096 return AMR_CLEANUP;
2097
2098 args.eflags = st->eflags & (VARE_UNDEFERR | VARE_WANTRES);
2099 prev_sep = st->sep;
2100 st->sep = ' '; /* XXX: should be st->sep for consistency */
2101 st->newVal = ModifyWords(st->val, ModifyWord_Loop, &args,
2102 st->oneBigWord, st->sep);
2103 st->sep = prev_sep;
2104 Var_Delete(args.tvar, st->ctxt);
2105 free(args.tvar);
2106 free(args.str);
2107 return AMR_OK;
2108 }
2109
2110 /* :Ddefined or :Uundefined */
2111 static ApplyModifierResult
2112 ApplyModifier_Defined(const char **pp, ApplyModifiersState *st)
2113 {
2114 Buffer buf;
2115 const char *p;
2116
2117 VarEvalFlags eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2118 if (st->eflags & VARE_WANTRES) {
2119 if ((**pp == 'D') == !(st->exprFlags & VEF_UNDEF))
2120 eflags |= VARE_WANTRES;
2121 }
2122
2123 Buf_Init(&buf, 0);
2124 p = *pp + 1;
2125 while (*p != st->endc && *p != ':' && *p != '\0') {
2126
2127 /* Escaped delimiter or other special character */
2128 if (*p == '\\') {
2129 char c = p[1];
2130 if (c == st->endc || c == ':' || c == '$' || c == '\\') {
2131 Buf_AddByte(&buf, c);
2132 p += 2;
2133 continue;
2134 }
2135 }
2136
2137 /* Nested variable expression */
2138 if (*p == '$') {
2139 const char *nested_val;
2140 void *nested_val_freeIt;
2141
2142 (void)Var_Parse(&p, st->ctxt, eflags,
2143 &nested_val, &nested_val_freeIt);
2144 /* TODO: handle errors */
2145 Buf_AddStr(&buf, nested_val);
2146 free(nested_val_freeIt);
2147 continue;
2148 }
2149
2150 /* Ordinary text */
2151 Buf_AddByte(&buf, *p);
2152 p++;
2153 }
2154 *pp = p;
2155
2156 ApplyModifiersState_Define(st);
2157
2158 if (eflags & VARE_WANTRES) {
2159 st->newVal = Buf_Destroy(&buf, FALSE);
2160 } else {
2161 st->newVal = st->val;
2162 Buf_Destroy(&buf, TRUE);
2163 }
2164 return AMR_OK;
2165 }
2166
2167 /* :L */
2168 static ApplyModifierResult
2169 ApplyModifier_Literal(const char **pp, ApplyModifiersState *st)
2170 {
2171 ApplyModifiersState_Define(st);
2172 st->newVal = bmake_strdup(st->v->name);
2173 (*pp)++;
2174 return AMR_OK;
2175 }
2176
2177 static Boolean
2178 TryParseTime(const char **pp, time_t *out_time)
2179 {
2180 char *end;
2181 unsigned long n;
2182
2183 if (!ch_isdigit(**pp))
2184 return FALSE;
2185
2186 errno = 0;
2187 n = strtoul(*pp, &end, 10);
2188 if (n == ULONG_MAX && errno == ERANGE)
2189 return FALSE;
2190
2191 *pp = end;
2192 *out_time = (time_t)n; /* ignore possible truncation for now */
2193 return TRUE;
2194 }
2195
2196 /* :gmtime */
2197 static ApplyModifierResult
2198 ApplyModifier_Gmtime(const char **pp, ApplyModifiersState *st)
2199 {
2200 time_t utc;
2201
2202 const char *mod = *pp;
2203 if (!ModMatchEq(mod, "gmtime", st->endc))
2204 return AMR_UNKNOWN;
2205
2206 if (mod[6] == '=') {
2207 const char *arg = mod + 7;
2208 if (!TryParseTime(&arg, &utc)) {
2209 Parse_Error(PARSE_FATAL, "Invalid time value: %s\n", mod + 7);
2210 return AMR_CLEANUP;
2211 }
2212 *pp = arg;
2213 } else {
2214 utc = 0;
2215 *pp = mod + 6;
2216 }
2217 st->newVal = VarStrftime(st->val, TRUE, utc);
2218 return AMR_OK;
2219 }
2220
2221 /* :localtime */
2222 static ApplyModifierResult
2223 ApplyModifier_Localtime(const char **pp, ApplyModifiersState *st)
2224 {
2225 time_t utc;
2226
2227 const char *mod = *pp;
2228 if (!ModMatchEq(mod, "localtime", st->endc))
2229 return AMR_UNKNOWN;
2230
2231 if (mod[9] == '=') {
2232 const char *arg = mod + 10;
2233 if (!TryParseTime(&arg, &utc)) {
2234 Parse_Error(PARSE_FATAL, "Invalid time value: %s\n", mod + 10);
2235 return AMR_CLEANUP;
2236 }
2237 *pp = arg;
2238 } else {
2239 utc = 0;
2240 *pp = mod + 9;
2241 }
2242 st->newVal = VarStrftime(st->val, FALSE, utc);
2243 return AMR_OK;
2244 }
2245
2246 /* :hash */
2247 static ApplyModifierResult
2248 ApplyModifier_Hash(const char **pp, ApplyModifiersState *st)
2249 {
2250 if (!ModMatch(*pp, "hash", st->endc))
2251 return AMR_UNKNOWN;
2252
2253 st->newVal = VarHash(st->val);
2254 *pp += 4;
2255 return AMR_OK;
2256 }
2257
2258 /* :P */
2259 static ApplyModifierResult
2260 ApplyModifier_Path(const char **pp, ApplyModifiersState *st)
2261 {
2262 GNode *gn;
2263 char *path;
2264
2265 ApplyModifiersState_Define(st);
2266
2267 gn = Targ_FindNode(st->v->name);
2268 if (gn == NULL || gn->type & OP_NOPATH) {
2269 path = NULL;
2270 } else if (gn->path != NULL) {
2271 path = bmake_strdup(gn->path);
2272 } else {
2273 SearchPath *searchPath = Suff_FindPath(gn);
2274 path = Dir_FindFile(st->v->name, searchPath);
2275 }
2276 if (path == NULL)
2277 path = bmake_strdup(st->v->name);
2278 st->newVal = path;
2279
2280 (*pp)++;
2281 return AMR_OK;
2282 }
2283
2284 /* :!cmd! */
2285 static ApplyModifierResult
2286 ApplyModifier_ShellCommand(const char **pp, ApplyModifiersState *st)
2287 {
2288 char *cmd;
2289 const char *errfmt;
2290 VarParseResult res;
2291
2292 (*pp)++;
2293 res = ParseModifierPart(pp, '!', st->eflags, st,
2294 &cmd, NULL, NULL, NULL);
2295 if (res != VPR_OK)
2296 return AMR_CLEANUP;
2297
2298 errfmt = NULL;
2299 if (st->eflags & VARE_WANTRES)
2300 st->newVal = Cmd_Exec(cmd, &errfmt);
2301 else
2302 st->newVal = emptyString;
2303 free(cmd);
2304
2305 if (errfmt != NULL)
2306 Error(errfmt, st->val); /* XXX: why still return AMR_OK? */
2307
2308 ApplyModifiersState_Define(st);
2309 return AMR_OK;
2310 }
2311
2312 /* The :range modifier generates an integer sequence as long as the words.
2313 * The :range=7 modifier generates an integer sequence from 1 to 7. */
2314 static ApplyModifierResult
2315 ApplyModifier_Range(const char **pp, ApplyModifiersState *st)
2316 {
2317 size_t n;
2318 Buffer buf;
2319 size_t i;
2320
2321 const char *mod = *pp;
2322 if (!ModMatchEq(mod, "range", st->endc))
2323 return AMR_UNKNOWN;
2324
2325 if (mod[5] == '=') {
2326 const char *p = mod + 6;
2327 if (!TryParseSize(&p, &n)) {
2328 Parse_Error(PARSE_FATAL, "Invalid number: %s\n", mod + 6);
2329 return AMR_CLEANUP;
2330 }
2331 *pp = p;
2332 } else {
2333 n = 0;
2334 *pp = mod + 5;
2335 }
2336
2337 if (n == 0) {
2338 Words words = Str_Words(st->val, FALSE);
2339 n = words.len;
2340 Words_Free(words);
2341 }
2342
2343 Buf_Init(&buf, 0);
2344
2345 for (i = 0; i < n; i++) {
2346 if (i != 0)
2347 Buf_AddByte(&buf, ' '); /* XXX: st->sep, for consistency */
2348 Buf_AddInt(&buf, 1 + (int)i);
2349 }
2350
2351 st->newVal = Buf_Destroy(&buf, FALSE);
2352 return AMR_OK;
2353 }
2354
2355 /* :Mpattern or :Npattern */
2356 static ApplyModifierResult
2357 ApplyModifier_Match(const char **pp, ApplyModifiersState *st)
2358 {
2359 const char *mod = *pp;
2360 Boolean copy = FALSE; /* pattern should be, or has been, copied */
2361 Boolean needSubst = FALSE;
2362 const char *endpat;
2363 char *pattern;
2364 ModifyWordsCallback callback;
2365
2366 /*
2367 * In the loop below, ignore ':' unless we are at (or back to) the
2368 * original brace level.
2369 * XXX This will likely not work right if $() and ${} are intermixed.
2370 */
2371 int nest = 0;
2372 const char *p;
2373 for (p = mod + 1; *p != '\0' && !(*p == ':' && nest == 0); p++) {
2374 if (*p == '\\' &&
2375 (p[1] == ':' || p[1] == st->endc || p[1] == st->startc)) {
2376 if (!needSubst)
2377 copy = TRUE;
2378 p++;
2379 continue;
2380 }
2381 if (*p == '$')
2382 needSubst = TRUE;
2383 if (*p == '(' || *p == '{')
2384 nest++;
2385 if (*p == ')' || *p == '}') {
2386 nest--;
2387 if (nest < 0)
2388 break;
2389 }
2390 }
2391 *pp = p;
2392 endpat = p;
2393
2394 if (copy) {
2395 char *dst;
2396 const char *src;
2397
2398 /* Compress the \:'s out of the pattern. */
2399 pattern = bmake_malloc((size_t)(endpat - (mod + 1)) + 1);
2400 dst = pattern;
2401 src = mod + 1;
2402 for (; src < endpat; src++, dst++) {
2403 if (src[0] == '\\' && src + 1 < endpat &&
2404 /* XXX: st->startc is missing here; see above */
2405 (src[1] == ':' || src[1] == st->endc))
2406 src++;
2407 *dst = *src;
2408 }
2409 *dst = '\0';
2410 endpat = dst;
2411 } else {
2412 pattern = bmake_strsedup(mod + 1, endpat);
2413 }
2414
2415 if (needSubst) {
2416 /* pattern contains embedded '$', so use Var_Subst to expand it. */
2417 char *old_pattern = pattern;
2418 (void)Var_Subst(pattern, st->ctxt, st->eflags, &pattern);
2419 /* TODO: handle errors */
2420 free(old_pattern);
2421 }
2422
2423 VAR_DEBUG3("Pattern[%s] for [%s] is [%s]\n", st->v->name, st->val, pattern);
2424
2425 callback = mod[0] == 'M' ? ModifyWord_Match : ModifyWord_NoMatch;
2426 st->newVal = ModifyWords(st->val, callback, pattern,
2427 st->oneBigWord, st->sep);
2428 free(pattern);
2429 return AMR_OK;
2430 }
2431
2432 /* :S,from,to, */
2433 static ApplyModifierResult
2434 ApplyModifier_Subst(const char **pp, ApplyModifiersState *st)
2435 {
2436 struct ModifyWord_SubstArgs args;
2437 char *lhs, *rhs;
2438 Boolean oneBigWord;
2439 VarParseResult res;
2440
2441 char delim = (*pp)[1];
2442 if (delim == '\0') {
2443 Error("Missing delimiter for :S modifier");
2444 (*pp)++;
2445 return AMR_CLEANUP;
2446 }
2447
2448 *pp += 2;
2449
2450 args.pflags = 0;
2451 args.matched = FALSE;
2452
2453 /*
2454 * If pattern begins with '^', it is anchored to the
2455 * start of the word -- skip over it and flag pattern.
2456 */
2457 if (**pp == '^') {
2458 args.pflags |= VARP_ANCHOR_START;
2459 (*pp)++;
2460 }
2461
2462 res = ParseModifierPart(pp, delim, st->eflags, st,
2463 &lhs, &args.lhsLen, &args.pflags, NULL);
2464 if (res != VPR_OK)
2465 return AMR_CLEANUP;
2466 args.lhs = lhs;
2467
2468 res = ParseModifierPart(pp, delim, st->eflags, st,
2469 &rhs, &args.rhsLen, NULL, &args);
2470 if (res != VPR_OK)
2471 return AMR_CLEANUP;
2472 args.rhs = rhs;
2473
2474 oneBigWord = st->oneBigWord;
2475 for (;; (*pp)++) {
2476 switch (**pp) {
2477 case 'g':
2478 args.pflags |= VARP_SUB_GLOBAL;
2479 continue;
2480 case '1':
2481 args.pflags |= VARP_SUB_ONE;
2482 continue;
2483 case 'W':
2484 oneBigWord = TRUE;
2485 continue;
2486 }
2487 break;
2488 }
2489
2490 st->newVal = ModifyWords(st->val, ModifyWord_Subst, &args,
2491 oneBigWord, st->sep);
2492
2493 free(lhs);
2494 free(rhs);
2495 return AMR_OK;
2496 }
2497
2498 #ifndef NO_REGEX
2499
2500 /* :C,from,to, */
2501 static ApplyModifierResult
2502 ApplyModifier_Regex(const char **pp, ApplyModifiersState *st)
2503 {
2504 char *re;
2505 struct ModifyWord_SubstRegexArgs args;
2506 Boolean oneBigWord;
2507 int error;
2508 VarParseResult res;
2509
2510 char delim = (*pp)[1];
2511 if (delim == '\0') {
2512 Error("Missing delimiter for :C modifier");
2513 (*pp)++;
2514 return AMR_CLEANUP;
2515 }
2516
2517 *pp += 2;
2518
2519 res = ParseModifierPart(pp, delim, st->eflags, st,
2520 &re, NULL, NULL, NULL);
2521 if (res != VPR_OK)
2522 return AMR_CLEANUP;
2523
2524 res = ParseModifierPart(pp, delim, st->eflags, st,
2525 &args.replace, NULL, NULL, NULL);
2526 if (args.replace == NULL) {
2527 free(re);
2528 return AMR_CLEANUP;
2529 }
2530
2531 args.pflags = 0;
2532 args.matched = FALSE;
2533 oneBigWord = st->oneBigWord;
2534 for (;; (*pp)++) {
2535 switch (**pp) {
2536 case 'g':
2537 args.pflags |= VARP_SUB_GLOBAL;
2538 continue;
2539 case '1':
2540 args.pflags |= VARP_SUB_ONE;
2541 continue;
2542 case 'W':
2543 oneBigWord = TRUE;
2544 continue;
2545 }
2546 break;
2547 }
2548
2549 error = regcomp(&args.re, re, REG_EXTENDED);
2550 free(re);
2551 if (error) {
2552 VarREError(error, &args.re, "Regex compilation error");
2553 free(args.replace);
2554 return AMR_CLEANUP;
2555 }
2556
2557 args.nsub = args.re.re_nsub + 1;
2558 if (args.nsub > 10)
2559 args.nsub = 10;
2560 st->newVal = ModifyWords(st->val, ModifyWord_SubstRegex, &args,
2561 oneBigWord, st->sep);
2562 regfree(&args.re);
2563 free(args.replace);
2564 return AMR_OK;
2565 }
2566 #endif
2567
2568 /* :Q, :q */
2569 static ApplyModifierResult
2570 ApplyModifier_Quote(const char **pp, ApplyModifiersState *st)
2571 {
2572 if ((*pp)[1] == st->endc || (*pp)[1] == ':') {
2573 st->newVal = VarQuote(st->val, **pp == 'q');
2574 (*pp)++;
2575 return AMR_OK;
2576 } else
2577 return AMR_UNKNOWN;
2578 }
2579
2580 static void
2581 ModifyWord_Copy(const char *word, SepBuf *buf, void *data MAKE_ATTR_UNUSED)
2582 {
2583 SepBuf_AddStr(buf, word);
2584 }
2585
2586 /* :ts<separator> */
2587 static ApplyModifierResult
2588 ApplyModifier_ToSep(const char **pp, ApplyModifiersState *st)
2589 {
2590 const char *sep = *pp + 2;
2591
2592 /* ":ts<any><endc>" or ":ts<any>:" */
2593 if (sep[0] != st->endc && (sep[1] == st->endc || sep[1] == ':')) {
2594 st->sep = sep[0];
2595 *pp = sep + 1;
2596 goto ok;
2597 }
2598
2599 /* ":ts<endc>" or ":ts:" */
2600 if (sep[0] == st->endc || sep[0] == ':') {
2601 st->sep = '\0'; /* no separator */
2602 *pp = sep;
2603 goto ok;
2604 }
2605
2606 /* ":ts<unrecognised><unrecognised>". */
2607 if (sep[0] != '\\') {
2608 (*pp)++; /* just for backwards compatibility */
2609 return AMR_BAD;
2610 }
2611
2612 /* ":ts\n" */
2613 if (sep[1] == 'n') {
2614 st->sep = '\n';
2615 *pp = sep + 2;
2616 goto ok;
2617 }
2618
2619 /* ":ts\t" */
2620 if (sep[1] == 't') {
2621 st->sep = '\t';
2622 *pp = sep + 2;
2623 goto ok;
2624 }
2625
2626 /* ":ts\x40" or ":ts\100" */
2627 {
2628 const char *p = sep + 1;
2629 int base = 8; /* assume octal */
2630
2631 if (sep[1] == 'x') {
2632 base = 16;
2633 p++;
2634 } else if (!ch_isdigit(sep[1])) {
2635 (*pp)++; /* just for backwards compatibility */
2636 return AMR_BAD; /* ":ts<backslash><unrecognised>". */
2637 }
2638
2639 if (!TryParseChar(&p, base, &st->sep)) {
2640 Parse_Error(PARSE_FATAL, "Invalid character number: %s\n", p);
2641 return AMR_CLEANUP;
2642 }
2643 if (*p != ':' && *p != st->endc) {
2644 (*pp)++; /* just for backwards compatibility */
2645 return AMR_BAD;
2646 }
2647
2648 *pp = p;
2649 }
2650
2651 ok:
2652 st->newVal = ModifyWords(st->val, ModifyWord_Copy, NULL,
2653 st->oneBigWord, st->sep);
2654 return AMR_OK;
2655 }
2656
2657 /* :tA, :tu, :tl, :ts<separator>, etc. */
2658 static ApplyModifierResult
2659 ApplyModifier_To(const char **pp, ApplyModifiersState *st)
2660 {
2661 const char *mod = *pp;
2662 assert(mod[0] == 't');
2663
2664 if (mod[1] == st->endc || mod[1] == ':' || mod[1] == '\0') {
2665 *pp = mod + 1;
2666 return AMR_BAD; /* Found ":t<endc>" or ":t:". */
2667 }
2668
2669 if (mod[1] == 's')
2670 return ApplyModifier_ToSep(pp, st);
2671
2672 if (mod[2] != st->endc && mod[2] != ':') {
2673 *pp = mod + 1;
2674 return AMR_BAD; /* Found ":t<unrecognised><unrecognised>". */
2675 }
2676
2677 /* Check for two-character options: ":tu", ":tl" */
2678 if (mod[1] == 'A') { /* absolute path */
2679 st->newVal = ModifyWords(st->val, ModifyWord_Realpath, NULL,
2680 st->oneBigWord, st->sep);
2681 *pp = mod + 2;
2682 return AMR_OK;
2683 }
2684
2685 if (mod[1] == 'u') { /* :tu */
2686 size_t i;
2687 size_t len = strlen(st->val);
2688 st->newVal = bmake_malloc(len + 1);
2689 for (i = 0; i < len + 1; i++)
2690 st->newVal[i] = ch_toupper(st->val[i]);
2691 *pp = mod + 2;
2692 return AMR_OK;
2693 }
2694
2695 if (mod[1] == 'l') { /* :tl */
2696 size_t i;
2697 size_t len = strlen(st->val);
2698 st->newVal = bmake_malloc(len + 1);
2699 for (i = 0; i < len + 1; i++)
2700 st->newVal[i] = ch_tolower(st->val[i]);
2701 *pp = mod + 2;
2702 return AMR_OK;
2703 }
2704
2705 if (mod[1] == 'W' || mod[1] == 'w') { /* :tW, :tw */
2706 st->oneBigWord = mod[1] == 'W';
2707 st->newVal = st->val;
2708 *pp = mod + 2;
2709 return AMR_OK;
2710 }
2711
2712 /* Found ":t<unrecognised>:" or ":t<unrecognised><endc>". */
2713 *pp = mod + 1;
2714 return AMR_BAD;
2715 }
2716
2717 /* :[#], :[1], :[-1..1], etc. */
2718 static ApplyModifierResult
2719 ApplyModifier_Words(const char **pp, ApplyModifiersState *st)
2720 {
2721 char *estr;
2722 int first, last;
2723 VarParseResult res;
2724 const char *p;
2725
2726 (*pp)++; /* skip the '[' */
2727 res = ParseModifierPart(pp, ']', st->eflags, st,
2728 &estr, NULL, NULL, NULL);
2729 if (res != VPR_OK)
2730 return AMR_CLEANUP;
2731
2732 /* now *pp points just after the closing ']' */
2733 if (**pp != ':' && **pp != st->endc)
2734 goto bad_modifier; /* Found junk after ']' */
2735
2736 if (estr[0] == '\0')
2737 goto bad_modifier; /* empty square brackets in ":[]". */
2738
2739 if (estr[0] == '#' && estr[1] == '\0') { /* Found ":[#]" */
2740 if (st->oneBigWord) {
2741 st->newVal = bmake_strdup("1");
2742 } else {
2743 Buffer buf;
2744
2745 Words words = Str_Words(st->val, FALSE);
2746 size_t ac = words.len;
2747 Words_Free(words);
2748
2749 Buf_Init(&buf, 4); /* 3 digits + '\0' is usually enough */
2750 Buf_AddInt(&buf, (int)ac);
2751 st->newVal = Buf_Destroy(&buf, FALSE);
2752 }
2753 goto ok;
2754 }
2755
2756 if (estr[0] == '*' && estr[1] == '\0') {
2757 /* Found ":[*]" */
2758 st->oneBigWord = TRUE;
2759 st->newVal = st->val;
2760 goto ok;
2761 }
2762
2763 if (estr[0] == '@' && estr[1] == '\0') {
2764 /* Found ":[@]" */
2765 st->oneBigWord = FALSE;
2766 st->newVal = st->val;
2767 goto ok;
2768 }
2769
2770 /*
2771 * We expect estr to contain a single integer for :[N], or two integers
2772 * separated by ".." for :[start..end].
2773 */
2774 p = estr;
2775 if (!TryParseIntBase0(&p, &first))
2776 goto bad_modifier; /* Found junk instead of a number */
2777
2778 if (p[0] == '\0') { /* Found only one integer in :[N] */
2779 last = first;
2780 } else if (p[0] == '.' && p[1] == '.' && p[2] != '\0') {
2781 /* Expecting another integer after ".." */
2782 p += 2;
2783 if (!TryParseIntBase0(&p, &last) || *p != '\0')
2784 goto bad_modifier; /* Found junk after ".." */
2785 } else
2786 goto bad_modifier; /* Found junk instead of ".." */
2787
2788 /*
2789 * Now first and last are properly filled in, but we still have to check
2790 * for 0 as a special case.
2791 */
2792 if (first == 0 && last == 0) {
2793 /* ":[0]" or perhaps ":[0..0]" */
2794 st->oneBigWord = TRUE;
2795 st->newVal = st->val;
2796 goto ok;
2797 }
2798
2799 /* ":[0..N]" or ":[N..0]" */
2800 if (first == 0 || last == 0)
2801 goto bad_modifier;
2802
2803 /* Normal case: select the words described by first and last. */
2804 st->newVal = VarSelectWords(st->sep, st->oneBigWord, st->val, first, last);
2805
2806 ok:
2807 free(estr);
2808 return AMR_OK;
2809
2810 bad_modifier:
2811 free(estr);
2812 return AMR_BAD;
2813 }
2814
2815 static int
2816 str_cmp_asc(const void *a, const void *b)
2817 {
2818 return strcmp(*(const char * const *)a, *(const char * const *)b);
2819 }
2820
2821 static int
2822 str_cmp_desc(const void *a, const void *b)
2823 {
2824 return strcmp(*(const char * const *)b, *(const char * const *)a);
2825 }
2826
2827 /* :O (order ascending) or :Or (order descending) or :Ox (shuffle) */
2828 static ApplyModifierResult
2829 ApplyModifier_Order(const char **pp, ApplyModifiersState *st)
2830 {
2831 const char *mod = (*pp)++; /* skip past the 'O' in any case */
2832
2833 Words words = Str_Words(st->val, FALSE);
2834
2835 if (mod[1] == st->endc || mod[1] == ':') {
2836 /* :O sorts ascending */
2837 qsort(words.words, words.len, sizeof(char *), str_cmp_asc);
2838
2839 } else if ((mod[1] == 'r' || mod[1] == 'x') &&
2840 (mod[2] == st->endc || mod[2] == ':')) {
2841 (*pp)++;
2842
2843 if (mod[1] == 'r') {
2844 /* :Or sorts descending */
2845 qsort(words.words, words.len, sizeof(char *), str_cmp_desc);
2846
2847 } else {
2848 /* :Ox shuffles
2849 *
2850 * We will use [ac..2] range for mod factors. This will produce
2851 * random numbers in [(ac-1)..0] interval, and minimal
2852 * reasonable value for mod factor is 2 (the mod 1 will produce
2853 * 0 with probability 1).
2854 */
2855 size_t i;
2856 for (i = words.len - 1; i > 0; i--) {
2857 size_t rndidx = (size_t)random() % (i + 1);
2858 char *t = words.words[i];
2859 words.words[i] = words.words[rndidx];
2860 words.words[rndidx] = t;
2861 }
2862 }
2863 } else {
2864 Words_Free(words);
2865 return AMR_BAD;
2866 }
2867
2868 st->newVal = Words_JoinFree(words);
2869 return AMR_OK;
2870 }
2871
2872 /* :? then : else */
2873 static ApplyModifierResult
2874 ApplyModifier_IfElse(const char **pp, ApplyModifiersState *st)
2875 {
2876 char *then_expr, *else_expr;
2877 VarParseResult res;
2878
2879 Boolean value = FALSE;
2880 VarEvalFlags then_eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2881 VarEvalFlags else_eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2882
2883 int cond_rc = COND_PARSE; /* anything other than COND_INVALID */
2884 if (st->eflags & VARE_WANTRES) {
2885 cond_rc = Cond_EvalCondition(st->v->name, &value);
2886 if (cond_rc != COND_INVALID && value)
2887 then_eflags |= VARE_WANTRES;
2888 if (cond_rc != COND_INVALID && !value)
2889 else_eflags |= VARE_WANTRES;
2890 }
2891
2892 (*pp)++; /* skip past the '?' */
2893 res = ParseModifierPart(pp, ':', then_eflags, st,
2894 &then_expr, NULL, NULL, NULL);
2895 if (res != VPR_OK)
2896 return AMR_CLEANUP;
2897
2898 res = ParseModifierPart(pp, st->endc, else_eflags, st,
2899 &else_expr, NULL, NULL, NULL);
2900 if (res != VPR_OK)
2901 return AMR_CLEANUP;
2902
2903 (*pp)--;
2904 if (cond_rc == COND_INVALID) {
2905 Error("Bad conditional expression `%s' in %s?%s:%s",
2906 st->v->name, st->v->name, then_expr, else_expr);
2907 return AMR_CLEANUP;
2908 }
2909
2910 if (value) {
2911 st->newVal = then_expr;
2912 free(else_expr);
2913 } else {
2914 st->newVal = else_expr;
2915 free(then_expr);
2916 }
2917 ApplyModifiersState_Define(st);
2918 return AMR_OK;
2919 }
2920
2921 /*
2922 * The ::= modifiers actually assign a value to the variable.
2923 * Their main purpose is in supporting modifiers of .for loop
2924 * iterators and other obscure uses. They always expand to
2925 * nothing. In a target rule that would otherwise expand to an
2926 * empty line they can be preceded with @: to keep make happy.
2927 * Eg.
2928 *
2929 * foo: .USE
2930 * .for i in ${.TARGET} ${.TARGET:R}.gz
2931 * @: ${t::=$i}
2932 * @echo blah ${t:T}
2933 * .endfor
2934 *
2935 * ::=<str> Assigns <str> as the new value of variable.
2936 * ::?=<str> Assigns <str> as value of variable if
2937 * it was not already set.
2938 * ::+=<str> Appends <str> to variable.
2939 * ::!=<cmd> Assigns output of <cmd> as the new value of
2940 * variable.
2941 */
2942 static ApplyModifierResult
2943 ApplyModifier_Assign(const char **pp, ApplyModifiersState *st)
2944 {
2945 GNode *v_ctxt;
2946 char delim;
2947 char *val;
2948 VarParseResult res;
2949
2950 const char *mod = *pp;
2951 const char *op = mod + 1;
2952
2953 if (op[0] == '=')
2954 goto ok;
2955 if ((op[0] == '!' || op[0] == '+' || op[0] == '?') && op[1] == '=')
2956 goto ok;
2957 return AMR_UNKNOWN; /* "::<unrecognised>" */
2958 ok:
2959
2960 if (st->v->name[0] == '\0') {
2961 *pp = mod + 1;
2962 return AMR_BAD;
2963 }
2964
2965 v_ctxt = st->ctxt; /* context where v belongs */
2966 if (!(st->exprFlags & VEF_UNDEF) && st->ctxt != VAR_GLOBAL) {
2967 Var *gv = VarFind(st->v->name, st->ctxt, 0);
2968 if (gv == NULL)
2969 v_ctxt = VAR_GLOBAL;
2970 else
2971 VarFreeEnv(gv, TRUE);
2972 }
2973
2974 switch (op[0]) {
2975 case '+':
2976 case '?':
2977 case '!':
2978 *pp = mod + 3;
2979 break;
2980 default:
2981 *pp = mod + 2;
2982 break;
2983 }
2984
2985 delim = st->startc == '(' ? ')' : '}';
2986 res = ParseModifierPart(pp, delim, st->eflags, st, &val, NULL, NULL, NULL);
2987 if (res != VPR_OK)
2988 return AMR_CLEANUP;
2989
2990 (*pp)--;
2991
2992 if (st->eflags & VARE_WANTRES) {
2993 switch (op[0]) {
2994 case '+':
2995 Var_Append(st->v->name, val, v_ctxt);
2996 break;
2997 case '!': {
2998 const char *errfmt;
2999 char *cmd_output = Cmd_Exec(val, &errfmt);
3000 if (errfmt)
3001 Error(errfmt, val);
3002 else
3003 Var_Set(st->v->name, cmd_output, v_ctxt);
3004 free(cmd_output);
3005 break;
3006 }
3007 case '?':
3008 if (!(st->exprFlags & VEF_UNDEF))
3009 break;
3010 /* FALLTHROUGH */
3011 default:
3012 Var_Set(st->v->name, val, v_ctxt);
3013 break;
3014 }
3015 }
3016 free(val);
3017 st->newVal = emptyString;
3018 return AMR_OK;
3019 }
3020
3021 /* :_=...
3022 * remember current value */
3023 static ApplyModifierResult
3024 ApplyModifier_Remember(const char **pp, ApplyModifiersState *st)
3025 {
3026 const char *mod = *pp;
3027 if (!ModMatchEq(mod, "_", st->endc))
3028 return AMR_UNKNOWN;
3029
3030 if (mod[1] == '=') {
3031 size_t n = strcspn(mod + 2, ":)}");
3032 char *name = bmake_strldup(mod + 2, n);
3033 Var_Set(name, st->val, st->ctxt);
3034 free(name);
3035 *pp = mod + 2 + n;
3036 } else {
3037 Var_Set("_", st->val, st->ctxt);
3038 *pp = mod + 1;
3039 }
3040 st->newVal = st->val;
3041 return AMR_OK;
3042 }
3043
3044 /* Apply the given function to each word of the variable value,
3045 * for a single-letter modifier such as :H, :T. */
3046 static ApplyModifierResult
3047 ApplyModifier_WordFunc(const char **pp, ApplyModifiersState *st,
3048 ModifyWordsCallback modifyWord)
3049 {
3050 char delim = (*pp)[1];
3051 if (delim != st->endc && delim != ':')
3052 return AMR_UNKNOWN;
3053
3054 st->newVal = ModifyWords(st->val, modifyWord, NULL,
3055 st->oneBigWord, st->sep);
3056 (*pp)++;
3057 return AMR_OK;
3058 }
3059
3060 static ApplyModifierResult
3061 ApplyModifier_Unique(const char **pp, ApplyModifiersState *st)
3062 {
3063 if ((*pp)[1] == st->endc || (*pp)[1] == ':') {
3064 st->newVal = VarUniq(st->val);
3065 (*pp)++;
3066 return AMR_OK;
3067 } else
3068 return AMR_UNKNOWN;
3069 }
3070
3071 #ifdef SYSVVARSUB
3072 /* :from=to */
3073 static ApplyModifierResult
3074 ApplyModifier_SysV(const char **pp, ApplyModifiersState *st)
3075 {
3076 char *lhs, *rhs;
3077 VarParseResult res;
3078
3079 const char *mod = *pp;
3080 Boolean eqFound = FALSE;
3081
3082 /*
3083 * First we make a pass through the string trying to verify it is a
3084 * SysV-make-style translation. It must be: <lhs>=<rhs>
3085 */
3086 int depth = 1;
3087 const char *p = mod;
3088 while (*p != '\0' && depth > 0) {
3089 if (*p == '=') { /* XXX: should also test depth == 1 */
3090 eqFound = TRUE;
3091 /* continue looking for st->endc */
3092 } else if (*p == st->endc)
3093 depth--;
3094 else if (*p == st->startc)
3095 depth++;
3096 if (depth > 0)
3097 p++;
3098 }
3099 if (*p != st->endc || !eqFound)
3100 return AMR_UNKNOWN;
3101
3102 *pp = mod;
3103 res = ParseModifierPart(pp, '=', st->eflags, st,
3104 &lhs, NULL, NULL, NULL);
3105 if (res != VPR_OK)
3106 return AMR_CLEANUP;
3107
3108 /* The SysV modifier lasts until the end of the variable expression. */
3109 res = ParseModifierPart(pp, st->endc, st->eflags, st,
3110 &rhs, NULL, NULL, NULL);
3111 if (res != VPR_OK)
3112 return AMR_CLEANUP;
3113
3114 (*pp)--;
3115 if (lhs[0] == '\0' && st->val[0] == '\0') {
3116 st->newVal = st->val; /* special case */
3117 } else {
3118 struct ModifyWord_SYSVSubstArgs args = {st->ctxt, lhs, rhs};
3119 st->newVal = ModifyWords(st->val, ModifyWord_SYSVSubst, &args,
3120 st->oneBigWord, st->sep);
3121 }
3122 free(lhs);
3123 free(rhs);
3124 return AMR_OK;
3125 }
3126 #endif
3127
3128 #ifdef SUNSHCMD
3129 /* :sh */
3130 static ApplyModifierResult
3131 ApplyModifier_SunShell(const char **pp, ApplyModifiersState *st)
3132 {
3133 const char *p = *pp;
3134 if (p[1] == 'h' && (p[2] == st->endc || p[2] == ':')) {
3135 if (st->eflags & VARE_WANTRES) {
3136 const char *errfmt;
3137 st->newVal = Cmd_Exec(st->val, &errfmt);
3138 if (errfmt)
3139 Error(errfmt, st->val);
3140 } else
3141 st->newVal = emptyString;
3142 *pp = p + 2;
3143 return AMR_OK;
3144 } else
3145 return AMR_UNKNOWN;
3146 }
3147 #endif
3148
3149 static void
3150 LogBeforeApply(const ApplyModifiersState *st, const char *mod, const char endc)
3151 {
3152 char eflags_str[VarEvalFlags_ToStringSize];
3153 char vflags_str[VarFlags_ToStringSize];
3154 char exprflags_str[VarExprFlags_ToStringSize];
3155 Boolean is_single_char = mod[0] != '\0' &&
3156 (mod[1] == endc || mod[1] == ':');
3157
3158 /* At this point, only the first character of the modifier can
3159 * be used since the end of the modifier is not yet known. */
3160 debug_printf("Applying ${%s:%c%s} to \"%s\" (%s, %s, %s)\n",
3161 st->v->name, mod[0], is_single_char ? "" : "...", st->val,
3162 Enum_FlagsToString(eflags_str, sizeof eflags_str,
3163 st->eflags, VarEvalFlags_ToStringSpecs),
3164 Enum_FlagsToString(vflags_str, sizeof vflags_str,
3165 st->v->flags, VarFlags_ToStringSpecs),
3166 Enum_FlagsToString(exprflags_str, sizeof exprflags_str,
3167 st->exprFlags,
3168 VarExprFlags_ToStringSpecs));
3169 }
3170
3171 static void
3172 LogAfterApply(ApplyModifiersState *st, const char *p, const char *mod)
3173 {
3174 char eflags_str[VarEvalFlags_ToStringSize];
3175 char vflags_str[VarFlags_ToStringSize];
3176 char exprflags_str[VarExprFlags_ToStringSize];
3177 const char *quot = st->newVal == var_Error ? "" : "\"";
3178 const char *newVal = st->newVal == var_Error ? "error" : st->newVal;
3179
3180 debug_printf("Result of ${%s:%.*s} is %s%s%s (%s, %s, %s)\n",
3181 st->v->name, (int)(p - mod), mod, quot, newVal, quot,
3182 Enum_FlagsToString(eflags_str, sizeof eflags_str,
3183 st->eflags, VarEvalFlags_ToStringSpecs),
3184 Enum_FlagsToString(vflags_str, sizeof vflags_str,
3185 st->v->flags, VarFlags_ToStringSpecs),
3186 Enum_FlagsToString(exprflags_str, sizeof exprflags_str,
3187 st->exprFlags,
3188 VarExprFlags_ToStringSpecs));
3189 }
3190
3191 static ApplyModifierResult
3192 ApplyModifier(const char **pp, ApplyModifiersState *st)
3193 {
3194 switch (**pp) {
3195 case ':':
3196 return ApplyModifier_Assign(pp, st);
3197 case '@':
3198 return ApplyModifier_Loop(pp, st);
3199 case '_':
3200 return ApplyModifier_Remember(pp, st);
3201 case 'D':
3202 case 'U':
3203 return ApplyModifier_Defined(pp, st);
3204 case 'L':
3205 return ApplyModifier_Literal(pp, st);
3206 case 'P':
3207 return ApplyModifier_Path(pp, st);
3208 case '!':
3209 return ApplyModifier_ShellCommand(pp, st);
3210 case '[':
3211 return ApplyModifier_Words(pp, st);
3212 case 'g':
3213 return ApplyModifier_Gmtime(pp, st);
3214 case 'h':
3215 return ApplyModifier_Hash(pp, st);
3216 case 'l':
3217 return ApplyModifier_Localtime(pp, st);
3218 case 't':
3219 return ApplyModifier_To(pp, st);
3220 case 'N':
3221 case 'M':
3222 return ApplyModifier_Match(pp, st);
3223 case 'S':
3224 return ApplyModifier_Subst(pp, st);
3225 case '?':
3226 return ApplyModifier_IfElse(pp, st);
3227 #ifndef NO_REGEX
3228 case 'C':
3229 return ApplyModifier_Regex(pp, st);
3230 #endif
3231 case 'q':
3232 case 'Q':
3233 return ApplyModifier_Quote(pp, st);
3234 case 'T':
3235 return ApplyModifier_WordFunc(pp, st, ModifyWord_Tail);
3236 case 'H':
3237 return ApplyModifier_WordFunc(pp, st, ModifyWord_Head);
3238 case 'E':
3239 return ApplyModifier_WordFunc(pp, st, ModifyWord_Suffix);
3240 case 'R':
3241 return ApplyModifier_WordFunc(pp, st, ModifyWord_Root);
3242 case 'r':
3243 return ApplyModifier_Range(pp, st);
3244 case 'O':
3245 return ApplyModifier_Order(pp, st);
3246 case 'u':
3247 return ApplyModifier_Unique(pp, st);
3248 #ifdef SUNSHCMD
3249 case 's':
3250 return ApplyModifier_SunShell(pp, st);
3251 #endif
3252 default:
3253 return AMR_UNKNOWN;
3254 }
3255 }
3256
3257 static char *ApplyModifiers(const char **, char *, char, char, Var *,
3258 VarExprFlags *, GNode *, VarEvalFlags, void **);
3259
3260 typedef enum ApplyModifiersIndirectResult {
3261 AMIR_CONTINUE,
3262 AMIR_APPLY_MODS,
3263 AMIR_OUT
3264 } ApplyModifiersIndirectResult;
3265
3266 /* While expanding a variable expression, expand and apply indirect
3267 * modifiers such as in ${VAR:${M_indirect}}. */
3268 static ApplyModifiersIndirectResult
3269 ApplyModifiersIndirect(
3270 ApplyModifiersState *const st,
3271 const char **const inout_p,
3272 void **const out_freeIt
3273 ) {
3274 const char *p = *inout_p;
3275 const char *mods;
3276 void *mods_freeIt;
3277
3278 (void)Var_Parse(&p, st->ctxt, st->eflags, &mods, &mods_freeIt);
3279 /* TODO: handle errors */
3280
3281 /*
3282 * If we have not parsed up to st->endc or ':', we are not
3283 * interested. This means the expression ${VAR:${M_1}${M_2}}
3284 * is not accepted, but ${VAR:${M_1}:${M_2}} is.
3285 */
3286 if (mods[0] != '\0' && *p != '\0' && *p != ':' && *p != st->endc) {
3287 if (DEBUG(LINT))
3288 Parse_Error(PARSE_FATAL,
3289 "Missing delimiter ':' after indirect modifier \"%.*s\"",
3290 (int)(p - *inout_p), *inout_p);
3291
3292 free(mods_freeIt);
3293 /* XXX: apply_mods doesn't sound like "not interested". */
3294 /* XXX: Why is the indirect modifier parsed once more by
3295 * apply_mods? If any, p should be advanced to nested_p. */
3296 return AMIR_APPLY_MODS;
3297 }
3298
3299 VAR_DEBUG3("Indirect modifier \"%s\" from \"%.*s\"\n",
3300 mods, (int)(p - *inout_p), *inout_p);
3301
3302 if (mods[0] != '\0') {
3303 const char *rval_pp = mods;
3304 st->val = ApplyModifiers(&rval_pp, st->val, '\0', '\0', st->v,
3305 &st->exprFlags, st->ctxt, st->eflags,
3306 out_freeIt);
3307 if (st->val == var_Error
3308 || (st->val == varUndefined && !(st->eflags & VARE_UNDEFERR))
3309 || *rval_pp != '\0') {
3310 free(mods_freeIt);
3311 *inout_p = p;
3312 return AMIR_OUT; /* error already reported */
3313 }
3314 }
3315 free(mods_freeIt);
3316
3317 if (*p == ':')
3318 p++;
3319 else if (*p == '\0' && st->endc != '\0') {
3320 Error("Unclosed variable specification after complex "
3321 "modifier (expecting '%c') for %s", st->endc, st->v->name);
3322 *inout_p = p;
3323 return AMIR_OUT;
3324 }
3325
3326 *inout_p = p;
3327 return AMIR_CONTINUE;
3328 }
3329
3330 /* Apply any modifiers (such as :Mpattern or :@var@loop@ or :Q or ::=value). */
3331 static char *
3332 ApplyModifiers(
3333 const char **const pp, /* the parsing position, updated upon return */
3334 char *const val, /* the current value of the expression */
3335 char const startc, /* '(' or '{', or '\0' for indirect modifiers */
3336 char const endc, /* ')' or '}', or '\0' for indirect modifiers */
3337 Var *const v,
3338 VarExprFlags *const exprFlags,
3339 GNode *const ctxt, /* for looking up and modifying variables */
3340 VarEvalFlags const eflags,
3341 void **const out_freeIt /* free this after using the return value */
3342 ) {
3343 ApplyModifiersState st = {
3344 startc, endc, v, ctxt, eflags,
3345 val, /* .val */
3346 var_Error, /* .newVal */
3347 ' ', /* .sep */
3348 FALSE, /* .oneBigWord */
3349 *exprFlags /* .exprFlags */
3350 };
3351 const char *p;
3352 const char *mod;
3353 ApplyModifierResult res;
3354
3355 assert(startc == '(' || startc == '{' || startc == '\0');
3356 assert(endc == ')' || endc == '}' || endc == '\0');
3357 assert(val != NULL);
3358
3359 p = *pp;
3360 while (*p != '\0' && *p != endc) {
3361
3362 if (*p == '$') {
3363 ApplyModifiersIndirectResult amir;
3364 amir = ApplyModifiersIndirect(&st, &p, out_freeIt);
3365 if (amir == AMIR_CONTINUE)
3366 continue;
3367 if (amir == AMIR_OUT)
3368 goto out;
3369 }
3370 st.newVal = var_Error; /* default value, in case of errors */
3371 mod = p;
3372
3373 if (DEBUG(VAR))
3374 LogBeforeApply(&st, mod, endc);
3375
3376 res = ApplyModifier(&p, &st);
3377
3378 #ifdef SYSVVARSUB
3379 if (res == AMR_UNKNOWN) {
3380 assert(p == mod);
3381 res = ApplyModifier_SysV(&p, &st);
3382 }
3383 #endif
3384
3385 if (res == AMR_UNKNOWN) {
3386 Error("Unknown modifier '%c'", *mod);
3387 for (p++; *p != ':' && *p != st.endc && *p != '\0'; p++)
3388 continue;
3389 st.newVal = var_Error;
3390 }
3391 if (res == AMR_CLEANUP)
3392 goto cleanup;
3393 if (res == AMR_BAD)
3394 goto bad_modifier;
3395
3396 if (DEBUG(VAR))
3397 LogAfterApply(&st, p, mod);
3398
3399 if (st.newVal != st.val) {
3400 if (*out_freeIt) {
3401 free(st.val);
3402 *out_freeIt = NULL;
3403 }
3404 st.val = st.newVal;
3405 if (st.val != var_Error && st.val != varUndefined &&
3406 st.val != emptyString) {
3407 *out_freeIt = st.val;
3408 }
3409 }
3410 if (*p == '\0' && st.endc != '\0') {
3411 Error("Unclosed variable specification (expecting '%c') "
3412 "for \"%s\" (value \"%s\") modifier %c",
3413 st.endc, st.v->name, st.val, *mod);
3414 } else if (*p == ':') {
3415 p++;
3416 } else if (DEBUG(LINT) && *p != '\0' && *p != endc) {
3417 Parse_Error(PARSE_FATAL,
3418 "Missing delimiter ':' after modifier \"%.*s\"",
3419 (int)(p - mod), mod);
3420 /* TODO: propagate parse error to the enclosing expression */
3421 }
3422 }
3423 out:
3424 *pp = p;
3425 assert(st.val != NULL); /* Use var_Error or varUndefined instead. */
3426 *exprFlags = st.exprFlags;
3427 return st.val;
3428
3429 bad_modifier:
3430 Error("Bad modifier `:%.*s' for %s",
3431 (int)strcspn(mod, ":)}"), mod, st.v->name);
3432
3433 cleanup:
3434 *pp = p;
3435 free(*out_freeIt);
3436 *out_freeIt = NULL;
3437 *exprFlags = st.exprFlags;
3438 return var_Error;
3439 }
3440
3441 /* Only four of the local variables are treated specially as they are the
3442 * only four that will be set when dynamic sources are expanded. */
3443 static Boolean
3444 VarnameIsDynamic(const char *name, size_t len)
3445 {
3446 if (len == 1 || (len == 2 && (name[1] == 'F' || name[1] == 'D'))) {
3447 switch (name[0]) {
3448 case '@':
3449 case '%':
3450 case '*':
3451 case '!':
3452 return TRUE;
3453 }
3454 return FALSE;
3455 }
3456
3457 if ((len == 7 || len == 8) && name[0] == '.' && ch_isupper(name[1])) {
3458 return strcmp(name, ".TARGET") == 0 ||
3459 strcmp(name, ".ARCHIVE") == 0 ||
3460 strcmp(name, ".PREFIX") == 0 ||
3461 strcmp(name, ".MEMBER") == 0;
3462 }
3463
3464 return FALSE;
3465 }
3466
3467 static const char *
3468 UndefinedShortVarValue(char varname, const GNode *ctxt, VarEvalFlags eflags)
3469 {
3470 if (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL) {
3471 /*
3472 * If substituting a local variable in a non-local context,
3473 * assume it's for dynamic source stuff. We have to handle
3474 * this specially and return the longhand for the variable
3475 * with the dollar sign escaped so it makes it back to the
3476 * caller. Only four of the local variables are treated
3477 * specially as they are the only four that will be set
3478 * when dynamic sources are expanded.
3479 */
3480 switch (varname) {
3481 case '@':
3482 return "$(.TARGET)";
3483 case '%':
3484 return "$(.MEMBER)";
3485 case '*':
3486 return "$(.PREFIX)";
3487 case '!':
3488 return "$(.ARCHIVE)";
3489 }
3490 }
3491 return eflags & VARE_UNDEFERR ? var_Error : varUndefined;
3492 }
3493
3494 /* Parse a variable name, until the end character or a colon, whichever
3495 * comes first. */
3496 static char *
3497 ParseVarname(const char **pp, char startc, char endc,
3498 GNode *ctxt, VarEvalFlags eflags,
3499 size_t *out_varname_len)
3500 {
3501 Buffer buf;
3502 const char *p = *pp;
3503 int depth = 1;
3504
3505 Buf_Init(&buf, 0);
3506
3507 while (*p != '\0') {
3508 /* Track depth so we can spot parse errors. */
3509 if (*p == startc)
3510 depth++;
3511 if (*p == endc) {
3512 if (--depth == 0)
3513 break;
3514 }
3515 if (*p == ':' && depth == 1)
3516 break;
3517
3518 /* A variable inside a variable, expand. */
3519 if (*p == '$') {
3520 void *freeIt;
3521 const char *rval;
3522 (void)Var_Parse(&p, ctxt, eflags, &rval, &freeIt);
3523 /* TODO: handle errors */
3524 Buf_AddStr(&buf, rval);
3525 free(freeIt);
3526 } else {
3527 Buf_AddByte(&buf, *p);
3528 p++;
3529 }
3530 }
3531 *pp = p;
3532 *out_varname_len = Buf_Len(&buf);
3533 return Buf_Destroy(&buf, FALSE);
3534 }
3535
3536 static Boolean
3537 ValidShortVarname(char varname, const char *start)
3538 {
3539 switch (varname) {
3540 case '\0':
3541 case ')':
3542 case '}':
3543 case ':':
3544 case '$':
3545 break; /* and continue below */
3546 default:
3547 return TRUE;
3548 }
3549
3550 if (!DEBUG(LINT))
3551 return FALSE;
3552
3553 if (varname == '$')
3554 Parse_Error(PARSE_FATAL,
3555 "To escape a dollar, use \\$, not $$, at \"%s\"", start);
3556 else if (varname == '\0')
3557 Parse_Error(PARSE_FATAL, "Dollar followed by nothing");
3558 else
3559 Parse_Error(PARSE_FATAL,
3560 "Invalid variable name '%c', at \"%s\"", varname, start);
3561
3562 return FALSE;
3563 }
3564
3565 /* Parse a single-character variable name such as $V or $@.
3566 * Return whether to continue parsing. */
3567 static Boolean
3568 ParseVarnameShort(
3569 char startc,
3570 const char **pp,
3571 GNode *ctxt,
3572 VarEvalFlags eflags,
3573 VarParseResult *out_FALSE_res,
3574 const char **out_FALSE_val,
3575 Var **out_TRUE_var
3576 ) {
3577 char name[2];
3578 Var *v;
3579
3580 /*
3581 * If it's not bounded by braces of some sort, life is much simpler.
3582 * We just need to check for the first character and return the
3583 * value if it exists.
3584 */
3585
3586 if (!ValidShortVarname(startc, *pp)) {
3587 (*pp)++;
3588 *out_FALSE_val = var_Error;
3589 *out_FALSE_res = VPR_PARSE_MSG;
3590 return FALSE;
3591 }
3592
3593 name[0] = startc;
3594 name[1] = '\0';
3595 v = VarFind(name, ctxt, TRUE);
3596 if (v == NULL) {
3597 *pp += 2;
3598
3599 *out_FALSE_val = UndefinedShortVarValue(startc, ctxt, eflags);
3600 if (DEBUG(LINT) && *out_FALSE_val == var_Error) {
3601 Parse_Error(PARSE_FATAL, "Variable \"%s\" is undefined", name);
3602 *out_FALSE_res = VPR_UNDEF_MSG;
3603 return FALSE;
3604 }
3605 *out_FALSE_res = eflags & VARE_UNDEFERR ? VPR_UNDEF_SILENT : VPR_OK;
3606 return FALSE;
3607 }
3608
3609 *out_TRUE_var = v;
3610 return TRUE;
3611 }
3612
3613 /* Parse a long variable name enclosed in braces or parentheses such as $(VAR)
3614 * or ${VAR}, up to the closing brace or parenthesis, or in the case of
3615 * ${VAR:Modifiers}, up to the ':' that starts the modifiers.
3616 * Return whether to continue parsing. */
3617 static Boolean
3618 ParseVarnameLong(
3619 const char **pp,
3620 char startc,
3621 GNode *ctxt,
3622 VarEvalFlags eflags,
3623
3624 VarParseResult *out_FALSE_res,
3625 const char **out_FALSE_val,
3626 void **out_FALSE_freePtr,
3627
3628 char *out_TRUE_endc,
3629 const char **out_TRUE_p,
3630 Var **out_TRUE_v,
3631 Boolean *out_TRUE_haveModifier,
3632 const char **out_TRUE_extraModifiers,
3633 Boolean *out_TRUE_dynamic,
3634 VarExprFlags *out_TRUE_exprFlags
3635 ) {
3636 size_t namelen;
3637 char *varname;
3638 Var *v;
3639 Boolean haveModifier;
3640 Boolean dynamic = FALSE;
3641
3642 const char *const start = *pp;
3643 char endc = startc == '(' ? ')' : '}';
3644
3645 const char *p = start + 2;
3646 varname = ParseVarname(&p, startc, endc, ctxt, eflags, &namelen);
3647
3648 if (*p == ':') {
3649 haveModifier = TRUE;
3650 } else if (*p == endc) {
3651 haveModifier = FALSE;
3652 } else {
3653 Parse_Error(PARSE_FATAL, "Unclosed variable \"%s\"", varname);
3654 *pp = p;
3655 free(varname);
3656 *out_FALSE_val = var_Error;
3657 *out_FALSE_res = VPR_PARSE_MSG;
3658 return FALSE;
3659 }
3660
3661 v = VarFind(varname, ctxt, TRUE);
3662
3663 /* At this point, p points just after the variable name,
3664 * either at ':' or at endc. */
3665
3666 /*
3667 * Check also for bogus D and F forms of local variables since we're
3668 * in a local context and the name is the right length.
3669 */
3670 if (v == NULL && ctxt != VAR_CMDLINE && ctxt != VAR_GLOBAL &&
3671 namelen == 2 && (varname[1] == 'F' || varname[1] == 'D') &&
3672 strchr("@%?*!<>", varname[0]) != NULL)
3673 {
3674 /*
3675 * Well, it's local -- go look for it.
3676 */
3677 char name[] = { varname[0], '\0' };
3678 v = VarFind(name, ctxt, 0);
3679
3680 if (v != NULL) {
3681 if (varname[1] == 'D') {
3682 *out_TRUE_extraModifiers = "H:";
3683 } else { /* F */
3684 *out_TRUE_extraModifiers = "T:";
3685 }
3686 }
3687 }
3688
3689 if (v == NULL) {
3690 /* Defer expansion of dynamic variables if they appear in non-local
3691 * context since they are not defined there. */
3692 dynamic = VarnameIsDynamic(varname, namelen) &&
3693 (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL);
3694
3695 if (!haveModifier) {
3696 p++; /* skip endc */
3697 *pp = p;
3698 if (dynamic) {
3699 char *pstr = bmake_strsedup(start, p);
3700 free(varname);
3701 *out_FALSE_res = VPR_OK;
3702 *out_FALSE_freePtr = pstr;
3703 *out_FALSE_val = pstr;
3704 return FALSE;
3705 }
3706
3707 if ((eflags & VARE_UNDEFERR) && (eflags & VARE_WANTRES) &&
3708 DEBUG(LINT))
3709 {
3710 Parse_Error(PARSE_FATAL, "Variable \"%s\" is undefined",
3711 varname);
3712 free(varname);
3713 *out_FALSE_res = VPR_UNDEF_MSG;
3714 *out_FALSE_val = var_Error;
3715 return FALSE;
3716 }
3717
3718 if (eflags & VARE_UNDEFERR) {
3719 free(varname);
3720 *out_FALSE_res = VPR_UNDEF_SILENT;
3721 *out_FALSE_val = var_Error;
3722 return FALSE;
3723 }
3724
3725 free(varname);
3726 *out_FALSE_res = VPR_OK;
3727 *out_FALSE_val = varUndefined;
3728 return FALSE;
3729 }
3730
3731 /* The variable expression is based on an undefined variable.
3732 * Nevertheless it needs a Var, for modifiers that access the
3733 * variable name, such as :L or :?.
3734 *
3735 * Most modifiers leave this expression in the "undefined" state
3736 * (VEF_UNDEF), only a few modifiers like :D, :U, :L, :P turn this
3737 * undefined expression into a defined expression (VEF_DEF).
3738 *
3739 * At the end, after applying all modifiers, if the expression
3740 * is still undefined, Var_Parse will return an empty string
3741 * instead of the actually computed value. */
3742 v = VarNew(varname, varname, "", 0);
3743 *out_TRUE_exprFlags = VEF_UNDEF;
3744 } else
3745 free(varname);
3746
3747 *out_TRUE_endc = endc;
3748 *out_TRUE_p = p;
3749 *out_TRUE_v = v;
3750 *out_TRUE_haveModifier = haveModifier;
3751 *out_TRUE_dynamic = dynamic;
3752 return TRUE;
3753 }
3754
3755 /*-
3756 *-----------------------------------------------------------------------
3757 * Var_Parse --
3758 * Given the start of a variable expression (such as $v, $(VAR),
3759 * ${VAR:Mpattern}), extract the variable name, possibly some
3760 * modifiers and find its value by applying the modifiers to the
3761 * original value.
3762 *
3763 * When parsing a condition in ParseEmptyArg, pp may also point to
3764 * the "y" of "empty(VARNAME:Modifiers)", which is syntactically
3765 * identical.
3766 *
3767 * Input:
3768 * str The string to parse
3769 * ctxt The context for the variable
3770 * flags Select the exact details of parsing
3771 * out_val_freeIt Must be freed by the caller after using out_val
3772 *
3773 * Results:
3774 * Returns the value of the variable expression, never NULL.
3775 * Returns var_Error if there was a parse error and VARE_UNDEFERR was
3776 * set.
3777 * Returns varUndefined if there was an undefined variable and
3778 * VARE_UNDEFERR was not set.
3779 *
3780 * Parsing should continue at *pp.
3781 * TODO: Document the value of *pp on parse errors. It might be advanced
3782 * by 0, or +1, or the index of the parse error, or the guessed end of the
3783 * variable expression.
3784 *
3785 * If var_Error is returned, a diagnostic may or may not have been
3786 * printed. XXX: This is inconsistent.
3787 *
3788 * If varUndefined is returned, a diagnostic may or may not have been
3789 * printed. XXX: This is inconsistent.
3790 *
3791 * After using the returned value, *out_val_freeIt must be freed,
3792 * preferably using bmake_free since it is NULL in most cases.
3793 *
3794 * Side Effects:
3795 * Any effects from the modifiers, such as :!cmd! or ::=value.
3796 *-----------------------------------------------------------------------
3797 */
3798 /* coverity[+alloc : arg-*4] */
3799 VarParseResult
3800 Var_Parse(const char **pp, GNode *ctxt, VarEvalFlags eflags,
3801 const char **out_val, void **out_val_freeIt)
3802 {
3803 const char *const start = *pp;
3804 const char *p;
3805 Boolean haveModifier; /* TRUE if have modifiers for the variable */
3806 char startc; /* Starting character if variable in parens
3807 * or braces */
3808 char endc; /* Ending character if variable in parens
3809 * or braces */
3810 Boolean dynamic; /* TRUE if the variable is local and we're
3811 * expanding it in a non-local context. This
3812 * is done to support dynamic sources. The
3813 * result is just the expression, unaltered */
3814 const char *extramodifiers;
3815 Var *v;
3816 char *nstr;
3817 char eflags_str[VarEvalFlags_ToStringSize];
3818 VarExprFlags exprFlags = 0;
3819
3820 VAR_DEBUG2("Var_Parse: %s with %s\n", start,
3821 Enum_FlagsToString(eflags_str, sizeof eflags_str, eflags,
3822 VarEvalFlags_ToStringSpecs));
3823
3824 *out_val_freeIt = NULL;
3825 extramodifiers = NULL; /* extra modifiers to apply first */
3826 dynamic = FALSE;
3827
3828 /* Appease GCC, which thinks that the variable might not be
3829 * initialized. */
3830 endc = '\0';
3831
3832 startc = start[1];
3833 if (startc != '(' && startc != '{') {
3834 VarParseResult res;
3835 if (!ParseVarnameShort(startc, pp, ctxt, eflags, &res, out_val, &v))
3836 return res;
3837 haveModifier = FALSE;
3838 p = start + 1;
3839 } else {
3840 VarParseResult res;
3841 if (!ParseVarnameLong(pp, startc, ctxt, eflags,
3842 &res, out_val, out_val_freeIt,
3843 &endc, &p, &v, &haveModifier, &extramodifiers,
3844 &dynamic, &exprFlags))
3845 return res;
3846 }
3847
3848 if (v->flags & VAR_IN_USE)
3849 Fatal("Variable %s is recursive.", v->name);
3850
3851 /*
3852 * Before doing any modification, we have to make sure the value
3853 * has been fully expanded. If it looks like recursion might be
3854 * necessary (there's a dollar sign somewhere in the variable's value)
3855 * we just call Var_Subst to do any other substitutions that are
3856 * necessary. Note that the value returned by Var_Subst will have
3857 * been dynamically-allocated, so it will need freeing when we
3858 * return.
3859 */
3860 nstr = Buf_GetAll(&v->val, NULL);
3861 if (strchr(nstr, '$') != NULL && (eflags & VARE_WANTRES)) {
3862 VarEvalFlags nested_eflags = eflags;
3863 if (DEBUG(LINT))
3864 nested_eflags &= ~(unsigned)VARE_UNDEFERR;
3865 v->flags |= VAR_IN_USE;
3866 (void)Var_Subst(nstr, ctxt, nested_eflags, &nstr);
3867 v->flags &= ~(unsigned)VAR_IN_USE;
3868 /* TODO: handle errors */
3869 *out_val_freeIt = nstr;
3870 }
3871
3872 if (haveModifier || extramodifiers != NULL) {
3873 void *extraFree;
3874
3875 extraFree = NULL;
3876 if (extramodifiers != NULL) {
3877 const char *em = extramodifiers;
3878 nstr = ApplyModifiers(&em, nstr, '(', ')',
3879 v, &exprFlags, ctxt, eflags, &extraFree);
3880 }
3881
3882 if (haveModifier) {
3883 /* Skip initial colon. */
3884 p++;
3885
3886 nstr = ApplyModifiers(&p, nstr, startc, endc,
3887 v, &exprFlags, ctxt, eflags, out_val_freeIt);
3888 free(extraFree);
3889 } else {
3890 *out_val_freeIt = extraFree;
3891 }
3892 }
3893
3894 if (*p != '\0') /* Skip past endc if possible. */
3895 p++;
3896
3897 *pp = p;
3898
3899 if (v->flags & VAR_FROM_ENV) {
3900 /* Free the environment variable now since we own it,
3901 * but don't free the variable value if it will be returned. */
3902 Boolean keepValue = nstr == Buf_GetAll(&v->val, NULL);
3903 if (keepValue)
3904 *out_val_freeIt = nstr;
3905 (void)VarFreeEnv(v, !keepValue);
3906
3907 } else if (exprFlags & VEF_UNDEF) {
3908 if (!(exprFlags & VEF_DEF)) {
3909 if (*out_val_freeIt != NULL) {
3910 free(*out_val_freeIt);
3911 *out_val_freeIt = NULL;
3912 }
3913 if (dynamic) {
3914 nstr = bmake_strsedup(start, p);
3915 *out_val_freeIt = nstr;
3916 } else {
3917 /* The expression is still undefined, therefore discard the
3918 * actual value and return an error marker instead. */
3919 nstr = (eflags & VARE_UNDEFERR) ? var_Error : varUndefined;
3920 }
3921 }
3922 if (nstr != Buf_GetAll(&v->val, NULL))
3923 Buf_Destroy(&v->val, TRUE);
3924 free(v->name_freeIt);
3925 free(v);
3926 }
3927 *out_val = nstr;
3928 return VPR_UNKNOWN;
3929 }
3930
3931 /* Substitute for all variables in the given string in the given context.
3932 *
3933 * If eflags & VARE_UNDEFERR, Parse_Error will be called when an undefined
3934 * variable is encountered.
3935 *
3936 * If eflags & VARE_WANTRES, any effects from the modifiers, such as ::=,
3937 * :sh or !cmd! take place.
3938 *
3939 * Input:
3940 * str the string which to substitute
3941 * ctxt the context wherein to find variables
3942 * eflags VARE_UNDEFERR if undefineds are an error
3943 * VARE_WANTRES if we actually want the result
3944 * VARE_ASSIGN if we are in a := assignment
3945 *
3946 * Results:
3947 * The resulting string.
3948 */
3949 VarParseResult
3950 Var_Subst(const char *str, GNode *ctxt, VarEvalFlags eflags, char **out_res)
3951 {
3952 const char *p = str;
3953 Buffer buf; /* Buffer for forming things */
3954
3955 /* Set true if an error has already been reported,
3956 * to prevent a plethora of messages when recursing */
3957 static Boolean errorReported;
3958
3959 Buf_Init(&buf, 0);
3960 errorReported = FALSE;
3961
3962 while (*p != '\0') {
3963 if (p[0] == '$' && p[1] == '$') {
3964 /* A dollar sign may be escaped with another dollar sign. */
3965 if (save_dollars && (eflags & VARE_ASSIGN))
3966 Buf_AddByte(&buf, '$');
3967 Buf_AddByte(&buf, '$');
3968 p += 2;
3969 } else if (*p != '$') {
3970 /*
3971 * Skip as many characters as possible -- either to the end of
3972 * the string or to the next dollar sign (variable expression).
3973 */
3974 const char *plainStart = p;
3975
3976 for (p++; *p != '$' && *p != '\0'; p++)
3977 continue;
3978 Buf_AddBytesBetween(&buf, plainStart, p);
3979 } else {
3980 const char *nested_p = p;
3981 void *freeIt;
3982 const char *val;
3983 (void)Var_Parse(&nested_p, ctxt, eflags, &val, &freeIt);
3984 /* TODO: handle errors */
3985
3986 if (val == var_Error || val == varUndefined) {
3987 /*
3988 * If performing old-time variable substitution, skip over
3989 * the variable and continue with the substitution. Otherwise,
3990 * store the dollar sign and advance str so we continue with
3991 * the string...
3992 */
3993 if (oldVars) {
3994 p = nested_p;
3995 } else if ((eflags & VARE_UNDEFERR) || val == var_Error) {
3996 /*
3997 * If variable is undefined, complain and skip the
3998 * variable. The complaint will stop us from doing anything
3999 * when the file is parsed.
4000 */
4001 if (!errorReported) {
4002 Parse_Error(PARSE_FATAL, "Undefined variable \"%.*s\"",
4003 (int)(size_t)(nested_p - p), p);
4004 }
4005 p = nested_p;
4006 errorReported = TRUE;
4007 } else {
4008 /* Copy the initial '$' of the undefined expression,
4009 * thereby deferring expansion of the expression, but
4010 * expand nested expressions if already possible.
4011 * See unit-tests/varparse-undef-partial.mk. */
4012 Buf_AddByte(&buf, *p);
4013 p++;
4014 }
4015 } else {
4016 p = nested_p;
4017 Buf_AddStr(&buf, val);
4018 }
4019 free(freeIt);
4020 freeIt = NULL;
4021 }
4022 }
4023
4024 *out_res = Buf_DestroyCompact(&buf);
4025 return VPR_OK;
4026 }
4027
4028 /* Initialize the variables module. */
4029 void
4030 Var_Init(void)
4031 {
4032 VAR_INTERNAL = Targ_NewGN("Internal");
4033 VAR_GLOBAL = Targ_NewGN("Global");
4034 VAR_CMDLINE = Targ_NewGN("Command");
4035 }
4036
4037 /* Clean up the variables module. */
4038 void
4039 Var_End(void)
4040 {
4041 Var_Stats();
4042 }
4043
4044 void
4045 Var_Stats(void)
4046 {
4047 HashTable_DebugStats(&VAR_GLOBAL->context, "VAR_GLOBAL");
4048 }
4049
4050 /* Print all variables in a context, sorted by name. */
4051 void
4052 Var_Dump(GNode *ctxt)
4053 {
4054 Vector /* of const char * */ vec;
4055 HashIter hi;
4056 size_t i;
4057 const char **varnames;
4058
4059 Vector_Init(&vec, sizeof(const char *));
4060
4061 HashIter_Init(&hi, &ctxt->context);
4062 while (HashIter_Next(&hi) != NULL)
4063 *(const char **)Vector_Push(&vec) = hi.entry->key;
4064 varnames = vec.items;
4065
4066 qsort(varnames, vec.len, sizeof varnames[0], str_cmp_asc);
4067
4068 for (i = 0; i < vec.len; i++) {
4069 const char *varname = varnames[i];
4070 Var *var = HashTable_FindValue(&ctxt->context, varname);
4071 debug_printf("%-16s = %s\n", varname, Buf_GetAll(&var->val, NULL));
4072 }
4073
4074 Vector_Done(&vec);
4075 }
4076