var.c revision 1.673 1 /* $NetBSD: var.c,v 1.673 2020/11/07 14:11:58 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 1.673 2020/11/07 14:11:58 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 /*
161 * Traditionally this make consumed $$ during := like any other expansion.
162 * Other make's do not, and this make follows straight since 2016-01-09.
163 *
164 * This knob allows controlling the behavior.
165 * FALSE to consume $$ during := assignment.
166 * TRUE to preserve $$ during := assignment.
167 */
168 #define MAKE_SAVE_DOLLARS ".MAKE.SAVE_DOLLARS"
169 static Boolean save_dollars = TRUE;
170
171 /*
172 * Internally, variables are contained in four different contexts.
173 * 1) the environment. They cannot be changed. If an environment
174 * variable is appended to, the result is placed in the global
175 * context.
176 * 2) the global context. Variables set in the makefiles are located
177 * here.
178 * 3) the command-line context. All variables set on the command line
179 * are placed in this context.
180 * 4) the local context. Each target has associated with it a context
181 * list. On this list are located the structures describing such
182 * local variables as $(@) and $(*)
183 * The four contexts are searched in the reverse order from which they are
184 * listed (but see opts.checkEnvFirst).
185 */
186 GNode *VAR_INTERNAL; /* variables from make itself */
187 GNode *VAR_GLOBAL; /* variables from the makefile */
188 GNode *VAR_CMDLINE; /* variables defined on the command-line */
189
190 typedef enum VarFlags {
191
192 /* The variable's value is currently being used by Var_Parse or Var_Subst.
193 * This marker is used to avoid endless recursion. */
194 VAR_IN_USE = 0x01,
195
196 /* The variable comes from the environment.
197 * These variables are not registered in any GNode, therefore they must
198 * be freed as soon as they are not used anymore. */
199 VAR_FROM_ENV = 0x02,
200
201 /* The variable is exported to the environment, to be used by child
202 * processes. */
203 VAR_EXPORTED = 0x10,
204
205 /* At the point where this variable was exported, it contained an
206 * unresolved reference to another variable. Before any child process is
207 * started, it needs to be exported again, in the hope that the referenced
208 * variable can then be resolved. */
209 VAR_REEXPORT = 0x20,
210
211 /* The variable came from the command line. */
212 VAR_FROM_CMD = 0x40,
213
214 /* The variable value cannot be changed anymore, and the variable cannot
215 * be deleted. Any attempts to do so are ignored. */
216 VAR_READONLY = 0x80
217 } VarFlags;
218
219 ENUM_FLAGS_RTTI_6(VarFlags,
220 VAR_IN_USE, VAR_FROM_ENV,
221 VAR_EXPORTED, VAR_REEXPORT, VAR_FROM_CMD, VAR_READONLY);
222
223 /* Variables are defined using one of the VAR=value assignments. Their
224 * value can be queried by expressions such as $V, ${VAR}, or with modifiers
225 * such as ${VAR:S,from,to,g:Q}.
226 *
227 * There are 3 kinds of variables: context variables, environment variables,
228 * undefined variables.
229 *
230 * Context variables are stored in a GNode.context. The only way to undefine
231 * a context variable is using the .undef directive. In particular, it must
232 * not be possible to undefine a variable during the evaluation of an
233 * expression, or Var.name might point nowhere.
234 *
235 * Environment variables are temporary. They are returned by VarFind, and
236 * after using them, they must be freed using VarFreeEnv.
237 *
238 * Undefined variables occur during evaluation of variable expressions such
239 * as ${UNDEF:Ufallback} in Var_Parse and ApplyModifiers.
240 */
241 typedef struct Var {
242 /* The name of the variable, once set, doesn't change anymore.
243 * For context variables, it aliases the corresponding HashEntry name.
244 * For environment and undefined variables, it is allocated. */
245 const char *name;
246 void *name_freeIt;
247
248 Buffer val; /* its value */
249 VarFlags flags; /* miscellaneous status flags */
250 } Var;
251
252 /*
253 * Exporting vars is expensive so skip it if we can
254 */
255 typedef enum VarExportedMode {
256 VAR_EXPORTED_NONE,
257 VAR_EXPORTED_SOME,
258 VAR_EXPORTED_ALL
259 } VarExportedMode;
260
261 static VarExportedMode var_exportedVars = VAR_EXPORTED_NONE;
262
263 typedef enum VarExportFlags {
264 /*
265 * We pass this to Var_Export when doing the initial export
266 * or after updating an exported var.
267 */
268 VAR_EXPORT_PARENT = 0x01,
269 /*
270 * We pass this to Var_Export1 to tell it to leave the value alone.
271 */
272 VAR_EXPORT_LITERAL = 0x02
273 } VarExportFlags;
274
275 /* Flags for pattern matching in the :S and :C modifiers */
276 typedef enum VarPatternFlags {
277 VARP_SUB_GLOBAL = 0x01, /* Replace as often as possible ('g') */
278 VARP_SUB_ONE = 0x02, /* Replace only once ('1') */
279 VARP_ANCHOR_START = 0x04, /* Match at start of word ('^') */
280 VARP_ANCHOR_END = 0x08 /* Match at end of word ('$') */
281 } VarPatternFlags;
282
283 static Var *
284 VarNew(const char *name, void *name_freeIt, const char *value, VarFlags flags)
285 {
286 size_t value_len = strlen(value);
287 Var *var = bmake_malloc(sizeof *var);
288 var->name = name;
289 var->name_freeIt = name_freeIt;
290 Buf_InitSize(&var->val, value_len + 1);
291 Buf_AddBytes(&var->val, value, value_len);
292 var->flags = flags;
293 return var;
294 }
295
296 static const char *
297 CanonicalVarname(const char *name)
298 {
299 if (*name == '.' && ch_isupper(name[1])) {
300 switch (name[1]) {
301 case 'A':
302 if (strcmp(name, ".ALLSRC") == 0)
303 name = ALLSRC;
304 if (strcmp(name, ".ARCHIVE") == 0)
305 name = ARCHIVE;
306 break;
307 case 'I':
308 if (strcmp(name, ".IMPSRC") == 0)
309 name = IMPSRC;
310 break;
311 case 'M':
312 if (strcmp(name, ".MEMBER") == 0)
313 name = MEMBER;
314 break;
315 case 'O':
316 if (strcmp(name, ".OODATE") == 0)
317 name = OODATE;
318 break;
319 case 'P':
320 if (strcmp(name, ".PREFIX") == 0)
321 name = PREFIX;
322 break;
323 case 'S':
324 if (strcmp(name, ".SHELL") == 0) {
325 if (!shellPath)
326 Shell_Init();
327 }
328 break;
329 case 'T':
330 if (strcmp(name, ".TARGET") == 0)
331 name = TARGET;
332 break;
333 }
334 }
335
336 /* GNU make has an additional alias $^ == ${.ALLSRC}. */
337
338 return name;
339 }
340
341 static Var *
342 GNode_FindVar(GNode *ctxt, const char *varname, unsigned int hash)
343 {
344 return HashTable_FindValueHash(&ctxt->context, varname, hash);
345 }
346
347 /* Find the variable in the context, and maybe in other contexts as well.
348 *
349 * Input:
350 * name name to find, is not expanded any further
351 * ctxt context in which to look first
352 * elsewhere TRUE to look in other contexts as well
353 *
354 * Results:
355 * The found variable, or NULL if the variable does not exist.
356 * If the variable is an environment variable, it must be freed using
357 * VarFreeEnv after use.
358 */
359 static Var *
360 VarFind(const char *name, GNode *ctxt, Boolean elsewhere)
361 {
362 Var *var;
363 unsigned int nameHash;
364
365 /*
366 * If the variable name begins with a '.', it could very well be one of
367 * the local ones. We check the name against all the local variables
368 * and substitute the short version in for 'name' if it matches one of
369 * them.
370 */
371 name = CanonicalVarname(name);
372 nameHash = Hash_Hash(name);
373
374 /* First look for the variable in the given context. */
375 var = GNode_FindVar(ctxt, name, nameHash);
376 if (!elsewhere)
377 return var;
378
379 /* The variable was not found in the given context. Now look for it in
380 * the other contexts as well. */
381 if (var == NULL && ctxt != VAR_CMDLINE)
382 var = GNode_FindVar(VAR_CMDLINE, name, nameHash);
383
384 if (!opts.checkEnvFirst && var == NULL && ctxt != VAR_GLOBAL) {
385 var = GNode_FindVar(VAR_GLOBAL, name, nameHash);
386 if (var == NULL && ctxt != VAR_INTERNAL) {
387 /* VAR_INTERNAL is subordinate to VAR_GLOBAL */
388 var = GNode_FindVar(VAR_INTERNAL, name, nameHash);
389 }
390 }
391
392 if (var == NULL) {
393 char *env;
394
395 if ((env = getenv(name)) != NULL) {
396 char *varname = bmake_strdup(name);
397 return VarNew(varname, varname, env, VAR_FROM_ENV);
398 }
399
400 if (opts.checkEnvFirst && ctxt != VAR_GLOBAL) {
401 var = GNode_FindVar(VAR_GLOBAL, name, nameHash);
402 if (var == NULL && ctxt != VAR_INTERNAL)
403 var = GNode_FindVar(VAR_INTERNAL, name, nameHash);
404 return var;
405 }
406
407 return NULL;
408 }
409
410 return var;
411 }
412
413 /* If the variable is an environment variable, free it.
414 *
415 * Input:
416 * v the variable
417 * freeValue true if the variable value should be freed as well
418 *
419 * Results:
420 * TRUE if it is an environment variable, FALSE otherwise.
421 */
422 static Boolean
423 VarFreeEnv(Var *v, Boolean freeValue)
424 {
425 if (!(v->flags & VAR_FROM_ENV))
426 return FALSE;
427
428 free(v->name_freeIt);
429 Buf_Destroy(&v->val, freeValue);
430 free(v);
431 return TRUE;
432 }
433
434 /* Add a new variable of the given name and value to the given context.
435 * The name and val arguments are duplicated so they may safely be freed. */
436 static void
437 VarAdd(const char *name, const char *val, GNode *ctxt, VarSetFlags flags)
438 {
439 HashEntry *he = HashTable_CreateEntry(&ctxt->context, name, NULL);
440 Var *v = VarNew(he->key /* aliased */, NULL, val,
441 flags & VAR_SET_READONLY ? VAR_READONLY : 0);
442 HashEntry_Set(he, v);
443 if (!(ctxt->flags & INTERNAL)) {
444 VAR_DEBUG3("%s:%s = %s\n", ctxt->name, name, val);
445 }
446 }
447
448 /* Remove a variable from a context, freeing all related memory as well.
449 * The variable name is expanded once. */
450 void
451 Var_Delete(const char *name, GNode *ctxt)
452 {
453 char *name_freeIt = NULL;
454 HashEntry *he;
455
456 if (strchr(name, '$') != NULL) {
457 (void)Var_Subst(name, VAR_GLOBAL, VARE_WANTRES, &name_freeIt);
458 /* TODO: handle errors */
459 name = name_freeIt;
460 }
461 he = HashTable_FindEntry(&ctxt->context, name);
462 VAR_DEBUG3("%s:delete %s%s\n",
463 ctxt->name, name, he != NULL ? "" : " (not found)");
464 free(name_freeIt);
465
466 if (he != NULL) {
467 Var *v = HashEntry_Get(he);
468 if (v->flags & VAR_EXPORTED)
469 unsetenv(v->name);
470 if (strcmp(v->name, MAKE_EXPORTED) == 0)
471 var_exportedVars = VAR_EXPORTED_NONE;
472 assert(v->name_freeIt == NULL);
473 HashTable_DeleteEntry(&ctxt->context, he);
474 Buf_Destroy(&v->val, TRUE);
475 free(v);
476 }
477 }
478
479 static Boolean
480 MayExport(const char *name)
481 {
482 if (name[0] == '.')
483 return FALSE; /* skip internals */
484 if (name[0] == '-')
485 return FALSE; /* skip misnamed variables */
486 if (name[1] == '\0') {
487 /*
488 * A single char.
489 * If it is one of the vars that should only appear in
490 * local context, skip it, else we can get Var_Subst
491 * into a loop.
492 */
493 switch (name[0]) {
494 case '@':
495 case '%':
496 case '*':
497 case '!':
498 return FALSE;
499 }
500 }
501 return TRUE;
502 }
503
504 /*
505 * Export a single variable.
506 * We ignore make internal variables (those which start with '.').
507 * Also we jump through some hoops to avoid calling setenv
508 * more than necessary since it can leak.
509 * We only manipulate flags of vars if 'parent' is set.
510 */
511 static Boolean
512 Var_Export1(const char *name, VarExportFlags flags)
513 {
514 VarExportFlags parent = flags & VAR_EXPORT_PARENT;
515 Var *v;
516 char *val;
517
518 if (!MayExport(name))
519 return FALSE;
520
521 v = VarFind(name, VAR_GLOBAL, FALSE);
522 if (v == NULL)
523 return FALSE;
524
525 if (!parent && (v->flags & VAR_EXPORTED) && !(v->flags & VAR_REEXPORT))
526 return FALSE; /* nothing to do */
527
528 val = Buf_GetAll(&v->val, NULL);
529 if (!(flags & VAR_EXPORT_LITERAL) && strchr(val, '$') != NULL) {
530 char *expr;
531
532 if (parent) {
533 /*
534 * Flag the variable as something we need to re-export.
535 * No point actually exporting it now though,
536 * the child process can do it at the last minute.
537 */
538 v->flags |= VAR_EXPORTED | VAR_REEXPORT;
539 return TRUE;
540 }
541 if (v->flags & VAR_IN_USE) {
542 /*
543 * We recursed while exporting in a child.
544 * This isn't going to end well, just skip it.
545 */
546 return FALSE;
547 }
548
549 /* XXX: name is injected without escaping it */
550 expr = str_concat3("${", name, "}");
551 (void)Var_Subst(expr, VAR_GLOBAL, VARE_WANTRES, &val);
552 /* TODO: handle errors */
553 setenv(name, val, 1);
554 free(val);
555 free(expr);
556 } else {
557 if (parent)
558 v->flags &= ~(unsigned)VAR_REEXPORT; /* once will do */
559 if (parent || !(v->flags & VAR_EXPORTED))
560 setenv(name, val, 1);
561 }
562
563 /*
564 * This is so Var_Set knows to call Var_Export again...
565 */
566 if (parent) {
567 v->flags |= VAR_EXPORTED;
568 }
569 return TRUE;
570 }
571
572 /*
573 * This gets called from our child processes.
574 */
575 void
576 Var_ExportVars(void)
577 {
578 char *val;
579
580 /*
581 * Several make's support this sort of mechanism for tracking
582 * recursion - but each uses a different name.
583 * We allow the makefiles to update MAKELEVEL and ensure
584 * children see a correctly incremented value.
585 */
586 char tmp[BUFSIZ];
587 snprintf(tmp, sizeof tmp, "%d", makelevel + 1);
588 setenv(MAKE_LEVEL_ENV, tmp, 1);
589
590 if (var_exportedVars == VAR_EXPORTED_NONE)
591 return;
592
593 if (var_exportedVars == VAR_EXPORTED_ALL) {
594 HashIter hi;
595
596 /* Ouch! Exporting all variables at once is crazy... */
597 HashIter_Init(&hi, &VAR_GLOBAL->context);
598 while (HashIter_Next(&hi) != NULL) {
599 Var *var = hi.entry->value;
600 Var_Export1(var->name, 0);
601 }
602 return;
603 }
604
605 (void)Var_Subst("${" MAKE_EXPORTED ":O:u}", VAR_GLOBAL, VARE_WANTRES, &val);
606 /* TODO: handle errors */
607 if (*val) {
608 Words words = Str_Words(val, FALSE);
609 size_t i;
610
611 for (i = 0; i < words.len; i++)
612 Var_Export1(words.words[i], 0);
613 Words_Free(words);
614 }
615 free(val);
616 }
617
618 /*
619 * This is called when .export is seen or .MAKE.EXPORTED is modified.
620 *
621 * It is also called when any exported variable is modified.
622 * XXX: Is it really?
623 *
624 * str has the format "[-env|-literal] varname...".
625 */
626 void
627 Var_Export(const char *str, Boolean isExport)
628 {
629 VarExportFlags flags;
630 char *val;
631
632 if (isExport && str[0] == '\0') {
633 var_exportedVars = VAR_EXPORTED_ALL; /* use with caution! */
634 return;
635 }
636
637 if (isExport && strncmp(str, "-env", 4) == 0) {
638 str += 4;
639 flags = 0;
640 } else if (isExport && strncmp(str, "-literal", 8) == 0) {
641 str += 8;
642 flags = VAR_EXPORT_LITERAL;
643 } else {
644 flags = VAR_EXPORT_PARENT;
645 }
646
647 (void)Var_Subst(str, VAR_GLOBAL, VARE_WANTRES, &val);
648 /* TODO: handle errors */
649 if (val[0] != '\0') {
650 Words words = Str_Words(val, FALSE);
651
652 size_t i;
653 for (i = 0; i < words.len; i++) {
654 const char *name = words.words[i];
655 if (Var_Export1(name, flags)) {
656 if (var_exportedVars == VAR_EXPORTED_NONE)
657 var_exportedVars = VAR_EXPORTED_SOME;
658 if (isExport && (flags & VAR_EXPORT_PARENT)) {
659 Var_Append(MAKE_EXPORTED, name, VAR_GLOBAL);
660 }
661 }
662 }
663 Words_Free(words);
664 }
665 free(val);
666 }
667
668
669 extern char **environ;
670
671 /*
672 * This is called when .unexport[-env] is seen.
673 *
674 * str must have the form "unexport[-env] varname...".
675 */
676 void
677 Var_UnExport(const char *str)
678 {
679 const char *varnames;
680 char *varnames_freeIt;
681 Boolean unexport_env;
682
683 varnames = NULL;
684 varnames_freeIt = NULL;
685
686 str += strlen("unexport");
687 unexport_env = strncmp(str, "-env", 4) == 0;
688 if (unexport_env) {
689 const char *cp;
690 char **newenv;
691
692 cp = getenv(MAKE_LEVEL_ENV); /* we should preserve this */
693 if (environ == savedEnv) {
694 /* we have been here before! */
695 newenv = bmake_realloc(environ, 2 * sizeof(char *));
696 } else {
697 if (savedEnv) {
698 free(savedEnv);
699 savedEnv = NULL;
700 }
701 newenv = bmake_malloc(2 * sizeof(char *));
702 }
703
704 /* Note: we cannot safely free() the original environ. */
705 environ = savedEnv = newenv;
706 newenv[0] = NULL;
707 newenv[1] = NULL;
708 if (cp && *cp)
709 setenv(MAKE_LEVEL_ENV, cp, 1);
710 } else {
711 cpp_skip_whitespace(&str);
712 if (str[0] != '\0')
713 varnames = str;
714 }
715
716 if (varnames == NULL) {
717 /* Using .MAKE.EXPORTED */
718 (void)Var_Subst("${" MAKE_EXPORTED ":O:u}", VAR_GLOBAL, VARE_WANTRES,
719 &varnames_freeIt);
720 /* TODO: handle errors */
721 varnames = varnames_freeIt;
722 }
723
724 {
725 Var *v;
726 size_t i;
727
728 Words words = Str_Words(varnames, FALSE);
729 for (i = 0; i < words.len; i++) {
730 const char *varname = words.words[i];
731 v = VarFind(varname, VAR_GLOBAL, FALSE);
732 if (v == NULL) {
733 VAR_DEBUG1("Not unexporting \"%s\" (not found)\n", varname);
734 continue;
735 }
736
737 VAR_DEBUG1("Unexporting \"%s\"\n", varname);
738 if (!unexport_env && (v->flags & VAR_EXPORTED) &&
739 !(v->flags & VAR_REEXPORT))
740 unsetenv(v->name);
741 v->flags &= ~(unsigned)(VAR_EXPORTED | VAR_REEXPORT);
742
743 /*
744 * If we are unexporting a list,
745 * remove each one from .MAKE.EXPORTED.
746 * If we are removing them all,
747 * just delete .MAKE.EXPORTED below.
748 */
749 if (varnames == str) {
750 /* XXX: v->name is injected without escaping it */
751 char *expr = str_concat3("${" MAKE_EXPORTED ":N", v->name, "}");
752 char *cp;
753 (void)Var_Subst(expr, VAR_GLOBAL, VARE_WANTRES, &cp);
754 /* TODO: handle errors */
755 Var_Set(MAKE_EXPORTED, cp, VAR_GLOBAL);
756 free(cp);
757 free(expr);
758 }
759 }
760 Words_Free(words);
761 if (varnames != str) {
762 Var_Delete(MAKE_EXPORTED, VAR_GLOBAL);
763 free(varnames_freeIt);
764 }
765 }
766 }
767
768 /* See Var_Set for documentation. */
769 void
770 Var_SetWithFlags(const char *name, const char *val, GNode *ctxt,
771 VarSetFlags flags)
772 {
773 const char *unexpanded_name = name;
774 char *name_freeIt = NULL;
775 Var *v;
776
777 assert(val != NULL);
778
779 if (strchr(name, '$') != NULL) {
780 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
781 /* TODO: handle errors */
782 name = name_freeIt;
783 }
784
785 if (name[0] == '\0') {
786 VAR_DEBUG2("Var_Set(\"%s\", \"%s\", ...) "
787 "name expands to empty string - ignored\n",
788 unexpanded_name, val);
789 free(name_freeIt);
790 return;
791 }
792
793 if (ctxt == VAR_GLOBAL) {
794 v = VarFind(name, VAR_CMDLINE, FALSE);
795 if (v != NULL) {
796 if (v->flags & VAR_FROM_CMD) {
797 VAR_DEBUG3("%s:%s = %s ignored!\n", ctxt->name, name, val);
798 goto out;
799 }
800 VarFreeEnv(v, TRUE);
801 }
802 }
803
804 /*
805 * We only look for a variable in the given context since anything set
806 * here will override anything in a lower context, so there's not much
807 * point in searching them all just to save a bit of memory...
808 */
809 v = VarFind(name, ctxt, FALSE);
810 if (v == NULL) {
811 if (ctxt == VAR_CMDLINE && !(flags & VAR_NO_EXPORT)) {
812 /*
813 * This var would normally prevent the same name being added
814 * to VAR_GLOBAL, so delete it from there if needed.
815 * Otherwise -V name may show the wrong value.
816 */
817 /* XXX: name is expanded for the second time */
818 Var_Delete(name, VAR_GLOBAL);
819 }
820 VarAdd(name, val, ctxt, flags);
821 } else {
822 if ((v->flags & VAR_READONLY) && !(flags & VAR_SET_READONLY)) {
823 VAR_DEBUG3("%s:%s = %s ignored (read-only)\n",
824 ctxt->name, name, val);
825 goto out;
826 }
827 Buf_Empty(&v->val);
828 Buf_AddStr(&v->val, val);
829
830 VAR_DEBUG3("%s:%s = %s\n", ctxt->name, name, val);
831 if (v->flags & VAR_EXPORTED) {
832 Var_Export1(name, VAR_EXPORT_PARENT);
833 }
834 }
835 /*
836 * Any variables given on the command line are automatically exported
837 * to the environment (as per POSIX standard)
838 * Other than internals.
839 */
840 if (ctxt == VAR_CMDLINE && !(flags & VAR_NO_EXPORT) && name[0] != '.') {
841 if (v == NULL)
842 v = VarFind(name, ctxt, FALSE); /* we just added it */
843 v->flags |= VAR_FROM_CMD;
844
845 /*
846 * If requested, don't export these in the environment
847 * individually. We still put them in MAKEOVERRIDES so
848 * that the command-line settings continue to override
849 * Makefile settings.
850 */
851 if (!opts.varNoExportEnv)
852 setenv(name, val, 1);
853
854 Var_Append(MAKEOVERRIDES, name, VAR_GLOBAL);
855 }
856 if (name[0] == '.' && strcmp(name, MAKE_SAVE_DOLLARS) == 0)
857 save_dollars = ParseBoolean(val, save_dollars);
858
859 out:
860 free(name_freeIt);
861 if (v != NULL)
862 VarFreeEnv(v, TRUE);
863 }
864
865 /*-
866 *-----------------------------------------------------------------------
867 * Var_Set --
868 * Set the variable name to the value val in the given context.
869 *
870 * If the variable doesn't yet exist, it is created.
871 * Otherwise the new value overwrites and replaces the old value.
872 *
873 * Input:
874 * name name of the variable to set, is expanded once
875 * val value to give to the variable
876 * ctxt context in which to set it
877 *
878 * Notes:
879 * The variable is searched for only in its context before being
880 * created in that context. I.e. if the context is VAR_GLOBAL,
881 * only VAR_GLOBAL->context is searched. Likewise if it is VAR_CMDLINE,
882 * only VAR_CMDLINE->context is searched. This is done to avoid the
883 * literally thousands of unnecessary strcmp's that used to be done to
884 * set, say, $(@) or $(<).
885 * If the context is VAR_GLOBAL though, we check if the variable
886 * was set in VAR_CMDLINE from the command line and skip it if so.
887 *-----------------------------------------------------------------------
888 */
889 void
890 Var_Set(const char *name, const char *val, GNode *ctxt)
891 {
892 Var_SetWithFlags(name, val, ctxt, 0);
893 }
894
895 /*-
896 *-----------------------------------------------------------------------
897 * Var_Append --
898 * The variable of the given name has the given value appended to it in
899 * the given context.
900 *
901 * If the variable doesn't exist, it is created. Otherwise the strings
902 * are concatenated, with a space in between.
903 *
904 * Input:
905 * name name of the variable to modify, is expanded once
906 * val string to append to it
907 * ctxt context in which this should occur
908 *
909 * Notes:
910 * Only if the variable is being sought in the global context is the
911 * environment searched.
912 * XXX: Knows its calling circumstances in that if called with ctxt
913 * an actual target, it will only search that context since only
914 * a local variable could be being appended to. This is actually
915 * a big win and must be tolerated.
916 *-----------------------------------------------------------------------
917 */
918 void
919 Var_Append(const char *name, const char *val, GNode *ctxt)
920 {
921 char *name_freeIt = NULL;
922 Var *v;
923
924 assert(val != NULL);
925
926 if (strchr(name, '$') != NULL) {
927 const char *unexpanded_name = name;
928 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
929 /* TODO: handle errors */
930 name = name_freeIt;
931 if (name[0] == '\0') {
932 VAR_DEBUG2("Var_Append(\"%s\", \"%s\", ...) "
933 "name expands to empty string - ignored\n",
934 unexpanded_name, val);
935 free(name_freeIt);
936 return;
937 }
938 }
939
940 v = VarFind(name, ctxt, ctxt == VAR_GLOBAL);
941
942 if (v == NULL) {
943 /* XXX: name is expanded for the second time */
944 Var_Set(name, val, ctxt);
945 } else if (v->flags & VAR_READONLY) {
946 VAR_DEBUG1("Ignoring append to %s since it is read-only\n", name);
947 } else if (ctxt == VAR_CMDLINE || !(v->flags & VAR_FROM_CMD)) {
948 Buf_AddByte(&v->val, ' ');
949 Buf_AddStr(&v->val, val);
950
951 VAR_DEBUG3("%s:%s = %s\n",
952 ctxt->name, name, Buf_GetAll(&v->val, NULL));
953
954 if (v->flags & VAR_FROM_ENV) {
955 HashEntry *h;
956
957 /*
958 * If the original variable came from the environment, we
959 * have to install it in the global context (we could place
960 * it in the environment, but then we should provide a way to
961 * export other variables...)
962 */
963 v->flags &= ~(unsigned)VAR_FROM_ENV;
964 h = HashTable_CreateEntry(&ctxt->context, name, NULL);
965 HashEntry_Set(h, v);
966 }
967 }
968 free(name_freeIt);
969 }
970
971 /* See if the given variable exists, in the given context or in other
972 * fallback contexts.
973 *
974 * Input:
975 * name Variable to find, is expanded once
976 * ctxt Context in which to start search
977 */
978 Boolean
979 Var_Exists(const char *name, GNode *ctxt)
980 {
981 char *name_freeIt = NULL;
982 Var *v;
983
984 if (strchr(name, '$') != NULL) {
985 (void)Var_Subst(name, ctxt, VARE_WANTRES, &name_freeIt);
986 /* TODO: handle errors */
987 name = name_freeIt;
988 }
989
990 v = VarFind(name, ctxt, TRUE);
991 free(name_freeIt);
992 if (v == NULL)
993 return FALSE;
994
995 (void)VarFreeEnv(v, TRUE);
996 return TRUE;
997 }
998
999 /*-
1000 *-----------------------------------------------------------------------
1001 * Var_Value --
1002 * Return the unexpanded value of the given variable in the given
1003 * context, or the usual contexts.
1004 *
1005 * Input:
1006 * name name to find, is not expanded any further
1007 * ctxt context in which to search for it
1008 *
1009 * Results:
1010 * The value if the variable exists, NULL if it doesn't.
1011 * If the returned value is not NULL, the caller must free
1012 * out_freeIt when the returned value is no longer needed.
1013 *-----------------------------------------------------------------------
1014 */
1015 const char *
1016 Var_Value(const char *name, GNode *ctxt, void **out_freeIt)
1017 {
1018 Var *v = VarFind(name, ctxt, TRUE);
1019 char *value;
1020
1021 *out_freeIt = NULL;
1022 if (v == NULL)
1023 return NULL;
1024
1025 value = Buf_GetAll(&v->val, NULL);
1026 if (VarFreeEnv(v, FALSE))
1027 *out_freeIt = value;
1028 return value;
1029 }
1030
1031 /* Return the unexpanded variable value from this node, without trying to look
1032 * up the variable in any other context. */
1033 const char *
1034 Var_ValueDirect(const char *name, GNode *ctxt)
1035 {
1036 Var *v = VarFind(name, ctxt, FALSE);
1037 return v != NULL ? Buf_GetAll(&v->val, NULL) : NULL;
1038 }
1039
1040
1041 /* SepBuf is a string being built from words, interleaved with separators. */
1042 typedef struct SepBuf {
1043 Buffer buf;
1044 Boolean needSep;
1045 char sep; /* usually ' ', but see the :ts modifier */
1046 } SepBuf;
1047
1048 static void
1049 SepBuf_Init(SepBuf *buf, char sep)
1050 {
1051 Buf_InitSize(&buf->buf, 32);
1052 buf->needSep = FALSE;
1053 buf->sep = sep;
1054 }
1055
1056 static void
1057 SepBuf_Sep(SepBuf *buf)
1058 {
1059 buf->needSep = TRUE;
1060 }
1061
1062 static void
1063 SepBuf_AddBytes(SepBuf *buf, const char *mem, size_t mem_size)
1064 {
1065 if (mem_size == 0)
1066 return;
1067 if (buf->needSep && buf->sep != '\0') {
1068 Buf_AddByte(&buf->buf, buf->sep);
1069 buf->needSep = FALSE;
1070 }
1071 Buf_AddBytes(&buf->buf, mem, mem_size);
1072 }
1073
1074 static void
1075 SepBuf_AddBytesBetween(SepBuf *buf, const char *start, const char *end)
1076 {
1077 SepBuf_AddBytes(buf, start, (size_t)(end - start));
1078 }
1079
1080 static void
1081 SepBuf_AddStr(SepBuf *buf, const char *str)
1082 {
1083 SepBuf_AddBytes(buf, str, strlen(str));
1084 }
1085
1086 static char *
1087 SepBuf_Destroy(SepBuf *buf, Boolean free_buf)
1088 {
1089 return Buf_Destroy(&buf->buf, free_buf);
1090 }
1091
1092
1093 /* This callback for ModifyWords gets a single word from a variable expression
1094 * and typically adds a modification of this word to the buffer. It may also
1095 * do nothing or add several words.
1096 *
1097 * For example, in ${:Ua b c:M*2}, the callback is called 3 times, once for
1098 * each word of "a b c". */
1099 typedef void (*ModifyWordsCallback)(const char *word, SepBuf *buf, void *data);
1100
1101
1102 /* Callback for ModifyWords to implement the :H modifier.
1103 * Add the dirname of the given word to the buffer. */
1104 static void
1105 ModifyWord_Head(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1106 {
1107 const char *slash = strrchr(word, '/');
1108 if (slash != NULL)
1109 SepBuf_AddBytesBetween(buf, word, slash);
1110 else
1111 SepBuf_AddStr(buf, ".");
1112 }
1113
1114 /* Callback for ModifyWords to implement the :T modifier.
1115 * Add the basename of the given word to the buffer. */
1116 static void
1117 ModifyWord_Tail(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1118 {
1119 const char *slash = strrchr(word, '/');
1120 const char *base = slash != NULL ? slash + 1 : word;
1121 SepBuf_AddStr(buf, base);
1122 }
1123
1124 /* Callback for ModifyWords to implement the :E modifier.
1125 * Add the filename suffix of the given word to the buffer, if it exists. */
1126 static void
1127 ModifyWord_Suffix(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1128 {
1129 const char *dot = strrchr(word, '.');
1130 if (dot != NULL)
1131 SepBuf_AddStr(buf, dot + 1);
1132 }
1133
1134 /* Callback for ModifyWords to implement the :R modifier.
1135 * Add the basename of the given word to the buffer. */
1136 static void
1137 ModifyWord_Root(const char *word, SepBuf *buf, void *dummy MAKE_ATTR_UNUSED)
1138 {
1139 const char *dot = strrchr(word, '.');
1140 size_t len = dot != NULL ? (size_t)(dot - word) : strlen(word);
1141 SepBuf_AddBytes(buf, word, len);
1142 }
1143
1144 /* Callback for ModifyWords to implement the :M modifier.
1145 * Place the word in the buffer if it matches the given pattern. */
1146 static void
1147 ModifyWord_Match(const char *word, SepBuf *buf, void *data)
1148 {
1149 const char *pattern = data;
1150 VAR_DEBUG2("VarMatch [%s] [%s]\n", word, pattern);
1151 if (Str_Match(word, pattern))
1152 SepBuf_AddStr(buf, word);
1153 }
1154
1155 /* Callback for ModifyWords to implement the :N modifier.
1156 * Place the word in the buffer if it doesn't match the given pattern. */
1157 static void
1158 ModifyWord_NoMatch(const char *word, SepBuf *buf, void *data)
1159 {
1160 const char *pattern = data;
1161 if (!Str_Match(word, pattern))
1162 SepBuf_AddStr(buf, word);
1163 }
1164
1165 #ifdef SYSVVARSUB
1166 /* Check word against pattern for a match (% is a wildcard).
1167 *
1168 * Input:
1169 * word Word to examine
1170 * pattern Pattern to examine against
1171 *
1172 * Results:
1173 * Returns the start of the match, or NULL.
1174 * out_match_len returns the length of the match, if any.
1175 * out_hasPercent returns whether the pattern contains a percent.
1176 */
1177 static const char *
1178 SysVMatch(const char *word, const char *pattern,
1179 size_t *out_match_len, Boolean *out_hasPercent)
1180 {
1181 const char *p = pattern;
1182 const char *w = word;
1183 const char *percent;
1184 size_t w_len;
1185 size_t p_len;
1186 const char *w_tail;
1187
1188 *out_hasPercent = FALSE;
1189 percent = strchr(p, '%');
1190 if (percent != NULL) { /* ${VAR:...%...=...} */
1191 *out_hasPercent = TRUE;
1192 if (*w == '\0')
1193 return NULL; /* empty word does not match pattern */
1194
1195 /* check that the prefix matches */
1196 for (; p != percent && *w != '\0' && *w == *p; w++, p++)
1197 continue;
1198 if (p != percent)
1199 return NULL; /* No match */
1200
1201 p++; /* Skip the percent */
1202 if (*p == '\0') {
1203 /* No more pattern, return the rest of the string */
1204 *out_match_len = strlen(w);
1205 return w;
1206 }
1207 }
1208
1209 /* Test whether the tail matches */
1210 w_len = strlen(w);
1211 p_len = strlen(p);
1212 if (w_len < p_len)
1213 return NULL;
1214
1215 w_tail = w + w_len - p_len;
1216 if (memcmp(p, w_tail, p_len) != 0)
1217 return NULL;
1218
1219 *out_match_len = (size_t)(w_tail - w);
1220 return w;
1221 }
1222
1223 struct ModifyWord_SYSVSubstArgs {
1224 GNode *ctx;
1225 const char *lhs;
1226 const char *rhs;
1227 };
1228
1229 /* Callback for ModifyWords to implement the :%.from=%.to modifier. */
1230 static void
1231 ModifyWord_SYSVSubst(const char *word, SepBuf *buf, void *data)
1232 {
1233 const struct ModifyWord_SYSVSubstArgs *args = data;
1234 char *rhs_expanded;
1235 const char *rhs;
1236 const char *percent;
1237
1238 size_t match_len;
1239 Boolean lhsPercent;
1240 const char *match = SysVMatch(word, args->lhs, &match_len, &lhsPercent);
1241 if (match == NULL) {
1242 SepBuf_AddStr(buf, word);
1243 return;
1244 }
1245
1246 /* Append rhs to the buffer, substituting the first '%' with the
1247 * match, but only if the lhs had a '%' as well. */
1248
1249 (void)Var_Subst(args->rhs, args->ctx, VARE_WANTRES, &rhs_expanded);
1250 /* TODO: handle errors */
1251
1252 rhs = rhs_expanded;
1253 percent = strchr(rhs, '%');
1254
1255 if (percent != NULL && lhsPercent) {
1256 /* Copy the prefix of the replacement pattern */
1257 SepBuf_AddBytesBetween(buf, rhs, percent);
1258 rhs = percent + 1;
1259 }
1260 if (percent != NULL || !lhsPercent)
1261 SepBuf_AddBytes(buf, match, match_len);
1262
1263 /* Append the suffix of the replacement pattern */
1264 SepBuf_AddStr(buf, rhs);
1265
1266 free(rhs_expanded);
1267 }
1268 #endif
1269
1270
1271 struct ModifyWord_SubstArgs {
1272 const char *lhs;
1273 size_t lhsLen;
1274 const char *rhs;
1275 size_t rhsLen;
1276 VarPatternFlags pflags;
1277 Boolean matched;
1278 };
1279
1280 /* Callback for ModifyWords to implement the :S,from,to, modifier.
1281 * Perform a string substitution on the given word. */
1282 static void
1283 ModifyWord_Subst(const char *word, SepBuf *buf, void *data)
1284 {
1285 size_t wordLen = strlen(word);
1286 struct ModifyWord_SubstArgs *args = data;
1287 const char *match;
1288
1289 if ((args->pflags & VARP_SUB_ONE) && args->matched)
1290 goto nosub;
1291
1292 if (args->pflags & VARP_ANCHOR_START) {
1293 if (wordLen < args->lhsLen ||
1294 memcmp(word, args->lhs, args->lhsLen) != 0)
1295 goto nosub;
1296
1297 if ((args->pflags & VARP_ANCHOR_END) && wordLen != args->lhsLen)
1298 goto nosub;
1299
1300 /* :S,^prefix,replacement, or :S,^whole$,replacement, */
1301 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1302 SepBuf_AddBytes(buf, word + args->lhsLen, wordLen - args->lhsLen);
1303 args->matched = TRUE;
1304 return;
1305 }
1306
1307 if (args->pflags & VARP_ANCHOR_END) {
1308 const char *start;
1309
1310 if (wordLen < args->lhsLen)
1311 goto nosub;
1312
1313 start = word + (wordLen - args->lhsLen);
1314 if (memcmp(start, args->lhs, args->lhsLen) != 0)
1315 goto nosub;
1316
1317 /* :S,suffix$,replacement, */
1318 SepBuf_AddBytesBetween(buf, word, start);
1319 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1320 args->matched = TRUE;
1321 return;
1322 }
1323
1324 if (args->lhs[0] == '\0')
1325 goto nosub;
1326
1327 /* unanchored case, may match more than once */
1328 while ((match = strstr(word, args->lhs)) != NULL) {
1329 SepBuf_AddBytesBetween(buf, word, match);
1330 SepBuf_AddBytes(buf, args->rhs, args->rhsLen);
1331 args->matched = TRUE;
1332 wordLen -= (size_t)(match - word) + args->lhsLen;
1333 word += (size_t)(match - word) + args->lhsLen;
1334 if (wordLen == 0 || !(args->pflags & VARP_SUB_GLOBAL))
1335 break;
1336 }
1337 nosub:
1338 SepBuf_AddBytes(buf, word, wordLen);
1339 }
1340
1341 #ifndef NO_REGEX
1342 /* Print the error caused by a regcomp or regexec call. */
1343 static void
1344 VarREError(int reerr, const regex_t *pat, const char *str)
1345 {
1346 size_t errlen = regerror(reerr, pat, NULL, 0);
1347 char *errbuf = bmake_malloc(errlen);
1348 regerror(reerr, pat, errbuf, errlen);
1349 Error("%s: %s", str, errbuf);
1350 free(errbuf);
1351 }
1352
1353 struct ModifyWord_SubstRegexArgs {
1354 regex_t re;
1355 size_t nsub;
1356 char *replace;
1357 VarPatternFlags pflags;
1358 Boolean matched;
1359 };
1360
1361 /* Callback for ModifyWords to implement the :C/from/to/ modifier.
1362 * Perform a regex substitution on the given word. */
1363 static void
1364 ModifyWord_SubstRegex(const char *word, SepBuf *buf, void *data)
1365 {
1366 struct ModifyWord_SubstRegexArgs *args = data;
1367 int xrv;
1368 const char *wp = word;
1369 char *rp;
1370 int flags = 0;
1371 regmatch_t m[10];
1372
1373 if ((args->pflags & VARP_SUB_ONE) && args->matched)
1374 goto nosub;
1375
1376 tryagain:
1377 xrv = regexec(&args->re, wp, args->nsub, m, flags);
1378
1379 switch (xrv) {
1380 case 0:
1381 args->matched = TRUE;
1382 SepBuf_AddBytes(buf, wp, (size_t)m[0].rm_so);
1383
1384 for (rp = args->replace; *rp; rp++) {
1385 if (*rp == '\\' && (rp[1] == '&' || rp[1] == '\\')) {
1386 SepBuf_AddBytes(buf, rp + 1, 1);
1387 rp++;
1388 continue;
1389 }
1390
1391 if (*rp == '&') {
1392 SepBuf_AddBytesBetween(buf, wp + m[0].rm_so, wp + m[0].rm_eo);
1393 continue;
1394 }
1395
1396 if (*rp != '\\' || !ch_isdigit(rp[1])) {
1397 SepBuf_AddBytes(buf, rp, 1);
1398 continue;
1399 }
1400
1401 { /* \0 to \9 backreference */
1402 size_t n = (size_t)(rp[1] - '0');
1403 rp++;
1404
1405 if (n >= args->nsub) {
1406 Error("No subexpression \\%zu", n);
1407 } else if (m[n].rm_so == -1) {
1408 Error("No match for subexpression \\%zu", n);
1409 } else {
1410 SepBuf_AddBytesBetween(buf, wp + m[n].rm_so,
1411 wp + m[n].rm_eo);
1412 }
1413 }
1414 }
1415
1416 wp += m[0].rm_eo;
1417 if (args->pflags & VARP_SUB_GLOBAL) {
1418 flags |= REG_NOTBOL;
1419 if (m[0].rm_so == 0 && m[0].rm_eo == 0) {
1420 SepBuf_AddBytes(buf, wp, 1);
1421 wp++;
1422 }
1423 if (*wp)
1424 goto tryagain;
1425 }
1426 if (*wp) {
1427 SepBuf_AddStr(buf, wp);
1428 }
1429 break;
1430 default:
1431 VarREError(xrv, &args->re, "Unexpected regex error");
1432 /* FALLTHROUGH */
1433 case REG_NOMATCH:
1434 nosub:
1435 SepBuf_AddStr(buf, wp);
1436 break;
1437 }
1438 }
1439 #endif
1440
1441
1442 struct ModifyWord_LoopArgs {
1443 GNode *ctx;
1444 char *tvar; /* name of temporary variable */
1445 char *str; /* string to expand */
1446 VarEvalFlags eflags;
1447 };
1448
1449 /* Callback for ModifyWords to implement the :@var (at) ...@ modifier of ODE make. */
1450 static void
1451 ModifyWord_Loop(const char *word, SepBuf *buf, void *data)
1452 {
1453 const struct ModifyWord_LoopArgs *args;
1454 char *s;
1455
1456 if (word[0] == '\0')
1457 return;
1458
1459 args = data;
1460 Var_SetWithFlags(args->tvar, word, args->ctx, VAR_NO_EXPORT);
1461 (void)Var_Subst(args->str, args->ctx, args->eflags, &s);
1462 /* TODO: handle errors */
1463
1464 VAR_DEBUG4("ModifyWord_Loop: "
1465 "in \"%s\", replace \"%s\" with \"%s\" to \"%s\"\n",
1466 word, args->tvar, args->str, s);
1467
1468 if (s[0] == '\n' || Buf_EndsWith(&buf->buf, '\n'))
1469 buf->needSep = FALSE;
1470 SepBuf_AddStr(buf, s);
1471 free(s);
1472 }
1473
1474
1475 /* The :[first..last] modifier selects words from the expression.
1476 * It can also reverse the words. */
1477 static char *
1478 VarSelectWords(char sep, Boolean oneBigWord, const char *str, int first,
1479 int last)
1480 {
1481 Words words;
1482 int len, start, end, step;
1483 int i;
1484
1485 SepBuf buf;
1486 SepBuf_Init(&buf, sep);
1487
1488 if (oneBigWord) {
1489 /* fake what Str_Words() would do if there were only one word */
1490 words.len = 1;
1491 words.words = bmake_malloc((words.len + 1) * sizeof(words.words[0]));
1492 words.freeIt = bmake_strdup(str);
1493 words.words[0] = words.freeIt;
1494 words.words[1] = NULL;
1495 } else {
1496 words = Str_Words(str, FALSE);
1497 }
1498
1499 /*
1500 * Now sanitize the given range.
1501 * If first or last are negative, convert them to the positive equivalents
1502 * (-1 gets converted to ac, -2 gets converted to (ac - 1), etc.).
1503 */
1504 len = (int)words.len;
1505 if (first < 0)
1506 first += len + 1;
1507 if (last < 0)
1508 last += len + 1;
1509
1510 /*
1511 * We avoid scanning more of the list than we need to.
1512 */
1513 if (first > last) {
1514 start = (first > len ? len : first) - 1;
1515 end = last < 1 ? 0 : last - 1;
1516 step = -1;
1517 } else {
1518 start = first < 1 ? 0 : first - 1;
1519 end = last > len ? len : last;
1520 step = 1;
1521 }
1522
1523 for (i = start; (step < 0) == (i >= end); i += step) {
1524 SepBuf_AddStr(&buf, words.words[i]);
1525 SepBuf_Sep(&buf);
1526 }
1527
1528 Words_Free(words);
1529
1530 return SepBuf_Destroy(&buf, FALSE);
1531 }
1532
1533
1534 /* Callback for ModifyWords to implement the :tA modifier.
1535 * Replace each word with the result of realpath() if successful. */
1536 static void
1537 ModifyWord_Realpath(const char *word, SepBuf *buf, void *data MAKE_ATTR_UNUSED)
1538 {
1539 struct stat st;
1540 char rbuf[MAXPATHLEN];
1541
1542 const char *rp = cached_realpath(word, rbuf);
1543 if (rp != NULL && *rp == '/' && stat(rp, &st) == 0)
1544 word = rp;
1545
1546 SepBuf_AddStr(buf, word);
1547 }
1548
1549 /* Modify each of the words of the passed string using the given function.
1550 *
1551 * Input:
1552 * str String whose words should be modified
1553 * modifyWord Function that modifies a single word
1554 * modifyWord_args Custom arguments for modifyWord
1555 *
1556 * Results:
1557 * A string of all the words modified appropriately.
1558 *-----------------------------------------------------------------------
1559 */
1560 static char *
1561 ModifyWords(const char *str,
1562 ModifyWordsCallback modifyWord, void *modifyWord_args,
1563 Boolean oneBigWord, char sep)
1564 {
1565 SepBuf result;
1566 Words words;
1567 size_t i;
1568
1569 if (oneBigWord) {
1570 SepBuf_Init(&result, sep);
1571 modifyWord(str, &result, modifyWord_args);
1572 return SepBuf_Destroy(&result, FALSE);
1573 }
1574
1575 SepBuf_Init(&result, sep);
1576
1577 words = Str_Words(str, FALSE);
1578
1579 VAR_DEBUG2("ModifyWords: split \"%s\" into %zu words\n", str, words.len);
1580
1581 for (i = 0; i < words.len; i++) {
1582 modifyWord(words.words[i], &result, modifyWord_args);
1583 if (Buf_Len(&result.buf) > 0)
1584 SepBuf_Sep(&result);
1585 }
1586
1587 Words_Free(words);
1588
1589 return SepBuf_Destroy(&result, FALSE);
1590 }
1591
1592
1593 static char *
1594 Words_JoinFree(Words words)
1595 {
1596 Buffer buf;
1597 size_t i;
1598
1599 Buf_Init(&buf);
1600
1601 for (i = 0; i < words.len; i++) {
1602 if (i != 0)
1603 Buf_AddByte(&buf, ' '); /* XXX: st->sep, for consistency */
1604 Buf_AddStr(&buf, words.words[i]);
1605 }
1606
1607 Words_Free(words);
1608
1609 return Buf_Destroy(&buf, FALSE);
1610 }
1611
1612 /* Remove adjacent duplicate words. */
1613 static char *
1614 VarUniq(const char *str)
1615 {
1616 Words words = Str_Words(str, FALSE);
1617
1618 if (words.len > 1) {
1619 size_t i, j;
1620 for (j = 0, i = 1; i < words.len; i++)
1621 if (strcmp(words.words[i], words.words[j]) != 0 && (++j != i))
1622 words.words[j] = words.words[i];
1623 words.len = j + 1;
1624 }
1625
1626 return Words_JoinFree(words);
1627 }
1628
1629
1630 /* Quote shell meta-characters and space characters in the string.
1631 * If quoteDollar is set, also quote and double any '$' characters. */
1632 static char *
1633 VarQuote(const char *str, Boolean quoteDollar)
1634 {
1635 Buffer buf;
1636 Buf_Init(&buf);
1637
1638 for (; *str != '\0'; str++) {
1639 if (*str == '\n') {
1640 const char *newline = Shell_GetNewline();
1641 if (newline == NULL)
1642 newline = "\\\n";
1643 Buf_AddStr(&buf, newline);
1644 continue;
1645 }
1646 if (ch_isspace(*str) || is_shell_metachar((unsigned char)*str))
1647 Buf_AddByte(&buf, '\\');
1648 Buf_AddByte(&buf, *str);
1649 if (quoteDollar && *str == '$')
1650 Buf_AddStr(&buf, "\\$");
1651 }
1652
1653 return Buf_Destroy(&buf, FALSE);
1654 }
1655
1656 /* Compute the 32-bit hash of the given string, using the MurmurHash3
1657 * algorithm. Output is encoded as 8 hex digits, in Little Endian order. */
1658 static char *
1659 VarHash(const char *str)
1660 {
1661 static const char hexdigits[16] = "0123456789abcdef";
1662 const unsigned char *ustr = (const unsigned char *)str;
1663
1664 uint32_t h = 0x971e137bU;
1665 uint32_t c1 = 0x95543787U;
1666 uint32_t c2 = 0x2ad7eb25U;
1667 size_t len2 = strlen(str);
1668
1669 char *buf;
1670 size_t i;
1671
1672 size_t len;
1673 for (len = len2; len; ) {
1674 uint32_t k = 0;
1675 switch (len) {
1676 default:
1677 k = ((uint32_t)ustr[3] << 24) |
1678 ((uint32_t)ustr[2] << 16) |
1679 ((uint32_t)ustr[1] << 8) |
1680 (uint32_t)ustr[0];
1681 len -= 4;
1682 ustr += 4;
1683 break;
1684 case 3:
1685 k |= (uint32_t)ustr[2] << 16;
1686 /* FALLTHROUGH */
1687 case 2:
1688 k |= (uint32_t)ustr[1] << 8;
1689 /* FALLTHROUGH */
1690 case 1:
1691 k |= (uint32_t)ustr[0];
1692 len = 0;
1693 }
1694 c1 = c1 * 5 + 0x7b7d159cU;
1695 c2 = c2 * 5 + 0x6bce6396U;
1696 k *= c1;
1697 k = (k << 11) ^ (k >> 21);
1698 k *= c2;
1699 h = (h << 13) ^ (h >> 19);
1700 h = h * 5 + 0x52dce729U;
1701 h ^= k;
1702 }
1703 h ^= (uint32_t)len2;
1704 h *= 0x85ebca6b;
1705 h ^= h >> 13;
1706 h *= 0xc2b2ae35;
1707 h ^= h >> 16;
1708
1709 buf = bmake_malloc(9);
1710 for (i = 0; i < 8; i++) {
1711 buf[i] = hexdigits[h & 0x0f];
1712 h >>= 4;
1713 }
1714 buf[8] = '\0';
1715 return buf;
1716 }
1717
1718 static char *
1719 VarStrftime(const char *fmt, Boolean zulu, time_t tim)
1720 {
1721 char buf[BUFSIZ];
1722
1723 if (!tim)
1724 time(&tim);
1725 if (!*fmt)
1726 fmt = "%c";
1727 strftime(buf, sizeof buf, fmt, zulu ? gmtime(&tim) : localtime(&tim));
1728
1729 buf[sizeof buf - 1] = '\0';
1730 return bmake_strdup(buf);
1731 }
1732
1733 /* The ApplyModifier functions all work in the same way. They get the
1734 * current parsing position (pp) and parse the modifier from there. The
1735 * modifier typically lasts until the next ':', or a closing '}' or ')'
1736 * (taken from st->endc), or the end of the string (parse error).
1737 *
1738 * The high-level behavior of these functions is:
1739 *
1740 * 1. parse the modifier
1741 * 2. evaluate the modifier
1742 * 3. housekeeping
1743 *
1744 * Parsing the modifier
1745 *
1746 * If parsing succeeds, the parsing position *pp is updated to point to the
1747 * first character following the modifier, which typically is either ':' or
1748 * st->endc.
1749 *
1750 * If parsing fails because of a missing delimiter (as in the :S, :C or :@
1751 * modifiers), return AMR_CLEANUP.
1752 *
1753 * If parsing fails because the modifier is unknown, return AMR_UNKNOWN to
1754 * try the SysV modifier ${VAR:from=to} as fallback. This should only be
1755 * done as long as there have been no side effects from evaluating nested
1756 * variables, to avoid evaluating them more than once. In this case, the
1757 * parsing position must not be updated. (XXX: Why not? The original parsing
1758 * position is well-known in ApplyModifiers.)
1759 *
1760 * If parsing fails and the SysV modifier ${VAR:from=to} should not be used
1761 * as a fallback, either issue an error message using Error or Parse_Error
1762 * and then return AMR_CLEANUP, or return AMR_BAD for the default error
1763 * message. Both of these return values will stop processing the variable
1764 * expression. (XXX: As of 2020-08-23, evaluation of the whole string
1765 * continues nevertheless after skipping a few bytes, which essentially is
1766 * undefined behavior. Not in the sense of C, but still it's impossible to
1767 * predict what happens in the parser.)
1768 *
1769 * Evaluating the modifier
1770 *
1771 * After parsing, the modifier is evaluated. The side effects from evaluating
1772 * nested variable expressions in the modifier text often already happen
1773 * during parsing though.
1774 *
1775 * Evaluating the modifier usually takes the current value of the variable
1776 * expression from st->val, or the variable name from st->v->name and stores
1777 * the result in st->newVal.
1778 *
1779 * If evaluating fails (as of 2020-08-23), an error message is printed using
1780 * Error. This function has no side-effects, it really just prints the error
1781 * message. Processing the expression continues as if everything were ok.
1782 * XXX: This should be fixed by adding proper error handling to Var_Subst,
1783 * Var_Parse, ApplyModifiers and ModifyWords.
1784 *
1785 * Housekeeping
1786 *
1787 * Some modifiers such as :D and :U turn undefined expressions into defined
1788 * expressions (see VEF_UNDEF, VEF_DEF).
1789 *
1790 * Some modifiers need to free some memory.
1791 */
1792
1793 typedef enum VarExprFlags {
1794 /* The variable expression is based on an undefined variable. */
1795 VEF_UNDEF = 0x01,
1796 /* The variable expression started as an undefined expression, but one
1797 * of the modifiers (such as :D or :U) has turned the expression from
1798 * undefined to defined. */
1799 VEF_DEF = 0x02
1800 } VarExprFlags;
1801
1802 ENUM_FLAGS_RTTI_2(VarExprFlags,
1803 VEF_UNDEF, VEF_DEF);
1804
1805
1806 typedef struct ApplyModifiersState {
1807 const char startc; /* '\0' or '{' or '(' */
1808 const char endc; /* '\0' or '}' or ')' */
1809 Var * const v;
1810 GNode * const ctxt;
1811 const VarEvalFlags eflags;
1812
1813 char *val; /* The old value of the expression,
1814 * before applying the modifier, never NULL */
1815 char *newVal; /* The new value of the expression,
1816 * after applying the modifier, never NULL */
1817 char sep; /* Word separator in expansions
1818 * (see the :ts modifier) */
1819 Boolean oneBigWord; /* TRUE if some modifiers that otherwise split
1820 * the variable value into words, like :S and
1821 * :C, treat the variable value as a single big
1822 * word, possibly containing spaces. */
1823 VarExprFlags exprFlags;
1824 } ApplyModifiersState;
1825
1826 static void
1827 ApplyModifiersState_Define(ApplyModifiersState *st)
1828 {
1829 if (st->exprFlags & VEF_UNDEF)
1830 st->exprFlags |= VEF_DEF;
1831 }
1832
1833 typedef enum ApplyModifierResult {
1834 AMR_OK, /* Continue parsing */
1835 AMR_UNKNOWN, /* Not a match, try other modifiers as well */
1836 AMR_BAD, /* Error out with "Bad modifier" message */
1837 AMR_CLEANUP /* Error out without error message */
1838 } ApplyModifierResult;
1839
1840 /* Allow backslashes to escape the delimiter, $, and \, but don't touch other
1841 * backslashes. */
1842 static Boolean
1843 IsEscapedModifierPart(const char *p, char delim,
1844 struct ModifyWord_SubstArgs *subst)
1845 {
1846 if (p[0] != '\\')
1847 return FALSE;
1848 if (p[1] == delim || p[1] == '\\' || p[1] == '$')
1849 return TRUE;
1850 return p[1] == '&' && subst != NULL;
1851 }
1852
1853 /*
1854 * Parse a part of a modifier such as the "from" and "to" in :S/from/to/ or
1855 * the "var" or "replacement ${var}" in :@var@replacement ${var}@, up to and
1856 * including the next unescaped delimiter. The delimiter, as well as the
1857 * backslash or the dollar, can be escaped with a backslash.
1858 *
1859 * Return the parsed (and possibly expanded) string, or NULL if no delimiter
1860 * was found. On successful return, the parsing position pp points right
1861 * after the delimiter. The delimiter is not included in the returned
1862 * value though.
1863 */
1864 static VarParseResult
1865 ParseModifierPart(
1866 const char **pp, /* The parsing position, updated upon return */
1867 char delim, /* Parsing stops at this delimiter */
1868 VarEvalFlags eflags, /* Flags for evaluating nested variables;
1869 * if VARE_WANTRES is not set, the text is
1870 * only parsed */
1871 ApplyModifiersState *st,
1872 char **out_part,
1873 size_t *out_length, /* Optionally stores the length of the returned
1874 * string, just to save another strlen call. */
1875 VarPatternFlags *out_pflags,/* For the first part of the :S modifier,
1876 * sets the VARP_ANCHOR_END flag if the last
1877 * character of the pattern is a $. */
1878 struct ModifyWord_SubstArgs *subst
1879 /* For the second part of the :S modifier,
1880 * allow ampersands to be escaped and replace
1881 * unescaped ampersands with subst->lhs. */
1882 ) {
1883 Buffer buf;
1884 const char *p;
1885
1886 Buf_Init(&buf);
1887
1888 /*
1889 * Skim through until the matching delimiter is found; pick up variable
1890 * expressions on the way.
1891 */
1892 p = *pp;
1893 while (*p != '\0' && *p != delim) {
1894 const char *varstart;
1895
1896 if (IsEscapedModifierPart(p, delim, subst)) {
1897 Buf_AddByte(&buf, p[1]);
1898 p += 2;
1899 continue;
1900 }
1901
1902 if (*p != '$') { /* Unescaped, simple text */
1903 if (subst != NULL && *p == '&')
1904 Buf_AddBytes(&buf, subst->lhs, subst->lhsLen);
1905 else
1906 Buf_AddByte(&buf, *p);
1907 p++;
1908 continue;
1909 }
1910
1911 if (p[1] == delim) { /* Unescaped $ at end of pattern */
1912 if (out_pflags != NULL)
1913 *out_pflags |= VARP_ANCHOR_END;
1914 else
1915 Buf_AddByte(&buf, *p);
1916 p++;
1917 continue;
1918 }
1919
1920 if (eflags & VARE_WANTRES) { /* Nested variable, evaluated */
1921 const char *nested_p = p;
1922 const char *nested_val;
1923 void *nested_val_freeIt;
1924 VarEvalFlags nested_eflags = eflags & ~(unsigned)VARE_ASSIGN;
1925
1926 (void)Var_Parse(&nested_p, st->ctxt, nested_eflags,
1927 &nested_val, &nested_val_freeIt);
1928 /* TODO: handle errors */
1929 Buf_AddStr(&buf, nested_val);
1930 free(nested_val_freeIt);
1931 p += nested_p - p;
1932 continue;
1933 }
1934
1935 /* XXX: This whole block is very similar to Var_Parse without
1936 * VARE_WANTRES. There may be subtle edge cases though that are
1937 * not yet covered in the unit tests and that are parsed differently,
1938 * depending on whether they are evaluated or not.
1939 *
1940 * This subtle difference is not documented in the manual page,
1941 * neither is the difference between parsing :D and :M documented.
1942 * No code should ever depend on these details, but who knows. */
1943
1944 varstart = p; /* Nested variable, only parsed */
1945 if (p[1] == '(' || p[1] == '{') {
1946 /*
1947 * Find the end of this variable reference
1948 * and suck it in without further ado.
1949 * It will be interpreted later.
1950 */
1951 char startc = p[1];
1952 int endc = startc == '(' ? ')' : '}';
1953 int depth = 1;
1954
1955 for (p += 2; *p != '\0' && depth > 0; p++) {
1956 if (p[-1] != '\\') {
1957 if (*p == startc)
1958 depth++;
1959 if (*p == endc)
1960 depth--;
1961 }
1962 }
1963 Buf_AddBytesBetween(&buf, varstart, p);
1964 } else {
1965 Buf_AddByte(&buf, *varstart);
1966 p++;
1967 }
1968 }
1969
1970 if (*p != delim) {
1971 *pp = p;
1972 Error("Unfinished modifier for %s ('%c' missing)", st->v->name, delim);
1973 *out_part = NULL;
1974 return VPR_PARSE_MSG;
1975 }
1976
1977 *pp = ++p;
1978 if (out_length != NULL)
1979 *out_length = Buf_Len(&buf);
1980
1981 *out_part = Buf_Destroy(&buf, FALSE);
1982 VAR_DEBUG1("Modifier part: \"%s\"\n", *out_part);
1983 return VPR_OK;
1984 }
1985
1986 /* Test whether mod starts with modname, followed by a delimiter. */
1987 static Boolean
1988 ModMatch(const char *mod, const char *modname, char endc)
1989 {
1990 size_t n = strlen(modname);
1991 return strncmp(mod, modname, n) == 0 &&
1992 (mod[n] == endc || mod[n] == ':');
1993 }
1994
1995 /* Test whether mod starts with modname, followed by a delimiter or '='. */
1996 static inline Boolean
1997 ModMatchEq(const char *mod, const char *modname, char endc)
1998 {
1999 size_t n = strlen(modname);
2000 return strncmp(mod, modname, n) == 0 &&
2001 (mod[n] == endc || mod[n] == ':' || mod[n] == '=');
2002 }
2003
2004 static Boolean
2005 TryParseIntBase0(const char **pp, int *out_num)
2006 {
2007 char *end;
2008 long n;
2009
2010 errno = 0;
2011 n = strtol(*pp, &end, 0);
2012 if ((n == LONG_MIN || n == LONG_MAX) && errno == ERANGE)
2013 return FALSE;
2014 if (n < INT_MIN || n > INT_MAX)
2015 return FALSE;
2016
2017 *pp = end;
2018 *out_num = (int)n;
2019 return TRUE;
2020 }
2021
2022 static Boolean
2023 TryParseSize(const char **pp, size_t *out_num)
2024 {
2025 char *end;
2026 unsigned long n;
2027
2028 if (!ch_isdigit(**pp))
2029 return FALSE;
2030
2031 errno = 0;
2032 n = strtoul(*pp, &end, 10);
2033 if (n == ULONG_MAX && errno == ERANGE)
2034 return FALSE;
2035 if (n > SIZE_MAX)
2036 return FALSE;
2037
2038 *pp = end;
2039 *out_num = (size_t)n;
2040 return TRUE;
2041 }
2042
2043 static Boolean
2044 TryParseChar(const char **pp, int base, char *out_ch)
2045 {
2046 char *end;
2047 unsigned long n;
2048
2049 if (!ch_isalnum(**pp))
2050 return FALSE;
2051
2052 errno = 0;
2053 n = strtoul(*pp, &end, base);
2054 if (n == ULONG_MAX && errno == ERANGE)
2055 return FALSE;
2056 if (n > UCHAR_MAX)
2057 return FALSE;
2058
2059 *pp = end;
2060 *out_ch = (char)n;
2061 return TRUE;
2062 }
2063
2064 /* :@var (at) ...${var}...@ */
2065 static ApplyModifierResult
2066 ApplyModifier_Loop(const char **pp, ApplyModifiersState *st)
2067 {
2068 struct ModifyWord_LoopArgs args;
2069 char prev_sep;
2070 VarEvalFlags eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2071 VarParseResult res;
2072
2073 args.ctx = st->ctxt;
2074
2075 (*pp)++; /* Skip the first '@' */
2076 res = ParseModifierPart(pp, '@', eflags, st,
2077 &args.tvar, NULL, NULL, NULL);
2078 if (res != VPR_OK)
2079 return AMR_CLEANUP;
2080 if (DEBUG(LINT) && strchr(args.tvar, '$') != NULL) {
2081 Parse_Error(PARSE_FATAL,
2082 "In the :@ modifier of \"%s\", the variable name \"%s\" "
2083 "must not contain a dollar.",
2084 st->v->name, args.tvar);
2085 return AMR_CLEANUP;
2086 }
2087
2088 res = ParseModifierPart(pp, '@', eflags, st,
2089 &args.str, NULL, NULL, NULL);
2090 if (res != VPR_OK)
2091 return AMR_CLEANUP;
2092
2093 args.eflags = st->eflags & (VARE_UNDEFERR | VARE_WANTRES);
2094 prev_sep = st->sep;
2095 st->sep = ' '; /* XXX: should be st->sep for consistency */
2096 st->newVal = ModifyWords(st->val, ModifyWord_Loop, &args,
2097 st->oneBigWord, st->sep);
2098 st->sep = prev_sep;
2099 Var_Delete(args.tvar, st->ctxt);
2100 free(args.tvar);
2101 free(args.str);
2102 return AMR_OK;
2103 }
2104
2105 /* :Ddefined or :Uundefined */
2106 static ApplyModifierResult
2107 ApplyModifier_Defined(const char **pp, ApplyModifiersState *st)
2108 {
2109 Buffer buf;
2110 const char *p;
2111
2112 VarEvalFlags eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2113 if (st->eflags & VARE_WANTRES) {
2114 if ((**pp == 'D') == !(st->exprFlags & VEF_UNDEF))
2115 eflags |= VARE_WANTRES;
2116 }
2117
2118 Buf_Init(&buf);
2119 p = *pp + 1;
2120 while (*p != st->endc && *p != ':' && *p != '\0') {
2121
2122 /* Escaped delimiter or other special character */
2123 if (*p == '\\') {
2124 char c = p[1];
2125 if (c == st->endc || c == ':' || c == '$' || c == '\\') {
2126 Buf_AddByte(&buf, c);
2127 p += 2;
2128 continue;
2129 }
2130 }
2131
2132 /* Nested variable expression */
2133 if (*p == '$') {
2134 const char *nested_val;
2135 void *nested_val_freeIt;
2136
2137 (void)Var_Parse(&p, st->ctxt, eflags,
2138 &nested_val, &nested_val_freeIt);
2139 /* TODO: handle errors */
2140 Buf_AddStr(&buf, nested_val);
2141 free(nested_val_freeIt);
2142 continue;
2143 }
2144
2145 /* Ordinary text */
2146 Buf_AddByte(&buf, *p);
2147 p++;
2148 }
2149 *pp = p;
2150
2151 ApplyModifiersState_Define(st);
2152
2153 if (eflags & VARE_WANTRES) {
2154 st->newVal = Buf_Destroy(&buf, FALSE);
2155 } else {
2156 st->newVal = st->val;
2157 Buf_Destroy(&buf, TRUE);
2158 }
2159 return AMR_OK;
2160 }
2161
2162 /* :L */
2163 static ApplyModifierResult
2164 ApplyModifier_Literal(const char **pp, ApplyModifiersState *st)
2165 {
2166 ApplyModifiersState_Define(st);
2167 st->newVal = bmake_strdup(st->v->name);
2168 (*pp)++;
2169 return AMR_OK;
2170 }
2171
2172 static Boolean
2173 TryParseTime(const char **pp, time_t *out_time)
2174 {
2175 char *end;
2176 unsigned long n;
2177
2178 if (!ch_isdigit(**pp))
2179 return FALSE;
2180
2181 errno = 0;
2182 n = strtoul(*pp, &end, 10);
2183 if (n == ULONG_MAX && errno == ERANGE)
2184 return FALSE;
2185
2186 *pp = end;
2187 *out_time = (time_t)n; /* ignore possible truncation for now */
2188 return TRUE;
2189 }
2190
2191 /* :gmtime */
2192 static ApplyModifierResult
2193 ApplyModifier_Gmtime(const char **pp, ApplyModifiersState *st)
2194 {
2195 time_t utc;
2196
2197 const char *mod = *pp;
2198 if (!ModMatchEq(mod, "gmtime", st->endc))
2199 return AMR_UNKNOWN;
2200
2201 if (mod[6] == '=') {
2202 const char *arg = mod + 7;
2203 if (!TryParseTime(&arg, &utc)) {
2204 Parse_Error(PARSE_FATAL, "Invalid time value: %s\n", mod + 7);
2205 return AMR_CLEANUP;
2206 }
2207 *pp = arg;
2208 } else {
2209 utc = 0;
2210 *pp = mod + 6;
2211 }
2212 st->newVal = VarStrftime(st->val, TRUE, utc);
2213 return AMR_OK;
2214 }
2215
2216 /* :localtime */
2217 static ApplyModifierResult
2218 ApplyModifier_Localtime(const char **pp, ApplyModifiersState *st)
2219 {
2220 time_t utc;
2221
2222 const char *mod = *pp;
2223 if (!ModMatchEq(mod, "localtime", st->endc))
2224 return AMR_UNKNOWN;
2225
2226 if (mod[9] == '=') {
2227 const char *arg = mod + 10;
2228 if (!TryParseTime(&arg, &utc)) {
2229 Parse_Error(PARSE_FATAL, "Invalid time value: %s\n", mod + 10);
2230 return AMR_CLEANUP;
2231 }
2232 *pp = arg;
2233 } else {
2234 utc = 0;
2235 *pp = mod + 9;
2236 }
2237 st->newVal = VarStrftime(st->val, FALSE, utc);
2238 return AMR_OK;
2239 }
2240
2241 /* :hash */
2242 static ApplyModifierResult
2243 ApplyModifier_Hash(const char **pp, ApplyModifiersState *st)
2244 {
2245 if (!ModMatch(*pp, "hash", st->endc))
2246 return AMR_UNKNOWN;
2247
2248 st->newVal = VarHash(st->val);
2249 *pp += 4;
2250 return AMR_OK;
2251 }
2252
2253 /* :P */
2254 static ApplyModifierResult
2255 ApplyModifier_Path(const char **pp, ApplyModifiersState *st)
2256 {
2257 GNode *gn;
2258 char *path;
2259
2260 ApplyModifiersState_Define(st);
2261
2262 gn = Targ_FindNode(st->v->name);
2263 if (gn == NULL || gn->type & OP_NOPATH) {
2264 path = NULL;
2265 } else if (gn->path != NULL) {
2266 path = bmake_strdup(gn->path);
2267 } else {
2268 SearchPath *searchPath = Suff_FindPath(gn);
2269 path = Dir_FindFile(st->v->name, searchPath);
2270 }
2271 if (path == NULL)
2272 path = bmake_strdup(st->v->name);
2273 st->newVal = path;
2274
2275 (*pp)++;
2276 return AMR_OK;
2277 }
2278
2279 /* :!cmd! */
2280 static ApplyModifierResult
2281 ApplyModifier_ShellCommand(const char **pp, ApplyModifiersState *st)
2282 {
2283 char *cmd;
2284 const char *errfmt;
2285 VarParseResult res;
2286
2287 (*pp)++;
2288 res = ParseModifierPart(pp, '!', st->eflags, st,
2289 &cmd, NULL, NULL, NULL);
2290 if (res != VPR_OK)
2291 return AMR_CLEANUP;
2292
2293 errfmt = NULL;
2294 if (st->eflags & VARE_WANTRES)
2295 st->newVal = Cmd_Exec(cmd, &errfmt);
2296 else
2297 st->newVal = bmake_strdup("");
2298 free(cmd);
2299
2300 if (errfmt != NULL)
2301 Error(errfmt, st->val); /* XXX: why still return AMR_OK? */
2302
2303 ApplyModifiersState_Define(st);
2304 return AMR_OK;
2305 }
2306
2307 /* The :range modifier generates an integer sequence as long as the words.
2308 * The :range=7 modifier generates an integer sequence from 1 to 7. */
2309 static ApplyModifierResult
2310 ApplyModifier_Range(const char **pp, ApplyModifiersState *st)
2311 {
2312 size_t n;
2313 Buffer buf;
2314 size_t i;
2315
2316 const char *mod = *pp;
2317 if (!ModMatchEq(mod, "range", st->endc))
2318 return AMR_UNKNOWN;
2319
2320 if (mod[5] == '=') {
2321 const char *p = mod + 6;
2322 if (!TryParseSize(&p, &n)) {
2323 Parse_Error(PARSE_FATAL, "Invalid number: %s\n", mod + 6);
2324 return AMR_CLEANUP;
2325 }
2326 *pp = p;
2327 } else {
2328 n = 0;
2329 *pp = mod + 5;
2330 }
2331
2332 if (n == 0) {
2333 Words words = Str_Words(st->val, FALSE);
2334 n = words.len;
2335 Words_Free(words);
2336 }
2337
2338 Buf_Init(&buf);
2339
2340 for (i = 0; i < n; i++) {
2341 if (i != 0)
2342 Buf_AddByte(&buf, ' '); /* XXX: st->sep, for consistency */
2343 Buf_AddInt(&buf, 1 + (int)i);
2344 }
2345
2346 st->newVal = Buf_Destroy(&buf, FALSE);
2347 return AMR_OK;
2348 }
2349
2350 /* :Mpattern or :Npattern */
2351 static ApplyModifierResult
2352 ApplyModifier_Match(const char **pp, ApplyModifiersState *st)
2353 {
2354 const char *mod = *pp;
2355 Boolean copy = FALSE; /* pattern should be, or has been, copied */
2356 Boolean needSubst = FALSE;
2357 const char *endpat;
2358 char *pattern;
2359 ModifyWordsCallback callback;
2360
2361 /*
2362 * In the loop below, ignore ':' unless we are at (or back to) the
2363 * original brace level.
2364 * XXX This will likely not work right if $() and ${} are intermixed.
2365 */
2366 int nest = 0;
2367 const char *p;
2368 for (p = mod + 1; *p != '\0' && !(*p == ':' && nest == 0); p++) {
2369 if (*p == '\\' &&
2370 (p[1] == ':' || p[1] == st->endc || p[1] == st->startc)) {
2371 if (!needSubst)
2372 copy = TRUE;
2373 p++;
2374 continue;
2375 }
2376 if (*p == '$')
2377 needSubst = TRUE;
2378 if (*p == '(' || *p == '{')
2379 nest++;
2380 if (*p == ')' || *p == '}') {
2381 nest--;
2382 if (nest < 0)
2383 break;
2384 }
2385 }
2386 *pp = p;
2387 endpat = p;
2388
2389 if (copy) {
2390 char *dst;
2391 const char *src;
2392
2393 /* Compress the \:'s out of the pattern. */
2394 pattern = bmake_malloc((size_t)(endpat - (mod + 1)) + 1);
2395 dst = pattern;
2396 src = mod + 1;
2397 for (; src < endpat; src++, dst++) {
2398 if (src[0] == '\\' && src + 1 < endpat &&
2399 /* XXX: st->startc is missing here; see above */
2400 (src[1] == ':' || src[1] == st->endc))
2401 src++;
2402 *dst = *src;
2403 }
2404 *dst = '\0';
2405 endpat = dst;
2406 } else {
2407 pattern = bmake_strsedup(mod + 1, endpat);
2408 }
2409
2410 if (needSubst) {
2411 /* pattern contains embedded '$', so use Var_Subst to expand it. */
2412 char *old_pattern = pattern;
2413 (void)Var_Subst(pattern, st->ctxt, st->eflags, &pattern);
2414 /* TODO: handle errors */
2415 free(old_pattern);
2416 }
2417
2418 VAR_DEBUG3("Pattern[%s] for [%s] is [%s]\n", st->v->name, st->val, pattern);
2419
2420 callback = mod[0] == 'M' ? ModifyWord_Match : ModifyWord_NoMatch;
2421 st->newVal = ModifyWords(st->val, callback, pattern,
2422 st->oneBigWord, st->sep);
2423 free(pattern);
2424 return AMR_OK;
2425 }
2426
2427 /* :S,from,to, */
2428 static ApplyModifierResult
2429 ApplyModifier_Subst(const char **pp, ApplyModifiersState *st)
2430 {
2431 struct ModifyWord_SubstArgs args;
2432 char *lhs, *rhs;
2433 Boolean oneBigWord;
2434 VarParseResult res;
2435
2436 char delim = (*pp)[1];
2437 if (delim == '\0') {
2438 Error("Missing delimiter for :S modifier");
2439 (*pp)++;
2440 return AMR_CLEANUP;
2441 }
2442
2443 *pp += 2;
2444
2445 args.pflags = 0;
2446 args.matched = FALSE;
2447
2448 /*
2449 * If pattern begins with '^', it is anchored to the
2450 * start of the word -- skip over it and flag pattern.
2451 */
2452 if (**pp == '^') {
2453 args.pflags |= VARP_ANCHOR_START;
2454 (*pp)++;
2455 }
2456
2457 res = ParseModifierPart(pp, delim, st->eflags, st,
2458 &lhs, &args.lhsLen, &args.pflags, NULL);
2459 if (res != VPR_OK)
2460 return AMR_CLEANUP;
2461 args.lhs = lhs;
2462
2463 res = ParseModifierPart(pp, delim, st->eflags, st,
2464 &rhs, &args.rhsLen, NULL, &args);
2465 if (res != VPR_OK)
2466 return AMR_CLEANUP;
2467 args.rhs = rhs;
2468
2469 oneBigWord = st->oneBigWord;
2470 for (;; (*pp)++) {
2471 switch (**pp) {
2472 case 'g':
2473 args.pflags |= VARP_SUB_GLOBAL;
2474 continue;
2475 case '1':
2476 args.pflags |= VARP_SUB_ONE;
2477 continue;
2478 case 'W':
2479 oneBigWord = TRUE;
2480 continue;
2481 }
2482 break;
2483 }
2484
2485 st->newVal = ModifyWords(st->val, ModifyWord_Subst, &args,
2486 oneBigWord, st->sep);
2487
2488 free(lhs);
2489 free(rhs);
2490 return AMR_OK;
2491 }
2492
2493 #ifndef NO_REGEX
2494
2495 /* :C,from,to, */
2496 static ApplyModifierResult
2497 ApplyModifier_Regex(const char **pp, ApplyModifiersState *st)
2498 {
2499 char *re;
2500 struct ModifyWord_SubstRegexArgs args;
2501 Boolean oneBigWord;
2502 int error;
2503 VarParseResult res;
2504
2505 char delim = (*pp)[1];
2506 if (delim == '\0') {
2507 Error("Missing delimiter for :C modifier");
2508 (*pp)++;
2509 return AMR_CLEANUP;
2510 }
2511
2512 *pp += 2;
2513
2514 res = ParseModifierPart(pp, delim, st->eflags, st,
2515 &re, NULL, NULL, NULL);
2516 if (res != VPR_OK)
2517 return AMR_CLEANUP;
2518
2519 res = ParseModifierPart(pp, delim, st->eflags, st,
2520 &args.replace, NULL, NULL, NULL);
2521 if (args.replace == NULL) {
2522 free(re);
2523 return AMR_CLEANUP;
2524 }
2525
2526 args.pflags = 0;
2527 args.matched = FALSE;
2528 oneBigWord = st->oneBigWord;
2529 for (;; (*pp)++) {
2530 switch (**pp) {
2531 case 'g':
2532 args.pflags |= VARP_SUB_GLOBAL;
2533 continue;
2534 case '1':
2535 args.pflags |= VARP_SUB_ONE;
2536 continue;
2537 case 'W':
2538 oneBigWord = TRUE;
2539 continue;
2540 }
2541 break;
2542 }
2543
2544 error = regcomp(&args.re, re, REG_EXTENDED);
2545 free(re);
2546 if (error) {
2547 VarREError(error, &args.re, "Regex compilation error");
2548 free(args.replace);
2549 return AMR_CLEANUP;
2550 }
2551
2552 args.nsub = args.re.re_nsub + 1;
2553 if (args.nsub > 10)
2554 args.nsub = 10;
2555 st->newVal = ModifyWords(st->val, ModifyWord_SubstRegex, &args,
2556 oneBigWord, st->sep);
2557 regfree(&args.re);
2558 free(args.replace);
2559 return AMR_OK;
2560 }
2561 #endif
2562
2563 /* :Q, :q */
2564 static ApplyModifierResult
2565 ApplyModifier_Quote(const char **pp, ApplyModifiersState *st)
2566 {
2567 if ((*pp)[1] == st->endc || (*pp)[1] == ':') {
2568 st->newVal = VarQuote(st->val, **pp == 'q');
2569 (*pp)++;
2570 return AMR_OK;
2571 } else
2572 return AMR_UNKNOWN;
2573 }
2574
2575 static void
2576 ModifyWord_Copy(const char *word, SepBuf *buf, void *data MAKE_ATTR_UNUSED)
2577 {
2578 SepBuf_AddStr(buf, word);
2579 }
2580
2581 /* :ts<separator> */
2582 static ApplyModifierResult
2583 ApplyModifier_ToSep(const char **pp, ApplyModifiersState *st)
2584 {
2585 const char *sep = *pp + 2;
2586
2587 /* ":ts<any><endc>" or ":ts<any>:" */
2588 if (sep[0] != st->endc && (sep[1] == st->endc || sep[1] == ':')) {
2589 st->sep = sep[0];
2590 *pp = sep + 1;
2591 goto ok;
2592 }
2593
2594 /* ":ts<endc>" or ":ts:" */
2595 if (sep[0] == st->endc || sep[0] == ':') {
2596 st->sep = '\0'; /* no separator */
2597 *pp = sep;
2598 goto ok;
2599 }
2600
2601 /* ":ts<unrecognised><unrecognised>". */
2602 if (sep[0] != '\\') {
2603 (*pp)++; /* just for backwards compatibility */
2604 return AMR_BAD;
2605 }
2606
2607 /* ":ts\n" */
2608 if (sep[1] == 'n') {
2609 st->sep = '\n';
2610 *pp = sep + 2;
2611 goto ok;
2612 }
2613
2614 /* ":ts\t" */
2615 if (sep[1] == 't') {
2616 st->sep = '\t';
2617 *pp = sep + 2;
2618 goto ok;
2619 }
2620
2621 /* ":ts\x40" or ":ts\100" */
2622 {
2623 const char *p = sep + 1;
2624 int base = 8; /* assume octal */
2625
2626 if (sep[1] == 'x') {
2627 base = 16;
2628 p++;
2629 } else if (!ch_isdigit(sep[1])) {
2630 (*pp)++; /* just for backwards compatibility */
2631 return AMR_BAD; /* ":ts<backslash><unrecognised>". */
2632 }
2633
2634 if (!TryParseChar(&p, base, &st->sep)) {
2635 Parse_Error(PARSE_FATAL, "Invalid character number: %s\n", p);
2636 return AMR_CLEANUP;
2637 }
2638 if (*p != ':' && *p != st->endc) {
2639 (*pp)++; /* just for backwards compatibility */
2640 return AMR_BAD;
2641 }
2642
2643 *pp = p;
2644 }
2645
2646 ok:
2647 st->newVal = ModifyWords(st->val, ModifyWord_Copy, NULL,
2648 st->oneBigWord, st->sep);
2649 return AMR_OK;
2650 }
2651
2652 /* :tA, :tu, :tl, :ts<separator>, etc. */
2653 static ApplyModifierResult
2654 ApplyModifier_To(const char **pp, ApplyModifiersState *st)
2655 {
2656 const char *mod = *pp;
2657 assert(mod[0] == 't');
2658
2659 if (mod[1] == st->endc || mod[1] == ':' || mod[1] == '\0') {
2660 *pp = mod + 1;
2661 return AMR_BAD; /* Found ":t<endc>" or ":t:". */
2662 }
2663
2664 if (mod[1] == 's')
2665 return ApplyModifier_ToSep(pp, st);
2666
2667 if (mod[2] != st->endc && mod[2] != ':') {
2668 *pp = mod + 1;
2669 return AMR_BAD; /* Found ":t<unrecognised><unrecognised>". */
2670 }
2671
2672 /* Check for two-character options: ":tu", ":tl" */
2673 if (mod[1] == 'A') { /* absolute path */
2674 st->newVal = ModifyWords(st->val, ModifyWord_Realpath, NULL,
2675 st->oneBigWord, st->sep);
2676 *pp = mod + 2;
2677 return AMR_OK;
2678 }
2679
2680 if (mod[1] == 'u') { /* :tu */
2681 size_t i;
2682 size_t len = strlen(st->val);
2683 st->newVal = bmake_malloc(len + 1);
2684 for (i = 0; i < len + 1; i++)
2685 st->newVal[i] = ch_toupper(st->val[i]);
2686 *pp = mod + 2;
2687 return AMR_OK;
2688 }
2689
2690 if (mod[1] == 'l') { /* :tl */
2691 size_t i;
2692 size_t len = strlen(st->val);
2693 st->newVal = bmake_malloc(len + 1);
2694 for (i = 0; i < len + 1; i++)
2695 st->newVal[i] = ch_tolower(st->val[i]);
2696 *pp = mod + 2;
2697 return AMR_OK;
2698 }
2699
2700 if (mod[1] == 'W' || mod[1] == 'w') { /* :tW, :tw */
2701 st->oneBigWord = mod[1] == 'W';
2702 st->newVal = st->val;
2703 *pp = mod + 2;
2704 return AMR_OK;
2705 }
2706
2707 /* Found ":t<unrecognised>:" or ":t<unrecognised><endc>". */
2708 *pp = mod + 1;
2709 return AMR_BAD;
2710 }
2711
2712 /* :[#], :[1], :[-1..1], etc. */
2713 static ApplyModifierResult
2714 ApplyModifier_Words(const char **pp, ApplyModifiersState *st)
2715 {
2716 char *estr;
2717 int first, last;
2718 VarParseResult res;
2719 const char *p;
2720
2721 (*pp)++; /* skip the '[' */
2722 res = ParseModifierPart(pp, ']', st->eflags, st,
2723 &estr, NULL, NULL, NULL);
2724 if (res != VPR_OK)
2725 return AMR_CLEANUP;
2726
2727 /* now *pp points just after the closing ']' */
2728 if (**pp != ':' && **pp != st->endc)
2729 goto bad_modifier; /* Found junk after ']' */
2730
2731 if (estr[0] == '\0')
2732 goto bad_modifier; /* empty square brackets in ":[]". */
2733
2734 if (estr[0] == '#' && estr[1] == '\0') { /* Found ":[#]" */
2735 if (st->oneBigWord) {
2736 st->newVal = bmake_strdup("1");
2737 } else {
2738 Buffer buf;
2739
2740 Words words = Str_Words(st->val, FALSE);
2741 size_t ac = words.len;
2742 Words_Free(words);
2743
2744 Buf_InitSize(&buf, 4); /* 3 digits + '\0' is usually enough */
2745 Buf_AddInt(&buf, (int)ac);
2746 st->newVal = Buf_Destroy(&buf, FALSE);
2747 }
2748 goto ok;
2749 }
2750
2751 if (estr[0] == '*' && estr[1] == '\0') {
2752 /* Found ":[*]" */
2753 st->oneBigWord = TRUE;
2754 st->newVal = st->val;
2755 goto ok;
2756 }
2757
2758 if (estr[0] == '@' && estr[1] == '\0') {
2759 /* Found ":[@]" */
2760 st->oneBigWord = FALSE;
2761 st->newVal = st->val;
2762 goto ok;
2763 }
2764
2765 /*
2766 * We expect estr to contain a single integer for :[N], or two integers
2767 * separated by ".." for :[start..end].
2768 */
2769 p = estr;
2770 if (!TryParseIntBase0(&p, &first))
2771 goto bad_modifier; /* Found junk instead of a number */
2772
2773 if (p[0] == '\0') { /* Found only one integer in :[N] */
2774 last = first;
2775 } else if (p[0] == '.' && p[1] == '.' && p[2] != '\0') {
2776 /* Expecting another integer after ".." */
2777 p += 2;
2778 if (!TryParseIntBase0(&p, &last) || *p != '\0')
2779 goto bad_modifier; /* Found junk after ".." */
2780 } else
2781 goto bad_modifier; /* Found junk instead of ".." */
2782
2783 /*
2784 * Now first and last are properly filled in, but we still have to check
2785 * for 0 as a special case.
2786 */
2787 if (first == 0 && last == 0) {
2788 /* ":[0]" or perhaps ":[0..0]" */
2789 st->oneBigWord = TRUE;
2790 st->newVal = st->val;
2791 goto ok;
2792 }
2793
2794 /* ":[0..N]" or ":[N..0]" */
2795 if (first == 0 || last == 0)
2796 goto bad_modifier;
2797
2798 /* Normal case: select the words described by first and last. */
2799 st->newVal = VarSelectWords(st->sep, st->oneBigWord, st->val, first, last);
2800
2801 ok:
2802 free(estr);
2803 return AMR_OK;
2804
2805 bad_modifier:
2806 free(estr);
2807 return AMR_BAD;
2808 }
2809
2810 static int
2811 str_cmp_asc(const void *a, const void *b)
2812 {
2813 return strcmp(*(const char * const *)a, *(const char * const *)b);
2814 }
2815
2816 static int
2817 str_cmp_desc(const void *a, const void *b)
2818 {
2819 return strcmp(*(const char * const *)b, *(const char * const *)a);
2820 }
2821
2822 /* :O (order ascending) or :Or (order descending) or :Ox (shuffle) */
2823 static ApplyModifierResult
2824 ApplyModifier_Order(const char **pp, ApplyModifiersState *st)
2825 {
2826 const char *mod = (*pp)++; /* skip past the 'O' in any case */
2827
2828 Words words = Str_Words(st->val, FALSE);
2829
2830 if (mod[1] == st->endc || mod[1] == ':') {
2831 /* :O sorts ascending */
2832 qsort(words.words, words.len, sizeof words.words[0], str_cmp_asc);
2833
2834 } else if ((mod[1] == 'r' || mod[1] == 'x') &&
2835 (mod[2] == st->endc || mod[2] == ':')) {
2836 (*pp)++;
2837
2838 if (mod[1] == 'r') {
2839 /* :Or sorts descending */
2840 qsort(words.words, words.len, sizeof words.words[0], str_cmp_desc);
2841
2842 } else {
2843 /* :Ox shuffles
2844 *
2845 * We will use [ac..2] range for mod factors. This will produce
2846 * random numbers in [(ac-1)..0] interval, and minimal
2847 * reasonable value for mod factor is 2 (the mod 1 will produce
2848 * 0 with probability 1).
2849 */
2850 size_t i;
2851 for (i = words.len - 1; i > 0; i--) {
2852 size_t rndidx = (size_t)random() % (i + 1);
2853 char *t = words.words[i];
2854 words.words[i] = words.words[rndidx];
2855 words.words[rndidx] = t;
2856 }
2857 }
2858 } else {
2859 Words_Free(words);
2860 return AMR_BAD;
2861 }
2862
2863 st->newVal = Words_JoinFree(words);
2864 return AMR_OK;
2865 }
2866
2867 /* :? then : else */
2868 static ApplyModifierResult
2869 ApplyModifier_IfElse(const char **pp, ApplyModifiersState *st)
2870 {
2871 char *then_expr, *else_expr;
2872 VarParseResult res;
2873
2874 Boolean value = FALSE;
2875 VarEvalFlags then_eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2876 VarEvalFlags else_eflags = st->eflags & ~(unsigned)VARE_WANTRES;
2877
2878 int cond_rc = COND_PARSE; /* anything other than COND_INVALID */
2879 if (st->eflags & VARE_WANTRES) {
2880 cond_rc = Cond_EvalCondition(st->v->name, &value);
2881 if (cond_rc != COND_INVALID && value)
2882 then_eflags |= VARE_WANTRES;
2883 if (cond_rc != COND_INVALID && !value)
2884 else_eflags |= VARE_WANTRES;
2885 }
2886
2887 (*pp)++; /* skip past the '?' */
2888 res = ParseModifierPart(pp, ':', then_eflags, st,
2889 &then_expr, NULL, NULL, NULL);
2890 if (res != VPR_OK)
2891 return AMR_CLEANUP;
2892
2893 res = ParseModifierPart(pp, st->endc, else_eflags, st,
2894 &else_expr, NULL, NULL, NULL);
2895 if (res != VPR_OK)
2896 return AMR_CLEANUP;
2897
2898 (*pp)--;
2899 if (cond_rc == COND_INVALID) {
2900 Error("Bad conditional expression `%s' in %s?%s:%s",
2901 st->v->name, st->v->name, then_expr, else_expr);
2902 return AMR_CLEANUP;
2903 }
2904
2905 if (value) {
2906 st->newVal = then_expr;
2907 free(else_expr);
2908 } else {
2909 st->newVal = else_expr;
2910 free(then_expr);
2911 }
2912 ApplyModifiersState_Define(st);
2913 return AMR_OK;
2914 }
2915
2916 /*
2917 * The ::= modifiers actually assign a value to the variable.
2918 * Their main purpose is in supporting modifiers of .for loop
2919 * iterators and other obscure uses. They always expand to
2920 * nothing. In a target rule that would otherwise expand to an
2921 * empty line they can be preceded with @: to keep make happy.
2922 * Eg.
2923 *
2924 * foo: .USE
2925 * .for i in ${.TARGET} ${.TARGET:R}.gz
2926 * @: ${t::=$i}
2927 * @echo blah ${t:T}
2928 * .endfor
2929 *
2930 * ::=<str> Assigns <str> as the new value of variable.
2931 * ::?=<str> Assigns <str> as value of variable if
2932 * it was not already set.
2933 * ::+=<str> Appends <str> to variable.
2934 * ::!=<cmd> Assigns output of <cmd> as the new value of
2935 * variable.
2936 */
2937 static ApplyModifierResult
2938 ApplyModifier_Assign(const char **pp, ApplyModifiersState *st)
2939 {
2940 GNode *v_ctxt;
2941 char delim;
2942 char *val;
2943 VarParseResult res;
2944
2945 const char *mod = *pp;
2946 const char *op = mod + 1;
2947
2948 if (op[0] == '=')
2949 goto ok;
2950 if ((op[0] == '!' || op[0] == '+' || op[0] == '?') && op[1] == '=')
2951 goto ok;
2952 return AMR_UNKNOWN; /* "::<unrecognised>" */
2953 ok:
2954
2955 if (st->v->name[0] == '\0') {
2956 *pp = mod + 1;
2957 return AMR_BAD;
2958 }
2959
2960 v_ctxt = st->ctxt; /* context where v belongs */
2961 if (!(st->exprFlags & VEF_UNDEF) && st->ctxt != VAR_GLOBAL) {
2962 Var *gv = VarFind(st->v->name, st->ctxt, FALSE);
2963 if (gv == NULL)
2964 v_ctxt = VAR_GLOBAL;
2965 else
2966 VarFreeEnv(gv, TRUE);
2967 }
2968
2969 switch (op[0]) {
2970 case '+':
2971 case '?':
2972 case '!':
2973 *pp = mod + 3;
2974 break;
2975 default:
2976 *pp = mod + 2;
2977 break;
2978 }
2979
2980 delim = st->startc == '(' ? ')' : '}';
2981 /* TODO: Add test for using the ::= modifier in a := assignment line.
2982 * Probably st->eflags should be passed down without VARE_ASSIGN here. */
2983 res = ParseModifierPart(pp, delim, st->eflags, st, &val, NULL, NULL, NULL);
2984 if (res != VPR_OK)
2985 return AMR_CLEANUP;
2986
2987 (*pp)--;
2988
2989 if (st->eflags & VARE_WANTRES) {
2990 switch (op[0]) {
2991 case '+':
2992 Var_Append(st->v->name, val, v_ctxt);
2993 break;
2994 case '!': {
2995 const char *errfmt;
2996 char *cmd_output = Cmd_Exec(val, &errfmt);
2997 if (errfmt)
2998 Error(errfmt, val);
2999 else
3000 Var_Set(st->v->name, cmd_output, v_ctxt);
3001 free(cmd_output);
3002 break;
3003 }
3004 case '?':
3005 if (!(st->exprFlags & VEF_UNDEF))
3006 break;
3007 /* FALLTHROUGH */
3008 default:
3009 Var_Set(st->v->name, val, v_ctxt);
3010 break;
3011 }
3012 }
3013 free(val);
3014 st->newVal = bmake_strdup("");
3015 return AMR_OK;
3016 }
3017
3018 /* :_=...
3019 * remember current value */
3020 static ApplyModifierResult
3021 ApplyModifier_Remember(const char **pp, ApplyModifiersState *st)
3022 {
3023 const char *mod = *pp;
3024 if (!ModMatchEq(mod, "_", st->endc))
3025 return AMR_UNKNOWN;
3026
3027 if (mod[1] == '=') {
3028 size_t n = strcspn(mod + 2, ":)}");
3029 char *name = bmake_strldup(mod + 2, n);
3030 Var_Set(name, st->val, st->ctxt);
3031 free(name);
3032 *pp = mod + 2 + n;
3033 } else {
3034 Var_Set("_", st->val, st->ctxt);
3035 *pp = mod + 1;
3036 }
3037 st->newVal = st->val;
3038 return AMR_OK;
3039 }
3040
3041 /* Apply the given function to each word of the variable value,
3042 * for a single-letter modifier such as :H, :T. */
3043 static ApplyModifierResult
3044 ApplyModifier_WordFunc(const char **pp, ApplyModifiersState *st,
3045 ModifyWordsCallback modifyWord)
3046 {
3047 char delim = (*pp)[1];
3048 if (delim != st->endc && delim != ':')
3049 return AMR_UNKNOWN;
3050
3051 st->newVal = ModifyWords(st->val, modifyWord, NULL,
3052 st->oneBigWord, st->sep);
3053 (*pp)++;
3054 return AMR_OK;
3055 }
3056
3057 static ApplyModifierResult
3058 ApplyModifier_Unique(const char **pp, ApplyModifiersState *st)
3059 {
3060 if ((*pp)[1] == st->endc || (*pp)[1] == ':') {
3061 st->newVal = VarUniq(st->val);
3062 (*pp)++;
3063 return AMR_OK;
3064 } else
3065 return AMR_UNKNOWN;
3066 }
3067
3068 #ifdef SYSVVARSUB
3069 /* :from=to */
3070 static ApplyModifierResult
3071 ApplyModifier_SysV(const char **pp, ApplyModifiersState *st)
3072 {
3073 char *lhs, *rhs;
3074 VarParseResult res;
3075
3076 const char *mod = *pp;
3077 Boolean eqFound = FALSE;
3078
3079 /*
3080 * First we make a pass through the string trying to verify it is a
3081 * SysV-make-style translation. It must be: <lhs>=<rhs>
3082 */
3083 int depth = 1;
3084 const char *p = mod;
3085 while (*p != '\0' && depth > 0) {
3086 if (*p == '=') { /* XXX: should also test depth == 1 */
3087 eqFound = TRUE;
3088 /* continue looking for st->endc */
3089 } else if (*p == st->endc)
3090 depth--;
3091 else if (*p == st->startc)
3092 depth++;
3093 if (depth > 0)
3094 p++;
3095 }
3096 if (*p != st->endc || !eqFound)
3097 return AMR_UNKNOWN;
3098
3099 *pp = mod;
3100 res = ParseModifierPart(pp, '=', st->eflags, st,
3101 &lhs, NULL, NULL, NULL);
3102 if (res != VPR_OK)
3103 return AMR_CLEANUP;
3104
3105 /* The SysV modifier lasts until the end of the variable expression. */
3106 res = ParseModifierPart(pp, st->endc, st->eflags, st,
3107 &rhs, NULL, NULL, NULL);
3108 if (res != VPR_OK)
3109 return AMR_CLEANUP;
3110
3111 (*pp)--;
3112 if (lhs[0] == '\0' && st->val[0] == '\0') {
3113 st->newVal = st->val; /* special case */
3114 } else {
3115 struct ModifyWord_SYSVSubstArgs args = {st->ctxt, lhs, rhs};
3116 st->newVal = ModifyWords(st->val, ModifyWord_SYSVSubst, &args,
3117 st->oneBigWord, st->sep);
3118 }
3119 free(lhs);
3120 free(rhs);
3121 return AMR_OK;
3122 }
3123 #endif
3124
3125 #ifdef SUNSHCMD
3126 /* :sh */
3127 static ApplyModifierResult
3128 ApplyModifier_SunShell(const char **pp, ApplyModifiersState *st)
3129 {
3130 const char *p = *pp;
3131 if (p[1] == 'h' && (p[2] == st->endc || p[2] == ':')) {
3132 if (st->eflags & VARE_WANTRES) {
3133 const char *errfmt;
3134 st->newVal = Cmd_Exec(st->val, &errfmt);
3135 if (errfmt)
3136 Error(errfmt, st->val);
3137 } else
3138 st->newVal = bmake_strdup("");
3139 *pp = p + 2;
3140 return AMR_OK;
3141 } else
3142 return AMR_UNKNOWN;
3143 }
3144 #endif
3145
3146 static void
3147 LogBeforeApply(const ApplyModifiersState *st, const char *mod, const char endc)
3148 {
3149 char eflags_str[VarEvalFlags_ToStringSize];
3150 char vflags_str[VarFlags_ToStringSize];
3151 char exprflags_str[VarExprFlags_ToStringSize];
3152 Boolean is_single_char = mod[0] != '\0' &&
3153 (mod[1] == endc || mod[1] == ':');
3154
3155 /* At this point, only the first character of the modifier can
3156 * be used since the end of the modifier is not yet known. */
3157 debug_printf("Applying ${%s:%c%s} to \"%s\" (%s, %s, %s)\n",
3158 st->v->name, mod[0], is_single_char ? "" : "...", st->val,
3159 Enum_FlagsToString(eflags_str, sizeof eflags_str,
3160 st->eflags, VarEvalFlags_ToStringSpecs),
3161 Enum_FlagsToString(vflags_str, sizeof vflags_str,
3162 st->v->flags, VarFlags_ToStringSpecs),
3163 Enum_FlagsToString(exprflags_str, sizeof exprflags_str,
3164 st->exprFlags,
3165 VarExprFlags_ToStringSpecs));
3166 }
3167
3168 static void
3169 LogAfterApply(ApplyModifiersState *st, const char *p, const char *mod)
3170 {
3171 char eflags_str[VarEvalFlags_ToStringSize];
3172 char vflags_str[VarFlags_ToStringSize];
3173 char exprflags_str[VarExprFlags_ToStringSize];
3174 const char *quot = st->newVal == var_Error ? "" : "\"";
3175 const char *newVal = st->newVal == var_Error ? "error" : st->newVal;
3176
3177 debug_printf("Result of ${%s:%.*s} is %s%s%s (%s, %s, %s)\n",
3178 st->v->name, (int)(p - mod), mod, quot, newVal, quot,
3179 Enum_FlagsToString(eflags_str, sizeof eflags_str,
3180 st->eflags, VarEvalFlags_ToStringSpecs),
3181 Enum_FlagsToString(vflags_str, sizeof vflags_str,
3182 st->v->flags, VarFlags_ToStringSpecs),
3183 Enum_FlagsToString(exprflags_str, sizeof exprflags_str,
3184 st->exprFlags,
3185 VarExprFlags_ToStringSpecs));
3186 }
3187
3188 static ApplyModifierResult
3189 ApplyModifier(const char **pp, ApplyModifiersState *st)
3190 {
3191 switch (**pp) {
3192 case ':':
3193 return ApplyModifier_Assign(pp, st);
3194 case '@':
3195 return ApplyModifier_Loop(pp, st);
3196 case '_':
3197 return ApplyModifier_Remember(pp, st);
3198 case 'D':
3199 case 'U':
3200 return ApplyModifier_Defined(pp, st);
3201 case 'L':
3202 return ApplyModifier_Literal(pp, st);
3203 case 'P':
3204 return ApplyModifier_Path(pp, st);
3205 case '!':
3206 return ApplyModifier_ShellCommand(pp, st);
3207 case '[':
3208 return ApplyModifier_Words(pp, st);
3209 case 'g':
3210 return ApplyModifier_Gmtime(pp, st);
3211 case 'h':
3212 return ApplyModifier_Hash(pp, st);
3213 case 'l':
3214 return ApplyModifier_Localtime(pp, st);
3215 case 't':
3216 return ApplyModifier_To(pp, st);
3217 case 'N':
3218 case 'M':
3219 return ApplyModifier_Match(pp, st);
3220 case 'S':
3221 return ApplyModifier_Subst(pp, st);
3222 case '?':
3223 return ApplyModifier_IfElse(pp, st);
3224 #ifndef NO_REGEX
3225 case 'C':
3226 return ApplyModifier_Regex(pp, st);
3227 #endif
3228 case 'q':
3229 case 'Q':
3230 return ApplyModifier_Quote(pp, st);
3231 case 'T':
3232 return ApplyModifier_WordFunc(pp, st, ModifyWord_Tail);
3233 case 'H':
3234 return ApplyModifier_WordFunc(pp, st, ModifyWord_Head);
3235 case 'E':
3236 return ApplyModifier_WordFunc(pp, st, ModifyWord_Suffix);
3237 case 'R':
3238 return ApplyModifier_WordFunc(pp, st, ModifyWord_Root);
3239 case 'r':
3240 return ApplyModifier_Range(pp, st);
3241 case 'O':
3242 return ApplyModifier_Order(pp, st);
3243 case 'u':
3244 return ApplyModifier_Unique(pp, st);
3245 #ifdef SUNSHCMD
3246 case 's':
3247 return ApplyModifier_SunShell(pp, st);
3248 #endif
3249 default:
3250 return AMR_UNKNOWN;
3251 }
3252 }
3253
3254 static char *ApplyModifiers(const char **, char *, char, char, Var *,
3255 VarExprFlags *, GNode *, VarEvalFlags, void **);
3256
3257 typedef enum ApplyModifiersIndirectResult {
3258 AMIR_CONTINUE,
3259 AMIR_APPLY_MODS,
3260 AMIR_OUT
3261 } ApplyModifiersIndirectResult;
3262
3263 /* While expanding a variable expression, expand and apply indirect
3264 * modifiers such as in ${VAR:${M_indirect}}. */
3265 static ApplyModifiersIndirectResult
3266 ApplyModifiersIndirect(
3267 ApplyModifiersState *const st,
3268 const char **const inout_p,
3269 void **const inout_freeIt
3270 ) {
3271 const char *p = *inout_p;
3272 const char *mods;
3273 void *mods_freeIt;
3274
3275 (void)Var_Parse(&p, st->ctxt, st->eflags, &mods, &mods_freeIt);
3276 /* TODO: handle errors */
3277
3278 /*
3279 * If we have not parsed up to st->endc or ':', we are not
3280 * interested. This means the expression ${VAR:${M_1}${M_2}}
3281 * is not accepted, but ${VAR:${M_1}:${M_2}} is.
3282 */
3283 if (mods[0] != '\0' && *p != '\0' && *p != ':' && *p != st->endc) {
3284 if (DEBUG(LINT))
3285 Parse_Error(PARSE_FATAL,
3286 "Missing delimiter ':' after indirect modifier \"%.*s\"",
3287 (int)(p - *inout_p), *inout_p);
3288
3289 free(mods_freeIt);
3290 /* XXX: apply_mods doesn't sound like "not interested". */
3291 /* XXX: Why is the indirect modifier parsed once more by
3292 * apply_mods? If any, p should be advanced to nested_p. */
3293 return AMIR_APPLY_MODS;
3294 }
3295
3296 VAR_DEBUG3("Indirect modifier \"%s\" from \"%.*s\"\n",
3297 mods, (int)(p - *inout_p), *inout_p);
3298
3299 if (mods[0] != '\0') {
3300 const char *rval_pp = mods;
3301 st->val = ApplyModifiers(&rval_pp, st->val, '\0', '\0', st->v,
3302 &st->exprFlags, st->ctxt, st->eflags,
3303 inout_freeIt);
3304 if (st->val == var_Error || st->val == varUndefined ||
3305 *rval_pp != '\0') {
3306 free(mods_freeIt);
3307 *inout_p = p;
3308 return AMIR_OUT; /* error already reported */
3309 }
3310 }
3311 free(mods_freeIt);
3312
3313 if (*p == ':')
3314 p++;
3315 else if (*p == '\0' && st->endc != '\0') {
3316 Error("Unclosed variable specification after complex "
3317 "modifier (expecting '%c') for %s", st->endc, st->v->name);
3318 *inout_p = p;
3319 return AMIR_OUT;
3320 }
3321
3322 *inout_p = p;
3323 return AMIR_CONTINUE;
3324 }
3325
3326 /* Apply any modifiers (such as :Mpattern or :@var@loop@ or :Q or ::=value). */
3327 static char *
3328 ApplyModifiers(
3329 const char **const pp, /* the parsing position, updated upon return */
3330 char *const val, /* the current value of the expression */
3331 char const startc, /* '(' or '{', or '\0' for indirect modifiers */
3332 char const endc, /* ')' or '}', or '\0' for indirect modifiers */
3333 Var *const v,
3334 VarExprFlags *const exprFlags,
3335 GNode *const ctxt, /* for looking up and modifying variables */
3336 VarEvalFlags const eflags,
3337 void **const inout_freeIt /* free this after using the return value */
3338 ) {
3339 ApplyModifiersState st = {
3340 startc, endc, v, ctxt, eflags,
3341 val, /* .val */
3342 var_Error, /* .newVal */
3343 ' ', /* .sep */
3344 FALSE, /* .oneBigWord */
3345 *exprFlags /* .exprFlags */
3346 };
3347 const char *p;
3348 const char *mod;
3349 ApplyModifierResult res;
3350
3351 assert(startc == '(' || startc == '{' || startc == '\0');
3352 assert(endc == ')' || endc == '}' || endc == '\0');
3353 assert(val != NULL);
3354
3355 p = *pp;
3356
3357 if (*p == '\0' && endc != '\0') {
3358 Error("Unclosed variable expression (expecting '%c') for \"%s\"",
3359 st.endc, st.v->name);
3360 goto cleanup;
3361 }
3362
3363 while (*p != '\0' && *p != endc) {
3364
3365 if (*p == '$') {
3366 ApplyModifiersIndirectResult amir;
3367 amir = ApplyModifiersIndirect(&st, &p, inout_freeIt);
3368 if (amir == AMIR_CONTINUE)
3369 continue;
3370 if (amir == AMIR_OUT)
3371 goto out;
3372 }
3373 st.newVal = var_Error; /* default value, in case of errors */
3374 mod = p;
3375
3376 if (DEBUG(VAR))
3377 LogBeforeApply(&st, mod, endc);
3378
3379 res = ApplyModifier(&p, &st);
3380
3381 #ifdef SYSVVARSUB
3382 if (res == AMR_UNKNOWN) {
3383 assert(p == mod);
3384 res = ApplyModifier_SysV(&p, &st);
3385 }
3386 #endif
3387
3388 if (res == AMR_UNKNOWN) {
3389 Error("Unknown modifier '%c'", *mod);
3390 /* Guess the end of the current modifier.
3391 * XXX: Skipping the rest of the modifier hides errors and leads
3392 * to wrong results. Parsing should rather stop here. */
3393 for (p++; *p != ':' && *p != st.endc && *p != '\0'; p++)
3394 continue;
3395 st.newVal = var_Error;
3396 }
3397 if (res == AMR_CLEANUP)
3398 goto cleanup;
3399 if (res == AMR_BAD)
3400 goto bad_modifier;
3401
3402 if (DEBUG(VAR))
3403 LogAfterApply(&st, p, mod);
3404
3405 if (st.newVal != st.val) {
3406 if (*inout_freeIt != NULL) {
3407 free(st.val);
3408 *inout_freeIt = NULL;
3409 }
3410 st.val = st.newVal;
3411 if (st.val != var_Error && st.val != varUndefined)
3412 *inout_freeIt = st.val;
3413 }
3414 if (*p == '\0' && st.endc != '\0') {
3415 Error("Unclosed variable specification (expecting '%c') "
3416 "for \"%s\" (value \"%s\") modifier %c",
3417 st.endc, st.v->name, st.val, *mod);
3418 } else if (*p == ':') {
3419 p++;
3420 } else if (DEBUG(LINT) && *p != '\0' && *p != endc) {
3421 Parse_Error(PARSE_FATAL,
3422 "Missing delimiter ':' after modifier \"%.*s\"",
3423 (int)(p - mod), mod);
3424 /* TODO: propagate parse error to the enclosing expression */
3425 }
3426 }
3427 out:
3428 *pp = p;
3429 assert(st.val != NULL); /* Use var_Error or varUndefined instead. */
3430 *exprFlags = st.exprFlags;
3431 return st.val;
3432
3433 bad_modifier:
3434 /* XXX: The modifier end is only guessed. */
3435 Error("Bad modifier `:%.*s' for %s",
3436 (int)strcspn(mod, ":)}"), mod, st.v->name);
3437
3438 cleanup:
3439 *pp = p;
3440 free(*inout_freeIt);
3441 *inout_freeIt = NULL;
3442 *exprFlags = st.exprFlags;
3443 return var_Error;
3444 }
3445
3446 /* Only four of the local variables are treated specially as they are the
3447 * only four that will be set when dynamic sources are expanded. */
3448 static Boolean
3449 VarnameIsDynamic(const char *name, size_t len)
3450 {
3451 if (len == 1 || (len == 2 && (name[1] == 'F' || name[1] == 'D'))) {
3452 switch (name[0]) {
3453 case '@':
3454 case '%':
3455 case '*':
3456 case '!':
3457 return TRUE;
3458 }
3459 return FALSE;
3460 }
3461
3462 if ((len == 7 || len == 8) && name[0] == '.' && ch_isupper(name[1])) {
3463 return strcmp(name, ".TARGET") == 0 ||
3464 strcmp(name, ".ARCHIVE") == 0 ||
3465 strcmp(name, ".PREFIX") == 0 ||
3466 strcmp(name, ".MEMBER") == 0;
3467 }
3468
3469 return FALSE;
3470 }
3471
3472 static const char *
3473 UndefinedShortVarValue(char varname, const GNode *ctxt, VarEvalFlags eflags)
3474 {
3475 if (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL) {
3476 /*
3477 * If substituting a local variable in a non-local context,
3478 * assume it's for dynamic source stuff. We have to handle
3479 * this specially and return the longhand for the variable
3480 * with the dollar sign escaped so it makes it back to the
3481 * caller. Only four of the local variables are treated
3482 * specially as they are the only four that will be set
3483 * when dynamic sources are expanded.
3484 */
3485 switch (varname) {
3486 case '@':
3487 return "$(.TARGET)";
3488 case '%':
3489 return "$(.MEMBER)";
3490 case '*':
3491 return "$(.PREFIX)";
3492 case '!':
3493 return "$(.ARCHIVE)";
3494 }
3495 }
3496 return eflags & VARE_UNDEFERR ? var_Error : varUndefined;
3497 }
3498
3499 /* Parse a variable name, until the end character or a colon, whichever
3500 * comes first. */
3501 static char *
3502 ParseVarname(const char **pp, char startc, char endc,
3503 GNode *ctxt, VarEvalFlags eflags,
3504 size_t *out_varname_len)
3505 {
3506 Buffer buf;
3507 const char *p = *pp;
3508 int depth = 1;
3509
3510 Buf_Init(&buf);
3511
3512 while (*p != '\0') {
3513 /* Track depth so we can spot parse errors. */
3514 if (*p == startc)
3515 depth++;
3516 if (*p == endc) {
3517 if (--depth == 0)
3518 break;
3519 }
3520 if (*p == ':' && depth == 1)
3521 break;
3522
3523 /* A variable inside a variable, expand. */
3524 if (*p == '$') {
3525 const char *nested_val;
3526 void *nested_val_freeIt;
3527 (void)Var_Parse(&p, ctxt, eflags, &nested_val, &nested_val_freeIt);
3528 /* TODO: handle errors */
3529 Buf_AddStr(&buf, nested_val);
3530 free(nested_val_freeIt);
3531 } else {
3532 Buf_AddByte(&buf, *p);
3533 p++;
3534 }
3535 }
3536 *pp = p;
3537 *out_varname_len = Buf_Len(&buf);
3538 return Buf_Destroy(&buf, FALSE);
3539 }
3540
3541 static VarParseResult
3542 ValidShortVarname(char varname, const char *start)
3543 {
3544 switch (varname) {
3545 case '\0':
3546 case ')':
3547 case '}':
3548 case ':':
3549 case '$':
3550 break; /* and continue below */
3551 default:
3552 return VPR_OK;
3553 }
3554
3555 if (!DEBUG(LINT))
3556 return VPR_PARSE_SILENT;
3557
3558 if (varname == '$')
3559 Parse_Error(PARSE_FATAL,
3560 "To escape a dollar, use \\$, not $$, at \"%s\"", start);
3561 else if (varname == '\0')
3562 Parse_Error(PARSE_FATAL, "Dollar followed by nothing");
3563 else
3564 Parse_Error(PARSE_FATAL,
3565 "Invalid variable name '%c', at \"%s\"", varname, start);
3566
3567 return VPR_PARSE_MSG;
3568 }
3569
3570 /* Parse a single-character variable name such as $V or $@.
3571 * Return whether to continue parsing. */
3572 static Boolean
3573 ParseVarnameShort(char startc, const char **pp, GNode *ctxt,
3574 VarEvalFlags eflags,
3575 VarParseResult *out_FALSE_res, const char **out_FALSE_val,
3576 Var **out_TRUE_var)
3577 {
3578 char name[2];
3579 Var *v;
3580 VarParseResult vpr;
3581
3582 /*
3583 * If it's not bounded by braces of some sort, life is much simpler.
3584 * We just need to check for the first character and return the
3585 * value if it exists.
3586 */
3587
3588 vpr = ValidShortVarname(startc, *pp);
3589 if (vpr != VPR_OK) {
3590 (*pp)++;
3591 *out_FALSE_val = var_Error;
3592 *out_FALSE_res = vpr;
3593 return FALSE;
3594 }
3595
3596 name[0] = startc;
3597 name[1] = '\0';
3598 v = VarFind(name, ctxt, TRUE);
3599 if (v == NULL) {
3600 *pp += 2;
3601
3602 *out_FALSE_val = UndefinedShortVarValue(startc, ctxt, eflags);
3603 if (DEBUG(LINT) && *out_FALSE_val == var_Error) {
3604 Parse_Error(PARSE_FATAL, "Variable \"%s\" is undefined", name);
3605 *out_FALSE_res = VPR_UNDEF_MSG;
3606 return FALSE;
3607 }
3608 *out_FALSE_res = eflags & VARE_UNDEFERR ? VPR_UNDEF_SILENT : VPR_OK;
3609 return FALSE;
3610 }
3611
3612 *out_TRUE_var = v;
3613 return TRUE;
3614 }
3615
3616 /* Find variables like @F or <D. */
3617 static Var *
3618 FindLocalLegacyVar(const char *varname, size_t namelen, GNode *ctxt,
3619 const char **out_extraModifiers)
3620 {
3621 /* Only resolve these variables if ctxt is a "real" target. */
3622 if (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL)
3623 return NULL;
3624
3625 if (namelen != 2)
3626 return NULL;
3627 if (varname[1] != 'F' && varname[1] != 'D')
3628 return NULL;
3629 if (strchr("@%?*!<>", varname[0]) == NULL)
3630 return NULL;
3631
3632 {
3633 char name[] = { varname[0], '\0' };
3634 Var *v = VarFind(name, ctxt, FALSE);
3635
3636 if (v != NULL) {
3637 if (varname[1] == 'D') {
3638 *out_extraModifiers = "H:";
3639 } else { /* F */
3640 *out_extraModifiers = "T:";
3641 }
3642 }
3643 return v;
3644 }
3645 }
3646
3647 static VarParseResult
3648 EvalUndefined(Boolean dynamic, const char *start, const char *p, char *varname,
3649 VarEvalFlags eflags,
3650 void **out_freeIt, const char **out_val)
3651 {
3652 if (dynamic) {
3653 char *pstr = bmake_strsedup(start, p);
3654 free(varname);
3655 *out_freeIt = pstr;
3656 *out_val = pstr;
3657 return VPR_OK;
3658 }
3659
3660 if ((eflags & VARE_UNDEFERR) && (eflags & VARE_WANTRES) &&
3661 DEBUG(LINT))
3662 {
3663 Parse_Error(PARSE_FATAL, "Variable \"%s\" is undefined", varname);
3664 free(varname);
3665 *out_val = var_Error;
3666 return VPR_UNDEF_MSG;
3667 }
3668
3669 if (eflags & VARE_UNDEFERR) {
3670 free(varname);
3671 *out_val = var_Error;
3672 return VPR_UNDEF_SILENT;
3673 }
3674
3675 free(varname);
3676 *out_val = varUndefined;
3677 return VPR_OK;
3678 }
3679
3680 /* Parse a long variable name enclosed in braces or parentheses such as $(VAR)
3681 * or ${VAR}, up to the closing brace or parenthesis, or in the case of
3682 * ${VAR:Modifiers}, up to the ':' that starts the modifiers.
3683 * Return whether to continue parsing. */
3684 static Boolean
3685 ParseVarnameLong(
3686 const char *p,
3687 char startc,
3688 GNode *ctxt,
3689 VarEvalFlags eflags,
3690
3691 const char **out_FALSE_pp,
3692 VarParseResult *out_FALSE_res,
3693 const char **out_FALSE_val,
3694 void **out_FALSE_freeIt,
3695
3696 char *out_TRUE_endc,
3697 const char **out_TRUE_p,
3698 Var **out_TRUE_v,
3699 Boolean *out_TRUE_haveModifier,
3700 const char **out_TRUE_extraModifiers,
3701 Boolean *out_TRUE_dynamic,
3702 VarExprFlags *out_TRUE_exprFlags
3703 ) {
3704 size_t namelen;
3705 char *varname;
3706 Var *v;
3707 Boolean haveModifier;
3708 Boolean dynamic = FALSE;
3709
3710 const char *const start = p;
3711 char endc = startc == '(' ? ')' : '}';
3712
3713 p += 2; /* skip "${" or "$(" or "y(" */
3714 varname = ParseVarname(&p, startc, endc, ctxt, eflags, &namelen);
3715
3716 if (*p == ':') {
3717 haveModifier = TRUE;
3718 } else if (*p == endc) {
3719 haveModifier = FALSE;
3720 } else {
3721 Parse_Error(PARSE_FATAL, "Unclosed variable \"%s\"", varname);
3722 free(varname);
3723 *out_FALSE_pp = p;
3724 *out_FALSE_val = var_Error;
3725 *out_FALSE_res = VPR_PARSE_MSG;
3726 return FALSE;
3727 }
3728
3729 v = VarFind(varname, ctxt, TRUE);
3730
3731 /* At this point, p points just after the variable name,
3732 * either at ':' or at endc. */
3733
3734 if (v == NULL)
3735 v = FindLocalLegacyVar(varname, namelen, ctxt, out_TRUE_extraModifiers);
3736
3737 if (v == NULL) {
3738 /* Defer expansion of dynamic variables if they appear in non-local
3739 * context since they are not defined there. */
3740 dynamic = VarnameIsDynamic(varname, namelen) &&
3741 (ctxt == VAR_CMDLINE || ctxt == VAR_GLOBAL);
3742
3743 if (!haveModifier) {
3744 p++; /* skip endc */
3745 *out_FALSE_pp = p;
3746 *out_FALSE_res = EvalUndefined(dynamic, start, p, varname, eflags,
3747 out_FALSE_freeIt, out_FALSE_val);
3748 return FALSE;
3749 }
3750
3751 /* The variable expression is based on an undefined variable.
3752 * Nevertheless it needs a Var, for modifiers that access the
3753 * variable name, such as :L or :?.
3754 *
3755 * Most modifiers leave this expression in the "undefined" state
3756 * (VEF_UNDEF), only a few modifiers like :D, :U, :L, :P turn this
3757 * undefined expression into a defined expression (VEF_DEF).
3758 *
3759 * At the end, after applying all modifiers, if the expression
3760 * is still undefined, Var_Parse will return an empty string
3761 * instead of the actually computed value. */
3762 v = VarNew(varname, varname, "", 0);
3763 *out_TRUE_exprFlags = VEF_UNDEF;
3764 } else
3765 free(varname);
3766
3767 *out_TRUE_endc = endc;
3768 *out_TRUE_p = p;
3769 *out_TRUE_v = v;
3770 *out_TRUE_haveModifier = haveModifier;
3771 *out_TRUE_dynamic = dynamic;
3772 return TRUE;
3773 }
3774
3775 /*
3776 * Given the start of a variable expression (such as $v, $(VAR),
3777 * ${VAR:Mpattern}), extract the variable name and value, and the modifiers,
3778 * if any. While doing that, apply the modifiers to the value of the
3779 * expression, forming its final value. A few of the modifiers such as :!cmd!
3780 * or ::= have side effects.
3781 *
3782 * Input:
3783 * *pp The string to parse.
3784 * When parsing a condition in ParseEmptyArg, it may also
3785 * point to the "y" of "empty(VARNAME:Modifiers)", which
3786 * is syntactically the same.
3787 * ctxt The context for finding variables
3788 * eflags Control the exact details of parsing
3789 *
3790 * Output:
3791 * *pp The position where to continue parsing.
3792 * TODO: After a parse error, the value of *pp is
3793 * unspecified. It may not have been updated at all,
3794 * point to some random character in the string, to the
3795 * location of the parse error, or at the end of the
3796 * string.
3797 * *out_val The value of the variable expression, never NULL.
3798 * *out_val var_Error if there was a parse error.
3799 * *out_val var_Error if the base variable of the expression was
3800 * undefined, eflags contains VARE_UNDEFERR, and none of
3801 * the modifiers turned the undefined expression into a
3802 * defined expression.
3803 * XXX: It is not guaranteed that an error message has
3804 * been printed.
3805 * *out_val varUndefined if the base variable of the expression
3806 * was undefined, eflags did not contain VARE_UNDEFERR,
3807 * and none of the modifiers turned the undefined
3808 * expression into a defined expression.
3809 * XXX: It is not guaranteed that an error message has
3810 * been printed.
3811 * *out_val_freeIt Must be freed by the caller after using *out_val.
3812 */
3813 /* coverity[+alloc : arg-*4] */
3814 VarParseResult
3815 Var_Parse(const char **pp, GNode *ctxt, VarEvalFlags eflags,
3816 const char **out_val, void **out_val_freeIt)
3817 {
3818 const char *p = *pp;
3819 const char *const start = p;
3820 Boolean haveModifier; /* TRUE if have modifiers for the variable */
3821 char startc; /* Starting character if variable in parens
3822 * or braces */
3823 char endc; /* Ending character if variable in parens
3824 * or braces */
3825 Boolean dynamic; /* TRUE if the variable is local and we're
3826 * expanding it in a non-local context. This
3827 * is done to support dynamic sources. The
3828 * result is just the expression, unaltered */
3829 const char *extramodifiers;
3830 Var *v;
3831 char *value;
3832 char eflags_str[VarEvalFlags_ToStringSize];
3833 VarExprFlags exprFlags = 0;
3834
3835 VAR_DEBUG2("Var_Parse: %s with %s\n", start,
3836 Enum_FlagsToString(eflags_str, sizeof eflags_str, eflags,
3837 VarEvalFlags_ToStringSpecs));
3838
3839 *out_val_freeIt = NULL;
3840 extramodifiers = NULL; /* extra modifiers to apply first */
3841 dynamic = FALSE;
3842
3843 /* Appease GCC, which thinks that the variable might not be
3844 * initialized. */
3845 endc = '\0';
3846
3847 startc = p[1];
3848 if (startc != '(' && startc != '{') {
3849 VarParseResult res;
3850 if (!ParseVarnameShort(startc, pp, ctxt, eflags, &res, out_val, &v))
3851 return res;
3852 haveModifier = FALSE;
3853 p++;
3854 } else {
3855 VarParseResult res;
3856 if (!ParseVarnameLong(p, startc, ctxt, eflags,
3857 pp, &res, out_val, out_val_freeIt,
3858 &endc, &p, &v, &haveModifier, &extramodifiers,
3859 &dynamic, &exprFlags))
3860 return res;
3861 }
3862
3863 if (v->flags & VAR_IN_USE)
3864 Fatal("Variable %s is recursive.", v->name);
3865
3866 value = Buf_GetAll(&v->val, NULL);
3867
3868 /* Before applying any modifiers, expand any nested expressions from the
3869 * variable value. */
3870 if (strchr(value, '$') != NULL && (eflags & VARE_WANTRES)) {
3871 VarEvalFlags nested_eflags = eflags;
3872 if (DEBUG(LINT))
3873 nested_eflags &= ~(unsigned)VARE_UNDEFERR;
3874 v->flags |= VAR_IN_USE;
3875 (void)Var_Subst(value, ctxt, nested_eflags, &value);
3876 v->flags &= ~(unsigned)VAR_IN_USE;
3877 /* TODO: handle errors */
3878 *out_val_freeIt = value;
3879 }
3880
3881 if (haveModifier || extramodifiers != NULL) {
3882 void *extraFree;
3883
3884 extraFree = NULL;
3885 if (extramodifiers != NULL) {
3886 const char *em = extramodifiers;
3887 value = ApplyModifiers(&em, value, '\0', '\0',
3888 v, &exprFlags, ctxt, eflags, &extraFree);
3889 }
3890
3891 if (haveModifier) {
3892 /* Skip initial colon. */
3893 p++;
3894
3895 value = ApplyModifiers(&p, value, startc, endc,
3896 v, &exprFlags, ctxt, eflags, out_val_freeIt);
3897 free(extraFree);
3898 } else {
3899 *out_val_freeIt = extraFree;
3900 }
3901 }
3902
3903 if (*p != '\0') /* Skip past endc if possible. */
3904 p++;
3905
3906 *pp = p;
3907
3908 if (v->flags & VAR_FROM_ENV) {
3909 /* Free the environment variable now since we own it,
3910 * but don't free the variable value if it will be returned. */
3911 Boolean keepValue = value == Buf_GetAll(&v->val, NULL);
3912 if (keepValue)
3913 *out_val_freeIt = value;
3914 (void)VarFreeEnv(v, !keepValue);
3915
3916 } else if (exprFlags & VEF_UNDEF) {
3917 if (!(exprFlags & VEF_DEF)) {
3918 if (*out_val_freeIt != NULL) {
3919 free(*out_val_freeIt);
3920 *out_val_freeIt = NULL;
3921 }
3922 if (dynamic) {
3923 value = bmake_strsedup(start, p);
3924 *out_val_freeIt = value;
3925 } else {
3926 /* The expression is still undefined, therefore discard the
3927 * actual value and return an error marker instead. */
3928 value = (eflags & VARE_UNDEFERR) ? var_Error : varUndefined;
3929 }
3930 }
3931 if (value != Buf_GetAll(&v->val, NULL))
3932 Buf_Destroy(&v->val, TRUE);
3933 free(v->name_freeIt);
3934 free(v);
3935 }
3936 *out_val = value;
3937 return VPR_UNKNOWN;
3938 }
3939
3940 /* Expand all variable expressions like $V, ${VAR}, $(VAR:Modifiers) in the
3941 * given string.
3942 *
3943 * Input:
3944 * str The string in which the variable expressions are
3945 * expanded.
3946 * ctxt The context in which to start searching for
3947 * variables. The other contexts are searched as well.
3948 * eflags VARE_UNDEFERR if undefineds are an error
3949 * VARE_WANTRES if we actually want the result
3950 * VARE_ASSIGN if we are in a := assignment
3951 */
3952 VarParseResult
3953 Var_Subst(const char *str, GNode *ctxt, VarEvalFlags eflags, char **out_res)
3954 {
3955 const char *p = str;
3956 Buffer buf; /* Buffer for forming things */
3957
3958 /* Set true if an error has already been reported,
3959 * to prevent a plethora of messages when recursing */
3960 static Boolean errorReported;
3961
3962 Buf_Init(&buf);
3963 errorReported = FALSE;
3964
3965 while (*p != '\0') {
3966 if (p[0] == '$' && p[1] == '$') {
3967 /* A dollar sign may be escaped with another dollar sign. */
3968 if (save_dollars && (eflags & VARE_ASSIGN))
3969 Buf_AddByte(&buf, '$');
3970 Buf_AddByte(&buf, '$');
3971 p += 2;
3972 } else if (*p != '$') {
3973 /*
3974 * Skip as many characters as possible -- either to the end of
3975 * the string or to the next dollar sign (variable expression).
3976 */
3977 const char *plainStart = p;
3978
3979 for (p++; *p != '$' && *p != '\0'; p++)
3980 continue;
3981 Buf_AddBytesBetween(&buf, plainStart, p);
3982 } else {
3983 const char *nested_p = p;
3984 void *freeIt;
3985 const char *val;
3986 (void)Var_Parse(&nested_p, ctxt, eflags, &val, &freeIt);
3987 /* TODO: handle errors */
3988
3989 if (val == var_Error || val == varUndefined) {
3990 if (!preserveUndefined) {
3991 p = nested_p;
3992 } else if ((eflags & VARE_UNDEFERR) || val == var_Error) {
3993 /* XXX: This condition is wrong. If val == var_Error,
3994 * this doesn't necessarily mean there was an undefined
3995 * variable. It could equally well be a parse error; see
3996 * unit-tests/varmod-order.exp. */
3997
3998 /*
3999 * If variable is undefined, complain and skip the
4000 * variable. The complaint will stop us from doing anything
4001 * when the file is parsed.
4002 */
4003 if (!errorReported) {
4004 Parse_Error(PARSE_FATAL, "Undefined variable \"%.*s\"",
4005 (int)(size_t)(nested_p - p), p);
4006 }
4007 p = nested_p;
4008 errorReported = TRUE;
4009 } else {
4010 /* Copy the initial '$' of the undefined expression,
4011 * thereby deferring expansion of the expression, but
4012 * expand nested expressions if already possible.
4013 * See unit-tests/varparse-undef-partial.mk. */
4014 Buf_AddByte(&buf, *p);
4015 p++;
4016 }
4017 } else {
4018 p = nested_p;
4019 Buf_AddStr(&buf, val);
4020 }
4021 free(freeIt);
4022 freeIt = NULL;
4023 }
4024 }
4025
4026 *out_res = Buf_DestroyCompact(&buf);
4027 return VPR_OK;
4028 }
4029
4030 /* Initialize the variables module. */
4031 void
4032 Var_Init(void)
4033 {
4034 VAR_INTERNAL = Targ_NewGN("Internal");
4035 VAR_GLOBAL = Targ_NewGN("Global");
4036 VAR_CMDLINE = Targ_NewGN("Command");
4037 }
4038
4039 /* Clean up the variables module. */
4040 void
4041 Var_End(void)
4042 {
4043 Var_Stats();
4044 }
4045
4046 void
4047 Var_Stats(void)
4048 {
4049 HashTable_DebugStats(&VAR_GLOBAL->context, "VAR_GLOBAL");
4050 }
4051
4052 /* Print all variables in a context, sorted by name. */
4053 void
4054 Var_Dump(GNode *ctxt)
4055 {
4056 Vector /* of const char * */ vec;
4057 HashIter hi;
4058 size_t i;
4059 const char **varnames;
4060
4061 Vector_Init(&vec, sizeof(const char *));
4062
4063 HashIter_Init(&hi, &ctxt->context);
4064 while (HashIter_Next(&hi) != NULL)
4065 *(const char **)Vector_Push(&vec) = hi.entry->key;
4066 varnames = vec.items;
4067
4068 qsort(varnames, vec.len, sizeof varnames[0], str_cmp_asc);
4069
4070 for (i = 0; i < vec.len; i++) {
4071 const char *varname = varnames[i];
4072 Var *var = HashTable_FindValue(&ctxt->context, varname);
4073 debug_printf("%-16s = %s\n", varname, Buf_GetAll(&var->val, NULL));
4074 }
4075
4076 Vector_Done(&vec);
4077 }
4078