disasm: Add EVEX decorator syntax
[nasm.git] / preproc.c
blob3a0f67b06871e401063051099060db6620d8e5c1
1 /* ----------------------------------------------------------------------- *
3 * Copyright 1996-2012 The NASM Authors - All Rights Reserved
4 * See the file AUTHORS included with the NASM distribution for
5 * the specific copyright holders.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following
9 * conditions are met:
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above
14 * copyright notice, this list of conditions and the following
15 * disclaimer in the documentation and/or other materials provided
16 * with the distribution.
18 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
19 * CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
20 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
21 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
22 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
23 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
25 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
26 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
29 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
30 * EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32 * ----------------------------------------------------------------------- */
35 * preproc.c macro preprocessor for the Netwide Assembler
38 /* Typical flow of text through preproc
40 * pp_getline gets tokenized lines, either
42 * from a macro expansion
44 * or
45 * {
46 * read_line gets raw text from stdmacpos, or predef, or current input file
47 * tokenize converts to tokens
48 * }
50 * expand_mmac_params is used to expand %1 etc., unless a macro is being
51 * defined or a false conditional is being processed
52 * (%0, %1, %+1, %-1, %%foo
54 * do_directive checks for directives
56 * expand_smacro is used to expand single line macros
58 * expand_mmacro is used to expand multi-line macros
60 * detoken is used to convert the line back to text
63 #include "compiler.h"
65 #include <stdio.h>
66 #include <stdarg.h>
67 #include <stdlib.h>
68 #include <stddef.h>
69 #include <string.h>
70 #include <ctype.h>
71 #include <limits.h>
72 #include <inttypes.h>
74 #include "nasm.h"
75 #include "nasmlib.h"
76 #include "preproc.h"
77 #include "hashtbl.h"
78 #include "quote.h"
79 #include "stdscan.h"
80 #include "eval.h"
81 #include "tokens.h"
82 #include "tables.h"
84 typedef struct SMacro SMacro;
85 typedef struct MMacro MMacro;
86 typedef struct MMacroInvocation MMacroInvocation;
87 typedef struct Context Context;
88 typedef struct Token Token;
89 typedef struct Blocks Blocks;
90 typedef struct Line Line;
91 typedef struct Include Include;
92 typedef struct Cond Cond;
93 typedef struct IncPath IncPath;
96 * Note on the storage of both SMacro and MMacros: the hash table
97 * indexes them case-insensitively, and we then have to go through a
98 * linked list of potential case aliases (and, for MMacros, parameter
99 * ranges); this is to preserve the matching semantics of the earlier
100 * code. If the number of case aliases for a specific macro is a
101 * performance issue, you may want to reconsider your coding style.
105 * Store the definition of a single-line macro.
107 struct SMacro {
108 SMacro *next;
109 char *name;
110 bool casesense;
111 bool in_progress;
112 unsigned int nparam;
113 Token *expansion;
117 * Store the definition of a multi-line macro. This is also used to
118 * store the interiors of `%rep...%endrep' blocks, which are
119 * effectively self-re-invoking multi-line macros which simply
120 * don't have a name or bother to appear in the hash tables. %rep
121 * blocks are signified by having a NULL `name' field.
123 * In a MMacro describing a `%rep' block, the `in_progress' field
124 * isn't merely boolean, but gives the number of repeats left to
125 * run.
127 * The `next' field is used for storing MMacros in hash tables; the
128 * `next_active' field is for stacking them on istk entries.
130 * When a MMacro is being expanded, `params', `iline', `nparam',
131 * `paramlen', `rotate' and `unique' are local to the invocation.
133 struct MMacro {
134 MMacro *next;
135 MMacroInvocation *prev; /* previous invocation */
136 char *name;
137 int nparam_min, nparam_max;
138 bool casesense;
139 bool plus; /* is the last parameter greedy? */
140 bool nolist; /* is this macro listing-inhibited? */
141 int64_t in_progress; /* is this macro currently being expanded? */
142 int32_t max_depth; /* maximum number of recursive expansions allowed */
143 Token *dlist; /* All defaults as one list */
144 Token **defaults; /* Parameter default pointers */
145 int ndefs; /* number of default parameters */
146 Line *expansion;
148 MMacro *next_active;
149 MMacro *rep_nest; /* used for nesting %rep */
150 Token **params; /* actual parameters */
151 Token *iline; /* invocation line */
152 unsigned int nparam, rotate;
153 int *paramlen;
154 uint64_t unique;
155 int lineno; /* Current line number on expansion */
156 uint64_t condcnt; /* number of if blocks... */
160 /* Store the definition of a multi-line macro, as defined in a
161 * previous recursive macro expansion.
163 struct MMacroInvocation {
164 MMacroInvocation *prev; /* previous invocation */
165 Token **params; /* actual parameters */
166 Token *iline; /* invocation line */
167 unsigned int nparam, rotate;
168 int *paramlen;
169 uint64_t unique;
170 uint64_t condcnt;
175 * The context stack is composed of a linked list of these.
177 struct Context {
178 Context *next;
179 char *name;
180 struct hash_table localmac;
181 uint32_t number;
185 * This is the internal form which we break input lines up into.
186 * Typically stored in linked lists.
188 * Note that `type' serves a double meaning: TOK_SMAC_PARAM is not
189 * necessarily used as-is, but is intended to denote the number of
190 * the substituted parameter. So in the definition
192 * %define a(x,y) ( (x) & ~(y) )
194 * the token representing `x' will have its type changed to
195 * TOK_SMAC_PARAM, but the one representing `y' will be
196 * TOK_SMAC_PARAM+1.
198 * TOK_INTERNAL_STRING is a dirty hack: it's a single string token
199 * which doesn't need quotes around it. Used in the pre-include
200 * mechanism as an alternative to trying to find a sensible type of
201 * quote to use on the filename we were passed.
203 enum pp_token_type {
204 TOK_NONE = 0, TOK_WHITESPACE, TOK_COMMENT, TOK_ID,
205 TOK_PREPROC_ID, TOK_STRING,
206 TOK_NUMBER, TOK_FLOAT, TOK_SMAC_END, TOK_OTHER,
207 TOK_INTERNAL_STRING,
208 TOK_PREPROC_Q, TOK_PREPROC_QQ,
209 TOK_PASTE, /* %+ */
210 TOK_INDIRECT, /* %[...] */
211 TOK_BRACE, /* \{...\} */
212 TOK_SMAC_PARAM, /* MUST BE LAST IN THE LIST!!! */
213 TOK_MAX = INT_MAX /* Keep compiler from reducing the range */
216 #define PP_CONCAT_MASK(x) (1 << (x))
217 #define PP_CONCAT_MATCH(t, mask) (PP_CONCAT_MASK((t)->type) & mask)
219 struct tokseq_match {
220 int mask_head;
221 int mask_tail;
224 struct Token {
225 Token *next;
226 char *text;
227 union {
228 SMacro *mac; /* associated macro for TOK_SMAC_END */
229 size_t len; /* scratch length field */
230 } a; /* Auxiliary data */
231 enum pp_token_type type;
235 * Multi-line macro definitions are stored as a linked list of
236 * these, which is essentially a container to allow several linked
237 * lists of Tokens.
239 * Note that in this module, linked lists are treated as stacks
240 * wherever possible. For this reason, Lines are _pushed_ on to the
241 * `expansion' field in MMacro structures, so that the linked list,
242 * if walked, would give the macro lines in reverse order; this
243 * means that we can walk the list when expanding a macro, and thus
244 * push the lines on to the `expansion' field in _istk_ in reverse
245 * order (so that when popped back off they are in the right
246 * order). It may seem cockeyed, and it relies on my design having
247 * an even number of steps in, but it works...
249 * Some of these structures, rather than being actual lines, are
250 * markers delimiting the end of the expansion of a given macro.
251 * This is for use in the cycle-tracking and %rep-handling code.
252 * Such structures have `finishes' non-NULL, and `first' NULL. All
253 * others have `finishes' NULL, but `first' may still be NULL if
254 * the line is blank.
256 struct Line {
257 Line *next;
258 MMacro *finishes;
259 Token *first;
263 * To handle an arbitrary level of file inclusion, we maintain a
264 * stack (ie linked list) of these things.
266 struct Include {
267 Include *next;
268 FILE *fp;
269 Cond *conds;
270 Line *expansion;
271 char *fname;
272 int lineno, lineinc;
273 MMacro *mstk; /* stack of active macros/reps */
277 * Include search path. This is simply a list of strings which get
278 * prepended, in turn, to the name of an include file, in an
279 * attempt to find the file if it's not in the current directory.
281 struct IncPath {
282 IncPath *next;
283 char *path;
287 * Conditional assembly: we maintain a separate stack of these for
288 * each level of file inclusion. (The only reason we keep the
289 * stacks separate is to ensure that a stray `%endif' in a file
290 * included from within the true branch of a `%if' won't terminate
291 * it and cause confusion: instead, rightly, it'll cause an error.)
293 struct Cond {
294 Cond *next;
295 int state;
297 enum {
299 * These states are for use just after %if or %elif: IF_TRUE
300 * means the condition has evaluated to truth so we are
301 * currently emitting, whereas IF_FALSE means we are not
302 * currently emitting but will start doing so if a %else comes
303 * up. In these states, all directives are admissible: %elif,
304 * %else and %endif. (And of course %if.)
306 COND_IF_TRUE, COND_IF_FALSE,
308 * These states come up after a %else: ELSE_TRUE means we're
309 * emitting, and ELSE_FALSE means we're not. In ELSE_* states,
310 * any %elif or %else will cause an error.
312 COND_ELSE_TRUE, COND_ELSE_FALSE,
314 * These states mean that we're not emitting now, and also that
315 * nothing until %endif will be emitted at all. COND_DONE is
316 * used when we've had our moment of emission
317 * and have now started seeing %elifs. COND_NEVER is used when
318 * the condition construct in question is contained within a
319 * non-emitting branch of a larger condition construct,
320 * or if there is an error.
322 COND_DONE, COND_NEVER
324 #define emitting(x) ( (x) == COND_IF_TRUE || (x) == COND_ELSE_TRUE )
327 * These defines are used as the possible return values for do_directive
329 #define NO_DIRECTIVE_FOUND 0
330 #define DIRECTIVE_FOUND 1
333 * This define sets the upper limit for smacro and recursive mmacro
334 * expansions
336 #define DEADMAN_LIMIT (1 << 20)
338 /* max reps */
339 #define REP_LIMIT ((INT64_C(1) << 62))
342 * Condition codes. Note that we use c_ prefix not C_ because C_ is
343 * used in nasm.h for the "real" condition codes. At _this_ level,
344 * we treat CXZ and ECXZ as condition codes, albeit non-invertible
345 * ones, so we need a different enum...
347 static const char * const conditions[] = {
348 "a", "ae", "b", "be", "c", "cxz", "e", "ecxz", "g", "ge", "l", "le",
349 "na", "nae", "nb", "nbe", "nc", "ne", "ng", "nge", "nl", "nle", "no",
350 "np", "ns", "nz", "o", "p", "pe", "po", "rcxz", "s", "z"
352 enum pp_conds {
353 c_A, c_AE, c_B, c_BE, c_C, c_CXZ, c_E, c_ECXZ, c_G, c_GE, c_L, c_LE,
354 c_NA, c_NAE, c_NB, c_NBE, c_NC, c_NE, c_NG, c_NGE, c_NL, c_NLE, c_NO,
355 c_NP, c_NS, c_NZ, c_O, c_P, c_PE, c_PO, c_RCXZ, c_S, c_Z,
356 c_none = -1
358 static const enum pp_conds inverse_ccs[] = {
359 c_NA, c_NAE, c_NB, c_NBE, c_NC, -1, c_NE, -1, c_NG, c_NGE, c_NL, c_NLE,
360 c_A, c_AE, c_B, c_BE, c_C, c_E, c_G, c_GE, c_L, c_LE, c_O, c_P, c_S,
361 c_Z, c_NO, c_NP, c_PO, c_PE, -1, c_NS, c_NZ
365 * Directive names.
367 /* If this is a an IF, ELIF, ELSE or ENDIF keyword */
368 static int is_condition(enum preproc_token arg)
370 return PP_IS_COND(arg) || (arg == PP_ELSE) || (arg == PP_ENDIF);
373 /* For TASM compatibility we need to be able to recognise TASM compatible
374 * conditional compilation directives. Using the NASM pre-processor does
375 * not work, so we look for them specifically from the following list and
376 * then jam in the equivalent NASM directive into the input stream.
379 enum {
380 TM_ARG, TM_ELIF, TM_ELSE, TM_ENDIF, TM_IF, TM_IFDEF, TM_IFDIFI,
381 TM_IFNDEF, TM_INCLUDE, TM_LOCAL
384 static const char * const tasm_directives[] = {
385 "arg", "elif", "else", "endif", "if", "ifdef", "ifdifi",
386 "ifndef", "include", "local"
389 static int StackSize = 4;
390 static char *StackPointer = "ebp";
391 static int ArgOffset = 8;
392 static int LocalOffset = 0;
394 static Context *cstk;
395 static Include *istk;
396 static IncPath *ipath = NULL;
398 static int pass; /* HACK: pass 0 = generate dependencies only */
399 static StrList **dephead, **deptail; /* Dependency list */
401 static uint64_t unique; /* unique identifier numbers */
403 static Line *predef = NULL;
404 static bool do_predef;
406 static ListGen *list;
409 * The current set of multi-line macros we have defined.
411 static struct hash_table mmacros;
414 * The current set of single-line macros we have defined.
416 static struct hash_table smacros;
419 * The multi-line macro we are currently defining, or the %rep
420 * block we are currently reading, if any.
422 static MMacro *defining;
424 static uint64_t nested_mac_count;
425 static uint64_t nested_rep_count;
428 * The number of macro parameters to allocate space for at a time.
430 #define PARAM_DELTA 16
433 * The standard macro set: defined in macros.c in the array nasm_stdmac.
434 * This gives our position in the macro set, when we're processing it.
436 static macros_t *stdmacpos;
439 * The extra standard macros that come from the object format, if
440 * any.
442 static macros_t *extrastdmac = NULL;
443 static bool any_extrastdmac;
446 * Tokens are allocated in blocks to improve speed
448 #define TOKEN_BLOCKSIZE 4096
449 static Token *freeTokens = NULL;
450 struct Blocks {
451 Blocks *next;
452 void *chunk;
455 static Blocks blocks = { NULL, NULL };
458 * Forward declarations.
460 static Token *expand_mmac_params(Token * tline);
461 static Token *expand_smacro(Token * tline);
462 static Token *expand_id(Token * tline);
463 static Context *get_ctx(const char *name, const char **namep);
464 static void make_tok_num(Token * tok, int64_t val);
465 static void error(int severity, const char *fmt, ...);
466 static void error_precond(int severity, const char *fmt, ...);
467 static void *new_Block(size_t size);
468 static void delete_Blocks(void);
469 static Token *new_Token(Token * next, enum pp_token_type type,
470 const char *text, int txtlen);
471 static Token *delete_Token(Token * t);
474 * Macros for safe checking of token pointers, avoid *(NULL)
476 #define tok_type_(x,t) ((x) && (x)->type == (t))
477 #define skip_white_(x) if (tok_type_((x), TOK_WHITESPACE)) (x)=(x)->next
478 #define tok_is_(x,v) (tok_type_((x), TOK_OTHER) && !strcmp((x)->text,(v)))
479 #define tok_isnt_(x,v) ((x) && ((x)->type!=TOK_OTHER || strcmp((x)->text,(v))))
482 * nasm_unquote with error if the string contains NUL characters.
483 * If the string contains NUL characters, issue an error and return
484 * the C len, i.e. truncate at the NUL.
486 static size_t nasm_unquote_cstr(char *qstr, enum preproc_token directive)
488 size_t len = nasm_unquote(qstr, NULL);
489 size_t clen = strlen(qstr);
491 if (len != clen)
492 error(ERR_NONFATAL, "NUL character in `%s' directive",
493 pp_directives[directive]);
495 return clen;
499 * In-place reverse a list of tokens.
501 static Token *reverse_tokens(Token *t)
503 Token *prev = NULL;
504 Token *next;
506 while (t) {
507 next = t->next;
508 t->next = prev;
509 prev = t;
510 t = next;
513 return prev;
517 * Handle TASM specific directives, which do not contain a % in
518 * front of them. We do it here because I could not find any other
519 * place to do it for the moment, and it is a hack (ideally it would
520 * be nice to be able to use the NASM pre-processor to do it).
522 static char *check_tasm_directive(char *line)
524 int32_t i, j, k, m, len;
525 char *p, *q, *oldline, oldchar;
527 p = nasm_skip_spaces(line);
529 /* Binary search for the directive name */
530 i = -1;
531 j = ARRAY_SIZE(tasm_directives);
532 q = nasm_skip_word(p);
533 len = q - p;
534 if (len) {
535 oldchar = p[len];
536 p[len] = 0;
537 while (j - i > 1) {
538 k = (j + i) / 2;
539 m = nasm_stricmp(p, tasm_directives[k]);
540 if (m == 0) {
541 /* We have found a directive, so jam a % in front of it
542 * so that NASM will then recognise it as one if it's own.
544 p[len] = oldchar;
545 len = strlen(p);
546 oldline = line;
547 line = nasm_malloc(len + 2);
548 line[0] = '%';
549 if (k == TM_IFDIFI) {
551 * NASM does not recognise IFDIFI, so we convert
552 * it to %if 0. This is not used in NASM
553 * compatible code, but does need to parse for the
554 * TASM macro package.
556 strcpy(line + 1, "if 0");
557 } else {
558 memcpy(line + 1, p, len + 1);
560 nasm_free(oldline);
561 return line;
562 } else if (m < 0) {
563 j = k;
564 } else
565 i = k;
567 p[len] = oldchar;
569 return line;
573 * The pre-preprocessing stage... This function translates line
574 * number indications as they emerge from GNU cpp (`# lineno "file"
575 * flags') into NASM preprocessor line number indications (`%line
576 * lineno file').
578 static char *prepreproc(char *line)
580 int lineno, fnlen;
581 char *fname, *oldline;
583 if (line[0] == '#' && line[1] == ' ') {
584 oldline = line;
585 fname = oldline + 2;
586 lineno = atoi(fname);
587 fname += strspn(fname, "0123456789 ");
588 if (*fname == '"')
589 fname++;
590 fnlen = strcspn(fname, "\"");
591 line = nasm_malloc(20 + fnlen);
592 snprintf(line, 20 + fnlen, "%%line %d %.*s", lineno, fnlen, fname);
593 nasm_free(oldline);
595 if (tasm_compatible_mode)
596 return check_tasm_directive(line);
597 return line;
601 * Free a linked list of tokens.
603 static void free_tlist(Token * list)
605 while (list)
606 list = delete_Token(list);
610 * Free a linked list of lines.
612 static void free_llist(Line * list)
614 Line *l, *tmp;
615 list_for_each_safe(l, tmp, list) {
616 free_tlist(l->first);
617 nasm_free(l);
622 * Free an MMacro
624 static void free_mmacro(MMacro * m)
626 nasm_free(m->name);
627 free_tlist(m->dlist);
628 nasm_free(m->defaults);
629 free_llist(m->expansion);
630 nasm_free(m);
634 * Free all currently defined macros, and free the hash tables
636 static void free_smacro_table(struct hash_table *smt)
638 SMacro *s, *tmp;
639 const char *key;
640 struct hash_tbl_node *it = NULL;
642 while ((s = hash_iterate(smt, &it, &key)) != NULL) {
643 nasm_free((void *)key);
644 list_for_each_safe(s, tmp, s) {
645 nasm_free(s->name);
646 free_tlist(s->expansion);
647 nasm_free(s);
650 hash_free(smt);
653 static void free_mmacro_table(struct hash_table *mmt)
655 MMacro *m, *tmp;
656 const char *key;
657 struct hash_tbl_node *it = NULL;
659 it = NULL;
660 while ((m = hash_iterate(mmt, &it, &key)) != NULL) {
661 nasm_free((void *)key);
662 list_for_each_safe(m ,tmp, m)
663 free_mmacro(m);
665 hash_free(mmt);
668 static void free_macros(void)
670 free_smacro_table(&smacros);
671 free_mmacro_table(&mmacros);
675 * Initialize the hash tables
677 static void init_macros(void)
679 hash_init(&smacros, HASH_LARGE);
680 hash_init(&mmacros, HASH_LARGE);
684 * Pop the context stack.
686 static void ctx_pop(void)
688 Context *c = cstk;
690 cstk = cstk->next;
691 free_smacro_table(&c->localmac);
692 nasm_free(c->name);
693 nasm_free(c);
697 * Search for a key in the hash index; adding it if necessary
698 * (in which case we initialize the data pointer to NULL.)
700 static void **
701 hash_findi_add(struct hash_table *hash, const char *str)
703 struct hash_insert hi;
704 void **r;
705 char *strx;
707 r = hash_findi(hash, str, &hi);
708 if (r)
709 return r;
711 strx = nasm_strdup(str); /* Use a more efficient allocator here? */
712 return hash_add(&hi, strx, NULL);
716 * Like hash_findi, but returns the data element rather than a pointer
717 * to it. Used only when not adding a new element, hence no third
718 * argument.
720 static void *
721 hash_findix(struct hash_table *hash, const char *str)
723 void **p;
725 p = hash_findi(hash, str, NULL);
726 return p ? *p : NULL;
730 * read line from standart macros set,
731 * if there no more left -- return NULL
733 static char *line_from_stdmac(void)
735 unsigned char c;
736 const unsigned char *p = stdmacpos;
737 char *line, *q;
738 size_t len = 0;
740 if (!stdmacpos)
741 return NULL;
743 while ((c = *p++)) {
744 if (c >= 0x80)
745 len += pp_directives_len[c - 0x80] + 1;
746 else
747 len++;
750 line = nasm_malloc(len + 1);
751 q = line;
752 while ((c = *stdmacpos++)) {
753 if (c >= 0x80) {
754 memcpy(q, pp_directives[c - 0x80], pp_directives_len[c - 0x80]);
755 q += pp_directives_len[c - 0x80];
756 *q++ = ' ';
757 } else {
758 *q++ = c;
761 stdmacpos = p;
762 *q = '\0';
764 if (!*stdmacpos) {
765 /* This was the last of the standard macro chain... */
766 stdmacpos = NULL;
767 if (any_extrastdmac) {
768 stdmacpos = extrastdmac;
769 any_extrastdmac = false;
770 } else if (do_predef) {
771 Line *pd, *l;
772 Token *head, **tail, *t;
775 * Nasty hack: here we push the contents of
776 * `predef' on to the top-level expansion stack,
777 * since this is the most convenient way to
778 * implement the pre-include and pre-define
779 * features.
781 list_for_each(pd, predef) {
782 head = NULL;
783 tail = &head;
784 list_for_each(t, pd->first) {
785 *tail = new_Token(NULL, t->type, t->text, 0);
786 tail = &(*tail)->next;
789 l = nasm_malloc(sizeof(Line));
790 l->next = istk->expansion;
791 l->first = head;
792 l->finishes = NULL;
794 istk->expansion = l;
796 do_predef = false;
800 return line;
803 static char *read_line(void)
805 unsigned int size, c, next;
806 const unsigned int delta = 512;
807 const unsigned int pad = 8;
808 unsigned int nr_cont = 0;
809 bool cont = false;
810 char *buffer, *p;
812 /* Standart macros set (predefined) goes first */
813 p = line_from_stdmac();
814 if (p)
815 return p;
817 size = delta;
818 p = buffer = nasm_malloc(size);
820 for (;;) {
821 c = fgetc(istk->fp);
822 if ((int)(c) == EOF) {
823 p[0] = 0;
824 break;
827 switch (c) {
828 case '\r':
829 next = fgetc(istk->fp);
830 if (next != '\n')
831 ungetc(next, istk->fp);
832 if (cont) {
833 cont = false;
834 continue;
836 break;
838 case '\n':
839 if (cont) {
840 cont = false;
841 continue;
843 break;
845 case '\\':
846 next = fgetc(istk->fp);
847 ungetc(next, istk->fp);
848 if (next == '\r' || next == '\n') {
849 cont = true;
850 nr_cont++;
851 continue;
853 break;
856 if (c == '\r' || c == '\n') {
857 *p++ = 0;
858 break;
861 if (p >= (buffer + size - pad)) {
862 buffer = nasm_realloc(buffer, size + delta);
863 p = buffer + size - pad;
864 size += delta;
867 *p++ = (unsigned char)c;
870 if (p == buffer) {
871 nasm_free(buffer);
872 return NULL;
875 src_set_linnum(src_get_linnum() + istk->lineinc +
876 (nr_cont * istk->lineinc));
879 * Handle spurious ^Z, which may be inserted into source files
880 * by some file transfer utilities.
882 buffer[strcspn(buffer, "\032")] = '\0';
884 list->line(LIST_READ, buffer);
886 return buffer;
890 * Tokenize a line of text. This is a very simple process since we
891 * don't need to parse the value out of e.g. numeric tokens: we
892 * simply split one string into many.
894 static Token *tokenize(char *line)
896 char c, *p = line;
897 enum pp_token_type type;
898 Token *list = NULL;
899 Token *t, **tail = &list;
901 while (*line) {
902 p = line;
903 if (*p == '%') {
904 p++;
905 if (*p == '+' && !nasm_isdigit(p[1])) {
906 p++;
907 type = TOK_PASTE;
908 } else if (nasm_isdigit(*p) ||
909 ((*p == '-' || *p == '+') && nasm_isdigit(p[1]))) {
910 do {
911 p++;
913 while (nasm_isdigit(*p));
914 type = TOK_PREPROC_ID;
915 } else if (*p == '{') {
916 p++;
917 while (*p) {
918 if (*p == '}')
919 break;
920 p[-1] = *p;
921 p++;
923 if (*p != '}')
924 error(ERR_WARNING | ERR_PASS1, "unterminated %{ construct");
925 p[-1] = '\0';
926 if (*p)
927 p++;
928 type = TOK_PREPROC_ID;
929 } else if (*p == '[') {
930 int lvl = 1;
931 line += 2; /* Skip the leading %[ */
932 p++;
933 while (lvl && (c = *p++)) {
934 switch (c) {
935 case ']':
936 lvl--;
937 break;
938 case '%':
939 if (*p == '[')
940 lvl++;
941 break;
942 case '\'':
943 case '\"':
944 case '`':
945 p = nasm_skip_string(p - 1) + 1;
946 break;
947 default:
948 break;
951 p--;
952 if (*p)
953 *p++ = '\0';
954 if (lvl)
955 error(ERR_NONFATAL, "unterminated %[ construct");
956 type = TOK_INDIRECT;
957 } else if (*p == '?') {
958 type = TOK_PREPROC_Q; /* %? */
959 p++;
960 if (*p == '?') {
961 type = TOK_PREPROC_QQ; /* %?? */
962 p++;
964 } else if (*p == '!') {
965 type = TOK_PREPROC_ID;
966 p++;
967 if (isidchar(*p)) {
968 do {
969 p++;
971 while (isidchar(*p));
972 } else if (*p == '\'' || *p == '\"' || *p == '`') {
973 p = nasm_skip_string(p);
974 if (*p)
975 p++;
976 else
977 error(ERR_NONFATAL|ERR_PASS1, "unterminated %! string");
978 } else {
979 /* %! without string or identifier */
980 type = TOK_OTHER; /* Legacy behavior... */
982 } else if (isidchar(*p) ||
983 ((*p == '!' || *p == '%' || *p == '$') &&
984 isidchar(p[1]))) {
985 do {
986 p++;
988 while (isidchar(*p));
989 type = TOK_PREPROC_ID;
990 } else {
991 type = TOK_OTHER;
992 if (*p == '%')
993 p++;
995 } else if (isidstart(*p) || (*p == '$' && isidstart(p[1]))) {
996 type = TOK_ID;
997 p++;
998 while (*p && isidchar(*p))
999 p++;
1000 } else if (*p == '\'' || *p == '"' || *p == '`') {
1002 * A string token.
1004 type = TOK_STRING;
1005 p = nasm_skip_string(p);
1007 if (*p) {
1008 p++;
1009 } else {
1010 error(ERR_WARNING|ERR_PASS1, "unterminated string");
1011 /* Handling unterminated strings by UNV */
1012 /* type = -1; */
1014 } else if (p[0] == '$' && p[1] == '$') {
1015 type = TOK_OTHER; /* TOKEN_BASE */
1016 p += 2;
1017 } else if (isnumstart(*p)) {
1018 bool is_hex = false;
1019 bool is_float = false;
1020 bool has_e = false;
1021 char c, *r;
1024 * A numeric token.
1027 if (*p == '$') {
1028 p++;
1029 is_hex = true;
1032 for (;;) {
1033 c = *p++;
1035 if (!is_hex && (c == 'e' || c == 'E')) {
1036 has_e = true;
1037 if (*p == '+' || *p == '-') {
1039 * e can only be followed by +/- if it is either a
1040 * prefixed hex number or a floating-point number
1042 p++;
1043 is_float = true;
1045 } else if (c == 'H' || c == 'h' || c == 'X' || c == 'x') {
1046 is_hex = true;
1047 } else if (c == 'P' || c == 'p') {
1048 is_float = true;
1049 if (*p == '+' || *p == '-')
1050 p++;
1051 } else if (isnumchar(c) || c == '_')
1052 ; /* just advance */
1053 else if (c == '.') {
1055 * we need to deal with consequences of the legacy
1056 * parser, like "1.nolist" being two tokens
1057 * (TOK_NUMBER, TOK_ID) here; at least give it
1058 * a shot for now. In the future, we probably need
1059 * a flex-based scanner with proper pattern matching
1060 * to do it as well as it can be done. Nothing in
1061 * the world is going to help the person who wants
1062 * 0x123.p16 interpreted as two tokens, though.
1064 r = p;
1065 while (*r == '_')
1066 r++;
1068 if (nasm_isdigit(*r) || (is_hex && nasm_isxdigit(*r)) ||
1069 (!is_hex && (*r == 'e' || *r == 'E')) ||
1070 (*r == 'p' || *r == 'P')) {
1071 p = r;
1072 is_float = true;
1073 } else
1074 break; /* Terminate the token */
1075 } else
1076 break;
1078 p--; /* Point to first character beyond number */
1080 if (p == line+1 && *line == '$') {
1081 type = TOK_OTHER; /* TOKEN_HERE */
1082 } else {
1083 if (has_e && !is_hex) {
1084 /* 1e13 is floating-point, but 1e13h is not */
1085 is_float = true;
1088 type = is_float ? TOK_FLOAT : TOK_NUMBER;
1090 } else if (nasm_isspace(*p)) {
1091 type = TOK_WHITESPACE;
1092 p = nasm_skip_spaces(p);
1094 * Whitespace just before end-of-line is discarded by
1095 * pretending it's a comment; whitespace just before a
1096 * comment gets lumped into the comment.
1098 if (!*p || *p == ';') {
1099 type = TOK_COMMENT;
1100 while (*p)
1101 p++;
1103 } else if (*p == ';') {
1104 type = TOK_COMMENT;
1105 while (*p)
1106 p++;
1107 } else if (p[0] == '\\' && (p[1] == '{' || p[1] == '}')) {
1108 type = TOK_BRACE;
1109 p += 2;
1110 line++;
1111 } else {
1113 * Anything else is an operator of some kind. We check
1114 * for all the double-character operators (>>, <<, //,
1115 * %%, <=, >=, ==, !=, <>, &&, ||, ^^), but anything
1116 * else is a single-character operator.
1118 type = TOK_OTHER;
1119 if ((p[0] == '>' && p[1] == '>') ||
1120 (p[0] == '<' && p[1] == '<') ||
1121 (p[0] == '/' && p[1] == '/') ||
1122 (p[0] == '<' && p[1] == '=') ||
1123 (p[0] == '>' && p[1] == '=') ||
1124 (p[0] == '=' && p[1] == '=') ||
1125 (p[0] == '!' && p[1] == '=') ||
1126 (p[0] == '<' && p[1] == '>') ||
1127 (p[0] == '&' && p[1] == '&') ||
1128 (p[0] == '|' && p[1] == '|') ||
1129 (p[0] == '^' && p[1] == '^')) {
1130 p++;
1132 p++;
1135 /* Handling unterminated string by UNV */
1136 /*if (type == -1)
1138 *tail = t = new_Token(NULL, TOK_STRING, line, p-line+1);
1139 t->text[p-line] = *line;
1140 tail = &t->next;
1142 else */
1143 if (type != TOK_COMMENT) {
1144 *tail = t = new_Token(NULL, type, line, p - line);
1145 tail = &t->next;
1147 line = p;
1149 return list;
1153 * this function allocates a new managed block of memory and
1154 * returns a pointer to the block. The managed blocks are
1155 * deleted only all at once by the delete_Blocks function.
1157 static void *new_Block(size_t size)
1159 Blocks *b = &blocks;
1161 /* first, get to the end of the linked list */
1162 while (b->next)
1163 b = b->next;
1164 /* now allocate the requested chunk */
1165 b->chunk = nasm_malloc(size);
1167 /* now allocate a new block for the next request */
1168 b->next = nasm_malloc(sizeof(Blocks));
1169 /* and initialize the contents of the new block */
1170 b->next->next = NULL;
1171 b->next->chunk = NULL;
1172 return b->chunk;
1176 * this function deletes all managed blocks of memory
1178 static void delete_Blocks(void)
1180 Blocks *a, *b = &blocks;
1183 * keep in mind that the first block, pointed to by blocks
1184 * is a static and not dynamically allocated, so we don't
1185 * free it.
1187 while (b) {
1188 if (b->chunk)
1189 nasm_free(b->chunk);
1190 a = b;
1191 b = b->next;
1192 if (a != &blocks)
1193 nasm_free(a);
1198 * this function creates a new Token and passes a pointer to it
1199 * back to the caller. It sets the type and text elements, and
1200 * also the a.mac and next elements to NULL.
1202 static Token *new_Token(Token * next, enum pp_token_type type,
1203 const char *text, int txtlen)
1205 Token *t;
1206 int i;
1208 if (!freeTokens) {
1209 freeTokens = (Token *) new_Block(TOKEN_BLOCKSIZE * sizeof(Token));
1210 for (i = 0; i < TOKEN_BLOCKSIZE - 1; i++)
1211 freeTokens[i].next = &freeTokens[i + 1];
1212 freeTokens[i].next = NULL;
1214 t = freeTokens;
1215 freeTokens = t->next;
1216 t->next = next;
1217 t->a.mac = NULL;
1218 t->type = type;
1219 if (type == TOK_WHITESPACE || !text) {
1220 t->text = NULL;
1221 } else {
1222 if (txtlen == 0)
1223 txtlen = strlen(text);
1224 t->text = nasm_malloc(txtlen+1);
1225 memcpy(t->text, text, txtlen);
1226 t->text[txtlen] = '\0';
1228 return t;
1231 static Token *delete_Token(Token * t)
1233 Token *next = t->next;
1234 nasm_free(t->text);
1235 t->next = freeTokens;
1236 freeTokens = t;
1237 return next;
1241 * Convert a line of tokens back into text.
1242 * If expand_locals is not zero, identifiers of the form "%$*xxx"
1243 * will be transformed into ..@ctxnum.xxx
1245 static char *detoken(Token * tlist, bool expand_locals)
1247 Token *t;
1248 char *line, *p;
1249 const char *q;
1250 int len = 0;
1252 list_for_each(t, tlist) {
1253 if (t->type == TOK_PREPROC_ID && t->text[1] == '!') {
1254 char *v;
1255 char *q = t->text;
1257 v = t->text + 2;
1258 if (*v == '\'' || *v == '\"' || *v == '`') {
1259 size_t len = nasm_unquote(v, NULL);
1260 size_t clen = strlen(v);
1262 if (len != clen) {
1263 error(ERR_NONFATAL | ERR_PASS1,
1264 "NUL character in %! string");
1265 v = NULL;
1269 if (v) {
1270 char *p = getenv(v);
1271 if (!p) {
1272 error(ERR_NONFATAL | ERR_PASS1,
1273 "nonexistent environment variable `%s'", v);
1274 p = "";
1276 t->text = nasm_strdup(p);
1278 nasm_free(q);
1281 /* Expand local macros here and not during preprocessing */
1282 if (expand_locals &&
1283 t->type == TOK_PREPROC_ID && t->text &&
1284 t->text[0] == '%' && t->text[1] == '$') {
1285 const char *q;
1286 char *p;
1287 Context *ctx = get_ctx(t->text, &q);
1288 if (ctx) {
1289 char buffer[40];
1290 snprintf(buffer, sizeof(buffer), "..@%"PRIu32".", ctx->number);
1291 p = nasm_strcat(buffer, q);
1292 nasm_free(t->text);
1293 t->text = p;
1296 if (t->type == TOK_WHITESPACE)
1297 len++;
1298 else if (t->text)
1299 len += strlen(t->text);
1302 p = line = nasm_malloc(len + 1);
1304 list_for_each(t, tlist) {
1305 if (t->type == TOK_WHITESPACE) {
1306 *p++ = ' ';
1307 } else if (t->text) {
1308 q = t->text;
1309 while (*q)
1310 *p++ = *q++;
1313 *p = '\0';
1315 return line;
1319 * A scanner, suitable for use by the expression evaluator, which
1320 * operates on a line of Tokens. Expects a pointer to a pointer to
1321 * the first token in the line to be passed in as its private_data
1322 * field.
1324 * FIX: This really needs to be unified with stdscan.
1326 static int ppscan(void *private_data, struct tokenval *tokval)
1328 Token **tlineptr = private_data;
1329 Token *tline;
1330 char ourcopy[MAX_KEYWORD+1], *p, *r, *s;
1332 do {
1333 tline = *tlineptr;
1334 *tlineptr = tline ? tline->next : NULL;
1335 } while (tline && (tline->type == TOK_WHITESPACE ||
1336 tline->type == TOK_COMMENT));
1338 if (!tline)
1339 return tokval->t_type = TOKEN_EOS;
1341 tokval->t_charptr = tline->text;
1343 if (tline->text[0] == '$' && !tline->text[1])
1344 return tokval->t_type = TOKEN_HERE;
1345 if (tline->text[0] == '$' && tline->text[1] == '$' && !tline->text[2])
1346 return tokval->t_type = TOKEN_BASE;
1348 if (tline->type == TOK_ID) {
1349 p = tokval->t_charptr = tline->text;
1350 if (p[0] == '$') {
1351 tokval->t_charptr++;
1352 return tokval->t_type = TOKEN_ID;
1355 for (r = p, s = ourcopy; *r; r++) {
1356 if (r >= p+MAX_KEYWORD)
1357 return tokval->t_type = TOKEN_ID; /* Not a keyword */
1358 *s++ = nasm_tolower(*r);
1360 *s = '\0';
1361 /* right, so we have an identifier sitting in temp storage. now,
1362 * is it actually a register or instruction name, or what? */
1363 return nasm_token_hash(ourcopy, tokval);
1366 if (tline->type == TOK_NUMBER) {
1367 bool rn_error;
1368 tokval->t_integer = readnum(tline->text, &rn_error);
1369 tokval->t_charptr = tline->text;
1370 if (rn_error)
1371 return tokval->t_type = TOKEN_ERRNUM;
1372 else
1373 return tokval->t_type = TOKEN_NUM;
1376 if (tline->type == TOK_FLOAT) {
1377 return tokval->t_type = TOKEN_FLOAT;
1380 if (tline->type == TOK_STRING) {
1381 char bq, *ep;
1383 bq = tline->text[0];
1384 tokval->t_charptr = tline->text;
1385 tokval->t_inttwo = nasm_unquote(tline->text, &ep);
1387 if (ep[0] != bq || ep[1] != '\0')
1388 return tokval->t_type = TOKEN_ERRSTR;
1389 else
1390 return tokval->t_type = TOKEN_STR;
1393 if (tline->type == TOK_OTHER) {
1394 if (!strcmp(tline->text, "<<"))
1395 return tokval->t_type = TOKEN_SHL;
1396 if (!strcmp(tline->text, ">>"))
1397 return tokval->t_type = TOKEN_SHR;
1398 if (!strcmp(tline->text, "//"))
1399 return tokval->t_type = TOKEN_SDIV;
1400 if (!strcmp(tline->text, "%%"))
1401 return tokval->t_type = TOKEN_SMOD;
1402 if (!strcmp(tline->text, "=="))
1403 return tokval->t_type = TOKEN_EQ;
1404 if (!strcmp(tline->text, "<>"))
1405 return tokval->t_type = TOKEN_NE;
1406 if (!strcmp(tline->text, "!="))
1407 return tokval->t_type = TOKEN_NE;
1408 if (!strcmp(tline->text, "<="))
1409 return tokval->t_type = TOKEN_LE;
1410 if (!strcmp(tline->text, ">="))
1411 return tokval->t_type = TOKEN_GE;
1412 if (!strcmp(tline->text, "&&"))
1413 return tokval->t_type = TOKEN_DBL_AND;
1414 if (!strcmp(tline->text, "^^"))
1415 return tokval->t_type = TOKEN_DBL_XOR;
1416 if (!strcmp(tline->text, "||"))
1417 return tokval->t_type = TOKEN_DBL_OR;
1421 * We have no other options: just return the first character of
1422 * the token text.
1424 return tokval->t_type = tline->text[0];
1428 * Compare a string to the name of an existing macro; this is a
1429 * simple wrapper which calls either strcmp or nasm_stricmp
1430 * depending on the value of the `casesense' parameter.
1432 static int mstrcmp(const char *p, const char *q, bool casesense)
1434 return casesense ? strcmp(p, q) : nasm_stricmp(p, q);
1438 * Compare a string to the name of an existing macro; this is a
1439 * simple wrapper which calls either strcmp or nasm_stricmp
1440 * depending on the value of the `casesense' parameter.
1442 static int mmemcmp(const char *p, const char *q, size_t l, bool casesense)
1444 return casesense ? memcmp(p, q, l) : nasm_memicmp(p, q, l);
1448 * Return the Context structure associated with a %$ token. Return
1449 * NULL, having _already_ reported an error condition, if the
1450 * context stack isn't deep enough for the supplied number of $
1451 * signs.
1453 * If "namep" is non-NULL, set it to the pointer to the macro name
1454 * tail, i.e. the part beyond %$...
1456 static Context *get_ctx(const char *name, const char **namep)
1458 Context *ctx;
1459 int i;
1461 if (namep)
1462 *namep = name;
1464 if (!name || name[0] != '%' || name[1] != '$')
1465 return NULL;
1467 if (!cstk) {
1468 error(ERR_NONFATAL, "`%s': context stack is empty", name);
1469 return NULL;
1472 name += 2;
1473 ctx = cstk;
1474 i = 0;
1475 while (ctx && *name == '$') {
1476 name++;
1477 i++;
1478 ctx = ctx->next;
1480 if (!ctx) {
1481 error(ERR_NONFATAL, "`%s': context stack is only"
1482 " %d level%s deep", name, i, (i == 1 ? "" : "s"));
1483 return NULL;
1486 if (namep)
1487 *namep = name;
1489 return ctx;
1493 * Check to see if a file is already in a string list
1495 static bool in_list(const StrList *list, const char *str)
1497 while (list) {
1498 if (!strcmp(list->str, str))
1499 return true;
1500 list = list->next;
1502 return false;
1506 * Open an include file. This routine must always return a valid
1507 * file pointer if it returns - it's responsible for throwing an
1508 * ERR_FATAL and bombing out completely if not. It should also try
1509 * the include path one by one until it finds the file or reaches
1510 * the end of the path.
1512 static FILE *inc_fopen(const char *file, StrList **dhead, StrList ***dtail,
1513 bool missing_ok)
1515 FILE *fp;
1516 char *prefix = "";
1517 IncPath *ip = ipath;
1518 int len = strlen(file);
1519 size_t prefix_len = 0;
1520 StrList *sl;
1522 while (1) {
1523 sl = nasm_malloc(prefix_len+len+1+sizeof sl->next);
1524 memcpy(sl->str, prefix, prefix_len);
1525 memcpy(sl->str+prefix_len, file, len+1);
1526 fp = fopen(sl->str, "r");
1527 if (fp && dhead && !in_list(*dhead, sl->str)) {
1528 sl->next = NULL;
1529 **dtail = sl;
1530 *dtail = &sl->next;
1531 } else {
1532 nasm_free(sl);
1534 if (fp)
1535 return fp;
1536 if (!ip) {
1537 if (!missing_ok)
1538 break;
1539 prefix = NULL;
1540 } else {
1541 prefix = ip->path;
1542 ip = ip->next;
1544 if (prefix) {
1545 prefix_len = strlen(prefix);
1546 } else {
1547 /* -MG given and file not found */
1548 if (dhead && !in_list(*dhead, file)) {
1549 sl = nasm_malloc(len+1+sizeof sl->next);
1550 sl->next = NULL;
1551 strcpy(sl->str, file);
1552 **dtail = sl;
1553 *dtail = &sl->next;
1555 return NULL;
1559 error(ERR_FATAL, "unable to open include file `%s'", file);
1560 return NULL;
1564 * Determine if we should warn on defining a single-line macro of
1565 * name `name', with `nparam' parameters. If nparam is 0 or -1, will
1566 * return true if _any_ single-line macro of that name is defined.
1567 * Otherwise, will return true if a single-line macro with either
1568 * `nparam' or no parameters is defined.
1570 * If a macro with precisely the right number of parameters is
1571 * defined, or nparam is -1, the address of the definition structure
1572 * will be returned in `defn'; otherwise NULL will be returned. If `defn'
1573 * is NULL, no action will be taken regarding its contents, and no
1574 * error will occur.
1576 * Note that this is also called with nparam zero to resolve
1577 * `ifdef'.
1579 * If you already know which context macro belongs to, you can pass
1580 * the context pointer as first parameter; if you won't but name begins
1581 * with %$ the context will be automatically computed. If all_contexts
1582 * is true, macro will be searched in outer contexts as well.
1584 static bool
1585 smacro_defined(Context * ctx, const char *name, int nparam, SMacro ** defn,
1586 bool nocase)
1588 struct hash_table *smtbl;
1589 SMacro *m;
1591 if (ctx) {
1592 smtbl = &ctx->localmac;
1593 } else if (name[0] == '%' && name[1] == '$') {
1594 if (cstk)
1595 ctx = get_ctx(name, &name);
1596 if (!ctx)
1597 return false; /* got to return _something_ */
1598 smtbl = &ctx->localmac;
1599 } else {
1600 smtbl = &smacros;
1602 m = (SMacro *) hash_findix(smtbl, name);
1604 while (m) {
1605 if (!mstrcmp(m->name, name, m->casesense && nocase) &&
1606 (nparam <= 0 || m->nparam == 0 || nparam == (int) m->nparam)) {
1607 if (defn) {
1608 if (nparam == (int) m->nparam || nparam == -1)
1609 *defn = m;
1610 else
1611 *defn = NULL;
1613 return true;
1615 m = m->next;
1618 return false;
1622 * Count and mark off the parameters in a multi-line macro call.
1623 * This is called both from within the multi-line macro expansion
1624 * code, and also to mark off the default parameters when provided
1625 * in a %macro definition line.
1627 static void count_mmac_params(Token * t, int *nparam, Token *** params)
1629 int paramsize, brace;
1631 *nparam = paramsize = 0;
1632 *params = NULL;
1633 while (t) {
1634 /* +1: we need space for the final NULL */
1635 if (*nparam+1 >= paramsize) {
1636 paramsize += PARAM_DELTA;
1637 *params = nasm_realloc(*params, sizeof(**params) * paramsize);
1639 skip_white_(t);
1640 brace = false;
1641 if (tok_is_(t, "{"))
1642 brace = true;
1643 (*params)[(*nparam)++] = t;
1644 while (tok_isnt_(t, brace ? "}" : ","))
1645 t = t->next;
1646 if (t) { /* got a comma/brace */
1647 t = t->next;
1648 if (brace) {
1650 * Now we've found the closing brace, look further
1651 * for the comma.
1653 skip_white_(t);
1654 if (tok_isnt_(t, ",")) {
1655 error(ERR_NONFATAL,
1656 "braces do not enclose all of macro parameter");
1657 while (tok_isnt_(t, ","))
1658 t = t->next;
1660 if (t)
1661 t = t->next; /* eat the comma */
1668 * Determine whether one of the various `if' conditions is true or
1669 * not.
1671 * We must free the tline we get passed.
1673 static bool if_condition(Token * tline, enum preproc_token ct)
1675 enum pp_conditional i = PP_COND(ct);
1676 bool j;
1677 Token *t, *tt, **tptr, *origline;
1678 struct tokenval tokval;
1679 expr *evalresult;
1680 enum pp_token_type needtype;
1681 char *p;
1683 origline = tline;
1685 switch (i) {
1686 case PPC_IFCTX:
1687 j = false; /* have we matched yet? */
1688 while (true) {
1689 skip_white_(tline);
1690 if (!tline)
1691 break;
1692 if (tline->type != TOK_ID) {
1693 error(ERR_NONFATAL,
1694 "`%s' expects context identifiers", pp_directives[ct]);
1695 free_tlist(origline);
1696 return -1;
1698 if (cstk && cstk->name && !nasm_stricmp(tline->text, cstk->name))
1699 j = true;
1700 tline = tline->next;
1702 break;
1704 case PPC_IFDEF:
1705 j = false; /* have we matched yet? */
1706 while (tline) {
1707 skip_white_(tline);
1708 if (!tline || (tline->type != TOK_ID &&
1709 (tline->type != TOK_PREPROC_ID ||
1710 tline->text[1] != '$'))) {
1711 error(ERR_NONFATAL,
1712 "`%s' expects macro identifiers", pp_directives[ct]);
1713 goto fail;
1715 if (smacro_defined(NULL, tline->text, 0, NULL, true))
1716 j = true;
1717 tline = tline->next;
1719 break;
1721 case PPC_IFENV:
1722 tline = expand_smacro(tline);
1723 j = false; /* have we matched yet? */
1724 while (tline) {
1725 skip_white_(tline);
1726 if (!tline || (tline->type != TOK_ID &&
1727 tline->type != TOK_STRING &&
1728 (tline->type != TOK_PREPROC_ID ||
1729 tline->text[1] != '!'))) {
1730 error(ERR_NONFATAL,
1731 "`%s' expects environment variable names",
1732 pp_directives[ct]);
1733 goto fail;
1735 p = tline->text;
1736 if (tline->type == TOK_PREPROC_ID)
1737 p += 2; /* Skip leading %! */
1738 if (*p == '\'' || *p == '\"' || *p == '`')
1739 nasm_unquote_cstr(p, ct);
1740 if (getenv(p))
1741 j = true;
1742 tline = tline->next;
1744 break;
1746 case PPC_IFIDN:
1747 case PPC_IFIDNI:
1748 tline = expand_smacro(tline);
1749 t = tt = tline;
1750 while (tok_isnt_(tt, ","))
1751 tt = tt->next;
1752 if (!tt) {
1753 error(ERR_NONFATAL,
1754 "`%s' expects two comma-separated arguments",
1755 pp_directives[ct]);
1756 goto fail;
1758 tt = tt->next;
1759 j = true; /* assume equality unless proved not */
1760 while ((t->type != TOK_OTHER || strcmp(t->text, ",")) && tt) {
1761 if (tt->type == TOK_OTHER && !strcmp(tt->text, ",")) {
1762 error(ERR_NONFATAL, "`%s': more than one comma on line",
1763 pp_directives[ct]);
1764 goto fail;
1766 if (t->type == TOK_WHITESPACE) {
1767 t = t->next;
1768 continue;
1770 if (tt->type == TOK_WHITESPACE) {
1771 tt = tt->next;
1772 continue;
1774 if (tt->type != t->type) {
1775 j = false; /* found mismatching tokens */
1776 break;
1778 /* When comparing strings, need to unquote them first */
1779 if (t->type == TOK_STRING) {
1780 size_t l1 = nasm_unquote(t->text, NULL);
1781 size_t l2 = nasm_unquote(tt->text, NULL);
1783 if (l1 != l2) {
1784 j = false;
1785 break;
1787 if (mmemcmp(t->text, tt->text, l1, i == PPC_IFIDN)) {
1788 j = false;
1789 break;
1791 } else if (mstrcmp(tt->text, t->text, i == PPC_IFIDN) != 0) {
1792 j = false; /* found mismatching tokens */
1793 break;
1796 t = t->next;
1797 tt = tt->next;
1799 if ((t->type != TOK_OTHER || strcmp(t->text, ",")) || tt)
1800 j = false; /* trailing gunk on one end or other */
1801 break;
1803 case PPC_IFMACRO:
1805 bool found = false;
1806 MMacro searching, *mmac;
1808 skip_white_(tline);
1809 tline = expand_id(tline);
1810 if (!tok_type_(tline, TOK_ID)) {
1811 error(ERR_NONFATAL,
1812 "`%s' expects a macro name", pp_directives[ct]);
1813 goto fail;
1815 searching.name = nasm_strdup(tline->text);
1816 searching.casesense = true;
1817 searching.plus = false;
1818 searching.nolist = false;
1819 searching.in_progress = 0;
1820 searching.max_depth = 0;
1821 searching.rep_nest = NULL;
1822 searching.nparam_min = 0;
1823 searching.nparam_max = INT_MAX;
1824 tline = expand_smacro(tline->next);
1825 skip_white_(tline);
1826 if (!tline) {
1827 } else if (!tok_type_(tline, TOK_NUMBER)) {
1828 error(ERR_NONFATAL,
1829 "`%s' expects a parameter count or nothing",
1830 pp_directives[ct]);
1831 } else {
1832 searching.nparam_min = searching.nparam_max =
1833 readnum(tline->text, &j);
1834 if (j)
1835 error(ERR_NONFATAL,
1836 "unable to parse parameter count `%s'",
1837 tline->text);
1839 if (tline && tok_is_(tline->next, "-")) {
1840 tline = tline->next->next;
1841 if (tok_is_(tline, "*"))
1842 searching.nparam_max = INT_MAX;
1843 else if (!tok_type_(tline, TOK_NUMBER))
1844 error(ERR_NONFATAL,
1845 "`%s' expects a parameter count after `-'",
1846 pp_directives[ct]);
1847 else {
1848 searching.nparam_max = readnum(tline->text, &j);
1849 if (j)
1850 error(ERR_NONFATAL,
1851 "unable to parse parameter count `%s'",
1852 tline->text);
1853 if (searching.nparam_min > searching.nparam_max)
1854 error(ERR_NONFATAL,
1855 "minimum parameter count exceeds maximum");
1858 if (tline && tok_is_(tline->next, "+")) {
1859 tline = tline->next;
1860 searching.plus = true;
1862 mmac = (MMacro *) hash_findix(&mmacros, searching.name);
1863 while (mmac) {
1864 if (!strcmp(mmac->name, searching.name) &&
1865 (mmac->nparam_min <= searching.nparam_max
1866 || searching.plus)
1867 && (searching.nparam_min <= mmac->nparam_max
1868 || mmac->plus)) {
1869 found = true;
1870 break;
1872 mmac = mmac->next;
1874 if (tline && tline->next)
1875 error(ERR_WARNING|ERR_PASS1,
1876 "trailing garbage after %%ifmacro ignored");
1877 nasm_free(searching.name);
1878 j = found;
1879 break;
1882 case PPC_IFID:
1883 needtype = TOK_ID;
1884 goto iftype;
1885 case PPC_IFNUM:
1886 needtype = TOK_NUMBER;
1887 goto iftype;
1888 case PPC_IFSTR:
1889 needtype = TOK_STRING;
1890 goto iftype;
1892 iftype:
1893 t = tline = expand_smacro(tline);
1895 while (tok_type_(t, TOK_WHITESPACE) ||
1896 (needtype == TOK_NUMBER &&
1897 tok_type_(t, TOK_OTHER) &&
1898 (t->text[0] == '-' || t->text[0] == '+') &&
1899 !t->text[1]))
1900 t = t->next;
1902 j = tok_type_(t, needtype);
1903 break;
1905 case PPC_IFTOKEN:
1906 t = tline = expand_smacro(tline);
1907 while (tok_type_(t, TOK_WHITESPACE))
1908 t = t->next;
1910 j = false;
1911 if (t) {
1912 t = t->next; /* Skip the actual token */
1913 while (tok_type_(t, TOK_WHITESPACE))
1914 t = t->next;
1915 j = !t; /* Should be nothing left */
1917 break;
1919 case PPC_IFEMPTY:
1920 t = tline = expand_smacro(tline);
1921 while (tok_type_(t, TOK_WHITESPACE))
1922 t = t->next;
1924 j = !t; /* Should be empty */
1925 break;
1927 case PPC_IF:
1928 t = tline = expand_smacro(tline);
1929 tptr = &t;
1930 tokval.t_type = TOKEN_INVALID;
1931 evalresult = evaluate(ppscan, tptr, &tokval,
1932 NULL, pass | CRITICAL, error, NULL);
1933 if (!evalresult)
1934 return -1;
1935 if (tokval.t_type)
1936 error(ERR_WARNING|ERR_PASS1,
1937 "trailing garbage after expression ignored");
1938 if (!is_simple(evalresult)) {
1939 error(ERR_NONFATAL,
1940 "non-constant value given to `%s'", pp_directives[ct]);
1941 goto fail;
1943 j = reloc_value(evalresult) != 0;
1944 break;
1946 default:
1947 error(ERR_FATAL,
1948 "preprocessor directive `%s' not yet implemented",
1949 pp_directives[ct]);
1950 goto fail;
1953 free_tlist(origline);
1954 return j ^ PP_NEGATIVE(ct);
1956 fail:
1957 free_tlist(origline);
1958 return -1;
1962 * Common code for defining an smacro
1964 static bool define_smacro(Context *ctx, const char *mname, bool casesense,
1965 int nparam, Token *expansion)
1967 SMacro *smac, **smhead;
1968 struct hash_table *smtbl;
1970 if (smacro_defined(ctx, mname, nparam, &smac, casesense)) {
1971 if (!smac) {
1972 error(ERR_WARNING|ERR_PASS1,
1973 "single-line macro `%s' defined both with and"
1974 " without parameters", mname);
1976 * Some instances of the old code considered this a failure,
1977 * some others didn't. What is the right thing to do here?
1979 free_tlist(expansion);
1980 return false; /* Failure */
1981 } else {
1983 * We're redefining, so we have to take over an
1984 * existing SMacro structure. This means freeing
1985 * what was already in it.
1987 nasm_free(smac->name);
1988 free_tlist(smac->expansion);
1990 } else {
1991 smtbl = ctx ? &ctx->localmac : &smacros;
1992 smhead = (SMacro **) hash_findi_add(smtbl, mname);
1993 smac = nasm_malloc(sizeof(SMacro));
1994 smac->next = *smhead;
1995 *smhead = smac;
1997 smac->name = nasm_strdup(mname);
1998 smac->casesense = casesense;
1999 smac->nparam = nparam;
2000 smac->expansion = expansion;
2001 smac->in_progress = false;
2002 return true; /* Success */
2006 * Undefine an smacro
2008 static void undef_smacro(Context *ctx, const char *mname)
2010 SMacro **smhead, *s, **sp;
2011 struct hash_table *smtbl;
2013 smtbl = ctx ? &ctx->localmac : &smacros;
2014 smhead = (SMacro **)hash_findi(smtbl, mname, NULL);
2016 if (smhead) {
2018 * We now have a macro name... go hunt for it.
2020 sp = smhead;
2021 while ((s = *sp) != NULL) {
2022 if (!mstrcmp(s->name, mname, s->casesense)) {
2023 *sp = s->next;
2024 nasm_free(s->name);
2025 free_tlist(s->expansion);
2026 nasm_free(s);
2027 } else {
2028 sp = &s->next;
2035 * Parse a mmacro specification.
2037 static bool parse_mmacro_spec(Token *tline, MMacro *def, const char *directive)
2039 bool err;
2041 tline = tline->next;
2042 skip_white_(tline);
2043 tline = expand_id(tline);
2044 if (!tok_type_(tline, TOK_ID)) {
2045 error(ERR_NONFATAL, "`%s' expects a macro name", directive);
2046 return false;
2049 def->prev = NULL;
2050 def->name = nasm_strdup(tline->text);
2051 def->plus = false;
2052 def->nolist = false;
2053 def->in_progress = 0;
2054 def->rep_nest = NULL;
2055 def->nparam_min = 0;
2056 def->nparam_max = 0;
2058 tline = expand_smacro(tline->next);
2059 skip_white_(tline);
2060 if (!tok_type_(tline, TOK_NUMBER)) {
2061 error(ERR_NONFATAL, "`%s' expects a parameter count", directive);
2062 } else {
2063 def->nparam_min = def->nparam_max =
2064 readnum(tline->text, &err);
2065 if (err)
2066 error(ERR_NONFATAL,
2067 "unable to parse parameter count `%s'", tline->text);
2069 if (tline && tok_is_(tline->next, "-")) {
2070 tline = tline->next->next;
2071 if (tok_is_(tline, "*")) {
2072 def->nparam_max = INT_MAX;
2073 } else if (!tok_type_(tline, TOK_NUMBER)) {
2074 error(ERR_NONFATAL,
2075 "`%s' expects a parameter count after `-'", directive);
2076 } else {
2077 def->nparam_max = readnum(tline->text, &err);
2078 if (err) {
2079 error(ERR_NONFATAL, "unable to parse parameter count `%s'",
2080 tline->text);
2082 if (def->nparam_min > def->nparam_max) {
2083 error(ERR_NONFATAL, "minimum parameter count exceeds maximum");
2087 if (tline && tok_is_(tline->next, "+")) {
2088 tline = tline->next;
2089 def->plus = true;
2091 if (tline && tok_type_(tline->next, TOK_ID) &&
2092 !nasm_stricmp(tline->next->text, ".nolist")) {
2093 tline = tline->next;
2094 def->nolist = true;
2098 * Handle default parameters.
2100 if (tline && tline->next) {
2101 def->dlist = tline->next;
2102 tline->next = NULL;
2103 count_mmac_params(def->dlist, &def->ndefs, &def->defaults);
2104 } else {
2105 def->dlist = NULL;
2106 def->defaults = NULL;
2108 def->expansion = NULL;
2110 if (def->defaults && def->ndefs > def->nparam_max - def->nparam_min &&
2111 !def->plus)
2112 error(ERR_WARNING|ERR_PASS1|ERR_WARN_MDP,
2113 "too many default macro parameters");
2115 return true;
2120 * Decode a size directive
2122 static int parse_size(const char *str) {
2123 static const char *size_names[] =
2124 { "byte", "dword", "oword", "qword", "tword", "word", "yword" };
2125 static const int sizes[] =
2126 { 0, 1, 4, 16, 8, 10, 2, 32 };
2128 return sizes[bsii(str, size_names, ARRAY_SIZE(size_names))+1];
2132 * find and process preprocessor directive in passed line
2133 * Find out if a line contains a preprocessor directive, and deal
2134 * with it if so.
2136 * If a directive _is_ found, it is the responsibility of this routine
2137 * (and not the caller) to free_tlist() the line.
2139 * @param tline a pointer to the current tokeninzed line linked list
2140 * @return DIRECTIVE_FOUND or NO_DIRECTIVE_FOUND
2143 static int do_directive(Token * tline)
2145 enum preproc_token i;
2146 int j;
2147 bool err;
2148 int nparam;
2149 bool nolist;
2150 bool casesense;
2151 int k, m;
2152 int offset;
2153 char *p, *pp;
2154 const char *mname;
2155 Include *inc;
2156 Context *ctx;
2157 Cond *cond;
2158 MMacro *mmac, **mmhead;
2159 Token *t = NULL, *tt, *param_start, *macro_start, *last, **tptr, *origline;
2160 Line *l;
2161 struct tokenval tokval;
2162 expr *evalresult;
2163 MMacro *tmp_defining; /* Used when manipulating rep_nest */
2164 int64_t count;
2165 size_t len;
2166 int severity;
2168 origline = tline;
2170 skip_white_(tline);
2171 if (!tline || !tok_type_(tline, TOK_PREPROC_ID) ||
2172 (tline->text[1] == '%' || tline->text[1] == '$'
2173 || tline->text[1] == '!'))
2174 return NO_DIRECTIVE_FOUND;
2176 i = pp_token_hash(tline->text);
2179 * FIXME: We zap execution of PP_RMACRO, PP_IRMACRO, PP_EXITMACRO
2180 * since they are known to be buggy at moment, we need to fix them
2181 * in future release (2.09-2.10)
2183 if (i == PP_RMACRO || i == PP_IRMACRO || i == PP_EXITMACRO) {
2184 error(ERR_NONFATAL, "unknown preprocessor directive `%s'",
2185 tline->text);
2186 return NO_DIRECTIVE_FOUND;
2190 * If we're in a non-emitting branch of a condition construct,
2191 * or walking to the end of an already terminated %rep block,
2192 * we should ignore all directives except for condition
2193 * directives.
2195 if (((istk->conds && !emitting(istk->conds->state)) ||
2196 (istk->mstk && !istk->mstk->in_progress)) && !is_condition(i)) {
2197 return NO_DIRECTIVE_FOUND;
2201 * If we're defining a macro or reading a %rep block, we should
2202 * ignore all directives except for %macro/%imacro (which nest),
2203 * %endm/%endmacro, and (only if we're in a %rep block) %endrep.
2204 * If we're in a %rep block, another %rep nests, so should be let through.
2206 if (defining && i != PP_MACRO && i != PP_IMACRO &&
2207 i != PP_RMACRO && i != PP_IRMACRO &&
2208 i != PP_ENDMACRO && i != PP_ENDM &&
2209 (defining->name || (i != PP_ENDREP && i != PP_REP))) {
2210 return NO_DIRECTIVE_FOUND;
2213 if (defining) {
2214 if (i == PP_MACRO || i == PP_IMACRO ||
2215 i == PP_RMACRO || i == PP_IRMACRO) {
2216 nested_mac_count++;
2217 return NO_DIRECTIVE_FOUND;
2218 } else if (nested_mac_count > 0) {
2219 if (i == PP_ENDMACRO) {
2220 nested_mac_count--;
2221 return NO_DIRECTIVE_FOUND;
2224 if (!defining->name) {
2225 if (i == PP_REP) {
2226 nested_rep_count++;
2227 return NO_DIRECTIVE_FOUND;
2228 } else if (nested_rep_count > 0) {
2229 if (i == PP_ENDREP) {
2230 nested_rep_count--;
2231 return NO_DIRECTIVE_FOUND;
2237 switch (i) {
2238 case PP_INVALID:
2239 error(ERR_NONFATAL, "unknown preprocessor directive `%s'",
2240 tline->text);
2241 return NO_DIRECTIVE_FOUND; /* didn't get it */
2243 case PP_STACKSIZE:
2244 /* Directive to tell NASM what the default stack size is. The
2245 * default is for a 16-bit stack, and this can be overriden with
2246 * %stacksize large.
2248 tline = tline->next;
2249 if (tline && tline->type == TOK_WHITESPACE)
2250 tline = tline->next;
2251 if (!tline || tline->type != TOK_ID) {
2252 error(ERR_NONFATAL, "`%%stacksize' missing size parameter");
2253 free_tlist(origline);
2254 return DIRECTIVE_FOUND;
2256 if (nasm_stricmp(tline->text, "flat") == 0) {
2257 /* All subsequent ARG directives are for a 32-bit stack */
2258 StackSize = 4;
2259 StackPointer = "ebp";
2260 ArgOffset = 8;
2261 LocalOffset = 0;
2262 } else if (nasm_stricmp(tline->text, "flat64") == 0) {
2263 /* All subsequent ARG directives are for a 64-bit stack */
2264 StackSize = 8;
2265 StackPointer = "rbp";
2266 ArgOffset = 16;
2267 LocalOffset = 0;
2268 } else if (nasm_stricmp(tline->text, "large") == 0) {
2269 /* All subsequent ARG directives are for a 16-bit stack,
2270 * far function call.
2272 StackSize = 2;
2273 StackPointer = "bp";
2274 ArgOffset = 4;
2275 LocalOffset = 0;
2276 } else if (nasm_stricmp(tline->text, "small") == 0) {
2277 /* All subsequent ARG directives are for a 16-bit stack,
2278 * far function call. We don't support near functions.
2280 StackSize = 2;
2281 StackPointer = "bp";
2282 ArgOffset = 6;
2283 LocalOffset = 0;
2284 } else {
2285 error(ERR_NONFATAL, "`%%stacksize' invalid size type");
2286 free_tlist(origline);
2287 return DIRECTIVE_FOUND;
2289 free_tlist(origline);
2290 return DIRECTIVE_FOUND;
2292 case PP_ARG:
2293 /* TASM like ARG directive to define arguments to functions, in
2294 * the following form:
2296 * ARG arg1:WORD, arg2:DWORD, arg4:QWORD
2298 offset = ArgOffset;
2299 do {
2300 char *arg, directive[256];
2301 int size = StackSize;
2303 /* Find the argument name */
2304 tline = tline->next;
2305 if (tline && tline->type == TOK_WHITESPACE)
2306 tline = tline->next;
2307 if (!tline || tline->type != TOK_ID) {
2308 error(ERR_NONFATAL, "`%%arg' missing argument parameter");
2309 free_tlist(origline);
2310 return DIRECTIVE_FOUND;
2312 arg = tline->text;
2314 /* Find the argument size type */
2315 tline = tline->next;
2316 if (!tline || tline->type != TOK_OTHER
2317 || tline->text[0] != ':') {
2318 error(ERR_NONFATAL,
2319 "Syntax error processing `%%arg' directive");
2320 free_tlist(origline);
2321 return DIRECTIVE_FOUND;
2323 tline = tline->next;
2324 if (!tline || tline->type != TOK_ID) {
2325 error(ERR_NONFATAL, "`%%arg' missing size type parameter");
2326 free_tlist(origline);
2327 return DIRECTIVE_FOUND;
2330 /* Allow macro expansion of type parameter */
2331 tt = tokenize(tline->text);
2332 tt = expand_smacro(tt);
2333 size = parse_size(tt->text);
2334 if (!size) {
2335 error(ERR_NONFATAL,
2336 "Invalid size type for `%%arg' missing directive");
2337 free_tlist(tt);
2338 free_tlist(origline);
2339 return DIRECTIVE_FOUND;
2341 free_tlist(tt);
2343 /* Round up to even stack slots */
2344 size = ALIGN(size, StackSize);
2346 /* Now define the macro for the argument */
2347 snprintf(directive, sizeof(directive), "%%define %s (%s+%d)",
2348 arg, StackPointer, offset);
2349 do_directive(tokenize(directive));
2350 offset += size;
2352 /* Move to the next argument in the list */
2353 tline = tline->next;
2354 if (tline && tline->type == TOK_WHITESPACE)
2355 tline = tline->next;
2356 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2357 ArgOffset = offset;
2358 free_tlist(origline);
2359 return DIRECTIVE_FOUND;
2361 case PP_LOCAL:
2362 /* TASM like LOCAL directive to define local variables for a
2363 * function, in the following form:
2365 * LOCAL local1:WORD, local2:DWORD, local4:QWORD = LocalSize
2367 * The '= LocalSize' at the end is ignored by NASM, but is
2368 * required by TASM to define the local parameter size (and used
2369 * by the TASM macro package).
2371 offset = LocalOffset;
2372 do {
2373 char *local, directive[256];
2374 int size = StackSize;
2376 /* Find the argument name */
2377 tline = tline->next;
2378 if (tline && tline->type == TOK_WHITESPACE)
2379 tline = tline->next;
2380 if (!tline || tline->type != TOK_ID) {
2381 error(ERR_NONFATAL,
2382 "`%%local' missing argument parameter");
2383 free_tlist(origline);
2384 return DIRECTIVE_FOUND;
2386 local = tline->text;
2388 /* Find the argument size type */
2389 tline = tline->next;
2390 if (!tline || tline->type != TOK_OTHER
2391 || tline->text[0] != ':') {
2392 error(ERR_NONFATAL,
2393 "Syntax error processing `%%local' directive");
2394 free_tlist(origline);
2395 return DIRECTIVE_FOUND;
2397 tline = tline->next;
2398 if (!tline || tline->type != TOK_ID) {
2399 error(ERR_NONFATAL,
2400 "`%%local' missing size type parameter");
2401 free_tlist(origline);
2402 return DIRECTIVE_FOUND;
2405 /* Allow macro expansion of type parameter */
2406 tt = tokenize(tline->text);
2407 tt = expand_smacro(tt);
2408 size = parse_size(tt->text);
2409 if (!size) {
2410 error(ERR_NONFATAL,
2411 "Invalid size type for `%%local' missing directive");
2412 free_tlist(tt);
2413 free_tlist(origline);
2414 return DIRECTIVE_FOUND;
2416 free_tlist(tt);
2418 /* Round up to even stack slots */
2419 size = ALIGN(size, StackSize);
2421 offset += size; /* Negative offset, increment before */
2423 /* Now define the macro for the argument */
2424 snprintf(directive, sizeof(directive), "%%define %s (%s-%d)",
2425 local, StackPointer, offset);
2426 do_directive(tokenize(directive));
2428 /* Now define the assign to setup the enter_c macro correctly */
2429 snprintf(directive, sizeof(directive),
2430 "%%assign %%$localsize %%$localsize+%d", size);
2431 do_directive(tokenize(directive));
2433 /* Move to the next argument in the list */
2434 tline = tline->next;
2435 if (tline && tline->type == TOK_WHITESPACE)
2436 tline = tline->next;
2437 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2438 LocalOffset = offset;
2439 free_tlist(origline);
2440 return DIRECTIVE_FOUND;
2442 case PP_CLEAR:
2443 if (tline->next)
2444 error(ERR_WARNING|ERR_PASS1,
2445 "trailing garbage after `%%clear' ignored");
2446 free_macros();
2447 init_macros();
2448 free_tlist(origline);
2449 return DIRECTIVE_FOUND;
2451 case PP_DEPEND:
2452 t = tline->next = expand_smacro(tline->next);
2453 skip_white_(t);
2454 if (!t || (t->type != TOK_STRING &&
2455 t->type != TOK_INTERNAL_STRING)) {
2456 error(ERR_NONFATAL, "`%%depend' expects a file name");
2457 free_tlist(origline);
2458 return DIRECTIVE_FOUND; /* but we did _something_ */
2460 if (t->next)
2461 error(ERR_WARNING|ERR_PASS1,
2462 "trailing garbage after `%%depend' ignored");
2463 p = t->text;
2464 if (t->type != TOK_INTERNAL_STRING)
2465 nasm_unquote_cstr(p, i);
2466 if (dephead && !in_list(*dephead, p)) {
2467 StrList *sl = nasm_malloc(strlen(p)+1+sizeof sl->next);
2468 sl->next = NULL;
2469 strcpy(sl->str, p);
2470 *deptail = sl;
2471 deptail = &sl->next;
2473 free_tlist(origline);
2474 return DIRECTIVE_FOUND;
2476 case PP_INCLUDE:
2477 t = tline->next = expand_smacro(tline->next);
2478 skip_white_(t);
2480 if (!t || (t->type != TOK_STRING &&
2481 t->type != TOK_INTERNAL_STRING)) {
2482 error(ERR_NONFATAL, "`%%include' expects a file name");
2483 free_tlist(origline);
2484 return DIRECTIVE_FOUND; /* but we did _something_ */
2486 if (t->next)
2487 error(ERR_WARNING|ERR_PASS1,
2488 "trailing garbage after `%%include' ignored");
2489 p = t->text;
2490 if (t->type != TOK_INTERNAL_STRING)
2491 nasm_unquote_cstr(p, i);
2492 inc = nasm_malloc(sizeof(Include));
2493 inc->next = istk;
2494 inc->conds = NULL;
2495 inc->fp = inc_fopen(p, dephead, &deptail, pass == 0);
2496 if (!inc->fp) {
2497 /* -MG given but file not found */
2498 nasm_free(inc);
2499 } else {
2500 inc->fname = src_set_fname(nasm_strdup(p));
2501 inc->lineno = src_set_linnum(0);
2502 inc->lineinc = 1;
2503 inc->expansion = NULL;
2504 inc->mstk = NULL;
2505 istk = inc;
2506 list->uplevel(LIST_INCLUDE);
2508 free_tlist(origline);
2509 return DIRECTIVE_FOUND;
2511 case PP_USE:
2513 static macros_t *use_pkg;
2514 const char *pkg_macro = NULL;
2516 tline = tline->next;
2517 skip_white_(tline);
2518 tline = expand_id(tline);
2520 if (!tline || (tline->type != TOK_STRING &&
2521 tline->type != TOK_INTERNAL_STRING &&
2522 tline->type != TOK_ID)) {
2523 error(ERR_NONFATAL, "`%%use' expects a package name");
2524 free_tlist(origline);
2525 return DIRECTIVE_FOUND; /* but we did _something_ */
2527 if (tline->next)
2528 error(ERR_WARNING|ERR_PASS1,
2529 "trailing garbage after `%%use' ignored");
2530 if (tline->type == TOK_STRING)
2531 nasm_unquote_cstr(tline->text, i);
2532 use_pkg = nasm_stdmac_find_package(tline->text);
2533 if (!use_pkg)
2534 error(ERR_NONFATAL, "unknown `%%use' package: %s", tline->text);
2535 else
2536 pkg_macro = (char *)use_pkg + 1; /* The first string will be <%define>__USE_*__ */
2537 if (use_pkg && ! smacro_defined(NULL, pkg_macro, 0, NULL, true)) {
2538 /* Not already included, go ahead and include it */
2539 stdmacpos = use_pkg;
2541 free_tlist(origline);
2542 return DIRECTIVE_FOUND;
2544 case PP_PUSH:
2545 case PP_REPL:
2546 case PP_POP:
2547 tline = tline->next;
2548 skip_white_(tline);
2549 tline = expand_id(tline);
2550 if (tline) {
2551 if (!tok_type_(tline, TOK_ID)) {
2552 error(ERR_NONFATAL, "`%s' expects a context identifier",
2553 pp_directives[i]);
2554 free_tlist(origline);
2555 return DIRECTIVE_FOUND; /* but we did _something_ */
2557 if (tline->next)
2558 error(ERR_WARNING|ERR_PASS1,
2559 "trailing garbage after `%s' ignored",
2560 pp_directives[i]);
2561 p = nasm_strdup(tline->text);
2562 } else {
2563 p = NULL; /* Anonymous */
2566 if (i == PP_PUSH) {
2567 ctx = nasm_malloc(sizeof(Context));
2568 ctx->next = cstk;
2569 hash_init(&ctx->localmac, HASH_SMALL);
2570 ctx->name = p;
2571 ctx->number = unique++;
2572 cstk = ctx;
2573 } else {
2574 /* %pop or %repl */
2575 if (!cstk) {
2576 error(ERR_NONFATAL, "`%s': context stack is empty",
2577 pp_directives[i]);
2578 } else if (i == PP_POP) {
2579 if (p && (!cstk->name || nasm_stricmp(p, cstk->name)))
2580 error(ERR_NONFATAL, "`%%pop' in wrong context: %s, "
2581 "expected %s",
2582 cstk->name ? cstk->name : "anonymous", p);
2583 else
2584 ctx_pop();
2585 } else {
2586 /* i == PP_REPL */
2587 nasm_free(cstk->name);
2588 cstk->name = p;
2589 p = NULL;
2591 nasm_free(p);
2593 free_tlist(origline);
2594 return DIRECTIVE_FOUND;
2595 case PP_FATAL:
2596 severity = ERR_FATAL;
2597 goto issue_error;
2598 case PP_ERROR:
2599 severity = ERR_NONFATAL;
2600 goto issue_error;
2601 case PP_WARNING:
2602 severity = ERR_WARNING|ERR_WARN_USER;
2603 goto issue_error;
2605 issue_error:
2607 /* Only error out if this is the final pass */
2608 if (pass != 2 && i != PP_FATAL)
2609 return DIRECTIVE_FOUND;
2611 tline->next = expand_smacro(tline->next);
2612 tline = tline->next;
2613 skip_white_(tline);
2614 t = tline ? tline->next : NULL;
2615 skip_white_(t);
2616 if (tok_type_(tline, TOK_STRING) && !t) {
2617 /* The line contains only a quoted string */
2618 p = tline->text;
2619 nasm_unquote(p, NULL); /* Ignore NUL character truncation */
2620 error(severity, "%s", p);
2621 } else {
2622 /* Not a quoted string, or more than a quoted string */
2623 p = detoken(tline, false);
2624 error(severity, "%s", p);
2625 nasm_free(p);
2627 free_tlist(origline);
2628 return DIRECTIVE_FOUND;
2631 CASE_PP_IF:
2632 if (istk->conds && !emitting(istk->conds->state))
2633 j = COND_NEVER;
2634 else {
2635 j = if_condition(tline->next, i);
2636 tline->next = NULL; /* it got freed */
2637 j = j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2639 cond = nasm_malloc(sizeof(Cond));
2640 cond->next = istk->conds;
2641 cond->state = j;
2642 istk->conds = cond;
2643 if(istk->mstk)
2644 istk->mstk->condcnt ++;
2645 free_tlist(origline);
2646 return DIRECTIVE_FOUND;
2648 CASE_PP_ELIF:
2649 if (!istk->conds)
2650 error(ERR_FATAL, "`%s': no matching `%%if'", pp_directives[i]);
2651 switch(istk->conds->state) {
2652 case COND_IF_TRUE:
2653 istk->conds->state = COND_DONE;
2654 break;
2656 case COND_DONE:
2657 case COND_NEVER:
2658 break;
2660 case COND_ELSE_TRUE:
2661 case COND_ELSE_FALSE:
2662 error_precond(ERR_WARNING|ERR_PASS1,
2663 "`%%elif' after `%%else' ignored");
2664 istk->conds->state = COND_NEVER;
2665 break;
2667 case COND_IF_FALSE:
2669 * IMPORTANT: In the case of %if, we will already have
2670 * called expand_mmac_params(); however, if we're
2671 * processing an %elif we must have been in a
2672 * non-emitting mode, which would have inhibited
2673 * the normal invocation of expand_mmac_params().
2674 * Therefore, we have to do it explicitly here.
2676 j = if_condition(expand_mmac_params(tline->next), i);
2677 tline->next = NULL; /* it got freed */
2678 istk->conds->state =
2679 j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2680 break;
2682 free_tlist(origline);
2683 return DIRECTIVE_FOUND;
2685 case PP_ELSE:
2686 if (tline->next)
2687 error_precond(ERR_WARNING|ERR_PASS1,
2688 "trailing garbage after `%%else' ignored");
2689 if (!istk->conds)
2690 error(ERR_FATAL, "`%%else': no matching `%%if'");
2691 switch(istk->conds->state) {
2692 case COND_IF_TRUE:
2693 case COND_DONE:
2694 istk->conds->state = COND_ELSE_FALSE;
2695 break;
2697 case COND_NEVER:
2698 break;
2700 case COND_IF_FALSE:
2701 istk->conds->state = COND_ELSE_TRUE;
2702 break;
2704 case COND_ELSE_TRUE:
2705 case COND_ELSE_FALSE:
2706 error_precond(ERR_WARNING|ERR_PASS1,
2707 "`%%else' after `%%else' ignored.");
2708 istk->conds->state = COND_NEVER;
2709 break;
2711 free_tlist(origline);
2712 return DIRECTIVE_FOUND;
2714 case PP_ENDIF:
2715 if (tline->next)
2716 error_precond(ERR_WARNING|ERR_PASS1,
2717 "trailing garbage after `%%endif' ignored");
2718 if (!istk->conds)
2719 error(ERR_FATAL, "`%%endif': no matching `%%if'");
2720 cond = istk->conds;
2721 istk->conds = cond->next;
2722 nasm_free(cond);
2723 if(istk->mstk)
2724 istk->mstk->condcnt --;
2725 free_tlist(origline);
2726 return DIRECTIVE_FOUND;
2728 case PP_RMACRO:
2729 case PP_IRMACRO:
2730 case PP_MACRO:
2731 case PP_IMACRO:
2732 if (defining) {
2733 error(ERR_FATAL, "`%s': already defining a macro",
2734 pp_directives[i]);
2735 return DIRECTIVE_FOUND;
2737 defining = nasm_malloc(sizeof(MMacro));
2738 defining->max_depth =
2739 (i == PP_RMACRO) || (i == PP_IRMACRO) ? DEADMAN_LIMIT : 0;
2740 defining->casesense = (i == PP_MACRO) || (i == PP_RMACRO);
2741 if (!parse_mmacro_spec(tline, defining, pp_directives[i])) {
2742 nasm_free(defining);
2743 defining = NULL;
2744 return DIRECTIVE_FOUND;
2747 mmac = (MMacro *) hash_findix(&mmacros, defining->name);
2748 while (mmac) {
2749 if (!strcmp(mmac->name, defining->name) &&
2750 (mmac->nparam_min <= defining->nparam_max
2751 || defining->plus)
2752 && (defining->nparam_min <= mmac->nparam_max
2753 || mmac->plus)) {
2754 error(ERR_WARNING|ERR_PASS1,
2755 "redefining multi-line macro `%s'", defining->name);
2756 return DIRECTIVE_FOUND;
2758 mmac = mmac->next;
2760 free_tlist(origline);
2761 return DIRECTIVE_FOUND;
2763 case PP_ENDM:
2764 case PP_ENDMACRO:
2765 if (! (defining && defining->name)) {
2766 error(ERR_NONFATAL, "`%s': not defining a macro", tline->text);
2767 return DIRECTIVE_FOUND;
2769 mmhead = (MMacro **) hash_findi_add(&mmacros, defining->name);
2770 defining->next = *mmhead;
2771 *mmhead = defining;
2772 defining = NULL;
2773 free_tlist(origline);
2774 return DIRECTIVE_FOUND;
2776 case PP_EXITMACRO:
2778 * We must search along istk->expansion until we hit a
2779 * macro-end marker for a macro with a name. Then we
2780 * bypass all lines between exitmacro and endmacro.
2782 list_for_each(l, istk->expansion)
2783 if (l->finishes && l->finishes->name)
2784 break;
2786 if (l) {
2788 * Remove all conditional entries relative to this
2789 * macro invocation. (safe to do in this context)
2791 for ( ; l->finishes->condcnt > 0; l->finishes->condcnt --) {
2792 cond = istk->conds;
2793 istk->conds = cond->next;
2794 nasm_free(cond);
2796 istk->expansion = l;
2797 } else {
2798 error(ERR_NONFATAL, "`%%exitmacro' not within `%%macro' block");
2800 free_tlist(origline);
2801 return DIRECTIVE_FOUND;
2803 case PP_UNMACRO:
2804 case PP_UNIMACRO:
2806 MMacro **mmac_p;
2807 MMacro spec;
2809 spec.casesense = (i == PP_UNMACRO);
2810 if (!parse_mmacro_spec(tline, &spec, pp_directives[i])) {
2811 return DIRECTIVE_FOUND;
2813 mmac_p = (MMacro **) hash_findi(&mmacros, spec.name, NULL);
2814 while (mmac_p && *mmac_p) {
2815 mmac = *mmac_p;
2816 if (mmac->casesense == spec.casesense &&
2817 !mstrcmp(mmac->name, spec.name, spec.casesense) &&
2818 mmac->nparam_min == spec.nparam_min &&
2819 mmac->nparam_max == spec.nparam_max &&
2820 mmac->plus == spec.plus) {
2821 *mmac_p = mmac->next;
2822 free_mmacro(mmac);
2823 } else {
2824 mmac_p = &mmac->next;
2827 free_tlist(origline);
2828 free_tlist(spec.dlist);
2829 return DIRECTIVE_FOUND;
2832 case PP_ROTATE:
2833 if (tline->next && tline->next->type == TOK_WHITESPACE)
2834 tline = tline->next;
2835 if (!tline->next) {
2836 free_tlist(origline);
2837 error(ERR_NONFATAL, "`%%rotate' missing rotate count");
2838 return DIRECTIVE_FOUND;
2840 t = expand_smacro(tline->next);
2841 tline->next = NULL;
2842 free_tlist(origline);
2843 tline = t;
2844 tptr = &t;
2845 tokval.t_type = TOKEN_INVALID;
2846 evalresult =
2847 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
2848 free_tlist(tline);
2849 if (!evalresult)
2850 return DIRECTIVE_FOUND;
2851 if (tokval.t_type)
2852 error(ERR_WARNING|ERR_PASS1,
2853 "trailing garbage after expression ignored");
2854 if (!is_simple(evalresult)) {
2855 error(ERR_NONFATAL, "non-constant value given to `%%rotate'");
2856 return DIRECTIVE_FOUND;
2858 mmac = istk->mstk;
2859 while (mmac && !mmac->name) /* avoid mistaking %reps for macros */
2860 mmac = mmac->next_active;
2861 if (!mmac) {
2862 error(ERR_NONFATAL, "`%%rotate' invoked outside a macro call");
2863 } else if (mmac->nparam == 0) {
2864 error(ERR_NONFATAL,
2865 "`%%rotate' invoked within macro without parameters");
2866 } else {
2867 int rotate = mmac->rotate + reloc_value(evalresult);
2869 rotate %= (int)mmac->nparam;
2870 if (rotate < 0)
2871 rotate += mmac->nparam;
2873 mmac->rotate = rotate;
2875 return DIRECTIVE_FOUND;
2877 case PP_REP:
2878 nolist = false;
2879 do {
2880 tline = tline->next;
2881 } while (tok_type_(tline, TOK_WHITESPACE));
2883 if (tok_type_(tline, TOK_ID) &&
2884 nasm_stricmp(tline->text, ".nolist") == 0) {
2885 nolist = true;
2886 do {
2887 tline = tline->next;
2888 } while (tok_type_(tline, TOK_WHITESPACE));
2891 if (tline) {
2892 t = expand_smacro(tline);
2893 tptr = &t;
2894 tokval.t_type = TOKEN_INVALID;
2895 evalresult =
2896 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
2897 if (!evalresult) {
2898 free_tlist(origline);
2899 return DIRECTIVE_FOUND;
2901 if (tokval.t_type)
2902 error(ERR_WARNING|ERR_PASS1,
2903 "trailing garbage after expression ignored");
2904 if (!is_simple(evalresult)) {
2905 error(ERR_NONFATAL, "non-constant value given to `%%rep'");
2906 return DIRECTIVE_FOUND;
2908 count = reloc_value(evalresult);
2909 if (count >= REP_LIMIT) {
2910 error(ERR_NONFATAL, "`%%rep' value exceeds limit");
2911 count = 0;
2912 } else
2913 count++;
2914 } else {
2915 error(ERR_NONFATAL, "`%%rep' expects a repeat count");
2916 count = 0;
2918 free_tlist(origline);
2920 tmp_defining = defining;
2921 defining = nasm_malloc(sizeof(MMacro));
2922 defining->prev = NULL;
2923 defining->name = NULL; /* flags this macro as a %rep block */
2924 defining->casesense = false;
2925 defining->plus = false;
2926 defining->nolist = nolist;
2927 defining->in_progress = count;
2928 defining->max_depth = 0;
2929 defining->nparam_min = defining->nparam_max = 0;
2930 defining->defaults = NULL;
2931 defining->dlist = NULL;
2932 defining->expansion = NULL;
2933 defining->next_active = istk->mstk;
2934 defining->rep_nest = tmp_defining;
2935 return DIRECTIVE_FOUND;
2937 case PP_ENDREP:
2938 if (!defining || defining->name) {
2939 error(ERR_NONFATAL, "`%%endrep': no matching `%%rep'");
2940 return DIRECTIVE_FOUND;
2944 * Now we have a "macro" defined - although it has no name
2945 * and we won't be entering it in the hash tables - we must
2946 * push a macro-end marker for it on to istk->expansion.
2947 * After that, it will take care of propagating itself (a
2948 * macro-end marker line for a macro which is really a %rep
2949 * block will cause the macro to be re-expanded, complete
2950 * with another macro-end marker to ensure the process
2951 * continues) until the whole expansion is forcibly removed
2952 * from istk->expansion by a %exitrep.
2954 l = nasm_malloc(sizeof(Line));
2955 l->next = istk->expansion;
2956 l->finishes = defining;
2957 l->first = NULL;
2958 istk->expansion = l;
2960 istk->mstk = defining;
2962 list->uplevel(defining->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
2963 tmp_defining = defining;
2964 defining = defining->rep_nest;
2965 free_tlist(origline);
2966 return DIRECTIVE_FOUND;
2968 case PP_EXITREP:
2970 * We must search along istk->expansion until we hit a
2971 * macro-end marker for a macro with no name. Then we set
2972 * its `in_progress' flag to 0.
2974 list_for_each(l, istk->expansion)
2975 if (l->finishes && !l->finishes->name)
2976 break;
2978 if (l)
2979 l->finishes->in_progress = 1;
2980 else
2981 error(ERR_NONFATAL, "`%%exitrep' not within `%%rep' block");
2982 free_tlist(origline);
2983 return DIRECTIVE_FOUND;
2985 case PP_XDEFINE:
2986 case PP_IXDEFINE:
2987 case PP_DEFINE:
2988 case PP_IDEFINE:
2989 casesense = (i == PP_DEFINE || i == PP_XDEFINE);
2991 tline = tline->next;
2992 skip_white_(tline);
2993 tline = expand_id(tline);
2994 if (!tline || (tline->type != TOK_ID &&
2995 (tline->type != TOK_PREPROC_ID ||
2996 tline->text[1] != '$'))) {
2997 error(ERR_NONFATAL, "`%s' expects a macro identifier",
2998 pp_directives[i]);
2999 free_tlist(origline);
3000 return DIRECTIVE_FOUND;
3003 ctx = get_ctx(tline->text, &mname);
3004 last = tline;
3005 param_start = tline = tline->next;
3006 nparam = 0;
3008 /* Expand the macro definition now for %xdefine and %ixdefine */
3009 if ((i == PP_XDEFINE) || (i == PP_IXDEFINE))
3010 tline = expand_smacro(tline);
3012 if (tok_is_(tline, "(")) {
3014 * This macro has parameters.
3017 tline = tline->next;
3018 while (1) {
3019 skip_white_(tline);
3020 if (!tline) {
3021 error(ERR_NONFATAL, "parameter identifier expected");
3022 free_tlist(origline);
3023 return DIRECTIVE_FOUND;
3025 if (tline->type != TOK_ID) {
3026 error(ERR_NONFATAL,
3027 "`%s': parameter identifier expected",
3028 tline->text);
3029 free_tlist(origline);
3030 return DIRECTIVE_FOUND;
3032 tline->type = TOK_SMAC_PARAM + nparam++;
3033 tline = tline->next;
3034 skip_white_(tline);
3035 if (tok_is_(tline, ",")) {
3036 tline = tline->next;
3037 } else {
3038 if (!tok_is_(tline, ")")) {
3039 error(ERR_NONFATAL,
3040 "`)' expected to terminate macro template");
3041 free_tlist(origline);
3042 return DIRECTIVE_FOUND;
3044 break;
3047 last = tline;
3048 tline = tline->next;
3050 if (tok_type_(tline, TOK_WHITESPACE))
3051 last = tline, tline = tline->next;
3052 macro_start = NULL;
3053 last->next = NULL;
3054 t = tline;
3055 while (t) {
3056 if (t->type == TOK_ID) {
3057 list_for_each(tt, param_start)
3058 if (tt->type >= TOK_SMAC_PARAM &&
3059 !strcmp(tt->text, t->text))
3060 t->type = tt->type;
3062 tt = t->next;
3063 t->next = macro_start;
3064 macro_start = t;
3065 t = tt;
3068 * Good. We now have a macro name, a parameter count, and a
3069 * token list (in reverse order) for an expansion. We ought
3070 * to be OK just to create an SMacro, store it, and let
3071 * free_tlist have the rest of the line (which we have
3072 * carefully re-terminated after chopping off the expansion
3073 * from the end).
3075 define_smacro(ctx, mname, casesense, nparam, macro_start);
3076 free_tlist(origline);
3077 return DIRECTIVE_FOUND;
3079 case PP_UNDEF:
3080 tline = tline->next;
3081 skip_white_(tline);
3082 tline = expand_id(tline);
3083 if (!tline || (tline->type != TOK_ID &&
3084 (tline->type != TOK_PREPROC_ID ||
3085 tline->text[1] != '$'))) {
3086 error(ERR_NONFATAL, "`%%undef' expects a macro identifier");
3087 free_tlist(origline);
3088 return DIRECTIVE_FOUND;
3090 if (tline->next) {
3091 error(ERR_WARNING|ERR_PASS1,
3092 "trailing garbage after macro name ignored");
3095 /* Find the context that symbol belongs to */
3096 ctx = get_ctx(tline->text, &mname);
3097 undef_smacro(ctx, mname);
3098 free_tlist(origline);
3099 return DIRECTIVE_FOUND;
3101 case PP_DEFSTR:
3102 case PP_IDEFSTR:
3103 casesense = (i == PP_DEFSTR);
3105 tline = tline->next;
3106 skip_white_(tline);
3107 tline = expand_id(tline);
3108 if (!tline || (tline->type != TOK_ID &&
3109 (tline->type != TOK_PREPROC_ID ||
3110 tline->text[1] != '$'))) {
3111 error(ERR_NONFATAL, "`%s' expects a macro identifier",
3112 pp_directives[i]);
3113 free_tlist(origline);
3114 return DIRECTIVE_FOUND;
3117 ctx = get_ctx(tline->text, &mname);
3118 last = tline;
3119 tline = expand_smacro(tline->next);
3120 last->next = NULL;
3122 while (tok_type_(tline, TOK_WHITESPACE))
3123 tline = delete_Token(tline);
3125 p = detoken(tline, false);
3126 macro_start = nasm_malloc(sizeof(*macro_start));
3127 macro_start->next = NULL;
3128 macro_start->text = nasm_quote(p, strlen(p));
3129 macro_start->type = TOK_STRING;
3130 macro_start->a.mac = NULL;
3131 nasm_free(p);
3134 * We now have a macro name, an implicit parameter count of
3135 * zero, and a string token to use as an expansion. Create
3136 * and store an SMacro.
3138 define_smacro(ctx, mname, casesense, 0, macro_start);
3139 free_tlist(origline);
3140 return DIRECTIVE_FOUND;
3142 case PP_DEFTOK:
3143 case PP_IDEFTOK:
3144 casesense = (i == PP_DEFTOK);
3146 tline = tline->next;
3147 skip_white_(tline);
3148 tline = expand_id(tline);
3149 if (!tline || (tline->type != TOK_ID &&
3150 (tline->type != TOK_PREPROC_ID ||
3151 tline->text[1] != '$'))) {
3152 error(ERR_NONFATAL,
3153 "`%s' expects a macro identifier as first parameter",
3154 pp_directives[i]);
3155 free_tlist(origline);
3156 return DIRECTIVE_FOUND;
3158 ctx = get_ctx(tline->text, &mname);
3159 last = tline;
3160 tline = expand_smacro(tline->next);
3161 last->next = NULL;
3163 t = tline;
3164 while (tok_type_(t, TOK_WHITESPACE))
3165 t = t->next;
3166 /* t should now point to the string */
3167 if (!tok_type_(t, TOK_STRING)) {
3168 error(ERR_NONFATAL,
3169 "`%s` requires string as second parameter",
3170 pp_directives[i]);
3171 free_tlist(tline);
3172 free_tlist(origline);
3173 return DIRECTIVE_FOUND;
3177 * Convert the string to a token stream. Note that smacros
3178 * are stored with the token stream reversed, so we have to
3179 * reverse the output of tokenize().
3181 nasm_unquote_cstr(t->text, i);
3182 macro_start = reverse_tokens(tokenize(t->text));
3185 * We now have a macro name, an implicit parameter count of
3186 * zero, and a numeric token to use as an expansion. Create
3187 * and store an SMacro.
3189 define_smacro(ctx, mname, casesense, 0, macro_start);
3190 free_tlist(tline);
3191 free_tlist(origline);
3192 return DIRECTIVE_FOUND;
3194 case PP_PATHSEARCH:
3196 FILE *fp;
3197 StrList *xsl = NULL;
3198 StrList **xst = &xsl;
3200 casesense = true;
3202 tline = tline->next;
3203 skip_white_(tline);
3204 tline = expand_id(tline);
3205 if (!tline || (tline->type != TOK_ID &&
3206 (tline->type != TOK_PREPROC_ID ||
3207 tline->text[1] != '$'))) {
3208 error(ERR_NONFATAL,
3209 "`%%pathsearch' expects a macro identifier as first parameter");
3210 free_tlist(origline);
3211 return DIRECTIVE_FOUND;
3213 ctx = get_ctx(tline->text, &mname);
3214 last = tline;
3215 tline = expand_smacro(tline->next);
3216 last->next = NULL;
3218 t = tline;
3219 while (tok_type_(t, TOK_WHITESPACE))
3220 t = t->next;
3222 if (!t || (t->type != TOK_STRING &&
3223 t->type != TOK_INTERNAL_STRING)) {
3224 error(ERR_NONFATAL, "`%%pathsearch' expects a file name");
3225 free_tlist(tline);
3226 free_tlist(origline);
3227 return DIRECTIVE_FOUND; /* but we did _something_ */
3229 if (t->next)
3230 error(ERR_WARNING|ERR_PASS1,
3231 "trailing garbage after `%%pathsearch' ignored");
3232 p = t->text;
3233 if (t->type != TOK_INTERNAL_STRING)
3234 nasm_unquote(p, NULL);
3236 fp = inc_fopen(p, &xsl, &xst, true);
3237 if (fp) {
3238 p = xsl->str;
3239 fclose(fp); /* Don't actually care about the file */
3241 macro_start = nasm_malloc(sizeof(*macro_start));
3242 macro_start->next = NULL;
3243 macro_start->text = nasm_quote(p, strlen(p));
3244 macro_start->type = TOK_STRING;
3245 macro_start->a.mac = NULL;
3246 if (xsl)
3247 nasm_free(xsl);
3250 * We now have a macro name, an implicit parameter count of
3251 * zero, and a string token to use as an expansion. Create
3252 * and store an SMacro.
3254 define_smacro(ctx, mname, casesense, 0, macro_start);
3255 free_tlist(tline);
3256 free_tlist(origline);
3257 return DIRECTIVE_FOUND;
3260 case PP_STRLEN:
3261 casesense = true;
3263 tline = tline->next;
3264 skip_white_(tline);
3265 tline = expand_id(tline);
3266 if (!tline || (tline->type != TOK_ID &&
3267 (tline->type != TOK_PREPROC_ID ||
3268 tline->text[1] != '$'))) {
3269 error(ERR_NONFATAL,
3270 "`%%strlen' expects a macro identifier as first parameter");
3271 free_tlist(origline);
3272 return DIRECTIVE_FOUND;
3274 ctx = get_ctx(tline->text, &mname);
3275 last = tline;
3276 tline = expand_smacro(tline->next);
3277 last->next = NULL;
3279 t = tline;
3280 while (tok_type_(t, TOK_WHITESPACE))
3281 t = t->next;
3282 /* t should now point to the string */
3283 if (!tok_type_(t, TOK_STRING)) {
3284 error(ERR_NONFATAL,
3285 "`%%strlen` requires string as second parameter");
3286 free_tlist(tline);
3287 free_tlist(origline);
3288 return DIRECTIVE_FOUND;
3291 macro_start = nasm_malloc(sizeof(*macro_start));
3292 macro_start->next = NULL;
3293 make_tok_num(macro_start, nasm_unquote(t->text, NULL));
3294 macro_start->a.mac = NULL;
3297 * We now have a macro name, an implicit parameter count of
3298 * zero, and a numeric token to use as an expansion. Create
3299 * and store an SMacro.
3301 define_smacro(ctx, mname, casesense, 0, macro_start);
3302 free_tlist(tline);
3303 free_tlist(origline);
3304 return DIRECTIVE_FOUND;
3306 case PP_STRCAT:
3307 casesense = true;
3309 tline = tline->next;
3310 skip_white_(tline);
3311 tline = expand_id(tline);
3312 if (!tline || (tline->type != TOK_ID &&
3313 (tline->type != TOK_PREPROC_ID ||
3314 tline->text[1] != '$'))) {
3315 error(ERR_NONFATAL,
3316 "`%%strcat' expects a macro identifier as first parameter");
3317 free_tlist(origline);
3318 return DIRECTIVE_FOUND;
3320 ctx = get_ctx(tline->text, &mname);
3321 last = tline;
3322 tline = expand_smacro(tline->next);
3323 last->next = NULL;
3325 len = 0;
3326 list_for_each(t, tline) {
3327 switch (t->type) {
3328 case TOK_WHITESPACE:
3329 break;
3330 case TOK_STRING:
3331 len += t->a.len = nasm_unquote(t->text, NULL);
3332 break;
3333 case TOK_OTHER:
3334 if (!strcmp(t->text, ",")) /* permit comma separators */
3335 break;
3336 /* else fall through */
3337 default:
3338 error(ERR_NONFATAL,
3339 "non-string passed to `%%strcat' (%d)", t->type);
3340 free_tlist(tline);
3341 free_tlist(origline);
3342 return DIRECTIVE_FOUND;
3346 p = pp = nasm_malloc(len);
3347 list_for_each(t, tline) {
3348 if (t->type == TOK_STRING) {
3349 memcpy(p, t->text, t->a.len);
3350 p += t->a.len;
3355 * We now have a macro name, an implicit parameter count of
3356 * zero, and a numeric token to use as an expansion. Create
3357 * and store an SMacro.
3359 macro_start = new_Token(NULL, TOK_STRING, NULL, 0);
3360 macro_start->text = nasm_quote(pp, len);
3361 nasm_free(pp);
3362 define_smacro(ctx, mname, casesense, 0, macro_start);
3363 free_tlist(tline);
3364 free_tlist(origline);
3365 return DIRECTIVE_FOUND;
3367 case PP_SUBSTR:
3369 int64_t start, count;
3370 size_t len;
3372 casesense = true;
3374 tline = tline->next;
3375 skip_white_(tline);
3376 tline = expand_id(tline);
3377 if (!tline || (tline->type != TOK_ID &&
3378 (tline->type != TOK_PREPROC_ID ||
3379 tline->text[1] != '$'))) {
3380 error(ERR_NONFATAL,
3381 "`%%substr' expects a macro identifier as first parameter");
3382 free_tlist(origline);
3383 return DIRECTIVE_FOUND;
3385 ctx = get_ctx(tline->text, &mname);
3386 last = tline;
3387 tline = expand_smacro(tline->next);
3388 last->next = NULL;
3390 if (tline) /* skip expanded id */
3391 t = tline->next;
3392 while (tok_type_(t, TOK_WHITESPACE))
3393 t = t->next;
3395 /* t should now point to the string */
3396 if (!tok_type_(t, TOK_STRING)) {
3397 error(ERR_NONFATAL,
3398 "`%%substr` requires string as second parameter");
3399 free_tlist(tline);
3400 free_tlist(origline);
3401 return DIRECTIVE_FOUND;
3404 tt = t->next;
3405 tptr = &tt;
3406 tokval.t_type = TOKEN_INVALID;
3407 evalresult = evaluate(ppscan, tptr, &tokval, NULL,
3408 pass, error, NULL);
3409 if (!evalresult) {
3410 free_tlist(tline);
3411 free_tlist(origline);
3412 return DIRECTIVE_FOUND;
3413 } else if (!is_simple(evalresult)) {
3414 error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3415 free_tlist(tline);
3416 free_tlist(origline);
3417 return DIRECTIVE_FOUND;
3419 start = evalresult->value - 1;
3421 while (tok_type_(tt, TOK_WHITESPACE))
3422 tt = tt->next;
3423 if (!tt) {
3424 count = 1; /* Backwards compatibility: one character */
3425 } else {
3426 tokval.t_type = TOKEN_INVALID;
3427 evalresult = evaluate(ppscan, tptr, &tokval, NULL,
3428 pass, error, NULL);
3429 if (!evalresult) {
3430 free_tlist(tline);
3431 free_tlist(origline);
3432 return DIRECTIVE_FOUND;
3433 } else if (!is_simple(evalresult)) {
3434 error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3435 free_tlist(tline);
3436 free_tlist(origline);
3437 return DIRECTIVE_FOUND;
3439 count = evalresult->value;
3442 len = nasm_unquote(t->text, NULL);
3444 /* make start and count being in range */
3445 if (start < 0)
3446 start = 0;
3447 if (count < 0)
3448 count = len + count + 1 - start;
3449 if (start + count > (int64_t)len)
3450 count = len - start;
3451 if (!len || count < 0 || start >=(int64_t)len)
3452 start = -1, count = 0; /* empty string */
3454 macro_start = nasm_malloc(sizeof(*macro_start));
3455 macro_start->next = NULL;
3456 macro_start->text = nasm_quote((start < 0) ? "" : t->text + start, count);
3457 macro_start->type = TOK_STRING;
3458 macro_start->a.mac = NULL;
3461 * We now have a macro name, an implicit parameter count of
3462 * zero, and a numeric token to use as an expansion. Create
3463 * and store an SMacro.
3465 define_smacro(ctx, mname, casesense, 0, macro_start);
3466 free_tlist(tline);
3467 free_tlist(origline);
3468 return DIRECTIVE_FOUND;
3471 case PP_ASSIGN:
3472 case PP_IASSIGN:
3473 casesense = (i == PP_ASSIGN);
3475 tline = tline->next;
3476 skip_white_(tline);
3477 tline = expand_id(tline);
3478 if (!tline || (tline->type != TOK_ID &&
3479 (tline->type != TOK_PREPROC_ID ||
3480 tline->text[1] != '$'))) {
3481 error(ERR_NONFATAL,
3482 "`%%%sassign' expects a macro identifier",
3483 (i == PP_IASSIGN ? "i" : ""));
3484 free_tlist(origline);
3485 return DIRECTIVE_FOUND;
3487 ctx = get_ctx(tline->text, &mname);
3488 last = tline;
3489 tline = expand_smacro(tline->next);
3490 last->next = NULL;
3492 t = tline;
3493 tptr = &t;
3494 tokval.t_type = TOKEN_INVALID;
3495 evalresult =
3496 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
3497 free_tlist(tline);
3498 if (!evalresult) {
3499 free_tlist(origline);
3500 return DIRECTIVE_FOUND;
3503 if (tokval.t_type)
3504 error(ERR_WARNING|ERR_PASS1,
3505 "trailing garbage after expression ignored");
3507 if (!is_simple(evalresult)) {
3508 error(ERR_NONFATAL,
3509 "non-constant value given to `%%%sassign'",
3510 (i == PP_IASSIGN ? "i" : ""));
3511 free_tlist(origline);
3512 return DIRECTIVE_FOUND;
3515 macro_start = nasm_malloc(sizeof(*macro_start));
3516 macro_start->next = NULL;
3517 make_tok_num(macro_start, reloc_value(evalresult));
3518 macro_start->a.mac = NULL;
3521 * We now have a macro name, an implicit parameter count of
3522 * zero, and a numeric token to use as an expansion. Create
3523 * and store an SMacro.
3525 define_smacro(ctx, mname, casesense, 0, macro_start);
3526 free_tlist(origline);
3527 return DIRECTIVE_FOUND;
3529 case PP_LINE:
3531 * Syntax is `%line nnn[+mmm] [filename]'
3533 tline = tline->next;
3534 skip_white_(tline);
3535 if (!tok_type_(tline, TOK_NUMBER)) {
3536 error(ERR_NONFATAL, "`%%line' expects line number");
3537 free_tlist(origline);
3538 return DIRECTIVE_FOUND;
3540 k = readnum(tline->text, &err);
3541 m = 1;
3542 tline = tline->next;
3543 if (tok_is_(tline, "+")) {
3544 tline = tline->next;
3545 if (!tok_type_(tline, TOK_NUMBER)) {
3546 error(ERR_NONFATAL, "`%%line' expects line increment");
3547 free_tlist(origline);
3548 return DIRECTIVE_FOUND;
3550 m = readnum(tline->text, &err);
3551 tline = tline->next;
3553 skip_white_(tline);
3554 src_set_linnum(k);
3555 istk->lineinc = m;
3556 if (tline) {
3557 nasm_free(src_set_fname(detoken(tline, false)));
3559 free_tlist(origline);
3560 return DIRECTIVE_FOUND;
3562 default:
3563 error(ERR_FATAL,
3564 "preprocessor directive `%s' not yet implemented",
3565 pp_directives[i]);
3566 return DIRECTIVE_FOUND;
3571 * Ensure that a macro parameter contains a condition code and
3572 * nothing else. Return the condition code index if so, or -1
3573 * otherwise.
3575 static int find_cc(Token * t)
3577 Token *tt;
3579 if (!t)
3580 return -1; /* Probably a %+ without a space */
3582 skip_white_(t);
3583 if (t->type != TOK_ID)
3584 return -1;
3585 tt = t->next;
3586 skip_white_(tt);
3587 if (tt && (tt->type != TOK_OTHER || strcmp(tt->text, ",")))
3588 return -1;
3590 return bsii(t->text, (const char **)conditions, ARRAY_SIZE(conditions));
3594 * This routines walks over tokens strem and hadnles tokens
3595 * pasting, if @handle_explicit passed then explicit pasting
3596 * term is handled, otherwise -- implicit pastings only.
3598 static bool paste_tokens(Token **head, const struct tokseq_match *m,
3599 size_t mnum, bool handle_explicit)
3601 Token *tok, *next, **prev_next, **prev_nonspace;
3602 bool pasted = false;
3603 char *buf, *p;
3604 size_t len, i;
3607 * The last token before pasting. We need it
3608 * to be able to connect new handled tokens.
3609 * In other words if there were a tokens stream
3611 * A -> B -> C -> D
3613 * and we've joined tokens B and C, the resulting
3614 * stream should be
3616 * A -> BC -> D
3618 tok = *head;
3619 prev_next = NULL;
3621 if (!tok_type_(tok, TOK_WHITESPACE) && !tok_type_(tok, TOK_PASTE))
3622 prev_nonspace = head;
3623 else
3624 prev_nonspace = NULL;
3626 while (tok && (next = tok->next)) {
3628 switch (tok->type) {
3629 case TOK_WHITESPACE:
3630 /* Zap redundant whitespaces */
3631 while (tok_type_(next, TOK_WHITESPACE))
3632 next = delete_Token(next);
3633 tok->next = next;
3634 break;
3636 case TOK_PASTE:
3637 /* Explicit pasting */
3638 if (!handle_explicit)
3639 break;
3640 next = delete_Token(tok);
3642 while (tok_type_(next, TOK_WHITESPACE))
3643 next = delete_Token(next);
3645 if (!pasted)
3646 pasted = true;
3648 /* Left pasting token is start of line */
3649 if (!prev_nonspace)
3650 error(ERR_FATAL, "No lvalue found on pasting");
3653 * No ending token, this might happen in two
3654 * cases
3656 * 1) There indeed no right token at all
3657 * 2) There is a bare "%define ID" statement,
3658 * and @ID does expand to whitespace.
3660 * So technically we need to do a grammar analysis
3661 * in another stage of parsing, but for now lets don't
3662 * change the behaviour people used to. Simply allow
3663 * whitespace after paste token.
3665 if (!next) {
3667 * Zap ending space tokens and that's all.
3669 tok = (*prev_nonspace)->next;
3670 while (tok_type_(tok, TOK_WHITESPACE))
3671 tok = delete_Token(tok);
3672 tok = *prev_nonspace;
3673 tok->next = NULL;
3674 break;
3677 tok = *prev_nonspace;
3678 while (tok_type_(tok, TOK_WHITESPACE))
3679 tok = delete_Token(tok);
3680 len = strlen(tok->text);
3681 len += strlen(next->text);
3683 p = buf = nasm_malloc(len + 1);
3684 strcpy(p, tok->text);
3685 p = strchr(p, '\0');
3686 strcpy(p, next->text);
3688 delete_Token(tok);
3690 tok = tokenize(buf);
3691 nasm_free(buf);
3693 *prev_nonspace = tok;
3694 while (tok && tok->next)
3695 tok = tok->next;
3697 tok->next = delete_Token(next);
3699 /* Restart from pasted tokens head */
3700 tok = *prev_nonspace;
3701 break;
3703 default:
3704 /* implicit pasting */
3705 for (i = 0; i < mnum; i++) {
3706 if (!(PP_CONCAT_MATCH(tok, m[i].mask_head)))
3707 continue;
3709 len = 0;
3710 while (next && PP_CONCAT_MATCH(next, m[i].mask_tail)) {
3711 len += strlen(next->text);
3712 next = next->next;
3715 /* No match */
3716 if (tok == next)
3717 break;
3719 len += strlen(tok->text);
3720 p = buf = nasm_malloc(len + 1);
3722 while (tok != next) {
3723 strcpy(p, tok->text);
3724 p = strchr(p, '\0');
3725 tok = delete_Token(tok);
3728 tok = tokenize(buf);
3729 nasm_free(buf);
3731 if (prev_next)
3732 *prev_next = tok;
3733 else
3734 *head = tok;
3737 * Connect pasted into original stream,
3738 * ie A -> new-tokens -> B
3740 while (tok && tok->next)
3741 tok = tok->next;
3742 tok->next = next;
3744 if (!pasted)
3745 pasted = true;
3747 /* Restart from pasted tokens head */
3748 tok = prev_next ? *prev_next : *head;
3751 break;
3754 prev_next = &tok->next;
3756 if (tok->next &&
3757 !tok_type_(tok->next, TOK_WHITESPACE) &&
3758 !tok_type_(tok->next, TOK_PASTE))
3759 prev_nonspace = prev_next;
3761 tok = tok->next;
3764 return pasted;
3768 * expands to a list of tokens from %{x:y}
3770 static Token *expand_mmac_params_range(MMacro *mac, Token *tline, Token ***last)
3772 Token *t = tline, **tt, *tm, *head;
3773 char *pos;
3774 int fst, lst, j, i;
3776 pos = strchr(tline->text, ':');
3777 nasm_assert(pos);
3779 lst = atoi(pos + 1);
3780 fst = atoi(tline->text + 1);
3783 * only macros params are accounted so
3784 * if someone passes %0 -- we reject such
3785 * value(s)
3787 if (lst == 0 || fst == 0)
3788 goto err;
3790 /* the values should be sane */
3791 if ((fst > (int)mac->nparam || fst < (-(int)mac->nparam)) ||
3792 (lst > (int)mac->nparam || lst < (-(int)mac->nparam)))
3793 goto err;
3795 fst = fst < 0 ? fst + (int)mac->nparam + 1: fst;
3796 lst = lst < 0 ? lst + (int)mac->nparam + 1: lst;
3798 /* counted from zero */
3799 fst--, lst--;
3802 * It will be at least one token. Note we
3803 * need to scan params until separator, otherwise
3804 * only first token will be passed.
3806 tm = mac->params[(fst + mac->rotate) % mac->nparam];
3807 head = new_Token(NULL, tm->type, tm->text, 0);
3808 tt = &head->next, tm = tm->next;
3809 while (tok_isnt_(tm, ",")) {
3810 t = new_Token(NULL, tm->type, tm->text, 0);
3811 *tt = t, tt = &t->next, tm = tm->next;
3814 if (fst < lst) {
3815 for (i = fst + 1; i <= lst; i++) {
3816 t = new_Token(NULL, TOK_OTHER, ",", 0);
3817 *tt = t, tt = &t->next;
3818 j = (i + mac->rotate) % mac->nparam;
3819 tm = mac->params[j];
3820 while (tok_isnt_(tm, ",")) {
3821 t = new_Token(NULL, tm->type, tm->text, 0);
3822 *tt = t, tt = &t->next, tm = tm->next;
3825 } else {
3826 for (i = fst - 1; i >= lst; i--) {
3827 t = new_Token(NULL, TOK_OTHER, ",", 0);
3828 *tt = t, tt = &t->next;
3829 j = (i + mac->rotate) % mac->nparam;
3830 tm = mac->params[j];
3831 while (tok_isnt_(tm, ",")) {
3832 t = new_Token(NULL, tm->type, tm->text, 0);
3833 *tt = t, tt = &t->next, tm = tm->next;
3838 *last = tt;
3839 return head;
3841 err:
3842 error(ERR_NONFATAL, "`%%{%s}': macro parameters out of range",
3843 &tline->text[1]);
3844 return tline;
3848 * Expand MMacro-local things: parameter references (%0, %n, %+n,
3849 * %-n) and MMacro-local identifiers (%%foo) as well as
3850 * macro indirection (%[...]) and range (%{..:..}).
3852 static Token *expand_mmac_params(Token * tline)
3854 Token *t, *tt, **tail, *thead;
3855 bool changed = false;
3856 char *pos;
3858 tail = &thead;
3859 thead = NULL;
3861 while (tline) {
3862 if (tline->type == TOK_PREPROC_ID &&
3863 (((tline->text[1] == '+' || tline->text[1] == '-') && tline->text[2]) ||
3864 (tline->text[1] >= '0' && tline->text[1] <= '9') ||
3865 tline->text[1] == '%')) {
3866 char *text = NULL;
3867 int type = 0, cc; /* type = 0 to placate optimisers */
3868 char tmpbuf[30];
3869 unsigned int n;
3870 int i;
3871 MMacro *mac;
3873 t = tline;
3874 tline = tline->next;
3876 mac = istk->mstk;
3877 while (mac && !mac->name) /* avoid mistaking %reps for macros */
3878 mac = mac->next_active;
3879 if (!mac) {
3880 error(ERR_NONFATAL, "`%s': not in a macro call", t->text);
3881 } else {
3882 pos = strchr(t->text, ':');
3883 if (!pos) {
3884 switch (t->text[1]) {
3886 * We have to make a substitution of one of the
3887 * forms %1, %-1, %+1, %%foo, %0.
3889 case '0':
3890 type = TOK_NUMBER;
3891 snprintf(tmpbuf, sizeof(tmpbuf), "%d", mac->nparam);
3892 text = nasm_strdup(tmpbuf);
3893 break;
3894 case '%':
3895 type = TOK_ID;
3896 snprintf(tmpbuf, sizeof(tmpbuf), "..@%"PRIu64".",
3897 mac->unique);
3898 text = nasm_strcat(tmpbuf, t->text + 2);
3899 break;
3900 case '-':
3901 n = atoi(t->text + 2) - 1;
3902 if (n >= mac->nparam)
3903 tt = NULL;
3904 else {
3905 if (mac->nparam > 1)
3906 n = (n + mac->rotate) % mac->nparam;
3907 tt = mac->params[n];
3909 cc = find_cc(tt);
3910 if (cc == -1) {
3911 error(ERR_NONFATAL,
3912 "macro parameter %d is not a condition code",
3913 n + 1);
3914 text = NULL;
3915 } else {
3916 type = TOK_ID;
3917 if (inverse_ccs[cc] == -1) {
3918 error(ERR_NONFATAL,
3919 "condition code `%s' is not invertible",
3920 conditions[cc]);
3921 text = NULL;
3922 } else
3923 text = nasm_strdup(conditions[inverse_ccs[cc]]);
3925 break;
3926 case '+':
3927 n = atoi(t->text + 2) - 1;
3928 if (n >= mac->nparam)
3929 tt = NULL;
3930 else {
3931 if (mac->nparam > 1)
3932 n = (n + mac->rotate) % mac->nparam;
3933 tt = mac->params[n];
3935 cc = find_cc(tt);
3936 if (cc == -1) {
3937 error(ERR_NONFATAL,
3938 "macro parameter %d is not a condition code",
3939 n + 1);
3940 text = NULL;
3941 } else {
3942 type = TOK_ID;
3943 text = nasm_strdup(conditions[cc]);
3945 break;
3946 default:
3947 n = atoi(t->text + 1) - 1;
3948 if (n >= mac->nparam)
3949 tt = NULL;
3950 else {
3951 if (mac->nparam > 1)
3952 n = (n + mac->rotate) % mac->nparam;
3953 tt = mac->params[n];
3955 if (tt) {
3956 for (i = 0; i < mac->paramlen[n]; i++) {
3957 *tail = new_Token(NULL, tt->type, tt->text, 0);
3958 tail = &(*tail)->next;
3959 tt = tt->next;
3962 text = NULL; /* we've done it here */
3963 break;
3965 } else {
3967 * seems we have a parameters range here
3969 Token *head, **last;
3970 head = expand_mmac_params_range(mac, t, &last);
3971 if (head != t) {
3972 *tail = head;
3973 *last = tline;
3974 tline = head;
3975 text = NULL;
3979 if (!text) {
3980 delete_Token(t);
3981 } else {
3982 *tail = t;
3983 tail = &t->next;
3984 t->type = type;
3985 nasm_free(t->text);
3986 t->text = text;
3987 t->a.mac = NULL;
3989 changed = true;
3990 continue;
3991 } else if (tline->type == TOK_INDIRECT) {
3992 t = tline;
3993 tline = tline->next;
3994 tt = tokenize(t->text);
3995 tt = expand_mmac_params(tt);
3996 tt = expand_smacro(tt);
3997 *tail = tt;
3998 while (tt) {
3999 tt->a.mac = NULL; /* Necessary? */
4000 tail = &tt->next;
4001 tt = tt->next;
4003 delete_Token(t);
4004 changed = true;
4005 } else {
4006 t = *tail = tline;
4007 tline = tline->next;
4008 t->a.mac = NULL;
4009 tail = &t->next;
4012 *tail = NULL;
4014 if (changed) {
4015 const struct tokseq_match t[] = {
4017 PP_CONCAT_MASK(TOK_ID) |
4018 PP_CONCAT_MASK(TOK_FLOAT), /* head */
4019 PP_CONCAT_MASK(TOK_ID) |
4020 PP_CONCAT_MASK(TOK_NUMBER) |
4021 PP_CONCAT_MASK(TOK_FLOAT) |
4022 PP_CONCAT_MASK(TOK_OTHER) /* tail */
4025 PP_CONCAT_MASK(TOK_NUMBER), /* head */
4026 PP_CONCAT_MASK(TOK_NUMBER) /* tail */
4029 paste_tokens(&thead, t, ARRAY_SIZE(t), false);
4032 return thead;
4036 * Expand all single-line macro calls made in the given line.
4037 * Return the expanded version of the line. The original is deemed
4038 * to be destroyed in the process. (In reality we'll just move
4039 * Tokens from input to output a lot of the time, rather than
4040 * actually bothering to destroy and replicate.)
4043 static Token *expand_smacro(Token * tline)
4045 Token *t, *tt, *mstart, **tail, *thead;
4046 SMacro *head = NULL, *m;
4047 Token **params;
4048 int *paramsize;
4049 unsigned int nparam, sparam;
4050 int brackets;
4051 Token *org_tline = tline;
4052 Context *ctx;
4053 const char *mname;
4054 int deadman = DEADMAN_LIMIT;
4055 bool expanded;
4058 * Trick: we should avoid changing the start token pointer since it can
4059 * be contained in "next" field of other token. Because of this
4060 * we allocate a copy of first token and work with it; at the end of
4061 * routine we copy it back
4063 if (org_tline) {
4064 tline = new_Token(org_tline->next, org_tline->type,
4065 org_tline->text, 0);
4066 tline->a.mac = org_tline->a.mac;
4067 nasm_free(org_tline->text);
4068 org_tline->text = NULL;
4071 expanded = true; /* Always expand %+ at least once */
4073 again:
4074 thead = NULL;
4075 tail = &thead;
4077 while (tline) { /* main token loop */
4078 if (!--deadman) {
4079 error(ERR_NONFATAL, "interminable macro recursion");
4080 goto err;
4083 if ((mname = tline->text)) {
4084 /* if this token is a local macro, look in local context */
4085 if (tline->type == TOK_ID) {
4086 head = (SMacro *)hash_findix(&smacros, mname);
4087 } else if (tline->type == TOK_PREPROC_ID) {
4088 ctx = get_ctx(mname, &mname);
4089 head = ctx ? (SMacro *)hash_findix(&ctx->localmac, mname) : NULL;
4090 } else
4091 head = NULL;
4094 * We've hit an identifier. As in is_mmacro below, we first
4095 * check whether the identifier is a single-line macro at
4096 * all, then think about checking for parameters if
4097 * necessary.
4099 list_for_each(m, head)
4100 if (!mstrcmp(m->name, mname, m->casesense))
4101 break;
4102 if (m) {
4103 mstart = tline;
4104 params = NULL;
4105 paramsize = NULL;
4106 if (m->nparam == 0) {
4108 * Simple case: the macro is parameterless. Discard the
4109 * one token that the macro call took, and push the
4110 * expansion back on the to-do stack.
4112 if (!m->expansion) {
4113 if (!strcmp("__FILE__", m->name)) {
4114 int32_t num = 0;
4115 char *file = NULL;
4116 src_get(&num, &file);
4117 tline->text = nasm_quote(file, strlen(file));
4118 tline->type = TOK_STRING;
4119 nasm_free(file);
4120 continue;
4122 if (!strcmp("__LINE__", m->name)) {
4123 nasm_free(tline->text);
4124 make_tok_num(tline, src_get_linnum());
4125 continue;
4127 if (!strcmp("__BITS__", m->name)) {
4128 nasm_free(tline->text);
4129 make_tok_num(tline, globalbits);
4130 continue;
4132 tline = delete_Token(tline);
4133 continue;
4135 } else {
4137 * Complicated case: at least one macro with this name
4138 * exists and takes parameters. We must find the
4139 * parameters in the call, count them, find the SMacro
4140 * that corresponds to that form of the macro call, and
4141 * substitute for the parameters when we expand. What a
4142 * pain.
4144 /*tline = tline->next;
4145 skip_white_(tline); */
4146 do {
4147 t = tline->next;
4148 while (tok_type_(t, TOK_SMAC_END)) {
4149 t->a.mac->in_progress = false;
4150 t->text = NULL;
4151 t = tline->next = delete_Token(t);
4153 tline = t;
4154 } while (tok_type_(tline, TOK_WHITESPACE));
4155 if (!tok_is_(tline, "(")) {
4157 * This macro wasn't called with parameters: ignore
4158 * the call. (Behaviour borrowed from gnu cpp.)
4160 tline = mstart;
4161 m = NULL;
4162 } else {
4163 int paren = 0;
4164 int white = 0;
4165 brackets = 0;
4166 nparam = 0;
4167 sparam = PARAM_DELTA;
4168 params = nasm_malloc(sparam * sizeof(Token *));
4169 params[0] = tline->next;
4170 paramsize = nasm_malloc(sparam * sizeof(int));
4171 paramsize[0] = 0;
4172 while (true) { /* parameter loop */
4174 * For some unusual expansions
4175 * which concatenates function call
4177 t = tline->next;
4178 while (tok_type_(t, TOK_SMAC_END)) {
4179 t->a.mac->in_progress = false;
4180 t->text = NULL;
4181 t = tline->next = delete_Token(t);
4183 tline = t;
4185 if (!tline) {
4186 error(ERR_NONFATAL,
4187 "macro call expects terminating `)'");
4188 break;
4190 if (tline->type == TOK_WHITESPACE
4191 && brackets <= 0) {
4192 if (paramsize[nparam])
4193 white++;
4194 else
4195 params[nparam] = tline->next;
4196 continue; /* parameter loop */
4198 if (tline->type == TOK_OTHER
4199 && tline->text[1] == 0) {
4200 char ch = tline->text[0];
4201 if (ch == ',' && !paren && brackets <= 0) {
4202 if (++nparam >= sparam) {
4203 sparam += PARAM_DELTA;
4204 params = nasm_realloc(params,
4205 sparam * sizeof(Token *));
4206 paramsize = nasm_realloc(paramsize,
4207 sparam * sizeof(int));
4209 params[nparam] = tline->next;
4210 paramsize[nparam] = 0;
4211 white = 0;
4212 continue; /* parameter loop */
4214 if (ch == '{' &&
4215 (brackets > 0 || (brackets == 0 &&
4216 !paramsize[nparam])))
4218 if (!(brackets++)) {
4219 params[nparam] = tline->next;
4220 continue; /* parameter loop */
4223 if (ch == '}' && brackets > 0)
4224 if (--brackets == 0) {
4225 brackets = -1;
4226 continue; /* parameter loop */
4228 if (ch == '(' && !brackets)
4229 paren++;
4230 if (ch == ')' && brackets <= 0)
4231 if (--paren < 0)
4232 break;
4234 if (brackets < 0) {
4235 brackets = 0;
4236 error(ERR_NONFATAL, "braces do not "
4237 "enclose all of macro parameter");
4239 paramsize[nparam] += white + 1;
4240 white = 0;
4241 } /* parameter loop */
4242 nparam++;
4243 while (m && (m->nparam != nparam ||
4244 mstrcmp(m->name, mname,
4245 m->casesense)))
4246 m = m->next;
4247 if (!m)
4248 error(ERR_WARNING|ERR_PASS1|ERR_WARN_MNP,
4249 "macro `%s' exists, "
4250 "but not taking %d parameters",
4251 mstart->text, nparam);
4254 if (m && m->in_progress)
4255 m = NULL;
4256 if (!m) { /* in progess or didn't find '(' or wrong nparam */
4258 * Design question: should we handle !tline, which
4259 * indicates missing ')' here, or expand those
4260 * macros anyway, which requires the (t) test a few
4261 * lines down?
4263 nasm_free(params);
4264 nasm_free(paramsize);
4265 tline = mstart;
4266 } else {
4268 * Expand the macro: we are placed on the last token of the
4269 * call, so that we can easily split the call from the
4270 * following tokens. We also start by pushing an SMAC_END
4271 * token for the cycle removal.
4273 t = tline;
4274 if (t) {
4275 tline = t->next;
4276 t->next = NULL;
4278 tt = new_Token(tline, TOK_SMAC_END, NULL, 0);
4279 tt->a.mac = m;
4280 m->in_progress = true;
4281 tline = tt;
4282 list_for_each(t, m->expansion) {
4283 if (t->type >= TOK_SMAC_PARAM) {
4284 Token *pcopy = tline, **ptail = &pcopy;
4285 Token *ttt, *pt;
4286 int i;
4288 ttt = params[t->type - TOK_SMAC_PARAM];
4289 i = paramsize[t->type - TOK_SMAC_PARAM];
4290 while (--i >= 0) {
4291 pt = *ptail = new_Token(tline, ttt->type,
4292 ttt->text, 0);
4293 ptail = &pt->next;
4294 ttt = ttt->next;
4296 tline = pcopy;
4297 } else if (t->type == TOK_PREPROC_Q) {
4298 tt = new_Token(tline, TOK_ID, mname, 0);
4299 tline = tt;
4300 } else if (t->type == TOK_PREPROC_QQ) {
4301 tt = new_Token(tline, TOK_ID, m->name, 0);
4302 tline = tt;
4303 } else {
4304 tt = new_Token(tline, t->type, t->text, 0);
4305 tline = tt;
4310 * Having done that, get rid of the macro call, and clean
4311 * up the parameters.
4313 nasm_free(params);
4314 nasm_free(paramsize);
4315 free_tlist(mstart);
4316 expanded = true;
4317 continue; /* main token loop */
4322 if (tline->type == TOK_SMAC_END) {
4323 tline->a.mac->in_progress = false;
4324 tline = delete_Token(tline);
4325 } else {
4326 t = *tail = tline;
4327 tline = tline->next;
4328 t->a.mac = NULL;
4329 t->next = NULL;
4330 tail = &t->next;
4335 * Now scan the entire line and look for successive TOK_IDs that resulted
4336 * after expansion (they can't be produced by tokenize()). The successive
4337 * TOK_IDs should be concatenated.
4338 * Also we look for %+ tokens and concatenate the tokens before and after
4339 * them (without white spaces in between).
4341 if (expanded) {
4342 const struct tokseq_match t[] = {
4344 PP_CONCAT_MASK(TOK_ID) |
4345 PP_CONCAT_MASK(TOK_PREPROC_ID), /* head */
4346 PP_CONCAT_MASK(TOK_ID) |
4347 PP_CONCAT_MASK(TOK_PREPROC_ID) |
4348 PP_CONCAT_MASK(TOK_NUMBER) /* tail */
4351 if (paste_tokens(&thead, t, ARRAY_SIZE(t), true)) {
4353 * If we concatenated something, *and* we had previously expanded
4354 * an actual macro, scan the lines again for macros...
4356 tline = thead;
4357 expanded = false;
4358 goto again;
4362 err:
4363 if (org_tline) {
4364 if (thead) {
4365 *org_tline = *thead;
4366 /* since we just gave text to org_line, don't free it */
4367 thead->text = NULL;
4368 delete_Token(thead);
4369 } else {
4370 /* the expression expanded to empty line;
4371 we can't return NULL for some reasons
4372 we just set the line to a single WHITESPACE token. */
4373 memset(org_tline, 0, sizeof(*org_tline));
4374 org_tline->text = NULL;
4375 org_tline->type = TOK_WHITESPACE;
4377 thead = org_tline;
4380 return thead;
4384 * Similar to expand_smacro but used exclusively with macro identifiers
4385 * right before they are fetched in. The reason is that there can be
4386 * identifiers consisting of several subparts. We consider that if there
4387 * are more than one element forming the name, user wants a expansion,
4388 * otherwise it will be left as-is. Example:
4390 * %define %$abc cde
4392 * the identifier %$abc will be left as-is so that the handler for %define
4393 * will suck it and define the corresponding value. Other case:
4395 * %define _%$abc cde
4397 * In this case user wants name to be expanded *before* %define starts
4398 * working, so we'll expand %$abc into something (if it has a value;
4399 * otherwise it will be left as-is) then concatenate all successive
4400 * PP_IDs into one.
4402 static Token *expand_id(Token * tline)
4404 Token *cur, *oldnext = NULL;
4406 if (!tline || !tline->next)
4407 return tline;
4409 cur = tline;
4410 while (cur->next &&
4411 (cur->next->type == TOK_ID ||
4412 cur->next->type == TOK_PREPROC_ID
4413 || cur->next->type == TOK_NUMBER))
4414 cur = cur->next;
4416 /* If identifier consists of just one token, don't expand */
4417 if (cur == tline)
4418 return tline;
4420 if (cur) {
4421 oldnext = cur->next; /* Detach the tail past identifier */
4422 cur->next = NULL; /* so that expand_smacro stops here */
4425 tline = expand_smacro(tline);
4427 if (cur) {
4428 /* expand_smacro possibly changhed tline; re-scan for EOL */
4429 cur = tline;
4430 while (cur && cur->next)
4431 cur = cur->next;
4432 if (cur)
4433 cur->next = oldnext;
4436 return tline;
4440 * Determine whether the given line constitutes a multi-line macro
4441 * call, and return the MMacro structure called if so. Doesn't have
4442 * to check for an initial label - that's taken care of in
4443 * expand_mmacro - but must check numbers of parameters. Guaranteed
4444 * to be called with tline->type == TOK_ID, so the putative macro
4445 * name is easy to find.
4447 static MMacro *is_mmacro(Token * tline, Token *** params_array)
4449 MMacro *head, *m;
4450 Token **params;
4451 int nparam;
4453 head = (MMacro *) hash_findix(&mmacros, tline->text);
4456 * Efficiency: first we see if any macro exists with the given
4457 * name. If not, we can return NULL immediately. _Then_ we
4458 * count the parameters, and then we look further along the
4459 * list if necessary to find the proper MMacro.
4461 list_for_each(m, head)
4462 if (!mstrcmp(m->name, tline->text, m->casesense))
4463 break;
4464 if (!m)
4465 return NULL;
4468 * OK, we have a potential macro. Count and demarcate the
4469 * parameters.
4471 count_mmac_params(tline->next, &nparam, &params);
4474 * So we know how many parameters we've got. Find the MMacro
4475 * structure that handles this number.
4477 while (m) {
4478 if (m->nparam_min <= nparam
4479 && (m->plus || nparam <= m->nparam_max)) {
4481 * This one is right. Just check if cycle removal
4482 * prohibits us using it before we actually celebrate...
4484 if (m->in_progress > m->max_depth) {
4485 if (m->max_depth > 0) {
4486 error(ERR_WARNING,
4487 "reached maximum recursion depth of %i",
4488 m->max_depth);
4490 nasm_free(params);
4491 return NULL;
4494 * It's right, and we can use it. Add its default
4495 * parameters to the end of our list if necessary.
4497 if (m->defaults && nparam < m->nparam_min + m->ndefs) {
4498 params =
4499 nasm_realloc(params,
4500 ((m->nparam_min + m->ndefs +
4501 1) * sizeof(*params)));
4502 while (nparam < m->nparam_min + m->ndefs) {
4503 params[nparam] = m->defaults[nparam - m->nparam_min];
4504 nparam++;
4508 * If we've gone over the maximum parameter count (and
4509 * we're in Plus mode), ignore parameters beyond
4510 * nparam_max.
4512 if (m->plus && nparam > m->nparam_max)
4513 nparam = m->nparam_max;
4515 * Then terminate the parameter list, and leave.
4517 if (!params) { /* need this special case */
4518 params = nasm_malloc(sizeof(*params));
4519 nparam = 0;
4521 params[nparam] = NULL;
4522 *params_array = params;
4523 return m;
4526 * This one wasn't right: look for the next one with the
4527 * same name.
4529 list_for_each(m, m->next)
4530 if (!mstrcmp(m->name, tline->text, m->casesense))
4531 break;
4535 * After all that, we didn't find one with the right number of
4536 * parameters. Issue a warning, and fail to expand the macro.
4538 error(ERR_WARNING|ERR_PASS1|ERR_WARN_MNP,
4539 "macro `%s' exists, but not taking %d parameters",
4540 tline->text, nparam);
4541 nasm_free(params);
4542 return NULL;
4547 * Save MMacro invocation specific fields in
4548 * preparation for a recursive macro expansion
4550 static void push_mmacro(MMacro *m)
4552 MMacroInvocation *i;
4554 i = nasm_malloc(sizeof(MMacroInvocation));
4555 i->prev = m->prev;
4556 i->params = m->params;
4557 i->iline = m->iline;
4558 i->nparam = m->nparam;
4559 i->rotate = m->rotate;
4560 i->paramlen = m->paramlen;
4561 i->unique = m->unique;
4562 i->condcnt = m->condcnt;
4563 m->prev = i;
4568 * Restore MMacro invocation specific fields that were
4569 * saved during a previous recursive macro expansion
4571 static void pop_mmacro(MMacro *m)
4573 MMacroInvocation *i;
4575 if (m->prev) {
4576 i = m->prev;
4577 m->prev = i->prev;
4578 m->params = i->params;
4579 m->iline = i->iline;
4580 m->nparam = i->nparam;
4581 m->rotate = i->rotate;
4582 m->paramlen = i->paramlen;
4583 m->unique = i->unique;
4584 m->condcnt = i->condcnt;
4585 nasm_free(i);
4591 * Expand the multi-line macro call made by the given line, if
4592 * there is one to be expanded. If there is, push the expansion on
4593 * istk->expansion and return 1. Otherwise return 0.
4595 static int expand_mmacro(Token * tline)
4597 Token *startline = tline;
4598 Token *label = NULL;
4599 int dont_prepend = 0;
4600 Token **params, *t, *tt;
4601 MMacro *m;
4602 Line *l, *ll;
4603 int i, nparam, *paramlen;
4604 const char *mname;
4606 t = tline;
4607 skip_white_(t);
4608 /* if (!tok_type_(t, TOK_ID)) Lino 02/25/02 */
4609 if (!tok_type_(t, TOK_ID) && !tok_type_(t, TOK_PREPROC_ID))
4610 return 0;
4611 m = is_mmacro(t, &params);
4612 if (m) {
4613 mname = t->text;
4614 } else {
4615 Token *last;
4617 * We have an id which isn't a macro call. We'll assume
4618 * it might be a label; we'll also check to see if a
4619 * colon follows it. Then, if there's another id after
4620 * that lot, we'll check it again for macro-hood.
4622 label = last = t;
4623 t = t->next;
4624 if (tok_type_(t, TOK_WHITESPACE))
4625 last = t, t = t->next;
4626 if (tok_is_(t, ":")) {
4627 dont_prepend = 1;
4628 last = t, t = t->next;
4629 if (tok_type_(t, TOK_WHITESPACE))
4630 last = t, t = t->next;
4632 if (!tok_type_(t, TOK_ID) || !(m = is_mmacro(t, &params)))
4633 return 0;
4634 last->next = NULL;
4635 mname = t->text;
4636 tline = t;
4640 * Fix up the parameters: this involves stripping leading and
4641 * trailing whitespace, then stripping braces if they are
4642 * present.
4644 for (nparam = 0; params[nparam]; nparam++) ;
4645 paramlen = nparam ? nasm_malloc(nparam * sizeof(*paramlen)) : NULL;
4647 for (i = 0; params[i]; i++) {
4648 int brace = false;
4649 int comma = (!m->plus || i < nparam - 1);
4651 t = params[i];
4652 skip_white_(t);
4653 if (tok_is_(t, "{"))
4654 t = t->next, brace = true, comma = false;
4655 params[i] = t;
4656 paramlen[i] = 0;
4657 while (t) {
4658 if (comma && t->type == TOK_OTHER && !strcmp(t->text, ","))
4659 break; /* ... because we have hit a comma */
4660 if (comma && t->type == TOK_WHITESPACE
4661 && tok_is_(t->next, ","))
4662 break; /* ... or a space then a comma */
4663 if (brace && t->type == TOK_OTHER && !strcmp(t->text, "}"))
4664 break; /* ... or a brace */
4665 t = t->next;
4666 paramlen[i]++;
4671 * OK, we have a MMacro structure together with a set of
4672 * parameters. We must now go through the expansion and push
4673 * copies of each Line on to istk->expansion. Substitution of
4674 * parameter tokens and macro-local tokens doesn't get done
4675 * until the single-line macro substitution process; this is
4676 * because delaying them allows us to change the semantics
4677 * later through %rotate.
4679 * First, push an end marker on to istk->expansion, mark this
4680 * macro as in progress, and set up its invocation-specific
4681 * variables.
4683 ll = nasm_malloc(sizeof(Line));
4684 ll->next = istk->expansion;
4685 ll->finishes = m;
4686 ll->first = NULL;
4687 istk->expansion = ll;
4690 * Save the previous MMacro expansion in the case of
4691 * macro recursion
4693 if (m->max_depth && m->in_progress)
4694 push_mmacro(m);
4696 m->in_progress ++;
4697 m->params = params;
4698 m->iline = tline;
4699 m->nparam = nparam;
4700 m->rotate = 0;
4701 m->paramlen = paramlen;
4702 m->unique = unique++;
4703 m->lineno = 0;
4704 m->condcnt = 0;
4706 m->next_active = istk->mstk;
4707 istk->mstk = m;
4709 list_for_each(l, m->expansion) {
4710 Token **tail;
4712 ll = nasm_malloc(sizeof(Line));
4713 ll->finishes = NULL;
4714 ll->next = istk->expansion;
4715 istk->expansion = ll;
4716 tail = &ll->first;
4718 list_for_each(t, l->first) {
4719 Token *x = t;
4720 switch (t->type) {
4721 case TOK_PREPROC_Q:
4722 tt = *tail = new_Token(NULL, TOK_ID, mname, 0);
4723 break;
4724 case TOK_PREPROC_QQ:
4725 tt = *tail = new_Token(NULL, TOK_ID, m->name, 0);
4726 break;
4727 case TOK_PREPROC_ID:
4728 if (t->text[1] == '0' && t->text[2] == '0') {
4729 dont_prepend = -1;
4730 x = label;
4731 if (!x)
4732 continue;
4734 /* fall through */
4735 default:
4736 tt = *tail = new_Token(NULL, x->type, x->text, 0);
4737 break;
4739 tail = &tt->next;
4741 *tail = NULL;
4745 * If we had a label, push it on as the first line of
4746 * the macro expansion.
4748 if (label) {
4749 if (dont_prepend < 0)
4750 free_tlist(startline);
4751 else {
4752 ll = nasm_malloc(sizeof(Line));
4753 ll->finishes = NULL;
4754 ll->next = istk->expansion;
4755 istk->expansion = ll;
4756 ll->first = startline;
4757 if (!dont_prepend) {
4758 while (label->next)
4759 label = label->next;
4760 label->next = tt = new_Token(NULL, TOK_OTHER, ":", 0);
4765 list->uplevel(m->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
4767 return 1;
4770 /* The function that actually does the error reporting */
4771 static void verror(int severity, const char *fmt, va_list arg)
4773 char buff[1024];
4774 MMacro *mmac = NULL;
4775 int delta = 0;
4777 vsnprintf(buff, sizeof(buff), fmt, arg);
4779 /* get %macro name */
4780 if (istk && istk->mstk) {
4781 mmac = istk->mstk;
4782 /* but %rep blocks should be skipped */
4783 while (mmac && !mmac->name)
4784 mmac = mmac->next_active, delta++;
4787 if (mmac)
4788 nasm_error(severity, "(%s:%d) %s",
4789 mmac->name, mmac->lineno - delta, buff);
4790 else
4791 nasm_error(severity, "%s", buff);
4795 * Since preprocessor always operate only on the line that didn't
4796 * arrived yet, we should always use ERR_OFFBY1.
4798 static void error(int severity, const char *fmt, ...)
4800 va_list arg;
4802 /* If we're in a dead branch of IF or something like it, ignore the error */
4803 if (istk && istk->conds && !emitting(istk->conds->state))
4804 return;
4806 va_start(arg, fmt);
4807 verror(severity, fmt, arg);
4808 va_end(arg);
4812 * Because %else etc are evaluated in the state context
4813 * of the previous branch, errors might get lost with error():
4814 * %if 0 ... %else trailing garbage ... %endif
4815 * So %else etc should report errors with this function.
4817 static void error_precond(int severity, const char *fmt, ...)
4819 va_list arg;
4821 /* Only ignore the error if it's really in a dead branch */
4822 if (istk && istk->conds && istk->conds->state == COND_NEVER)
4823 return;
4825 va_start(arg, fmt);
4826 verror(severity, fmt, arg);
4827 va_end(arg);
4830 static void
4831 pp_reset(char *file, int apass, ListGen * listgen, StrList **deplist)
4833 Token *t;
4835 cstk = NULL;
4836 istk = nasm_malloc(sizeof(Include));
4837 istk->next = NULL;
4838 istk->conds = NULL;
4839 istk->expansion = NULL;
4840 istk->mstk = NULL;
4841 istk->fp = fopen(file, "r");
4842 istk->fname = NULL;
4843 src_set_fname(nasm_strdup(file));
4844 src_set_linnum(0);
4845 istk->lineinc = 1;
4846 if (!istk->fp)
4847 error(ERR_FATAL|ERR_NOFILE, "unable to open input file `%s'",
4848 file);
4849 defining = NULL;
4850 nested_mac_count = 0;
4851 nested_rep_count = 0;
4852 init_macros();
4853 unique = 0;
4854 if (tasm_compatible_mode) {
4855 stdmacpos = nasm_stdmac;
4856 } else {
4857 stdmacpos = nasm_stdmac_after_tasm;
4859 any_extrastdmac = extrastdmac && *extrastdmac;
4860 do_predef = true;
4861 list = listgen;
4864 * 0 for dependencies, 1 for preparatory passes, 2 for final pass.
4865 * The caller, however, will also pass in 3 for preprocess-only so
4866 * we can set __PASS__ accordingly.
4868 pass = apass > 2 ? 2 : apass;
4870 dephead = deptail = deplist;
4871 if (deplist) {
4872 StrList *sl = nasm_malloc(strlen(file)+1+sizeof sl->next);
4873 sl->next = NULL;
4874 strcpy(sl->str, file);
4875 *deptail = sl;
4876 deptail = &sl->next;
4880 * Define the __PASS__ macro. This is defined here unlike
4881 * all the other builtins, because it is special -- it varies between
4882 * passes.
4884 t = nasm_malloc(sizeof(*t));
4885 t->next = NULL;
4886 make_tok_num(t, apass);
4887 t->a.mac = NULL;
4888 define_smacro(NULL, "__PASS__", true, 0, t);
4891 static char *pp_getline(void)
4893 char *line;
4894 Token *tline;
4896 while (1) {
4898 * Fetch a tokenized line, either from the macro-expansion
4899 * buffer or from the input file.
4901 tline = NULL;
4902 while (istk->expansion && istk->expansion->finishes) {
4903 Line *l = istk->expansion;
4904 if (!l->finishes->name && l->finishes->in_progress > 1) {
4905 Line *ll;
4908 * This is a macro-end marker for a macro with no
4909 * name, which means it's not really a macro at all
4910 * but a %rep block, and the `in_progress' field is
4911 * more than 1, meaning that we still need to
4912 * repeat. (1 means the natural last repetition; 0
4913 * means termination by %exitrep.) We have
4914 * therefore expanded up to the %endrep, and must
4915 * push the whole block on to the expansion buffer
4916 * again. We don't bother to remove the macro-end
4917 * marker: we'd only have to generate another one
4918 * if we did.
4920 l->finishes->in_progress--;
4921 list_for_each(l, l->finishes->expansion) {
4922 Token *t, *tt, **tail;
4924 ll = nasm_malloc(sizeof(Line));
4925 ll->next = istk->expansion;
4926 ll->finishes = NULL;
4927 ll->first = NULL;
4928 tail = &ll->first;
4930 list_for_each(t, l->first) {
4931 if (t->text || t->type == TOK_WHITESPACE) {
4932 tt = *tail = new_Token(NULL, t->type, t->text, 0);
4933 tail = &tt->next;
4937 istk->expansion = ll;
4939 } else {
4941 * Check whether a `%rep' was started and not ended
4942 * within this macro expansion. This can happen and
4943 * should be detected. It's a fatal error because
4944 * I'm too confused to work out how to recover
4945 * sensibly from it.
4947 if (defining) {
4948 if (defining->name)
4949 error(ERR_PANIC,
4950 "defining with name in expansion");
4951 else if (istk->mstk->name)
4952 error(ERR_FATAL,
4953 "`%%rep' without `%%endrep' within"
4954 " expansion of macro `%s'",
4955 istk->mstk->name);
4959 * FIXME: investigate the relationship at this point between
4960 * istk->mstk and l->finishes
4963 MMacro *m = istk->mstk;
4964 istk->mstk = m->next_active;
4965 if (m->name) {
4967 * This was a real macro call, not a %rep, and
4968 * therefore the parameter information needs to
4969 * be freed.
4971 if (m->prev) {
4972 pop_mmacro(m);
4973 l->finishes->in_progress --;
4974 } else {
4975 nasm_free(m->params);
4976 free_tlist(m->iline);
4977 nasm_free(m->paramlen);
4978 l->finishes->in_progress = 0;
4980 } else
4981 free_mmacro(m);
4983 istk->expansion = l->next;
4984 nasm_free(l);
4985 list->downlevel(LIST_MACRO);
4988 while (1) { /* until we get a line we can use */
4990 if (istk->expansion) { /* from a macro expansion */
4991 char *p;
4992 Line *l = istk->expansion;
4993 if (istk->mstk)
4994 istk->mstk->lineno++;
4995 tline = l->first;
4996 istk->expansion = l->next;
4997 nasm_free(l);
4998 p = detoken(tline, false);
4999 list->line(LIST_MACRO, p);
5000 nasm_free(p);
5001 break;
5003 line = read_line();
5004 if (line) { /* from the current input file */
5005 line = prepreproc(line);
5006 tline = tokenize(line);
5007 nasm_free(line);
5008 break;
5011 * The current file has ended; work down the istk
5014 Include *i = istk;
5015 fclose(i->fp);
5016 if (i->conds) {
5017 /* nasm_error can't be conditionally suppressed */
5018 nasm_error(ERR_FATAL,
5019 "expected `%%endif' before end of file");
5021 /* only set line and file name if there's a next node */
5022 if (i->next) {
5023 src_set_linnum(i->lineno);
5024 nasm_free(src_set_fname(nasm_strdup(i->fname)));
5026 istk = i->next;
5027 list->downlevel(LIST_INCLUDE);
5028 nasm_free(i);
5029 if (!istk)
5030 return NULL;
5031 if (istk->expansion && istk->expansion->finishes)
5032 break;
5037 * We must expand MMacro parameters and MMacro-local labels
5038 * _before_ we plunge into directive processing, to cope
5039 * with things like `%define something %1' such as STRUC
5040 * uses. Unless we're _defining_ a MMacro, in which case
5041 * those tokens should be left alone to go into the
5042 * definition; and unless we're in a non-emitting
5043 * condition, in which case we don't want to meddle with
5044 * anything.
5046 if (!defining && !(istk->conds && !emitting(istk->conds->state))
5047 && !(istk->mstk && !istk->mstk->in_progress)) {
5048 tline = expand_mmac_params(tline);
5052 * Check the line to see if it's a preprocessor directive.
5054 if (do_directive(tline) == DIRECTIVE_FOUND) {
5055 continue;
5056 } else if (defining) {
5058 * We're defining a multi-line macro. We emit nothing
5059 * at all, and just
5060 * shove the tokenized line on to the macro definition.
5062 Line *l = nasm_malloc(sizeof(Line));
5063 l->next = defining->expansion;
5064 l->first = tline;
5065 l->finishes = NULL;
5066 defining->expansion = l;
5067 continue;
5068 } else if (istk->conds && !emitting(istk->conds->state)) {
5070 * We're in a non-emitting branch of a condition block.
5071 * Emit nothing at all, not even a blank line: when we
5072 * emerge from the condition we'll give a line-number
5073 * directive so we keep our place correctly.
5075 free_tlist(tline);
5076 continue;
5077 } else if (istk->mstk && !istk->mstk->in_progress) {
5079 * We're in a %rep block which has been terminated, so
5080 * we're walking through to the %endrep without
5081 * emitting anything. Emit nothing at all, not even a
5082 * blank line: when we emerge from the %rep block we'll
5083 * give a line-number directive so we keep our place
5084 * correctly.
5086 free_tlist(tline);
5087 continue;
5088 } else {
5089 tline = expand_smacro(tline);
5090 if (!expand_mmacro(tline)) {
5092 * De-tokenize the line again, and emit it.
5094 line = detoken(tline, true);
5095 free_tlist(tline);
5096 break;
5097 } else {
5098 continue; /* expand_mmacro calls free_tlist */
5103 return line;
5106 static void pp_cleanup(int pass)
5108 if (defining) {
5109 if (defining->name) {
5110 error(ERR_NONFATAL,
5111 "end of file while still defining macro `%s'",
5112 defining->name);
5113 } else {
5114 error(ERR_NONFATAL, "end of file while still in %%rep");
5117 free_mmacro(defining);
5118 defining = NULL;
5120 while (cstk)
5121 ctx_pop();
5122 free_macros();
5123 while (istk) {
5124 Include *i = istk;
5125 istk = istk->next;
5126 fclose(i->fp);
5127 nasm_free(i->fname);
5128 nasm_free(i);
5130 while (cstk)
5131 ctx_pop();
5132 nasm_free(src_set_fname(NULL));
5133 if (pass == 0) {
5134 IncPath *i;
5135 free_llist(predef);
5136 delete_Blocks();
5137 while ((i = ipath)) {
5138 ipath = i->next;
5139 if (i->path)
5140 nasm_free(i->path);
5141 nasm_free(i);
5146 static void pp_include_path(char *path)
5148 IncPath *i;
5150 i = nasm_malloc(sizeof(IncPath));
5151 i->path = path ? nasm_strdup(path) : NULL;
5152 i->next = NULL;
5154 if (ipath) {
5155 IncPath *j = ipath;
5156 while (j->next)
5157 j = j->next;
5158 j->next = i;
5159 } else {
5160 ipath = i;
5164 static void pp_pre_include(char *fname)
5166 Token *inc, *space, *name;
5167 Line *l;
5169 name = new_Token(NULL, TOK_INTERNAL_STRING, fname, 0);
5170 space = new_Token(name, TOK_WHITESPACE, NULL, 0);
5171 inc = new_Token(space, TOK_PREPROC_ID, "%include", 0);
5173 l = nasm_malloc(sizeof(Line));
5174 l->next = predef;
5175 l->first = inc;
5176 l->finishes = NULL;
5177 predef = l;
5180 static void pp_pre_define(char *definition)
5182 Token *def, *space;
5183 Line *l;
5184 char *equals;
5186 equals = strchr(definition, '=');
5187 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
5188 def = new_Token(space, TOK_PREPROC_ID, "%define", 0);
5189 if (equals)
5190 *equals = ' ';
5191 space->next = tokenize(definition);
5192 if (equals)
5193 *equals = '=';
5195 l = nasm_malloc(sizeof(Line));
5196 l->next = predef;
5197 l->first = def;
5198 l->finishes = NULL;
5199 predef = l;
5202 static void pp_pre_undefine(char *definition)
5204 Token *def, *space;
5205 Line *l;
5207 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
5208 def = new_Token(space, TOK_PREPROC_ID, "%undef", 0);
5209 space->next = tokenize(definition);
5211 l = nasm_malloc(sizeof(Line));
5212 l->next = predef;
5213 l->first = def;
5214 l->finishes = NULL;
5215 predef = l;
5218 static void pp_extra_stdmac(macros_t *macros)
5220 extrastdmac = macros;
5223 static void make_tok_num(Token * tok, int64_t val)
5225 char numbuf[32];
5226 snprintf(numbuf, sizeof(numbuf), "%"PRId64"", val);
5227 tok->text = nasm_strdup(numbuf);
5228 tok->type = TOK_NUMBER;
5231 struct preproc_ops nasmpp = {
5232 pp_reset,
5233 pp_getline,
5234 pp_cleanup,
5235 pp_extra_stdmac,
5236 pp_pre_define,
5237 pp_pre_undefine,
5238 pp_pre_include,
5239 pp_include_path