Merge branch 'master' into elfmerge
[nasm.git] / preproc.c
blobd2b9bb7f2a31c2fe892bc58a7a3c9a9b5b786a86
1 /* ----------------------------------------------------------------------- *
3 * Copyright 1996-2016 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"
83 #include "listing.h"
85 typedef struct SMacro SMacro;
86 typedef struct MMacro MMacro;
87 typedef struct MMacroInvocation MMacroInvocation;
88 typedef struct Context Context;
89 typedef struct Token Token;
90 typedef struct Blocks Blocks;
91 typedef struct Line Line;
92 typedef struct Include Include;
93 typedef struct Cond Cond;
94 typedef struct IncPath IncPath;
97 * Note on the storage of both SMacro and MMacros: the hash table
98 * indexes them case-insensitively, and we then have to go through a
99 * linked list of potential case aliases (and, for MMacros, parameter
100 * ranges); this is to preserve the matching semantics of the earlier
101 * code. If the number of case aliases for a specific macro is a
102 * performance issue, you may want to reconsider your coding style.
106 * Store the definition of a single-line macro.
108 struct SMacro {
109 SMacro *next;
110 char *name;
111 bool casesense;
112 bool in_progress;
113 unsigned int nparam;
114 Token *expansion;
118 * Store the definition of a multi-line macro. This is also used to
119 * store the interiors of `%rep...%endrep' blocks, which are
120 * effectively self-re-invoking multi-line macros which simply
121 * don't have a name or bother to appear in the hash tables. %rep
122 * blocks are signified by having a NULL `name' field.
124 * In a MMacro describing a `%rep' block, the `in_progress' field
125 * isn't merely boolean, but gives the number of repeats left to
126 * run.
128 * The `next' field is used for storing MMacros in hash tables; the
129 * `next_active' field is for stacking them on istk entries.
131 * When a MMacro is being expanded, `params', `iline', `nparam',
132 * `paramlen', `rotate' and `unique' are local to the invocation.
134 struct MMacro {
135 MMacro *next;
136 MMacroInvocation *prev; /* previous invocation */
137 char *name;
138 int nparam_min, nparam_max;
139 bool casesense;
140 bool plus; /* is the last parameter greedy? */
141 bool nolist; /* is this macro listing-inhibited? */
142 int64_t in_progress; /* is this macro currently being expanded? */
143 int32_t max_depth; /* maximum number of recursive expansions allowed */
144 Token *dlist; /* All defaults as one list */
145 Token **defaults; /* Parameter default pointers */
146 int ndefs; /* number of default parameters */
147 Line *expansion;
149 MMacro *next_active;
150 MMacro *rep_nest; /* used for nesting %rep */
151 Token **params; /* actual parameters */
152 Token *iline; /* invocation line */
153 unsigned int nparam, rotate;
154 int *paramlen;
155 uint64_t unique;
156 int lineno; /* Current line number on expansion */
157 uint64_t condcnt; /* number of if blocks... */
161 /* Store the definition of a multi-line macro, as defined in a
162 * previous recursive macro expansion.
164 struct MMacroInvocation {
165 MMacroInvocation *prev; /* previous invocation */
166 Token **params; /* actual parameters */
167 Token *iline; /* invocation line */
168 unsigned int nparam, rotate;
169 int *paramlen;
170 uint64_t unique;
171 uint64_t condcnt;
176 * The context stack is composed of a linked list of these.
178 struct Context {
179 Context *next;
180 char *name;
181 struct hash_table localmac;
182 uint32_t number;
186 * This is the internal form which we break input lines up into.
187 * Typically stored in linked lists.
189 * Note that `type' serves a double meaning: TOK_SMAC_PARAM is not
190 * necessarily used as-is, but is intended to denote the number of
191 * the substituted parameter. So in the definition
193 * %define a(x,y) ( (x) & ~(y) )
195 * the token representing `x' will have its type changed to
196 * TOK_SMAC_PARAM, but the one representing `y' will be
197 * TOK_SMAC_PARAM+1.
199 * TOK_INTERNAL_STRING is a dirty hack: it's a single string token
200 * which doesn't need quotes around it. Used in the pre-include
201 * mechanism as an alternative to trying to find a sensible type of
202 * quote to use on the filename we were passed.
204 enum pp_token_type {
205 TOK_NONE = 0, TOK_WHITESPACE, TOK_COMMENT, TOK_ID,
206 TOK_PREPROC_ID, TOK_STRING,
207 TOK_NUMBER, TOK_FLOAT, TOK_SMAC_END, TOK_OTHER,
208 TOK_INTERNAL_STRING,
209 TOK_PREPROC_Q, TOK_PREPROC_QQ,
210 TOK_PASTE, /* %+ */
211 TOK_INDIRECT, /* %[...] */
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;
407 * The current set of multi-line macros we have defined.
409 static struct hash_table mmacros;
412 * The current set of single-line macros we have defined.
414 static struct hash_table smacros;
417 * The multi-line macro we are currently defining, or the %rep
418 * block we are currently reading, if any.
420 static MMacro *defining;
422 static uint64_t nested_mac_count;
423 static uint64_t nested_rep_count;
426 * The number of macro parameters to allocate space for at a time.
428 #define PARAM_DELTA 16
431 * The standard macro set: defined in macros.c in the array nasm_stdmac.
432 * This gives our position in the macro set, when we're processing it.
434 static macros_t *stdmacpos;
437 * The extra standard macros that come from the object format, if
438 * any.
440 static macros_t *extrastdmac = NULL;
441 static bool any_extrastdmac;
444 * Tokens are allocated in blocks to improve speed
446 #define TOKEN_BLOCKSIZE 4096
447 static Token *freeTokens = NULL;
448 struct Blocks {
449 Blocks *next;
450 void *chunk;
453 static Blocks blocks = { NULL, NULL };
456 * Forward declarations.
458 static Token *expand_mmac_params(Token * tline);
459 static Token *expand_smacro(Token * tline);
460 static Token *expand_id(Token * tline);
461 static Context *get_ctx(const char *name, const char **namep);
462 static void make_tok_num(Token * tok, int64_t val);
463 static void pp_verror(int severity, const char *fmt, va_list ap);
464 static vefunc real_verror;
465 static void *new_Block(size_t size);
466 static void delete_Blocks(void);
467 static Token *new_Token(Token * next, enum pp_token_type type,
468 const char *text, int txtlen);
469 static Token *delete_Token(Token * t);
472 * Macros for safe checking of token pointers, avoid *(NULL)
474 #define tok_type_(x,t) ((x) && (x)->type == (t))
475 #define skip_white_(x) if (tok_type_((x), TOK_WHITESPACE)) (x)=(x)->next
476 #define tok_is_(x,v) (tok_type_((x), TOK_OTHER) && !strcmp((x)->text,(v)))
477 #define tok_isnt_(x,v) ((x) && ((x)->type!=TOK_OTHER || strcmp((x)->text,(v))))
480 * nasm_unquote with error if the string contains NUL characters.
481 * If the string contains NUL characters, issue an error and return
482 * the C len, i.e. truncate at the NUL.
484 static size_t nasm_unquote_cstr(char *qstr, enum preproc_token directive)
486 size_t len = nasm_unquote(qstr, NULL);
487 size_t clen = strlen(qstr);
489 if (len != clen)
490 nasm_error(ERR_NONFATAL, "NUL character in `%s' directive",
491 pp_directives[directive]);
493 return clen;
497 * In-place reverse a list of tokens.
499 static Token *reverse_tokens(Token *t)
501 Token *prev = NULL;
502 Token *next;
504 while (t) {
505 next = t->next;
506 t->next = prev;
507 prev = t;
508 t = next;
511 return prev;
515 * Handle TASM specific directives, which do not contain a % in
516 * front of them. We do it here because I could not find any other
517 * place to do it for the moment, and it is a hack (ideally it would
518 * be nice to be able to use the NASM pre-processor to do it).
520 static char *check_tasm_directive(char *line)
522 int32_t i, j, k, m, len;
523 char *p, *q, *oldline, oldchar;
525 p = nasm_skip_spaces(line);
527 /* Binary search for the directive name */
528 i = -1;
529 j = ARRAY_SIZE(tasm_directives);
530 q = nasm_skip_word(p);
531 len = q - p;
532 if (len) {
533 oldchar = p[len];
534 p[len] = 0;
535 while (j - i > 1) {
536 k = (j + i) / 2;
537 m = nasm_stricmp(p, tasm_directives[k]);
538 if (m == 0) {
539 /* We have found a directive, so jam a % in front of it
540 * so that NASM will then recognise it as one if it's own.
542 p[len] = oldchar;
543 len = strlen(p);
544 oldline = line;
545 line = nasm_malloc(len + 2);
546 line[0] = '%';
547 if (k == TM_IFDIFI) {
549 * NASM does not recognise IFDIFI, so we convert
550 * it to %if 0. This is not used in NASM
551 * compatible code, but does need to parse for the
552 * TASM macro package.
554 strcpy(line + 1, "if 0");
555 } else {
556 memcpy(line + 1, p, len + 1);
558 nasm_free(oldline);
559 return line;
560 } else if (m < 0) {
561 j = k;
562 } else
563 i = k;
565 p[len] = oldchar;
567 return line;
571 * The pre-preprocessing stage... This function translates line
572 * number indications as they emerge from GNU cpp (`# lineno "file"
573 * flags') into NASM preprocessor line number indications (`%line
574 * lineno file').
576 static char *prepreproc(char *line)
578 int lineno, fnlen;
579 char *fname, *oldline;
581 if (line[0] == '#' && line[1] == ' ') {
582 oldline = line;
583 fname = oldline + 2;
584 lineno = atoi(fname);
585 fname += strspn(fname, "0123456789 ");
586 if (*fname == '"')
587 fname++;
588 fnlen = strcspn(fname, "\"");
589 line = nasm_malloc(20 + fnlen);
590 snprintf(line, 20 + fnlen, "%%line %d %.*s", lineno, fnlen, fname);
591 nasm_free(oldline);
593 if (tasm_compatible_mode)
594 return check_tasm_directive(line);
595 return line;
599 * Free a linked list of tokens.
601 static void free_tlist(Token * list)
603 while (list)
604 list = delete_Token(list);
608 * Free a linked list of lines.
610 static void free_llist(Line * list)
612 Line *l, *tmp;
613 list_for_each_safe(l, tmp, list) {
614 free_tlist(l->first);
615 nasm_free(l);
620 * Free an MMacro
622 static void free_mmacro(MMacro * m)
624 nasm_free(m->name);
625 free_tlist(m->dlist);
626 nasm_free(m->defaults);
627 free_llist(m->expansion);
628 nasm_free(m);
632 * Free all currently defined macros, and free the hash tables
634 static void free_smacro_table(struct hash_table *smt)
636 SMacro *s, *tmp;
637 const char *key;
638 struct hash_tbl_node *it = NULL;
640 while ((s = hash_iterate(smt, &it, &key)) != NULL) {
641 nasm_free((void *)key);
642 list_for_each_safe(s, tmp, s) {
643 nasm_free(s->name);
644 free_tlist(s->expansion);
645 nasm_free(s);
648 hash_free(smt);
651 static void free_mmacro_table(struct hash_table *mmt)
653 MMacro *m, *tmp;
654 const char *key;
655 struct hash_tbl_node *it = NULL;
657 it = NULL;
658 while ((m = hash_iterate(mmt, &it, &key)) != NULL) {
659 nasm_free((void *)key);
660 list_for_each_safe(m ,tmp, m)
661 free_mmacro(m);
663 hash_free(mmt);
666 static void free_macros(void)
668 free_smacro_table(&smacros);
669 free_mmacro_table(&mmacros);
673 * Initialize the hash tables
675 static void init_macros(void)
677 hash_init(&smacros, HASH_LARGE);
678 hash_init(&mmacros, HASH_LARGE);
682 * Pop the context stack.
684 static void ctx_pop(void)
686 Context *c = cstk;
688 cstk = cstk->next;
689 free_smacro_table(&c->localmac);
690 nasm_free(c->name);
691 nasm_free(c);
695 * Search for a key in the hash index; adding it if necessary
696 * (in which case we initialize the data pointer to NULL.)
698 static void **
699 hash_findi_add(struct hash_table *hash, const char *str)
701 struct hash_insert hi;
702 void **r;
703 char *strx;
705 r = hash_findi(hash, str, &hi);
706 if (r)
707 return r;
709 strx = nasm_strdup(str); /* Use a more efficient allocator here? */
710 return hash_add(&hi, strx, NULL);
714 * Like hash_findi, but returns the data element rather than a pointer
715 * to it. Used only when not adding a new element, hence no third
716 * argument.
718 static void *
719 hash_findix(struct hash_table *hash, const char *str)
721 void **p;
723 p = hash_findi(hash, str, NULL);
724 return p ? *p : NULL;
728 * read line from standart macros set,
729 * if there no more left -- return NULL
731 static char *line_from_stdmac(void)
733 unsigned char c;
734 const unsigned char *p = stdmacpos;
735 char *line, *q;
736 size_t len = 0;
738 if (!stdmacpos)
739 return NULL;
741 while ((c = *p++)) {
742 if (c >= 0x80)
743 len += pp_directives_len[c - 0x80] + 1;
744 else
745 len++;
748 line = nasm_malloc(len + 1);
749 q = line;
750 while ((c = *stdmacpos++)) {
751 if (c >= 0x80) {
752 memcpy(q, pp_directives[c - 0x80], pp_directives_len[c - 0x80]);
753 q += pp_directives_len[c - 0x80];
754 *q++ = ' ';
755 } else {
756 *q++ = c;
759 stdmacpos = p;
760 *q = '\0';
762 if (!*stdmacpos) {
763 /* This was the last of the standard macro chain... */
764 stdmacpos = NULL;
765 if (any_extrastdmac) {
766 stdmacpos = extrastdmac;
767 any_extrastdmac = false;
768 } else if (do_predef) {
769 Line *pd, *l;
770 Token *head, **tail, *t;
773 * Nasty hack: here we push the contents of
774 * `predef' on to the top-level expansion stack,
775 * since this is the most convenient way to
776 * implement the pre-include and pre-define
777 * features.
779 list_for_each(pd, predef) {
780 head = NULL;
781 tail = &head;
782 list_for_each(t, pd->first) {
783 *tail = new_Token(NULL, t->type, t->text, 0);
784 tail = &(*tail)->next;
787 l = nasm_malloc(sizeof(Line));
788 l->next = istk->expansion;
789 l->first = head;
790 l->finishes = NULL;
792 istk->expansion = l;
794 do_predef = false;
798 return line;
801 static char *read_line(void)
803 unsigned int size, c, next;
804 const unsigned int delta = 512;
805 const unsigned int pad = 8;
806 unsigned int nr_cont = 0;
807 bool cont = false;
808 char *buffer, *p;
810 /* Standart macros set (predefined) goes first */
811 p = line_from_stdmac();
812 if (p)
813 return p;
815 size = delta;
816 p = buffer = nasm_malloc(size);
818 for (;;) {
819 c = fgetc(istk->fp);
820 if ((int)(c) == EOF) {
821 p[0] = 0;
822 break;
825 switch (c) {
826 case '\r':
827 next = fgetc(istk->fp);
828 if (next != '\n')
829 ungetc(next, istk->fp);
830 if (cont) {
831 cont = false;
832 continue;
834 break;
836 case '\n':
837 if (cont) {
838 cont = false;
839 continue;
841 break;
843 case '\\':
844 next = fgetc(istk->fp);
845 ungetc(next, istk->fp);
846 if (next == '\r' || next == '\n') {
847 cont = true;
848 nr_cont++;
849 continue;
851 break;
854 if (c == '\r' || c == '\n') {
855 *p++ = 0;
856 break;
859 if (p >= (buffer + size - pad)) {
860 buffer = nasm_realloc(buffer, size + delta);
861 p = buffer + size - pad;
862 size += delta;
865 *p++ = (unsigned char)c;
868 if (p == buffer) {
869 nasm_free(buffer);
870 return NULL;
873 src_set_linnum(src_get_linnum() + istk->lineinc +
874 (nr_cont * istk->lineinc));
877 * Handle spurious ^Z, which may be inserted into source files
878 * by some file transfer utilities.
880 buffer[strcspn(buffer, "\032")] = '\0';
882 lfmt->line(LIST_READ, buffer);
884 return buffer;
888 * Tokenize a line of text. This is a very simple process since we
889 * don't need to parse the value out of e.g. numeric tokens: we
890 * simply split one string into many.
892 static Token *tokenize(char *line)
894 char c, *p = line;
895 enum pp_token_type type;
896 Token *list = NULL;
897 Token *t, **tail = &list;
899 while (*line) {
900 p = line;
901 if (*p == '%') {
902 p++;
903 if (*p == '+' && !nasm_isdigit(p[1])) {
904 p++;
905 type = TOK_PASTE;
906 } else if (nasm_isdigit(*p) ||
907 ((*p == '-' || *p == '+') && nasm_isdigit(p[1]))) {
908 do {
909 p++;
911 while (nasm_isdigit(*p));
912 type = TOK_PREPROC_ID;
913 } else if (*p == '{') {
914 p++;
915 while (*p) {
916 if (*p == '}')
917 break;
918 p[-1] = *p;
919 p++;
921 if (*p != '}')
922 nasm_error(ERR_WARNING | ERR_PASS1,
923 "unterminated %%{ construct");
924 p[-1] = '\0';
925 if (*p)
926 p++;
927 type = TOK_PREPROC_ID;
928 } else if (*p == '[') {
929 int lvl = 1;
930 line += 2; /* Skip the leading %[ */
931 p++;
932 while (lvl && (c = *p++)) {
933 switch (c) {
934 case ']':
935 lvl--;
936 break;
937 case '%':
938 if (*p == '[')
939 lvl++;
940 break;
941 case '\'':
942 case '\"':
943 case '`':
944 p = nasm_skip_string(p - 1) + 1;
945 break;
946 default:
947 break;
950 p--;
951 if (*p)
952 *p++ = '\0';
953 if (lvl)
954 nasm_error(ERR_NONFATAL|ERR_PASS1,
955 "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 nasm_error(ERR_NONFATAL|ERR_PASS1,
978 "unterminated %%! string");
979 } else {
980 /* %! without string or identifier */
981 type = TOK_OTHER; /* Legacy behavior... */
983 } else if (isidchar(*p) ||
984 ((*p == '!' || *p == '%' || *p == '$') &&
985 isidchar(p[1]))) {
986 do {
987 p++;
989 while (isidchar(*p));
990 type = TOK_PREPROC_ID;
991 } else {
992 type = TOK_OTHER;
993 if (*p == '%')
994 p++;
996 } else if (isidstart(*p) || (*p == '$' && isidstart(p[1]))) {
997 type = TOK_ID;
998 p++;
999 while (*p && isidchar(*p))
1000 p++;
1001 } else if (*p == '\'' || *p == '"' || *p == '`') {
1003 * A string token.
1005 type = TOK_STRING;
1006 p = nasm_skip_string(p);
1008 if (*p) {
1009 p++;
1010 } else {
1011 nasm_error(ERR_WARNING|ERR_PASS1, "unterminated string");
1012 /* Handling unterminated strings by UNV */
1013 /* type = -1; */
1015 } else if (p[0] == '$' && p[1] == '$') {
1016 type = TOK_OTHER; /* TOKEN_BASE */
1017 p += 2;
1018 } else if (isnumstart(*p)) {
1019 bool is_hex = false;
1020 bool is_float = false;
1021 bool has_e = false;
1022 char c, *r;
1025 * A numeric token.
1028 if (*p == '$') {
1029 p++;
1030 is_hex = true;
1033 for (;;) {
1034 c = *p++;
1036 if (!is_hex && (c == 'e' || c == 'E')) {
1037 has_e = true;
1038 if (*p == '+' || *p == '-') {
1040 * e can only be followed by +/- if it is either a
1041 * prefixed hex number or a floating-point number
1043 p++;
1044 is_float = true;
1046 } else if (c == 'H' || c == 'h' || c == 'X' || c == 'x') {
1047 is_hex = true;
1048 } else if (c == 'P' || c == 'p') {
1049 is_float = true;
1050 if (*p == '+' || *p == '-')
1051 p++;
1052 } else if (isnumchar(c) || c == '_')
1053 ; /* just advance */
1054 else if (c == '.') {
1056 * we need to deal with consequences of the legacy
1057 * parser, like "1.nolist" being two tokens
1058 * (TOK_NUMBER, TOK_ID) here; at least give it
1059 * a shot for now. In the future, we probably need
1060 * a flex-based scanner with proper pattern matching
1061 * to do it as well as it can be done. Nothing in
1062 * the world is going to help the person who wants
1063 * 0x123.p16 interpreted as two tokens, though.
1065 r = p;
1066 while (*r == '_')
1067 r++;
1069 if (nasm_isdigit(*r) || (is_hex && nasm_isxdigit(*r)) ||
1070 (!is_hex && (*r == 'e' || *r == 'E')) ||
1071 (*r == 'p' || *r == 'P')) {
1072 p = r;
1073 is_float = true;
1074 } else
1075 break; /* Terminate the token */
1076 } else
1077 break;
1079 p--; /* Point to first character beyond number */
1081 if (p == line+1 && *line == '$') {
1082 type = TOK_OTHER; /* TOKEN_HERE */
1083 } else {
1084 if (has_e && !is_hex) {
1085 /* 1e13 is floating-point, but 1e13h is not */
1086 is_float = true;
1089 type = is_float ? TOK_FLOAT : TOK_NUMBER;
1091 } else if (nasm_isspace(*p)) {
1092 type = TOK_WHITESPACE;
1093 p = nasm_skip_spaces(p);
1095 * Whitespace just before end-of-line is discarded by
1096 * pretending it's a comment; whitespace just before a
1097 * comment gets lumped into the comment.
1099 if (!*p || *p == ';') {
1100 type = TOK_COMMENT;
1101 while (*p)
1102 p++;
1104 } else if (*p == ';') {
1105 type = TOK_COMMENT;
1106 while (*p)
1107 p++;
1108 } else {
1110 * Anything else is an operator of some kind. We check
1111 * for all the double-character operators (>>, <<, //,
1112 * %%, <=, >=, ==, !=, <>, &&, ||, ^^), but anything
1113 * else is a single-character operator.
1115 type = TOK_OTHER;
1116 if ((p[0] == '>' && p[1] == '>') ||
1117 (p[0] == '<' && p[1] == '<') ||
1118 (p[0] == '/' && p[1] == '/') ||
1119 (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++;
1129 p++;
1132 /* Handling unterminated string by UNV */
1133 /*if (type == -1)
1135 *tail = t = new_Token(NULL, TOK_STRING, line, p-line+1);
1136 t->text[p-line] = *line;
1137 tail = &t->next;
1139 else */
1140 if (type != TOK_COMMENT) {
1141 *tail = t = new_Token(NULL, type, line, p - line);
1142 tail = &t->next;
1144 line = p;
1146 return list;
1150 * this function allocates a new managed block of memory and
1151 * returns a pointer to the block. The managed blocks are
1152 * deleted only all at once by the delete_Blocks function.
1154 static void *new_Block(size_t size)
1156 Blocks *b = &blocks;
1158 /* first, get to the end of the linked list */
1159 while (b->next)
1160 b = b->next;
1161 /* now allocate the requested chunk */
1162 b->chunk = nasm_malloc(size);
1164 /* now allocate a new block for the next request */
1165 b->next = nasm_zalloc(sizeof(Blocks));
1166 return b->chunk;
1170 * this function deletes all managed blocks of memory
1172 static void delete_Blocks(void)
1174 Blocks *a, *b = &blocks;
1177 * keep in mind that the first block, pointed to by blocks
1178 * is a static and not dynamically allocated, so we don't
1179 * free it.
1181 while (b) {
1182 if (b->chunk)
1183 nasm_free(b->chunk);
1184 a = b;
1185 b = b->next;
1186 if (a != &blocks)
1187 nasm_free(a);
1189 memset(&blocks, 0, sizeof(blocks));
1193 * this function creates a new Token and passes a pointer to it
1194 * back to the caller. It sets the type and text elements, and
1195 * also the a.mac and next elements to NULL.
1197 static Token *new_Token(Token * next, enum pp_token_type type,
1198 const char *text, int txtlen)
1200 Token *t;
1201 int i;
1203 if (!freeTokens) {
1204 freeTokens = (Token *) new_Block(TOKEN_BLOCKSIZE * sizeof(Token));
1205 for (i = 0; i < TOKEN_BLOCKSIZE - 1; i++)
1206 freeTokens[i].next = &freeTokens[i + 1];
1207 freeTokens[i].next = NULL;
1209 t = freeTokens;
1210 freeTokens = t->next;
1211 t->next = next;
1212 t->a.mac = NULL;
1213 t->type = type;
1214 if (type == TOK_WHITESPACE || !text) {
1215 t->text = NULL;
1216 } else {
1217 if (txtlen == 0)
1218 txtlen = strlen(text);
1219 t->text = nasm_malloc(txtlen+1);
1220 memcpy(t->text, text, txtlen);
1221 t->text[txtlen] = '\0';
1223 return t;
1226 static Token *delete_Token(Token * t)
1228 Token *next = t->next;
1229 nasm_free(t->text);
1230 t->next = freeTokens;
1231 freeTokens = t;
1232 return next;
1236 * Convert a line of tokens back into text.
1237 * If expand_locals is not zero, identifiers of the form "%$*xxx"
1238 * will be transformed into ..@ctxnum.xxx
1240 static char *detoken(Token * tlist, bool expand_locals)
1242 Token *t;
1243 char *line, *p;
1244 const char *q;
1245 int len = 0;
1247 list_for_each(t, tlist) {
1248 if (t->type == TOK_PREPROC_ID && t->text[1] == '!') {
1249 char *v;
1250 char *q = t->text;
1252 v = t->text + 2;
1253 if (*v == '\'' || *v == '\"' || *v == '`') {
1254 size_t len = nasm_unquote(v, NULL);
1255 size_t clen = strlen(v);
1257 if (len != clen) {
1258 nasm_error(ERR_NONFATAL | ERR_PASS1,
1259 "NUL character in %%! string");
1260 v = NULL;
1264 if (v) {
1265 char *p = getenv(v);
1266 if (!p) {
1267 nasm_error(ERR_NONFATAL | ERR_PASS1,
1268 "nonexistent environment variable `%s'", v);
1269 p = "";
1271 t->text = nasm_strdup(p);
1273 nasm_free(q);
1276 /* Expand local macros here and not during preprocessing */
1277 if (expand_locals &&
1278 t->type == TOK_PREPROC_ID && t->text &&
1279 t->text[0] == '%' && t->text[1] == '$') {
1280 const char *q;
1281 char *p;
1282 Context *ctx = get_ctx(t->text, &q);
1283 if (ctx) {
1284 char buffer[40];
1285 snprintf(buffer, sizeof(buffer), "..@%"PRIu32".", ctx->number);
1286 p = nasm_strcat(buffer, q);
1287 nasm_free(t->text);
1288 t->text = p;
1291 if (t->type == TOK_WHITESPACE)
1292 len++;
1293 else if (t->text)
1294 len += strlen(t->text);
1297 p = line = nasm_malloc(len + 1);
1299 list_for_each(t, tlist) {
1300 if (t->type == TOK_WHITESPACE) {
1301 *p++ = ' ';
1302 } else if (t->text) {
1303 q = t->text;
1304 while (*q)
1305 *p++ = *q++;
1308 *p = '\0';
1310 return line;
1314 * A scanner, suitable for use by the expression evaluator, which
1315 * operates on a line of Tokens. Expects a pointer to a pointer to
1316 * the first token in the line to be passed in as its private_data
1317 * field.
1319 * FIX: This really needs to be unified with stdscan.
1321 static int ppscan(void *private_data, struct tokenval *tokval)
1323 Token **tlineptr = private_data;
1324 Token *tline;
1325 char ourcopy[MAX_KEYWORD+1], *p, *r, *s;
1327 do {
1328 tline = *tlineptr;
1329 *tlineptr = tline ? tline->next : NULL;
1330 } while (tline && (tline->type == TOK_WHITESPACE ||
1331 tline->type == TOK_COMMENT));
1333 if (!tline)
1334 return tokval->t_type = TOKEN_EOS;
1336 tokval->t_charptr = tline->text;
1338 if (tline->text[0] == '$' && !tline->text[1])
1339 return tokval->t_type = TOKEN_HERE;
1340 if (tline->text[0] == '$' && tline->text[1] == '$' && !tline->text[2])
1341 return tokval->t_type = TOKEN_BASE;
1343 if (tline->type == TOK_ID) {
1344 p = tokval->t_charptr = tline->text;
1345 if (p[0] == '$') {
1346 tokval->t_charptr++;
1347 return tokval->t_type = TOKEN_ID;
1350 for (r = p, s = ourcopy; *r; r++) {
1351 if (r >= p+MAX_KEYWORD)
1352 return tokval->t_type = TOKEN_ID; /* Not a keyword */
1353 *s++ = nasm_tolower(*r);
1355 *s = '\0';
1356 /* right, so we have an identifier sitting in temp storage. now,
1357 * is it actually a register or instruction name, or what? */
1358 return nasm_token_hash(ourcopy, tokval);
1361 if (tline->type == TOK_NUMBER) {
1362 bool rn_error;
1363 tokval->t_integer = readnum(tline->text, &rn_error);
1364 tokval->t_charptr = tline->text;
1365 if (rn_error)
1366 return tokval->t_type = TOKEN_ERRNUM;
1367 else
1368 return tokval->t_type = TOKEN_NUM;
1371 if (tline->type == TOK_FLOAT) {
1372 return tokval->t_type = TOKEN_FLOAT;
1375 if (tline->type == TOK_STRING) {
1376 char bq, *ep;
1378 bq = tline->text[0];
1379 tokval->t_charptr = tline->text;
1380 tokval->t_inttwo = nasm_unquote(tline->text, &ep);
1382 if (ep[0] != bq || ep[1] != '\0')
1383 return tokval->t_type = TOKEN_ERRSTR;
1384 else
1385 return tokval->t_type = TOKEN_STR;
1388 if (tline->type == TOK_OTHER) {
1389 if (!strcmp(tline->text, "<<"))
1390 return tokval->t_type = TOKEN_SHL;
1391 if (!strcmp(tline->text, ">>"))
1392 return tokval->t_type = TOKEN_SHR;
1393 if (!strcmp(tline->text, "//"))
1394 return tokval->t_type = TOKEN_SDIV;
1395 if (!strcmp(tline->text, "%%"))
1396 return tokval->t_type = TOKEN_SMOD;
1397 if (!strcmp(tline->text, "=="))
1398 return tokval->t_type = TOKEN_EQ;
1399 if (!strcmp(tline->text, "<>"))
1400 return tokval->t_type = TOKEN_NE;
1401 if (!strcmp(tline->text, "!="))
1402 return tokval->t_type = TOKEN_NE;
1403 if (!strcmp(tline->text, "<="))
1404 return tokval->t_type = TOKEN_LE;
1405 if (!strcmp(tline->text, ">="))
1406 return tokval->t_type = TOKEN_GE;
1407 if (!strcmp(tline->text, "&&"))
1408 return tokval->t_type = TOKEN_DBL_AND;
1409 if (!strcmp(tline->text, "^^"))
1410 return tokval->t_type = TOKEN_DBL_XOR;
1411 if (!strcmp(tline->text, "||"))
1412 return tokval->t_type = TOKEN_DBL_OR;
1416 * We have no other options: just return the first character of
1417 * the token text.
1419 return tokval->t_type = tline->text[0];
1423 * Compare a string to the name of an existing macro; this is a
1424 * simple wrapper which calls either strcmp or nasm_stricmp
1425 * depending on the value of the `casesense' parameter.
1427 static int mstrcmp(const char *p, const char *q, bool casesense)
1429 return casesense ? strcmp(p, q) : nasm_stricmp(p, q);
1433 * Compare a string to the name of an existing macro; this is a
1434 * simple wrapper which calls either strcmp or nasm_stricmp
1435 * depending on the value of the `casesense' parameter.
1437 static int mmemcmp(const char *p, const char *q, size_t l, bool casesense)
1439 return casesense ? memcmp(p, q, l) : nasm_memicmp(p, q, l);
1443 * Return the Context structure associated with a %$ token. Return
1444 * NULL, having _already_ reported an error condition, if the
1445 * context stack isn't deep enough for the supplied number of $
1446 * signs.
1448 * If "namep" is non-NULL, set it to the pointer to the macro name
1449 * tail, i.e. the part beyond %$...
1451 static Context *get_ctx(const char *name, const char **namep)
1453 Context *ctx;
1454 int i;
1456 if (namep)
1457 *namep = name;
1459 if (!name || name[0] != '%' || name[1] != '$')
1460 return NULL;
1462 if (!cstk) {
1463 nasm_error(ERR_NONFATAL, "`%s': context stack is empty", name);
1464 return NULL;
1467 name += 2;
1468 ctx = cstk;
1469 i = 0;
1470 while (ctx && *name == '$') {
1471 name++;
1472 i++;
1473 ctx = ctx->next;
1475 if (!ctx) {
1476 nasm_error(ERR_NONFATAL, "`%s': context stack is only"
1477 " %d level%s deep", name, i, (i == 1 ? "" : "s"));
1478 return NULL;
1481 if (namep)
1482 *namep = name;
1484 return ctx;
1488 * Check to see if a file is already in a string list
1490 static bool in_list(const StrList *list, const char *str)
1492 while (list) {
1493 if (!strcmp(list->str, str))
1494 return true;
1495 list = list->next;
1497 return false;
1501 * Open an include file. This routine must always return a valid
1502 * file pointer if it returns - it's responsible for throwing an
1503 * ERR_FATAL and bombing out completely if not. It should also try
1504 * the include path one by one until it finds the file or reaches
1505 * the end of the path.
1507 static FILE *inc_fopen(const char *file, StrList **dhead, StrList ***dtail,
1508 bool missing_ok)
1510 FILE *fp;
1511 char *prefix = "";
1512 IncPath *ip = ipath;
1513 int len = strlen(file);
1514 size_t prefix_len = 0;
1515 StrList *sl;
1517 while (1) {
1518 sl = nasm_malloc(prefix_len+len+1+sizeof sl->next);
1519 memcpy(sl->str, prefix, prefix_len);
1520 memcpy(sl->str+prefix_len, file, len+1);
1521 fp = fopen(sl->str, "r");
1522 if (fp && dhead && !in_list(*dhead, sl->str)) {
1523 sl->next = NULL;
1524 **dtail = sl;
1525 *dtail = &sl->next;
1526 } else {
1527 nasm_free(sl);
1529 if (fp)
1530 return fp;
1531 if (!ip) {
1532 if (!missing_ok)
1533 break;
1534 prefix = NULL;
1535 } else {
1536 prefix = ip->path;
1537 ip = ip->next;
1539 if (prefix) {
1540 prefix_len = strlen(prefix);
1541 } else {
1542 /* -MG given and file not found */
1543 if (dhead && !in_list(*dhead, file)) {
1544 sl = nasm_malloc(len+1+sizeof sl->next);
1545 sl->next = NULL;
1546 strcpy(sl->str, file);
1547 **dtail = sl;
1548 *dtail = &sl->next;
1550 return NULL;
1554 nasm_error(ERR_FATAL, "unable to open include file `%s'", file);
1555 return NULL;
1559 * Determine if we should warn on defining a single-line macro of
1560 * name `name', with `nparam' parameters. If nparam is 0 or -1, will
1561 * return true if _any_ single-line macro of that name is defined.
1562 * Otherwise, will return true if a single-line macro with either
1563 * `nparam' or no parameters is defined.
1565 * If a macro with precisely the right number of parameters is
1566 * defined, or nparam is -1, the address of the definition structure
1567 * will be returned in `defn'; otherwise NULL will be returned. If `defn'
1568 * is NULL, no action will be taken regarding its contents, and no
1569 * error will occur.
1571 * Note that this is also called with nparam zero to resolve
1572 * `ifdef'.
1574 * If you already know which context macro belongs to, you can pass
1575 * the context pointer as first parameter; if you won't but name begins
1576 * with %$ the context will be automatically computed. If all_contexts
1577 * is true, macro will be searched in outer contexts as well.
1579 static bool
1580 smacro_defined(Context * ctx, const char *name, int nparam, SMacro ** defn,
1581 bool nocase)
1583 struct hash_table *smtbl;
1584 SMacro *m;
1586 if (ctx) {
1587 smtbl = &ctx->localmac;
1588 } else if (name[0] == '%' && name[1] == '$') {
1589 if (cstk)
1590 ctx = get_ctx(name, &name);
1591 if (!ctx)
1592 return false; /* got to return _something_ */
1593 smtbl = &ctx->localmac;
1594 } else {
1595 smtbl = &smacros;
1597 m = (SMacro *) hash_findix(smtbl, name);
1599 while (m) {
1600 if (!mstrcmp(m->name, name, m->casesense && nocase) &&
1601 (nparam <= 0 || m->nparam == 0 || nparam == (int) m->nparam)) {
1602 if (defn) {
1603 if (nparam == (int) m->nparam || nparam == -1)
1604 *defn = m;
1605 else
1606 *defn = NULL;
1608 return true;
1610 m = m->next;
1613 return false;
1617 * Count and mark off the parameters in a multi-line macro call.
1618 * This is called both from within the multi-line macro expansion
1619 * code, and also to mark off the default parameters when provided
1620 * in a %macro definition line.
1622 static void count_mmac_params(Token * t, int *nparam, Token *** params)
1624 int paramsize, brace;
1626 *nparam = paramsize = 0;
1627 *params = NULL;
1628 while (t) {
1629 /* +1: we need space for the final NULL */
1630 if (*nparam+1 >= paramsize) {
1631 paramsize += PARAM_DELTA;
1632 *params = nasm_realloc(*params, sizeof(**params) * paramsize);
1634 skip_white_(t);
1635 brace = 0;
1636 if (tok_is_(t, "{"))
1637 brace++;
1638 (*params)[(*nparam)++] = t;
1639 if (brace) {
1640 while (brace && (t = t->next) != NULL) {
1641 if (tok_is_(t, "{"))
1642 brace++;
1643 else if (tok_is_(t, "}"))
1644 brace--;
1647 if (t) {
1649 * Now we've found the closing brace, look further
1650 * for the comma.
1652 t = t->next;
1653 skip_white_(t);
1654 if (tok_isnt_(t, ",")) {
1655 nasm_error(ERR_NONFATAL,
1656 "braces do not enclose all of macro parameter");
1657 while (tok_isnt_(t, ","))
1658 t = t->next;
1661 } else {
1662 while (tok_isnt_(t, ","))
1663 t = t->next;
1665 if (t) { /* got a comma/brace */
1666 t = t->next; /* eat the comma */
1672 * Determine whether one of the various `if' conditions is true or
1673 * not.
1675 * We must free the tline we get passed.
1677 static bool if_condition(Token * tline, enum preproc_token ct)
1679 enum pp_conditional i = PP_COND(ct);
1680 bool j;
1681 Token *t, *tt, **tptr, *origline;
1682 struct tokenval tokval;
1683 expr *evalresult;
1684 enum pp_token_type needtype;
1685 char *p;
1687 origline = tline;
1689 switch (i) {
1690 case PPC_IFCTX:
1691 j = false; /* have we matched yet? */
1692 while (true) {
1693 skip_white_(tline);
1694 if (!tline)
1695 break;
1696 if (tline->type != TOK_ID) {
1697 nasm_error(ERR_NONFATAL,
1698 "`%s' expects context identifiers", pp_directives[ct]);
1699 free_tlist(origline);
1700 return -1;
1702 if (cstk && cstk->name && !nasm_stricmp(tline->text, cstk->name))
1703 j = true;
1704 tline = tline->next;
1706 break;
1708 case PPC_IFDEF:
1709 j = false; /* have we matched yet? */
1710 while (tline) {
1711 skip_white_(tline);
1712 if (!tline || (tline->type != TOK_ID &&
1713 (tline->type != TOK_PREPROC_ID ||
1714 tline->text[1] != '$'))) {
1715 nasm_error(ERR_NONFATAL,
1716 "`%s' expects macro identifiers", pp_directives[ct]);
1717 goto fail;
1719 if (smacro_defined(NULL, tline->text, 0, NULL, true))
1720 j = true;
1721 tline = tline->next;
1723 break;
1725 case PPC_IFENV:
1726 tline = expand_smacro(tline);
1727 j = false; /* have we matched yet? */
1728 while (tline) {
1729 skip_white_(tline);
1730 if (!tline || (tline->type != TOK_ID &&
1731 tline->type != TOK_STRING &&
1732 (tline->type != TOK_PREPROC_ID ||
1733 tline->text[1] != '!'))) {
1734 nasm_error(ERR_NONFATAL,
1735 "`%s' expects environment variable names",
1736 pp_directives[ct]);
1737 goto fail;
1739 p = tline->text;
1740 if (tline->type == TOK_PREPROC_ID)
1741 p += 2; /* Skip leading %! */
1742 if (*p == '\'' || *p == '\"' || *p == '`')
1743 nasm_unquote_cstr(p, ct);
1744 if (getenv(p))
1745 j = true;
1746 tline = tline->next;
1748 break;
1750 case PPC_IFIDN:
1751 case PPC_IFIDNI:
1752 tline = expand_smacro(tline);
1753 t = tt = tline;
1754 while (tok_isnt_(tt, ","))
1755 tt = tt->next;
1756 if (!tt) {
1757 nasm_error(ERR_NONFATAL,
1758 "`%s' expects two comma-separated arguments",
1759 pp_directives[ct]);
1760 goto fail;
1762 tt = tt->next;
1763 j = true; /* assume equality unless proved not */
1764 while ((t->type != TOK_OTHER || strcmp(t->text, ",")) && tt) {
1765 if (tt->type == TOK_OTHER && !strcmp(tt->text, ",")) {
1766 nasm_error(ERR_NONFATAL, "`%s': more than one comma on line",
1767 pp_directives[ct]);
1768 goto fail;
1770 if (t->type == TOK_WHITESPACE) {
1771 t = t->next;
1772 continue;
1774 if (tt->type == TOK_WHITESPACE) {
1775 tt = tt->next;
1776 continue;
1778 if (tt->type != t->type) {
1779 j = false; /* found mismatching tokens */
1780 break;
1782 /* When comparing strings, need to unquote them first */
1783 if (t->type == TOK_STRING) {
1784 size_t l1 = nasm_unquote(t->text, NULL);
1785 size_t l2 = nasm_unquote(tt->text, NULL);
1787 if (l1 != l2) {
1788 j = false;
1789 break;
1791 if (mmemcmp(t->text, tt->text, l1, i == PPC_IFIDN)) {
1792 j = false;
1793 break;
1795 } else if (mstrcmp(tt->text, t->text, i == PPC_IFIDN) != 0) {
1796 j = false; /* found mismatching tokens */
1797 break;
1800 t = t->next;
1801 tt = tt->next;
1803 if ((t->type != TOK_OTHER || strcmp(t->text, ",")) || tt)
1804 j = false; /* trailing gunk on one end or other */
1805 break;
1807 case PPC_IFMACRO:
1809 bool found = false;
1810 MMacro searching, *mmac;
1812 skip_white_(tline);
1813 tline = expand_id(tline);
1814 if (!tok_type_(tline, TOK_ID)) {
1815 nasm_error(ERR_NONFATAL,
1816 "`%s' expects a macro name", pp_directives[ct]);
1817 goto fail;
1819 searching.name = nasm_strdup(tline->text);
1820 searching.casesense = true;
1821 searching.plus = false;
1822 searching.nolist = false;
1823 searching.in_progress = 0;
1824 searching.max_depth = 0;
1825 searching.rep_nest = NULL;
1826 searching.nparam_min = 0;
1827 searching.nparam_max = INT_MAX;
1828 tline = expand_smacro(tline->next);
1829 skip_white_(tline);
1830 if (!tline) {
1831 } else if (!tok_type_(tline, TOK_NUMBER)) {
1832 nasm_error(ERR_NONFATAL,
1833 "`%s' expects a parameter count or nothing",
1834 pp_directives[ct]);
1835 } else {
1836 searching.nparam_min = searching.nparam_max =
1837 readnum(tline->text, &j);
1838 if (j)
1839 nasm_error(ERR_NONFATAL,
1840 "unable to parse parameter count `%s'",
1841 tline->text);
1843 if (tline && tok_is_(tline->next, "-")) {
1844 tline = tline->next->next;
1845 if (tok_is_(tline, "*"))
1846 searching.nparam_max = INT_MAX;
1847 else if (!tok_type_(tline, TOK_NUMBER))
1848 nasm_error(ERR_NONFATAL,
1849 "`%s' expects a parameter count after `-'",
1850 pp_directives[ct]);
1851 else {
1852 searching.nparam_max = readnum(tline->text, &j);
1853 if (j)
1854 nasm_error(ERR_NONFATAL,
1855 "unable to parse parameter count `%s'",
1856 tline->text);
1857 if (searching.nparam_min > searching.nparam_max)
1858 nasm_error(ERR_NONFATAL,
1859 "minimum parameter count exceeds maximum");
1862 if (tline && tok_is_(tline->next, "+")) {
1863 tline = tline->next;
1864 searching.plus = true;
1866 mmac = (MMacro *) hash_findix(&mmacros, searching.name);
1867 while (mmac) {
1868 if (!strcmp(mmac->name, searching.name) &&
1869 (mmac->nparam_min <= searching.nparam_max
1870 || searching.plus)
1871 && (searching.nparam_min <= mmac->nparam_max
1872 || mmac->plus)) {
1873 found = true;
1874 break;
1876 mmac = mmac->next;
1878 if (tline && tline->next)
1879 nasm_error(ERR_WARNING|ERR_PASS1,
1880 "trailing garbage after %%ifmacro ignored");
1881 nasm_free(searching.name);
1882 j = found;
1883 break;
1886 case PPC_IFID:
1887 needtype = TOK_ID;
1888 goto iftype;
1889 case PPC_IFNUM:
1890 needtype = TOK_NUMBER;
1891 goto iftype;
1892 case PPC_IFSTR:
1893 needtype = TOK_STRING;
1894 goto iftype;
1896 iftype:
1897 t = tline = expand_smacro(tline);
1899 while (tok_type_(t, TOK_WHITESPACE) ||
1900 (needtype == TOK_NUMBER &&
1901 tok_type_(t, TOK_OTHER) &&
1902 (t->text[0] == '-' || t->text[0] == '+') &&
1903 !t->text[1]))
1904 t = t->next;
1906 j = tok_type_(t, needtype);
1907 break;
1909 case PPC_IFTOKEN:
1910 t = tline = expand_smacro(tline);
1911 while (tok_type_(t, TOK_WHITESPACE))
1912 t = t->next;
1914 j = false;
1915 if (t) {
1916 t = t->next; /* Skip the actual token */
1917 while (tok_type_(t, TOK_WHITESPACE))
1918 t = t->next;
1919 j = !t; /* Should be nothing left */
1921 break;
1923 case PPC_IFEMPTY:
1924 t = tline = expand_smacro(tline);
1925 while (tok_type_(t, TOK_WHITESPACE))
1926 t = t->next;
1928 j = !t; /* Should be empty */
1929 break;
1931 case PPC_IF:
1932 t = tline = expand_smacro(tline);
1933 tptr = &t;
1934 tokval.t_type = TOKEN_INVALID;
1935 evalresult = evaluate(ppscan, tptr, &tokval,
1936 NULL, pass | CRITICAL, NULL);
1937 if (!evalresult)
1938 return -1;
1939 if (tokval.t_type)
1940 nasm_error(ERR_WARNING|ERR_PASS1,
1941 "trailing garbage after expression ignored");
1942 if (!is_simple(evalresult)) {
1943 nasm_error(ERR_NONFATAL,
1944 "non-constant value given to `%s'", pp_directives[ct]);
1945 goto fail;
1947 j = reloc_value(evalresult) != 0;
1948 break;
1950 default:
1951 nasm_error(ERR_FATAL,
1952 "preprocessor directive `%s' not yet implemented",
1953 pp_directives[ct]);
1954 goto fail;
1957 free_tlist(origline);
1958 return j ^ PP_NEGATIVE(ct);
1960 fail:
1961 free_tlist(origline);
1962 return -1;
1966 * Common code for defining an smacro
1968 static bool define_smacro(Context *ctx, const char *mname, bool casesense,
1969 int nparam, Token *expansion)
1971 SMacro *smac, **smhead;
1972 struct hash_table *smtbl;
1974 if (smacro_defined(ctx, mname, nparam, &smac, casesense)) {
1975 if (!smac) {
1976 nasm_error(ERR_WARNING|ERR_PASS1,
1977 "single-line macro `%s' defined both with and"
1978 " without parameters", mname);
1980 * Some instances of the old code considered this a failure,
1981 * some others didn't. What is the right thing to do here?
1983 free_tlist(expansion);
1984 return false; /* Failure */
1985 } else {
1987 * We're redefining, so we have to take over an
1988 * existing SMacro structure. This means freeing
1989 * what was already in it.
1991 nasm_free(smac->name);
1992 free_tlist(smac->expansion);
1994 } else {
1995 smtbl = ctx ? &ctx->localmac : &smacros;
1996 smhead = (SMacro **) hash_findi_add(smtbl, mname);
1997 smac = nasm_malloc(sizeof(SMacro));
1998 smac->next = *smhead;
1999 *smhead = smac;
2001 smac->name = nasm_strdup(mname);
2002 smac->casesense = casesense;
2003 smac->nparam = nparam;
2004 smac->expansion = expansion;
2005 smac->in_progress = false;
2006 return true; /* Success */
2010 * Undefine an smacro
2012 static void undef_smacro(Context *ctx, const char *mname)
2014 SMacro **smhead, *s, **sp;
2015 struct hash_table *smtbl;
2017 smtbl = ctx ? &ctx->localmac : &smacros;
2018 smhead = (SMacro **)hash_findi(smtbl, mname, NULL);
2020 if (smhead) {
2022 * We now have a macro name... go hunt for it.
2024 sp = smhead;
2025 while ((s = *sp) != NULL) {
2026 if (!mstrcmp(s->name, mname, s->casesense)) {
2027 *sp = s->next;
2028 nasm_free(s->name);
2029 free_tlist(s->expansion);
2030 nasm_free(s);
2031 } else {
2032 sp = &s->next;
2039 * Parse a mmacro specification.
2041 static bool parse_mmacro_spec(Token *tline, MMacro *def, const char *directive)
2043 bool err;
2045 tline = tline->next;
2046 skip_white_(tline);
2047 tline = expand_id(tline);
2048 if (!tok_type_(tline, TOK_ID)) {
2049 nasm_error(ERR_NONFATAL, "`%s' expects a macro name", directive);
2050 return false;
2053 def->prev = NULL;
2054 def->name = nasm_strdup(tline->text);
2055 def->plus = false;
2056 def->nolist = false;
2057 def->in_progress = 0;
2058 def->rep_nest = NULL;
2059 def->nparam_min = 0;
2060 def->nparam_max = 0;
2062 tline = expand_smacro(tline->next);
2063 skip_white_(tline);
2064 if (!tok_type_(tline, TOK_NUMBER)) {
2065 nasm_error(ERR_NONFATAL, "`%s' expects a parameter count", directive);
2066 } else {
2067 def->nparam_min = def->nparam_max =
2068 readnum(tline->text, &err);
2069 if (err)
2070 nasm_error(ERR_NONFATAL,
2071 "unable to parse parameter count `%s'", tline->text);
2073 if (tline && tok_is_(tline->next, "-")) {
2074 tline = tline->next->next;
2075 if (tok_is_(tline, "*")) {
2076 def->nparam_max = INT_MAX;
2077 } else if (!tok_type_(tline, TOK_NUMBER)) {
2078 nasm_error(ERR_NONFATAL,
2079 "`%s' expects a parameter count after `-'", directive);
2080 } else {
2081 def->nparam_max = readnum(tline->text, &err);
2082 if (err) {
2083 nasm_error(ERR_NONFATAL, "unable to parse parameter count `%s'",
2084 tline->text);
2086 if (def->nparam_min > def->nparam_max) {
2087 nasm_error(ERR_NONFATAL, "minimum parameter count exceeds maximum");
2091 if (tline && tok_is_(tline->next, "+")) {
2092 tline = tline->next;
2093 def->plus = true;
2095 if (tline && tok_type_(tline->next, TOK_ID) &&
2096 !nasm_stricmp(tline->next->text, ".nolist")) {
2097 tline = tline->next;
2098 def->nolist = true;
2102 * Handle default parameters.
2104 if (tline && tline->next) {
2105 def->dlist = tline->next;
2106 tline->next = NULL;
2107 count_mmac_params(def->dlist, &def->ndefs, &def->defaults);
2108 } else {
2109 def->dlist = NULL;
2110 def->defaults = NULL;
2112 def->expansion = NULL;
2114 if (def->defaults && def->ndefs > def->nparam_max - def->nparam_min &&
2115 !def->plus)
2116 nasm_error(ERR_WARNING|ERR_PASS1|ERR_WARN_MDP,
2117 "too many default macro parameters");
2119 return true;
2124 * Decode a size directive
2126 static int parse_size(const char *str) {
2127 static const char *size_names[] =
2128 { "byte", "dword", "oword", "qword", "tword", "word", "yword" };
2129 static const int sizes[] =
2130 { 0, 1, 4, 16, 8, 10, 2, 32 };
2132 return sizes[bsii(str, size_names, ARRAY_SIZE(size_names))+1];
2136 * find and process preprocessor directive in passed line
2137 * Find out if a line contains a preprocessor directive, and deal
2138 * with it if so.
2140 * If a directive _is_ found, it is the responsibility of this routine
2141 * (and not the caller) to free_tlist() the line.
2143 * @param tline a pointer to the current tokeninzed line linked list
2144 * @return DIRECTIVE_FOUND or NO_DIRECTIVE_FOUND
2147 static int do_directive(Token * tline)
2149 enum preproc_token i;
2150 int j;
2151 bool err;
2152 int nparam;
2153 bool nolist;
2154 bool casesense;
2155 int k, m;
2156 int offset;
2157 char *p, *pp;
2158 const char *mname;
2159 Include *inc;
2160 Context *ctx;
2161 Cond *cond;
2162 MMacro *mmac, **mmhead;
2163 Token *t = NULL, *tt, *param_start, *macro_start, *last, **tptr, *origline;
2164 Line *l;
2165 struct tokenval tokval;
2166 expr *evalresult;
2167 MMacro *tmp_defining; /* Used when manipulating rep_nest */
2168 int64_t count;
2169 size_t len;
2170 int severity;
2172 origline = tline;
2174 skip_white_(tline);
2175 if (!tline || !tok_type_(tline, TOK_PREPROC_ID) ||
2176 (tline->text[1] == '%' || tline->text[1] == '$'
2177 || tline->text[1] == '!'))
2178 return NO_DIRECTIVE_FOUND;
2180 i = pp_token_hash(tline->text);
2183 * FIXME: We zap execution of PP_RMACRO, PP_IRMACRO, PP_EXITMACRO
2184 * since they are known to be buggy at moment, we need to fix them
2185 * in future release (2.09-2.10)
2187 if (i == PP_RMACRO || i == PP_IRMACRO || i == PP_EXITMACRO) {
2188 nasm_error(ERR_NONFATAL, "unknown preprocessor directive `%s'",
2189 tline->text);
2190 return NO_DIRECTIVE_FOUND;
2194 * If we're in a non-emitting branch of a condition construct,
2195 * or walking to the end of an already terminated %rep block,
2196 * we should ignore all directives except for condition
2197 * directives.
2199 if (((istk->conds && !emitting(istk->conds->state)) ||
2200 (istk->mstk && !istk->mstk->in_progress)) && !is_condition(i)) {
2201 return NO_DIRECTIVE_FOUND;
2205 * If we're defining a macro or reading a %rep block, we should
2206 * ignore all directives except for %macro/%imacro (which nest),
2207 * %endm/%endmacro, and (only if we're in a %rep block) %endrep.
2208 * If we're in a %rep block, another %rep nests, so should be let through.
2210 if (defining && i != PP_MACRO && i != PP_IMACRO &&
2211 i != PP_RMACRO && i != PP_IRMACRO &&
2212 i != PP_ENDMACRO && i != PP_ENDM &&
2213 (defining->name || (i != PP_ENDREP && i != PP_REP))) {
2214 return NO_DIRECTIVE_FOUND;
2217 if (defining) {
2218 if (i == PP_MACRO || i == PP_IMACRO ||
2219 i == PP_RMACRO || i == PP_IRMACRO) {
2220 nested_mac_count++;
2221 return NO_DIRECTIVE_FOUND;
2222 } else if (nested_mac_count > 0) {
2223 if (i == PP_ENDMACRO) {
2224 nested_mac_count--;
2225 return NO_DIRECTIVE_FOUND;
2228 if (!defining->name) {
2229 if (i == PP_REP) {
2230 nested_rep_count++;
2231 return NO_DIRECTIVE_FOUND;
2232 } else if (nested_rep_count > 0) {
2233 if (i == PP_ENDREP) {
2234 nested_rep_count--;
2235 return NO_DIRECTIVE_FOUND;
2241 switch (i) {
2242 case PP_INVALID:
2243 nasm_error(ERR_NONFATAL, "unknown preprocessor directive `%s'",
2244 tline->text);
2245 return NO_DIRECTIVE_FOUND; /* didn't get it */
2247 case PP_STACKSIZE:
2248 /* Directive to tell NASM what the default stack size is. The
2249 * default is for a 16-bit stack, and this can be overriden with
2250 * %stacksize large.
2252 tline = tline->next;
2253 if (tline && tline->type == TOK_WHITESPACE)
2254 tline = tline->next;
2255 if (!tline || tline->type != TOK_ID) {
2256 nasm_error(ERR_NONFATAL, "`%%stacksize' missing size parameter");
2257 free_tlist(origline);
2258 return DIRECTIVE_FOUND;
2260 if (nasm_stricmp(tline->text, "flat") == 0) {
2261 /* All subsequent ARG directives are for a 32-bit stack */
2262 StackSize = 4;
2263 StackPointer = "ebp";
2264 ArgOffset = 8;
2265 LocalOffset = 0;
2266 } else if (nasm_stricmp(tline->text, "flat64") == 0) {
2267 /* All subsequent ARG directives are for a 64-bit stack */
2268 StackSize = 8;
2269 StackPointer = "rbp";
2270 ArgOffset = 16;
2271 LocalOffset = 0;
2272 } else if (nasm_stricmp(tline->text, "large") == 0) {
2273 /* All subsequent ARG directives are for a 16-bit stack,
2274 * far function call.
2276 StackSize = 2;
2277 StackPointer = "bp";
2278 ArgOffset = 4;
2279 LocalOffset = 0;
2280 } else if (nasm_stricmp(tline->text, "small") == 0) {
2281 /* All subsequent ARG directives are for a 16-bit stack,
2282 * far function call. We don't support near functions.
2284 StackSize = 2;
2285 StackPointer = "bp";
2286 ArgOffset = 6;
2287 LocalOffset = 0;
2288 } else {
2289 nasm_error(ERR_NONFATAL, "`%%stacksize' invalid size type");
2290 free_tlist(origline);
2291 return DIRECTIVE_FOUND;
2293 free_tlist(origline);
2294 return DIRECTIVE_FOUND;
2296 case PP_ARG:
2297 /* TASM like ARG directive to define arguments to functions, in
2298 * the following form:
2300 * ARG arg1:WORD, arg2:DWORD, arg4:QWORD
2302 offset = ArgOffset;
2303 do {
2304 char *arg, directive[256];
2305 int size = StackSize;
2307 /* Find the argument name */
2308 tline = tline->next;
2309 if (tline && tline->type == TOK_WHITESPACE)
2310 tline = tline->next;
2311 if (!tline || tline->type != TOK_ID) {
2312 nasm_error(ERR_NONFATAL, "`%%arg' missing argument parameter");
2313 free_tlist(origline);
2314 return DIRECTIVE_FOUND;
2316 arg = tline->text;
2318 /* Find the argument size type */
2319 tline = tline->next;
2320 if (!tline || tline->type != TOK_OTHER
2321 || tline->text[0] != ':') {
2322 nasm_error(ERR_NONFATAL,
2323 "Syntax error processing `%%arg' directive");
2324 free_tlist(origline);
2325 return DIRECTIVE_FOUND;
2327 tline = tline->next;
2328 if (!tline || tline->type != TOK_ID) {
2329 nasm_error(ERR_NONFATAL, "`%%arg' missing size type parameter");
2330 free_tlist(origline);
2331 return DIRECTIVE_FOUND;
2334 /* Allow macro expansion of type parameter */
2335 tt = tokenize(tline->text);
2336 tt = expand_smacro(tt);
2337 size = parse_size(tt->text);
2338 if (!size) {
2339 nasm_error(ERR_NONFATAL,
2340 "Invalid size type for `%%arg' missing directive");
2341 free_tlist(tt);
2342 free_tlist(origline);
2343 return DIRECTIVE_FOUND;
2345 free_tlist(tt);
2347 /* Round up to even stack slots */
2348 size = ALIGN(size, StackSize);
2350 /* Now define the macro for the argument */
2351 snprintf(directive, sizeof(directive), "%%define %s (%s+%d)",
2352 arg, StackPointer, offset);
2353 do_directive(tokenize(directive));
2354 offset += size;
2356 /* Move to the next argument in the list */
2357 tline = tline->next;
2358 if (tline && tline->type == TOK_WHITESPACE)
2359 tline = tline->next;
2360 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2361 ArgOffset = offset;
2362 free_tlist(origline);
2363 return DIRECTIVE_FOUND;
2365 case PP_LOCAL:
2366 /* TASM like LOCAL directive to define local variables for a
2367 * function, in the following form:
2369 * LOCAL local1:WORD, local2:DWORD, local4:QWORD = LocalSize
2371 * The '= LocalSize' at the end is ignored by NASM, but is
2372 * required by TASM to define the local parameter size (and used
2373 * by the TASM macro package).
2375 offset = LocalOffset;
2376 do {
2377 char *local, directive[256];
2378 int size = StackSize;
2380 /* Find the argument name */
2381 tline = tline->next;
2382 if (tline && tline->type == TOK_WHITESPACE)
2383 tline = tline->next;
2384 if (!tline || tline->type != TOK_ID) {
2385 nasm_error(ERR_NONFATAL,
2386 "`%%local' missing argument parameter");
2387 free_tlist(origline);
2388 return DIRECTIVE_FOUND;
2390 local = tline->text;
2392 /* Find the argument size type */
2393 tline = tline->next;
2394 if (!tline || tline->type != TOK_OTHER
2395 || tline->text[0] != ':') {
2396 nasm_error(ERR_NONFATAL,
2397 "Syntax error processing `%%local' directive");
2398 free_tlist(origline);
2399 return DIRECTIVE_FOUND;
2401 tline = tline->next;
2402 if (!tline || tline->type != TOK_ID) {
2403 nasm_error(ERR_NONFATAL,
2404 "`%%local' missing size type parameter");
2405 free_tlist(origline);
2406 return DIRECTIVE_FOUND;
2409 /* Allow macro expansion of type parameter */
2410 tt = tokenize(tline->text);
2411 tt = expand_smacro(tt);
2412 size = parse_size(tt->text);
2413 if (!size) {
2414 nasm_error(ERR_NONFATAL,
2415 "Invalid size type for `%%local' missing directive");
2416 free_tlist(tt);
2417 free_tlist(origline);
2418 return DIRECTIVE_FOUND;
2420 free_tlist(tt);
2422 /* Round up to even stack slots */
2423 size = ALIGN(size, StackSize);
2425 offset += size; /* Negative offset, increment before */
2427 /* Now define the macro for the argument */
2428 snprintf(directive, sizeof(directive), "%%define %s (%s-%d)",
2429 local, StackPointer, offset);
2430 do_directive(tokenize(directive));
2432 /* Now define the assign to setup the enter_c macro correctly */
2433 snprintf(directive, sizeof(directive),
2434 "%%assign %%$localsize %%$localsize+%d", size);
2435 do_directive(tokenize(directive));
2437 /* Move to the next argument in the list */
2438 tline = tline->next;
2439 if (tline && tline->type == TOK_WHITESPACE)
2440 tline = tline->next;
2441 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2442 LocalOffset = offset;
2443 free_tlist(origline);
2444 return DIRECTIVE_FOUND;
2446 case PP_CLEAR:
2447 if (tline->next)
2448 nasm_error(ERR_WARNING|ERR_PASS1,
2449 "trailing garbage after `%%clear' ignored");
2450 free_macros();
2451 init_macros();
2452 free_tlist(origline);
2453 return DIRECTIVE_FOUND;
2455 case PP_DEPEND:
2456 t = tline->next = expand_smacro(tline->next);
2457 skip_white_(t);
2458 if (!t || (t->type != TOK_STRING &&
2459 t->type != TOK_INTERNAL_STRING)) {
2460 nasm_error(ERR_NONFATAL, "`%%depend' expects a file name");
2461 free_tlist(origline);
2462 return DIRECTIVE_FOUND; /* but we did _something_ */
2464 if (t->next)
2465 nasm_error(ERR_WARNING|ERR_PASS1,
2466 "trailing garbage after `%%depend' ignored");
2467 p = t->text;
2468 if (t->type != TOK_INTERNAL_STRING)
2469 nasm_unquote_cstr(p, i);
2470 if (dephead && !in_list(*dephead, p)) {
2471 StrList *sl = nasm_malloc(strlen(p)+1+sizeof sl->next);
2472 sl->next = NULL;
2473 strcpy(sl->str, p);
2474 *deptail = sl;
2475 deptail = &sl->next;
2477 free_tlist(origline);
2478 return DIRECTIVE_FOUND;
2480 case PP_INCLUDE:
2481 t = tline->next = expand_smacro(tline->next);
2482 skip_white_(t);
2484 if (!t || (t->type != TOK_STRING &&
2485 t->type != TOK_INTERNAL_STRING)) {
2486 nasm_error(ERR_NONFATAL, "`%%include' expects a file name");
2487 free_tlist(origline);
2488 return DIRECTIVE_FOUND; /* but we did _something_ */
2490 if (t->next)
2491 nasm_error(ERR_WARNING|ERR_PASS1,
2492 "trailing garbage after `%%include' ignored");
2493 p = t->text;
2494 if (t->type != TOK_INTERNAL_STRING)
2495 nasm_unquote_cstr(p, i);
2496 inc = nasm_malloc(sizeof(Include));
2497 inc->next = istk;
2498 inc->conds = NULL;
2499 inc->fp = inc_fopen(p, dephead, &deptail, pass == 0);
2500 if (!inc->fp) {
2501 /* -MG given but file not found */
2502 nasm_free(inc);
2503 } else {
2504 inc->fname = src_set_fname(nasm_strdup(p));
2505 inc->lineno = src_set_linnum(0);
2506 inc->lineinc = 1;
2507 inc->expansion = NULL;
2508 inc->mstk = NULL;
2509 istk = inc;
2510 lfmt->uplevel(LIST_INCLUDE);
2512 free_tlist(origline);
2513 return DIRECTIVE_FOUND;
2515 case PP_USE:
2517 static macros_t *use_pkg;
2518 const char *pkg_macro = NULL;
2520 tline = tline->next;
2521 skip_white_(tline);
2522 tline = expand_id(tline);
2524 if (!tline || (tline->type != TOK_STRING &&
2525 tline->type != TOK_INTERNAL_STRING &&
2526 tline->type != TOK_ID)) {
2527 nasm_error(ERR_NONFATAL, "`%%use' expects a package name");
2528 free_tlist(origline);
2529 return DIRECTIVE_FOUND; /* but we did _something_ */
2531 if (tline->next)
2532 nasm_error(ERR_WARNING|ERR_PASS1,
2533 "trailing garbage after `%%use' ignored");
2534 if (tline->type == TOK_STRING)
2535 nasm_unquote_cstr(tline->text, i);
2536 use_pkg = nasm_stdmac_find_package(tline->text);
2537 if (!use_pkg)
2538 nasm_error(ERR_NONFATAL, "unknown `%%use' package: %s", tline->text);
2539 else
2540 pkg_macro = (char *)use_pkg + 1; /* The first string will be <%define>__USE_*__ */
2541 if (use_pkg && ! smacro_defined(NULL, pkg_macro, 0, NULL, true)) {
2542 /* Not already included, go ahead and include it */
2543 stdmacpos = use_pkg;
2545 free_tlist(origline);
2546 return DIRECTIVE_FOUND;
2548 case PP_PUSH:
2549 case PP_REPL:
2550 case PP_POP:
2551 tline = tline->next;
2552 skip_white_(tline);
2553 tline = expand_id(tline);
2554 if (tline) {
2555 if (!tok_type_(tline, TOK_ID)) {
2556 nasm_error(ERR_NONFATAL, "`%s' expects a context identifier",
2557 pp_directives[i]);
2558 free_tlist(origline);
2559 return DIRECTIVE_FOUND; /* but we did _something_ */
2561 if (tline->next)
2562 nasm_error(ERR_WARNING|ERR_PASS1,
2563 "trailing garbage after `%s' ignored",
2564 pp_directives[i]);
2565 p = nasm_strdup(tline->text);
2566 } else {
2567 p = NULL; /* Anonymous */
2570 if (i == PP_PUSH) {
2571 ctx = nasm_malloc(sizeof(Context));
2572 ctx->next = cstk;
2573 hash_init(&ctx->localmac, HASH_SMALL);
2574 ctx->name = p;
2575 ctx->number = unique++;
2576 cstk = ctx;
2577 } else {
2578 /* %pop or %repl */
2579 if (!cstk) {
2580 nasm_error(ERR_NONFATAL, "`%s': context stack is empty",
2581 pp_directives[i]);
2582 } else if (i == PP_POP) {
2583 if (p && (!cstk->name || nasm_stricmp(p, cstk->name)))
2584 nasm_error(ERR_NONFATAL, "`%%pop' in wrong context: %s, "
2585 "expected %s",
2586 cstk->name ? cstk->name : "anonymous", p);
2587 else
2588 ctx_pop();
2589 } else {
2590 /* i == PP_REPL */
2591 nasm_free(cstk->name);
2592 cstk->name = p;
2593 p = NULL;
2595 nasm_free(p);
2597 free_tlist(origline);
2598 return DIRECTIVE_FOUND;
2599 case PP_FATAL:
2600 severity = ERR_FATAL;
2601 goto issue_error;
2602 case PP_ERROR:
2603 severity = ERR_NONFATAL;
2604 goto issue_error;
2605 case PP_WARNING:
2606 severity = ERR_WARNING|ERR_WARN_USER;
2607 goto issue_error;
2609 issue_error:
2611 /* Only error out if this is the final pass */
2612 if (pass != 2 && i != PP_FATAL)
2613 return DIRECTIVE_FOUND;
2615 tline->next = expand_smacro(tline->next);
2616 tline = tline->next;
2617 skip_white_(tline);
2618 t = tline ? tline->next : NULL;
2619 skip_white_(t);
2620 if (tok_type_(tline, TOK_STRING) && !t) {
2621 /* The line contains only a quoted string */
2622 p = tline->text;
2623 nasm_unquote(p, NULL); /* Ignore NUL character truncation */
2624 nasm_error(severity, "%s", p);
2625 } else {
2626 /* Not a quoted string, or more than a quoted string */
2627 p = detoken(tline, false);
2628 nasm_error(severity, "%s", p);
2629 nasm_free(p);
2631 free_tlist(origline);
2632 return DIRECTIVE_FOUND;
2635 CASE_PP_IF:
2636 if (istk->conds && !emitting(istk->conds->state))
2637 j = COND_NEVER;
2638 else {
2639 j = if_condition(tline->next, i);
2640 tline->next = NULL; /* it got freed */
2641 j = j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2643 cond = nasm_malloc(sizeof(Cond));
2644 cond->next = istk->conds;
2645 cond->state = j;
2646 istk->conds = cond;
2647 if(istk->mstk)
2648 istk->mstk->condcnt ++;
2649 free_tlist(origline);
2650 return DIRECTIVE_FOUND;
2652 CASE_PP_ELIF:
2653 if (!istk->conds)
2654 nasm_error(ERR_FATAL, "`%s': no matching `%%if'", pp_directives[i]);
2655 switch(istk->conds->state) {
2656 case COND_IF_TRUE:
2657 istk->conds->state = COND_DONE;
2658 break;
2660 case COND_DONE:
2661 case COND_NEVER:
2662 break;
2664 case COND_ELSE_TRUE:
2665 case COND_ELSE_FALSE:
2666 nasm_error(ERR_WARNING|ERR_PASS1|ERR_PP_PRECOND,
2667 "`%%elif' after `%%else' ignored");
2668 istk->conds->state = COND_NEVER;
2669 break;
2671 case COND_IF_FALSE:
2673 * IMPORTANT: In the case of %if, we will already have
2674 * called expand_mmac_params(); however, if we're
2675 * processing an %elif we must have been in a
2676 * non-emitting mode, which would have inhibited
2677 * the normal invocation of expand_mmac_params().
2678 * Therefore, we have to do it explicitly here.
2680 j = if_condition(expand_mmac_params(tline->next), i);
2681 tline->next = NULL; /* it got freed */
2682 istk->conds->state =
2683 j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2684 break;
2686 free_tlist(origline);
2687 return DIRECTIVE_FOUND;
2689 case PP_ELSE:
2690 if (tline->next)
2691 nasm_error(ERR_WARNING|ERR_PASS1|ERR_PP_PRECOND,
2692 "trailing garbage after `%%else' ignored");
2693 if (!istk->conds)
2694 nasm_fatal(0, "`%%else: no matching `%%if'");
2695 switch(istk->conds->state) {
2696 case COND_IF_TRUE:
2697 case COND_DONE:
2698 istk->conds->state = COND_ELSE_FALSE;
2699 break;
2701 case COND_NEVER:
2702 break;
2704 case COND_IF_FALSE:
2705 istk->conds->state = COND_ELSE_TRUE;
2706 break;
2708 case COND_ELSE_TRUE:
2709 case COND_ELSE_FALSE:
2710 nasm_error(ERR_WARNING|ERR_PASS1|ERR_PP_PRECOND,
2711 "`%%else' after `%%else' ignored.");
2712 istk->conds->state = COND_NEVER;
2713 break;
2715 free_tlist(origline);
2716 return DIRECTIVE_FOUND;
2718 case PP_ENDIF:
2719 if (tline->next)
2720 nasm_error(ERR_WARNING|ERR_PASS1|ERR_PP_PRECOND,
2721 "trailing garbage after `%%endif' ignored");
2722 if (!istk->conds)
2723 nasm_error(ERR_FATAL, "`%%endif': no matching `%%if'");
2724 cond = istk->conds;
2725 istk->conds = cond->next;
2726 nasm_free(cond);
2727 if(istk->mstk)
2728 istk->mstk->condcnt --;
2729 free_tlist(origline);
2730 return DIRECTIVE_FOUND;
2732 case PP_RMACRO:
2733 case PP_IRMACRO:
2734 case PP_MACRO:
2735 case PP_IMACRO:
2736 if (defining) {
2737 nasm_error(ERR_FATAL, "`%s': already defining a macro",
2738 pp_directives[i]);
2739 return DIRECTIVE_FOUND;
2741 defining = nasm_malloc(sizeof(MMacro));
2742 defining->max_depth =
2743 (i == PP_RMACRO) || (i == PP_IRMACRO) ? DEADMAN_LIMIT : 0;
2744 defining->casesense = (i == PP_MACRO) || (i == PP_RMACRO);
2745 if (!parse_mmacro_spec(tline, defining, pp_directives[i])) {
2746 nasm_free(defining);
2747 defining = NULL;
2748 return DIRECTIVE_FOUND;
2751 mmac = (MMacro *) hash_findix(&mmacros, defining->name);
2752 while (mmac) {
2753 if (!strcmp(mmac->name, defining->name) &&
2754 (mmac->nparam_min <= defining->nparam_max
2755 || defining->plus)
2756 && (defining->nparam_min <= mmac->nparam_max
2757 || mmac->plus)) {
2758 nasm_error(ERR_WARNING|ERR_PASS1,
2759 "redefining multi-line macro `%s'", defining->name);
2760 return DIRECTIVE_FOUND;
2762 mmac = mmac->next;
2764 free_tlist(origline);
2765 return DIRECTIVE_FOUND;
2767 case PP_ENDM:
2768 case PP_ENDMACRO:
2769 if (! (defining && defining->name)) {
2770 nasm_error(ERR_NONFATAL, "`%s': not defining a macro", tline->text);
2771 return DIRECTIVE_FOUND;
2773 mmhead = (MMacro **) hash_findi_add(&mmacros, defining->name);
2774 defining->next = *mmhead;
2775 *mmhead = defining;
2776 defining = NULL;
2777 free_tlist(origline);
2778 return DIRECTIVE_FOUND;
2780 case PP_EXITMACRO:
2782 * We must search along istk->expansion until we hit a
2783 * macro-end marker for a macro with a name. Then we
2784 * bypass all lines between exitmacro and endmacro.
2786 list_for_each(l, istk->expansion)
2787 if (l->finishes && l->finishes->name)
2788 break;
2790 if (l) {
2792 * Remove all conditional entries relative to this
2793 * macro invocation. (safe to do in this context)
2795 for ( ; l->finishes->condcnt > 0; l->finishes->condcnt --) {
2796 cond = istk->conds;
2797 istk->conds = cond->next;
2798 nasm_free(cond);
2800 istk->expansion = l;
2801 } else {
2802 nasm_error(ERR_NONFATAL, "`%%exitmacro' not within `%%macro' block");
2804 free_tlist(origline);
2805 return DIRECTIVE_FOUND;
2807 case PP_UNMACRO:
2808 case PP_UNIMACRO:
2810 MMacro **mmac_p;
2811 MMacro spec;
2813 spec.casesense = (i == PP_UNMACRO);
2814 if (!parse_mmacro_spec(tline, &spec, pp_directives[i])) {
2815 return DIRECTIVE_FOUND;
2817 mmac_p = (MMacro **) hash_findi(&mmacros, spec.name, NULL);
2818 while (mmac_p && *mmac_p) {
2819 mmac = *mmac_p;
2820 if (mmac->casesense == spec.casesense &&
2821 !mstrcmp(mmac->name, spec.name, spec.casesense) &&
2822 mmac->nparam_min == spec.nparam_min &&
2823 mmac->nparam_max == spec.nparam_max &&
2824 mmac->plus == spec.plus) {
2825 *mmac_p = mmac->next;
2826 free_mmacro(mmac);
2827 } else {
2828 mmac_p = &mmac->next;
2831 free_tlist(origline);
2832 free_tlist(spec.dlist);
2833 return DIRECTIVE_FOUND;
2836 case PP_ROTATE:
2837 if (tline->next && tline->next->type == TOK_WHITESPACE)
2838 tline = tline->next;
2839 if (!tline->next) {
2840 free_tlist(origline);
2841 nasm_error(ERR_NONFATAL, "`%%rotate' missing rotate count");
2842 return DIRECTIVE_FOUND;
2844 t = expand_smacro(tline->next);
2845 tline->next = NULL;
2846 free_tlist(origline);
2847 tline = t;
2848 tptr = &t;
2849 tokval.t_type = TOKEN_INVALID;
2850 evalresult =
2851 evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
2852 free_tlist(tline);
2853 if (!evalresult)
2854 return DIRECTIVE_FOUND;
2855 if (tokval.t_type)
2856 nasm_error(ERR_WARNING|ERR_PASS1,
2857 "trailing garbage after expression ignored");
2858 if (!is_simple(evalresult)) {
2859 nasm_error(ERR_NONFATAL, "non-constant value given to `%%rotate'");
2860 return DIRECTIVE_FOUND;
2862 mmac = istk->mstk;
2863 while (mmac && !mmac->name) /* avoid mistaking %reps for macros */
2864 mmac = mmac->next_active;
2865 if (!mmac) {
2866 nasm_error(ERR_NONFATAL, "`%%rotate' invoked outside a macro call");
2867 } else if (mmac->nparam == 0) {
2868 nasm_error(ERR_NONFATAL,
2869 "`%%rotate' invoked within macro without parameters");
2870 } else {
2871 int rotate = mmac->rotate + reloc_value(evalresult);
2873 rotate %= (int)mmac->nparam;
2874 if (rotate < 0)
2875 rotate += mmac->nparam;
2877 mmac->rotate = rotate;
2879 return DIRECTIVE_FOUND;
2881 case PP_REP:
2882 nolist = false;
2883 do {
2884 tline = tline->next;
2885 } while (tok_type_(tline, TOK_WHITESPACE));
2887 if (tok_type_(tline, TOK_ID) &&
2888 nasm_stricmp(tline->text, ".nolist") == 0) {
2889 nolist = true;
2890 do {
2891 tline = tline->next;
2892 } while (tok_type_(tline, TOK_WHITESPACE));
2895 if (tline) {
2896 t = expand_smacro(tline);
2897 tptr = &t;
2898 tokval.t_type = TOKEN_INVALID;
2899 evalresult =
2900 evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
2901 if (!evalresult) {
2902 free_tlist(origline);
2903 return DIRECTIVE_FOUND;
2905 if (tokval.t_type)
2906 nasm_error(ERR_WARNING|ERR_PASS1,
2907 "trailing garbage after expression ignored");
2908 if (!is_simple(evalresult)) {
2909 nasm_error(ERR_NONFATAL, "non-constant value given to `%%rep'");
2910 return DIRECTIVE_FOUND;
2912 count = reloc_value(evalresult);
2913 if (count >= REP_LIMIT) {
2914 nasm_error(ERR_NONFATAL, "`%%rep' value exceeds limit");
2915 count = 0;
2916 } else
2917 count++;
2918 } else {
2919 nasm_error(ERR_NONFATAL, "`%%rep' expects a repeat count");
2920 count = 0;
2922 free_tlist(origline);
2924 tmp_defining = defining;
2925 defining = nasm_malloc(sizeof(MMacro));
2926 defining->prev = NULL;
2927 defining->name = NULL; /* flags this macro as a %rep block */
2928 defining->casesense = false;
2929 defining->plus = false;
2930 defining->nolist = nolist;
2931 defining->in_progress = count;
2932 defining->max_depth = 0;
2933 defining->nparam_min = defining->nparam_max = 0;
2934 defining->defaults = NULL;
2935 defining->dlist = NULL;
2936 defining->expansion = NULL;
2937 defining->next_active = istk->mstk;
2938 defining->rep_nest = tmp_defining;
2939 return DIRECTIVE_FOUND;
2941 case PP_ENDREP:
2942 if (!defining || defining->name) {
2943 nasm_error(ERR_NONFATAL, "`%%endrep': no matching `%%rep'");
2944 return DIRECTIVE_FOUND;
2948 * Now we have a "macro" defined - although it has no name
2949 * and we won't be entering it in the hash tables - we must
2950 * push a macro-end marker for it on to istk->expansion.
2951 * After that, it will take care of propagating itself (a
2952 * macro-end marker line for a macro which is really a %rep
2953 * block will cause the macro to be re-expanded, complete
2954 * with another macro-end marker to ensure the process
2955 * continues) until the whole expansion is forcibly removed
2956 * from istk->expansion by a %exitrep.
2958 l = nasm_malloc(sizeof(Line));
2959 l->next = istk->expansion;
2960 l->finishes = defining;
2961 l->first = NULL;
2962 istk->expansion = l;
2964 istk->mstk = defining;
2966 lfmt->uplevel(defining->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
2967 tmp_defining = defining;
2968 defining = defining->rep_nest;
2969 free_tlist(origline);
2970 return DIRECTIVE_FOUND;
2972 case PP_EXITREP:
2974 * We must search along istk->expansion until we hit a
2975 * macro-end marker for a macro with no name. Then we set
2976 * its `in_progress' flag to 0.
2978 list_for_each(l, istk->expansion)
2979 if (l->finishes && !l->finishes->name)
2980 break;
2982 if (l)
2983 l->finishes->in_progress = 1;
2984 else
2985 nasm_error(ERR_NONFATAL, "`%%exitrep' not within `%%rep' block");
2986 free_tlist(origline);
2987 return DIRECTIVE_FOUND;
2989 case PP_XDEFINE:
2990 case PP_IXDEFINE:
2991 case PP_DEFINE:
2992 case PP_IDEFINE:
2993 casesense = (i == PP_DEFINE || i == PP_XDEFINE);
2995 tline = tline->next;
2996 skip_white_(tline);
2997 tline = expand_id(tline);
2998 if (!tline || (tline->type != TOK_ID &&
2999 (tline->type != TOK_PREPROC_ID ||
3000 tline->text[1] != '$'))) {
3001 nasm_error(ERR_NONFATAL, "`%s' expects a macro identifier",
3002 pp_directives[i]);
3003 free_tlist(origline);
3004 return DIRECTIVE_FOUND;
3007 ctx = get_ctx(tline->text, &mname);
3008 last = tline;
3009 param_start = tline = tline->next;
3010 nparam = 0;
3012 /* Expand the macro definition now for %xdefine and %ixdefine */
3013 if ((i == PP_XDEFINE) || (i == PP_IXDEFINE))
3014 tline = expand_smacro(tline);
3016 if (tok_is_(tline, "(")) {
3018 * This macro has parameters.
3021 tline = tline->next;
3022 while (1) {
3023 skip_white_(tline);
3024 if (!tline) {
3025 nasm_error(ERR_NONFATAL, "parameter identifier expected");
3026 free_tlist(origline);
3027 return DIRECTIVE_FOUND;
3029 if (tline->type != TOK_ID) {
3030 nasm_error(ERR_NONFATAL,
3031 "`%s': parameter identifier expected",
3032 tline->text);
3033 free_tlist(origline);
3034 return DIRECTIVE_FOUND;
3036 tline->type = TOK_SMAC_PARAM + nparam++;
3037 tline = tline->next;
3038 skip_white_(tline);
3039 if (tok_is_(tline, ",")) {
3040 tline = tline->next;
3041 } else {
3042 if (!tok_is_(tline, ")")) {
3043 nasm_error(ERR_NONFATAL,
3044 "`)' expected to terminate macro template");
3045 free_tlist(origline);
3046 return DIRECTIVE_FOUND;
3048 break;
3051 last = tline;
3052 tline = tline->next;
3054 if (tok_type_(tline, TOK_WHITESPACE))
3055 last = tline, tline = tline->next;
3056 macro_start = NULL;
3057 last->next = NULL;
3058 t = tline;
3059 while (t) {
3060 if (t->type == TOK_ID) {
3061 list_for_each(tt, param_start)
3062 if (tt->type >= TOK_SMAC_PARAM &&
3063 !strcmp(tt->text, t->text))
3064 t->type = tt->type;
3066 tt = t->next;
3067 t->next = macro_start;
3068 macro_start = t;
3069 t = tt;
3072 * Good. We now have a macro name, a parameter count, and a
3073 * token list (in reverse order) for an expansion. We ought
3074 * to be OK just to create an SMacro, store it, and let
3075 * free_tlist have the rest of the line (which we have
3076 * carefully re-terminated after chopping off the expansion
3077 * from the end).
3079 define_smacro(ctx, mname, casesense, nparam, macro_start);
3080 free_tlist(origline);
3081 return DIRECTIVE_FOUND;
3083 case PP_UNDEF:
3084 tline = tline->next;
3085 skip_white_(tline);
3086 tline = expand_id(tline);
3087 if (!tline || (tline->type != TOK_ID &&
3088 (tline->type != TOK_PREPROC_ID ||
3089 tline->text[1] != '$'))) {
3090 nasm_error(ERR_NONFATAL, "`%%undef' expects a macro identifier");
3091 free_tlist(origline);
3092 return DIRECTIVE_FOUND;
3094 if (tline->next) {
3095 nasm_error(ERR_WARNING|ERR_PASS1,
3096 "trailing garbage after macro name ignored");
3099 /* Find the context that symbol belongs to */
3100 ctx = get_ctx(tline->text, &mname);
3101 undef_smacro(ctx, mname);
3102 free_tlist(origline);
3103 return DIRECTIVE_FOUND;
3105 case PP_DEFSTR:
3106 case PP_IDEFSTR:
3107 casesense = (i == PP_DEFSTR);
3109 tline = tline->next;
3110 skip_white_(tline);
3111 tline = expand_id(tline);
3112 if (!tline || (tline->type != TOK_ID &&
3113 (tline->type != TOK_PREPROC_ID ||
3114 tline->text[1] != '$'))) {
3115 nasm_error(ERR_NONFATAL, "`%s' expects a macro identifier",
3116 pp_directives[i]);
3117 free_tlist(origline);
3118 return DIRECTIVE_FOUND;
3121 ctx = get_ctx(tline->text, &mname);
3122 last = tline;
3123 tline = expand_smacro(tline->next);
3124 last->next = NULL;
3126 while (tok_type_(tline, TOK_WHITESPACE))
3127 tline = delete_Token(tline);
3129 p = detoken(tline, false);
3130 macro_start = nasm_malloc(sizeof(*macro_start));
3131 macro_start->next = NULL;
3132 macro_start->text = nasm_quote(p, strlen(p));
3133 macro_start->type = TOK_STRING;
3134 macro_start->a.mac = NULL;
3135 nasm_free(p);
3138 * We now have a macro name, an implicit parameter count of
3139 * zero, and a string token to use as an expansion. Create
3140 * and store an SMacro.
3142 define_smacro(ctx, mname, casesense, 0, macro_start);
3143 free_tlist(origline);
3144 return DIRECTIVE_FOUND;
3146 case PP_DEFTOK:
3147 case PP_IDEFTOK:
3148 casesense = (i == PP_DEFTOK);
3150 tline = tline->next;
3151 skip_white_(tline);
3152 tline = expand_id(tline);
3153 if (!tline || (tline->type != TOK_ID &&
3154 (tline->type != TOK_PREPROC_ID ||
3155 tline->text[1] != '$'))) {
3156 nasm_error(ERR_NONFATAL,
3157 "`%s' expects a macro identifier as first parameter",
3158 pp_directives[i]);
3159 free_tlist(origline);
3160 return DIRECTIVE_FOUND;
3162 ctx = get_ctx(tline->text, &mname);
3163 last = tline;
3164 tline = expand_smacro(tline->next);
3165 last->next = NULL;
3167 t = tline;
3168 while (tok_type_(t, TOK_WHITESPACE))
3169 t = t->next;
3170 /* t should now point to the string */
3171 if (!tok_type_(t, TOK_STRING)) {
3172 nasm_error(ERR_NONFATAL,
3173 "`%s` requires string as second parameter",
3174 pp_directives[i]);
3175 free_tlist(tline);
3176 free_tlist(origline);
3177 return DIRECTIVE_FOUND;
3181 * Convert the string to a token stream. Note that smacros
3182 * are stored with the token stream reversed, so we have to
3183 * reverse the output of tokenize().
3185 nasm_unquote_cstr(t->text, i);
3186 macro_start = reverse_tokens(tokenize(t->text));
3189 * We now have a macro name, an implicit parameter count of
3190 * zero, and a numeric token to use as an expansion. Create
3191 * and store an SMacro.
3193 define_smacro(ctx, mname, casesense, 0, macro_start);
3194 free_tlist(tline);
3195 free_tlist(origline);
3196 return DIRECTIVE_FOUND;
3198 case PP_PATHSEARCH:
3200 FILE *fp;
3201 StrList *xsl = NULL;
3202 StrList **xst = &xsl;
3204 casesense = true;
3206 tline = tline->next;
3207 skip_white_(tline);
3208 tline = expand_id(tline);
3209 if (!tline || (tline->type != TOK_ID &&
3210 (tline->type != TOK_PREPROC_ID ||
3211 tline->text[1] != '$'))) {
3212 nasm_error(ERR_NONFATAL,
3213 "`%%pathsearch' expects a macro identifier as first parameter");
3214 free_tlist(origline);
3215 return DIRECTIVE_FOUND;
3217 ctx = get_ctx(tline->text, &mname);
3218 last = tline;
3219 tline = expand_smacro(tline->next);
3220 last->next = NULL;
3222 t = tline;
3223 while (tok_type_(t, TOK_WHITESPACE))
3224 t = t->next;
3226 if (!t || (t->type != TOK_STRING &&
3227 t->type != TOK_INTERNAL_STRING)) {
3228 nasm_error(ERR_NONFATAL, "`%%pathsearch' expects a file name");
3229 free_tlist(tline);
3230 free_tlist(origline);
3231 return DIRECTIVE_FOUND; /* but we did _something_ */
3233 if (t->next)
3234 nasm_error(ERR_WARNING|ERR_PASS1,
3235 "trailing garbage after `%%pathsearch' ignored");
3236 p = t->text;
3237 if (t->type != TOK_INTERNAL_STRING)
3238 nasm_unquote(p, NULL);
3240 fp = inc_fopen(p, &xsl, &xst, true);
3241 if (fp) {
3242 p = xsl->str;
3243 fclose(fp); /* Don't actually care about the file */
3245 macro_start = nasm_malloc(sizeof(*macro_start));
3246 macro_start->next = NULL;
3247 macro_start->text = nasm_quote(p, strlen(p));
3248 macro_start->type = TOK_STRING;
3249 macro_start->a.mac = NULL;
3250 if (xsl)
3251 nasm_free(xsl);
3254 * We now have a macro name, an implicit parameter count of
3255 * zero, and a string token to use as an expansion. Create
3256 * and store an SMacro.
3258 define_smacro(ctx, mname, casesense, 0, macro_start);
3259 free_tlist(tline);
3260 free_tlist(origline);
3261 return DIRECTIVE_FOUND;
3264 case PP_STRLEN:
3265 casesense = true;
3267 tline = tline->next;
3268 skip_white_(tline);
3269 tline = expand_id(tline);
3270 if (!tline || (tline->type != TOK_ID &&
3271 (tline->type != TOK_PREPROC_ID ||
3272 tline->text[1] != '$'))) {
3273 nasm_error(ERR_NONFATAL,
3274 "`%%strlen' expects a macro identifier as first parameter");
3275 free_tlist(origline);
3276 return DIRECTIVE_FOUND;
3278 ctx = get_ctx(tline->text, &mname);
3279 last = tline;
3280 tline = expand_smacro(tline->next);
3281 last->next = NULL;
3283 t = tline;
3284 while (tok_type_(t, TOK_WHITESPACE))
3285 t = t->next;
3286 /* t should now point to the string */
3287 if (!tok_type_(t, TOK_STRING)) {
3288 nasm_error(ERR_NONFATAL,
3289 "`%%strlen` requires string as second parameter");
3290 free_tlist(tline);
3291 free_tlist(origline);
3292 return DIRECTIVE_FOUND;
3295 macro_start = nasm_malloc(sizeof(*macro_start));
3296 macro_start->next = NULL;
3297 make_tok_num(macro_start, nasm_unquote(t->text, NULL));
3298 macro_start->a.mac = NULL;
3301 * We now have a macro name, an implicit parameter count of
3302 * zero, and a numeric token to use as an expansion. Create
3303 * and store an SMacro.
3305 define_smacro(ctx, mname, casesense, 0, macro_start);
3306 free_tlist(tline);
3307 free_tlist(origline);
3308 return DIRECTIVE_FOUND;
3310 case PP_STRCAT:
3311 casesense = true;
3313 tline = tline->next;
3314 skip_white_(tline);
3315 tline = expand_id(tline);
3316 if (!tline || (tline->type != TOK_ID &&
3317 (tline->type != TOK_PREPROC_ID ||
3318 tline->text[1] != '$'))) {
3319 nasm_error(ERR_NONFATAL,
3320 "`%%strcat' expects a macro identifier as first parameter");
3321 free_tlist(origline);
3322 return DIRECTIVE_FOUND;
3324 ctx = get_ctx(tline->text, &mname);
3325 last = tline;
3326 tline = expand_smacro(tline->next);
3327 last->next = NULL;
3329 len = 0;
3330 list_for_each(t, tline) {
3331 switch (t->type) {
3332 case TOK_WHITESPACE:
3333 break;
3334 case TOK_STRING:
3335 len += t->a.len = nasm_unquote(t->text, NULL);
3336 break;
3337 case TOK_OTHER:
3338 if (!strcmp(t->text, ",")) /* permit comma separators */
3339 break;
3340 /* else fall through */
3341 default:
3342 nasm_error(ERR_NONFATAL,
3343 "non-string passed to `%%strcat' (%d)", t->type);
3344 free_tlist(tline);
3345 free_tlist(origline);
3346 return DIRECTIVE_FOUND;
3350 p = pp = nasm_malloc(len);
3351 list_for_each(t, tline) {
3352 if (t->type == TOK_STRING) {
3353 memcpy(p, t->text, t->a.len);
3354 p += t->a.len;
3359 * We now have a macro name, an implicit parameter count of
3360 * zero, and a numeric token to use as an expansion. Create
3361 * and store an SMacro.
3363 macro_start = new_Token(NULL, TOK_STRING, NULL, 0);
3364 macro_start->text = nasm_quote(pp, len);
3365 nasm_free(pp);
3366 define_smacro(ctx, mname, casesense, 0, macro_start);
3367 free_tlist(tline);
3368 free_tlist(origline);
3369 return DIRECTIVE_FOUND;
3371 case PP_SUBSTR:
3373 int64_t start, count;
3374 size_t len;
3376 casesense = true;
3378 tline = tline->next;
3379 skip_white_(tline);
3380 tline = expand_id(tline);
3381 if (!tline || (tline->type != TOK_ID &&
3382 (tline->type != TOK_PREPROC_ID ||
3383 tline->text[1] != '$'))) {
3384 nasm_error(ERR_NONFATAL,
3385 "`%%substr' expects a macro identifier as first parameter");
3386 free_tlist(origline);
3387 return DIRECTIVE_FOUND;
3389 ctx = get_ctx(tline->text, &mname);
3390 last = tline;
3391 tline = expand_smacro(tline->next);
3392 last->next = NULL;
3394 if (tline) /* skip expanded id */
3395 t = tline->next;
3396 while (tok_type_(t, TOK_WHITESPACE))
3397 t = t->next;
3399 /* t should now point to the string */
3400 if (!tok_type_(t, TOK_STRING)) {
3401 nasm_error(ERR_NONFATAL,
3402 "`%%substr` requires string as second parameter");
3403 free_tlist(tline);
3404 free_tlist(origline);
3405 return DIRECTIVE_FOUND;
3408 tt = t->next;
3409 tptr = &tt;
3410 tokval.t_type = TOKEN_INVALID;
3411 evalresult = evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
3412 if (!evalresult) {
3413 free_tlist(tline);
3414 free_tlist(origline);
3415 return DIRECTIVE_FOUND;
3416 } else if (!is_simple(evalresult)) {
3417 nasm_error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3418 free_tlist(tline);
3419 free_tlist(origline);
3420 return DIRECTIVE_FOUND;
3422 start = evalresult->value - 1;
3424 while (tok_type_(tt, TOK_WHITESPACE))
3425 tt = tt->next;
3426 if (!tt) {
3427 count = 1; /* Backwards compatibility: one character */
3428 } else {
3429 tokval.t_type = TOKEN_INVALID;
3430 evalresult = evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
3431 if (!evalresult) {
3432 free_tlist(tline);
3433 free_tlist(origline);
3434 return DIRECTIVE_FOUND;
3435 } else if (!is_simple(evalresult)) {
3436 nasm_error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3437 free_tlist(tline);
3438 free_tlist(origline);
3439 return DIRECTIVE_FOUND;
3441 count = evalresult->value;
3444 len = nasm_unquote(t->text, NULL);
3446 /* make start and count being in range */
3447 if (start < 0)
3448 start = 0;
3449 if (count < 0)
3450 count = len + count + 1 - start;
3451 if (start + count > (int64_t)len)
3452 count = len - start;
3453 if (!len || count < 0 || start >=(int64_t)len)
3454 start = -1, count = 0; /* empty string */
3456 macro_start = nasm_malloc(sizeof(*macro_start));
3457 macro_start->next = NULL;
3458 macro_start->text = nasm_quote((start < 0) ? "" : t->text + start, count);
3459 macro_start->type = TOK_STRING;
3460 macro_start->a.mac = NULL;
3463 * We now have a macro name, an implicit parameter count of
3464 * zero, and a numeric token to use as an expansion. Create
3465 * and store an SMacro.
3467 define_smacro(ctx, mname, casesense, 0, macro_start);
3468 free_tlist(tline);
3469 free_tlist(origline);
3470 return DIRECTIVE_FOUND;
3473 case PP_ASSIGN:
3474 case PP_IASSIGN:
3475 casesense = (i == PP_ASSIGN);
3477 tline = tline->next;
3478 skip_white_(tline);
3479 tline = expand_id(tline);
3480 if (!tline || (tline->type != TOK_ID &&
3481 (tline->type != TOK_PREPROC_ID ||
3482 tline->text[1] != '$'))) {
3483 nasm_error(ERR_NONFATAL,
3484 "`%%%sassign' expects a macro identifier",
3485 (i == PP_IASSIGN ? "i" : ""));
3486 free_tlist(origline);
3487 return DIRECTIVE_FOUND;
3489 ctx = get_ctx(tline->text, &mname);
3490 last = tline;
3491 tline = expand_smacro(tline->next);
3492 last->next = NULL;
3494 t = tline;
3495 tptr = &t;
3496 tokval.t_type = TOKEN_INVALID;
3497 evalresult = evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
3498 free_tlist(tline);
3499 if (!evalresult) {
3500 free_tlist(origline);
3501 return DIRECTIVE_FOUND;
3504 if (tokval.t_type)
3505 nasm_error(ERR_WARNING|ERR_PASS1,
3506 "trailing garbage after expression ignored");
3508 if (!is_simple(evalresult)) {
3509 nasm_error(ERR_NONFATAL,
3510 "non-constant value given to `%%%sassign'",
3511 (i == PP_IASSIGN ? "i" : ""));
3512 free_tlist(origline);
3513 return DIRECTIVE_FOUND;
3516 macro_start = nasm_malloc(sizeof(*macro_start));
3517 macro_start->next = NULL;
3518 make_tok_num(macro_start, reloc_value(evalresult));
3519 macro_start->a.mac = NULL;
3522 * We now have a macro name, an implicit parameter count of
3523 * zero, and a numeric token to use as an expansion. Create
3524 * and store an SMacro.
3526 define_smacro(ctx, mname, casesense, 0, macro_start);
3527 free_tlist(origline);
3528 return DIRECTIVE_FOUND;
3530 case PP_LINE:
3532 * Syntax is `%line nnn[+mmm] [filename]'
3534 tline = tline->next;
3535 skip_white_(tline);
3536 if (!tok_type_(tline, TOK_NUMBER)) {
3537 nasm_error(ERR_NONFATAL, "`%%line' expects line number");
3538 free_tlist(origline);
3539 return DIRECTIVE_FOUND;
3541 k = readnum(tline->text, &err);
3542 m = 1;
3543 tline = tline->next;
3544 if (tok_is_(tline, "+")) {
3545 tline = tline->next;
3546 if (!tok_type_(tline, TOK_NUMBER)) {
3547 nasm_error(ERR_NONFATAL, "`%%line' expects line increment");
3548 free_tlist(origline);
3549 return DIRECTIVE_FOUND;
3551 m = readnum(tline->text, &err);
3552 tline = tline->next;
3554 skip_white_(tline);
3555 src_set_linnum(k);
3556 istk->lineinc = m;
3557 if (tline) {
3558 nasm_free(src_set_fname(detoken(tline, false)));
3560 free_tlist(origline);
3561 return DIRECTIVE_FOUND;
3563 default:
3564 nasm_error(ERR_FATAL,
3565 "preprocessor directive `%s' not yet implemented",
3566 pp_directives[i]);
3567 return DIRECTIVE_FOUND;
3572 * Ensure that a macro parameter contains a condition code and
3573 * nothing else. Return the condition code index if so, or -1
3574 * otherwise.
3576 static int find_cc(Token * t)
3578 Token *tt;
3580 if (!t)
3581 return -1; /* Probably a %+ without a space */
3583 skip_white_(t);
3584 if (t->type != TOK_ID)
3585 return -1;
3586 tt = t->next;
3587 skip_white_(tt);
3588 if (tt && (tt->type != TOK_OTHER || strcmp(tt->text, ",")))
3589 return -1;
3591 return bsii(t->text, (const char **)conditions, ARRAY_SIZE(conditions));
3595 * This routines walks over tokens strem and hadnles tokens
3596 * pasting, if @handle_explicit passed then explicit pasting
3597 * term is handled, otherwise -- implicit pastings only.
3599 static bool paste_tokens(Token **head, const struct tokseq_match *m,
3600 size_t mnum, bool handle_explicit)
3602 Token *tok, *next, **prev_next, **prev_nonspace;
3603 bool pasted = false;
3604 char *buf, *p;
3605 size_t len, i;
3608 * The last token before pasting. We need it
3609 * to be able to connect new handled tokens.
3610 * In other words if there were a tokens stream
3612 * A -> B -> C -> D
3614 * and we've joined tokens B and C, the resulting
3615 * stream should be
3617 * A -> BC -> D
3619 tok = *head;
3620 prev_next = NULL;
3622 if (!tok_type_(tok, TOK_WHITESPACE) && !tok_type_(tok, TOK_PASTE))
3623 prev_nonspace = head;
3624 else
3625 prev_nonspace = NULL;
3627 while (tok && (next = tok->next)) {
3629 switch (tok->type) {
3630 case TOK_WHITESPACE:
3631 /* Zap redundant whitespaces */
3632 while (tok_type_(next, TOK_WHITESPACE))
3633 next = delete_Token(next);
3634 tok->next = next;
3635 break;
3637 case TOK_PASTE:
3638 /* Explicit pasting */
3639 if (!handle_explicit)
3640 break;
3641 next = delete_Token(tok);
3643 while (tok_type_(next, TOK_WHITESPACE))
3644 next = delete_Token(next);
3646 if (!pasted)
3647 pasted = true;
3649 /* Left pasting token is start of line */
3650 if (!prev_nonspace)
3651 nasm_error(ERR_FATAL, "No lvalue found on pasting");
3654 * No ending token, this might happen in two
3655 * cases
3657 * 1) There indeed no right token at all
3658 * 2) There is a bare "%define ID" statement,
3659 * and @ID does expand to whitespace.
3661 * So technically we need to do a grammar analysis
3662 * in another stage of parsing, but for now lets don't
3663 * change the behaviour people used to. Simply allow
3664 * whitespace after paste token.
3666 if (!next) {
3668 * Zap ending space tokens and that's all.
3670 tok = (*prev_nonspace)->next;
3671 while (tok_type_(tok, TOK_WHITESPACE))
3672 tok = delete_Token(tok);
3673 tok = *prev_nonspace;
3674 tok->next = NULL;
3675 break;
3678 tok = *prev_nonspace;
3679 while (tok_type_(tok, TOK_WHITESPACE))
3680 tok = delete_Token(tok);
3681 len = strlen(tok->text);
3682 len += strlen(next->text);
3684 p = buf = nasm_malloc(len + 1);
3685 strcpy(p, tok->text);
3686 p = strchr(p, '\0');
3687 strcpy(p, next->text);
3689 delete_Token(tok);
3691 tok = tokenize(buf);
3692 nasm_free(buf);
3694 *prev_nonspace = tok;
3695 while (tok && tok->next)
3696 tok = tok->next;
3698 tok->next = delete_Token(next);
3700 /* Restart from pasted tokens head */
3701 tok = *prev_nonspace;
3702 break;
3704 default:
3705 /* implicit pasting */
3706 for (i = 0; i < mnum; i++) {
3707 if (!(PP_CONCAT_MATCH(tok, m[i].mask_head)))
3708 continue;
3710 len = 0;
3711 while (next && PP_CONCAT_MATCH(next, m[i].mask_tail)) {
3712 len += strlen(next->text);
3713 next = next->next;
3716 /* No match */
3717 if (tok == next)
3718 break;
3720 len += strlen(tok->text);
3721 p = buf = nasm_malloc(len + 1);
3723 while (tok != next) {
3724 strcpy(p, tok->text);
3725 p = strchr(p, '\0');
3726 tok = delete_Token(tok);
3729 tok = tokenize(buf);
3730 nasm_free(buf);
3732 if (prev_next)
3733 *prev_next = tok;
3734 else
3735 *head = tok;
3738 * Connect pasted into original stream,
3739 * ie A -> new-tokens -> B
3741 while (tok && tok->next)
3742 tok = tok->next;
3743 tok->next = next;
3745 if (!pasted)
3746 pasted = true;
3748 /* Restart from pasted tokens head */
3749 tok = prev_next ? *prev_next : *head;
3752 break;
3755 prev_next = &tok->next;
3757 if (tok->next &&
3758 !tok_type_(tok->next, TOK_WHITESPACE) &&
3759 !tok_type_(tok->next, TOK_PASTE))
3760 prev_nonspace = prev_next;
3762 tok = tok->next;
3765 return pasted;
3769 * expands to a list of tokens from %{x:y}
3771 static Token *expand_mmac_params_range(MMacro *mac, Token *tline, Token ***last)
3773 Token *t = tline, **tt, *tm, *head;
3774 char *pos;
3775 int fst, lst, j, i;
3777 pos = strchr(tline->text, ':');
3778 nasm_assert(pos);
3780 lst = atoi(pos + 1);
3781 fst = atoi(tline->text + 1);
3784 * only macros params are accounted so
3785 * if someone passes %0 -- we reject such
3786 * value(s)
3788 if (lst == 0 || fst == 0)
3789 goto err;
3791 /* the values should be sane */
3792 if ((fst > (int)mac->nparam || fst < (-(int)mac->nparam)) ||
3793 (lst > (int)mac->nparam || lst < (-(int)mac->nparam)))
3794 goto err;
3796 fst = fst < 0 ? fst + (int)mac->nparam + 1: fst;
3797 lst = lst < 0 ? lst + (int)mac->nparam + 1: lst;
3799 /* counted from zero */
3800 fst--, lst--;
3803 * It will be at least one token. Note we
3804 * need to scan params until separator, otherwise
3805 * only first token will be passed.
3807 tm = mac->params[(fst + mac->rotate) % mac->nparam];
3808 head = new_Token(NULL, tm->type, tm->text, 0);
3809 tt = &head->next, tm = tm->next;
3810 while (tok_isnt_(tm, ",")) {
3811 t = new_Token(NULL, tm->type, tm->text, 0);
3812 *tt = t, tt = &t->next, tm = tm->next;
3815 if (fst < lst) {
3816 for (i = fst + 1; i <= lst; i++) {
3817 t = new_Token(NULL, TOK_OTHER, ",", 0);
3818 *tt = t, tt = &t->next;
3819 j = (i + mac->rotate) % mac->nparam;
3820 tm = mac->params[j];
3821 while (tok_isnt_(tm, ",")) {
3822 t = new_Token(NULL, tm->type, tm->text, 0);
3823 *tt = t, tt = &t->next, tm = tm->next;
3826 } else {
3827 for (i = fst - 1; i >= lst; i--) {
3828 t = new_Token(NULL, TOK_OTHER, ",", 0);
3829 *tt = t, tt = &t->next;
3830 j = (i + mac->rotate) % mac->nparam;
3831 tm = mac->params[j];
3832 while (tok_isnt_(tm, ",")) {
3833 t = new_Token(NULL, tm->type, tm->text, 0);
3834 *tt = t, tt = &t->next, tm = tm->next;
3839 *last = tt;
3840 return head;
3842 err:
3843 nasm_error(ERR_NONFATAL, "`%%{%s}': macro parameters out of range",
3844 &tline->text[1]);
3845 return tline;
3849 * Expand MMacro-local things: parameter references (%0, %n, %+n,
3850 * %-n) and MMacro-local identifiers (%%foo) as well as
3851 * macro indirection (%[...]) and range (%{..:..}).
3853 static Token *expand_mmac_params(Token * tline)
3855 Token *t, *tt, **tail, *thead;
3856 bool changed = false;
3857 char *pos;
3859 tail = &thead;
3860 thead = NULL;
3862 while (tline) {
3863 if (tline->type == TOK_PREPROC_ID &&
3864 (((tline->text[1] == '+' || tline->text[1] == '-') && tline->text[2]) ||
3865 (tline->text[1] >= '0' && tline->text[1] <= '9') ||
3866 tline->text[1] == '%')) {
3867 char *text = NULL;
3868 int type = 0, cc; /* type = 0 to placate optimisers */
3869 char tmpbuf[30];
3870 unsigned int n;
3871 int i;
3872 MMacro *mac;
3874 t = tline;
3875 tline = tline->next;
3877 mac = istk->mstk;
3878 while (mac && !mac->name) /* avoid mistaking %reps for macros */
3879 mac = mac->next_active;
3880 if (!mac) {
3881 nasm_error(ERR_NONFATAL, "`%s': not in a macro call", t->text);
3882 } else {
3883 pos = strchr(t->text, ':');
3884 if (!pos) {
3885 switch (t->text[1]) {
3887 * We have to make a substitution of one of the
3888 * forms %1, %-1, %+1, %%foo, %0.
3890 case '0':
3891 type = TOK_NUMBER;
3892 snprintf(tmpbuf, sizeof(tmpbuf), "%d", mac->nparam);
3893 text = nasm_strdup(tmpbuf);
3894 break;
3895 case '%':
3896 type = TOK_ID;
3897 snprintf(tmpbuf, sizeof(tmpbuf), "..@%"PRIu64".",
3898 mac->unique);
3899 text = nasm_strcat(tmpbuf, t->text + 2);
3900 break;
3901 case '-':
3902 n = atoi(t->text + 2) - 1;
3903 if (n >= mac->nparam)
3904 tt = NULL;
3905 else {
3906 if (mac->nparam > 1)
3907 n = (n + mac->rotate) % mac->nparam;
3908 tt = mac->params[n];
3910 cc = find_cc(tt);
3911 if (cc == -1) {
3912 nasm_error(ERR_NONFATAL,
3913 "macro parameter %d is not a condition code",
3914 n + 1);
3915 text = NULL;
3916 } else {
3917 type = TOK_ID;
3918 if (inverse_ccs[cc] == -1) {
3919 nasm_error(ERR_NONFATAL,
3920 "condition code `%s' is not invertible",
3921 conditions[cc]);
3922 text = NULL;
3923 } else
3924 text = nasm_strdup(conditions[inverse_ccs[cc]]);
3926 break;
3927 case '+':
3928 n = atoi(t->text + 2) - 1;
3929 if (n >= mac->nparam)
3930 tt = NULL;
3931 else {
3932 if (mac->nparam > 1)
3933 n = (n + mac->rotate) % mac->nparam;
3934 tt = mac->params[n];
3936 cc = find_cc(tt);
3937 if (cc == -1) {
3938 nasm_error(ERR_NONFATAL,
3939 "macro parameter %d is not a condition code",
3940 n + 1);
3941 text = NULL;
3942 } else {
3943 type = TOK_ID;
3944 text = nasm_strdup(conditions[cc]);
3946 break;
3947 default:
3948 n = atoi(t->text + 1) - 1;
3949 if (n >= mac->nparam)
3950 tt = NULL;
3951 else {
3952 if (mac->nparam > 1)
3953 n = (n + mac->rotate) % mac->nparam;
3954 tt = mac->params[n];
3956 if (tt) {
3957 for (i = 0; i < mac->paramlen[n]; i++) {
3958 *tail = new_Token(NULL, tt->type, tt->text, 0);
3959 tail = &(*tail)->next;
3960 tt = tt->next;
3963 text = NULL; /* we've done it here */
3964 break;
3966 } else {
3968 * seems we have a parameters range here
3970 Token *head, **last;
3971 head = expand_mmac_params_range(mac, t, &last);
3972 if (head != t) {
3973 *tail = head;
3974 *last = tline;
3975 tline = head;
3976 text = NULL;
3980 if (!text) {
3981 delete_Token(t);
3982 } else {
3983 *tail = t;
3984 tail = &t->next;
3985 t->type = type;
3986 nasm_free(t->text);
3987 t->text = text;
3988 t->a.mac = NULL;
3990 changed = true;
3991 continue;
3992 } else if (tline->type == TOK_INDIRECT) {
3993 t = tline;
3994 tline = tline->next;
3995 tt = tokenize(t->text);
3996 tt = expand_mmac_params(tt);
3997 tt = expand_smacro(tt);
3998 *tail = tt;
3999 while (tt) {
4000 tt->a.mac = NULL; /* Necessary? */
4001 tail = &tt->next;
4002 tt = tt->next;
4004 delete_Token(t);
4005 changed = true;
4006 } else {
4007 t = *tail = tline;
4008 tline = tline->next;
4009 t->a.mac = NULL;
4010 tail = &t->next;
4013 *tail = NULL;
4015 if (changed) {
4016 const struct tokseq_match t[] = {
4018 PP_CONCAT_MASK(TOK_ID) |
4019 PP_CONCAT_MASK(TOK_FLOAT), /* head */
4020 PP_CONCAT_MASK(TOK_ID) |
4021 PP_CONCAT_MASK(TOK_NUMBER) |
4022 PP_CONCAT_MASK(TOK_FLOAT) |
4023 PP_CONCAT_MASK(TOK_OTHER) /* tail */
4026 PP_CONCAT_MASK(TOK_NUMBER), /* head */
4027 PP_CONCAT_MASK(TOK_NUMBER) /* tail */
4030 paste_tokens(&thead, t, ARRAY_SIZE(t), false);
4033 return thead;
4037 * Expand all single-line macro calls made in the given line.
4038 * Return the expanded version of the line. The original is deemed
4039 * to be destroyed in the process. (In reality we'll just move
4040 * Tokens from input to output a lot of the time, rather than
4041 * actually bothering to destroy and replicate.)
4044 static Token *expand_smacro(Token * tline)
4046 Token *t, *tt, *mstart, **tail, *thead;
4047 SMacro *head = NULL, *m;
4048 Token **params;
4049 int *paramsize;
4050 unsigned int nparam, sparam;
4051 int brackets;
4052 Token *org_tline = tline;
4053 Context *ctx;
4054 const char *mname;
4055 int deadman = DEADMAN_LIMIT;
4056 bool expanded;
4059 * Trick: we should avoid changing the start token pointer since it can
4060 * be contained in "next" field of other token. Because of this
4061 * we allocate a copy of first token and work with it; at the end of
4062 * routine we copy it back
4064 if (org_tline) {
4065 tline = new_Token(org_tline->next, org_tline->type,
4066 org_tline->text, 0);
4067 tline->a.mac = org_tline->a.mac;
4068 nasm_free(org_tline->text);
4069 org_tline->text = NULL;
4072 expanded = true; /* Always expand %+ at least once */
4074 again:
4075 thead = NULL;
4076 tail = &thead;
4078 while (tline) { /* main token loop */
4079 if (!--deadman) {
4080 nasm_error(ERR_NONFATAL, "interminable macro recursion");
4081 goto err;
4084 if ((mname = tline->text)) {
4085 /* if this token is a local macro, look in local context */
4086 if (tline->type == TOK_ID) {
4087 head = (SMacro *)hash_findix(&smacros, mname);
4088 } else if (tline->type == TOK_PREPROC_ID) {
4089 ctx = get_ctx(mname, &mname);
4090 head = ctx ? (SMacro *)hash_findix(&ctx->localmac, mname) : NULL;
4091 } else
4092 head = NULL;
4095 * We've hit an identifier. As in is_mmacro below, we first
4096 * check whether the identifier is a single-line macro at
4097 * all, then think about checking for parameters if
4098 * necessary.
4100 list_for_each(m, head)
4101 if (!mstrcmp(m->name, mname, m->casesense))
4102 break;
4103 if (m) {
4104 mstart = tline;
4105 params = NULL;
4106 paramsize = NULL;
4107 if (m->nparam == 0) {
4109 * Simple case: the macro is parameterless. Discard the
4110 * one token that the macro call took, and push the
4111 * expansion back on the to-do stack.
4113 if (!m->expansion) {
4114 if (!strcmp("__FILE__", m->name)) {
4115 int32_t num = 0;
4116 char *file = NULL;
4117 src_get(&num, &file);
4118 tline->text = nasm_quote(file, strlen(file));
4119 tline->type = TOK_STRING;
4120 nasm_free(file);
4121 continue;
4123 if (!strcmp("__LINE__", m->name)) {
4124 nasm_free(tline->text);
4125 make_tok_num(tline, src_get_linnum());
4126 continue;
4128 if (!strcmp("__BITS__", m->name)) {
4129 nasm_free(tline->text);
4130 make_tok_num(tline, globalbits);
4131 continue;
4133 tline = delete_Token(tline);
4134 continue;
4136 } else {
4138 * Complicated case: at least one macro with this name
4139 * exists and takes parameters. We must find the
4140 * parameters in the call, count them, find the SMacro
4141 * that corresponds to that form of the macro call, and
4142 * substitute for the parameters when we expand. What a
4143 * pain.
4145 /*tline = tline->next;
4146 skip_white_(tline); */
4147 do {
4148 t = tline->next;
4149 while (tok_type_(t, TOK_SMAC_END)) {
4150 t->a.mac->in_progress = false;
4151 t->text = NULL;
4152 t = tline->next = delete_Token(t);
4154 tline = t;
4155 } while (tok_type_(tline, TOK_WHITESPACE));
4156 if (!tok_is_(tline, "(")) {
4158 * This macro wasn't called with parameters: ignore
4159 * the call. (Behaviour borrowed from gnu cpp.)
4161 tline = mstart;
4162 m = NULL;
4163 } else {
4164 int paren = 0;
4165 int white = 0;
4166 brackets = 0;
4167 nparam = 0;
4168 sparam = PARAM_DELTA;
4169 params = nasm_malloc(sparam * sizeof(Token *));
4170 params[0] = tline->next;
4171 paramsize = nasm_malloc(sparam * sizeof(int));
4172 paramsize[0] = 0;
4173 while (true) { /* parameter loop */
4175 * For some unusual expansions
4176 * which concatenates function call
4178 t = tline->next;
4179 while (tok_type_(t, TOK_SMAC_END)) {
4180 t->a.mac->in_progress = false;
4181 t->text = NULL;
4182 t = tline->next = delete_Token(t);
4184 tline = t;
4186 if (!tline) {
4187 nasm_error(ERR_NONFATAL,
4188 "macro call expects terminating `)'");
4189 break;
4191 if (tline->type == TOK_WHITESPACE
4192 && brackets <= 0) {
4193 if (paramsize[nparam])
4194 white++;
4195 else
4196 params[nparam] = tline->next;
4197 continue; /* parameter loop */
4199 if (tline->type == TOK_OTHER
4200 && tline->text[1] == 0) {
4201 char ch = tline->text[0];
4202 if (ch == ',' && !paren && brackets <= 0) {
4203 if (++nparam >= sparam) {
4204 sparam += PARAM_DELTA;
4205 params = nasm_realloc(params,
4206 sparam * sizeof(Token *));
4207 paramsize = nasm_realloc(paramsize,
4208 sparam * sizeof(int));
4210 params[nparam] = tline->next;
4211 paramsize[nparam] = 0;
4212 white = 0;
4213 continue; /* parameter loop */
4215 if (ch == '{' &&
4216 (brackets > 0 || (brackets == 0 &&
4217 !paramsize[nparam])))
4219 if (!(brackets++)) {
4220 params[nparam] = tline->next;
4221 continue; /* parameter loop */
4224 if (ch == '}' && brackets > 0)
4225 if (--brackets == 0) {
4226 brackets = -1;
4227 continue; /* parameter loop */
4229 if (ch == '(' && !brackets)
4230 paren++;
4231 if (ch == ')' && brackets <= 0)
4232 if (--paren < 0)
4233 break;
4235 if (brackets < 0) {
4236 brackets = 0;
4237 nasm_error(ERR_NONFATAL, "braces do not "
4238 "enclose all of macro parameter");
4240 paramsize[nparam] += white + 1;
4241 white = 0;
4242 } /* parameter loop */
4243 nparam++;
4244 while (m && (m->nparam != nparam ||
4245 mstrcmp(m->name, mname,
4246 m->casesense)))
4247 m = m->next;
4248 if (!m)
4249 nasm_error(ERR_WARNING|ERR_PASS1|ERR_WARN_MNP,
4250 "macro `%s' exists, "
4251 "but not taking %d parameters",
4252 mstart->text, nparam);
4255 if (m && m->in_progress)
4256 m = NULL;
4257 if (!m) { /* in progess or didn't find '(' or wrong nparam */
4259 * Design question: should we handle !tline, which
4260 * indicates missing ')' here, or expand those
4261 * macros anyway, which requires the (t) test a few
4262 * lines down?
4264 nasm_free(params);
4265 nasm_free(paramsize);
4266 tline = mstart;
4267 } else {
4269 * Expand the macro: we are placed on the last token of the
4270 * call, so that we can easily split the call from the
4271 * following tokens. We also start by pushing an SMAC_END
4272 * token for the cycle removal.
4274 t = tline;
4275 if (t) {
4276 tline = t->next;
4277 t->next = NULL;
4279 tt = new_Token(tline, TOK_SMAC_END, NULL, 0);
4280 tt->a.mac = m;
4281 m->in_progress = true;
4282 tline = tt;
4283 list_for_each(t, m->expansion) {
4284 if (t->type >= TOK_SMAC_PARAM) {
4285 Token *pcopy = tline, **ptail = &pcopy;
4286 Token *ttt, *pt;
4287 int i;
4289 ttt = params[t->type - TOK_SMAC_PARAM];
4290 i = paramsize[t->type - TOK_SMAC_PARAM];
4291 while (--i >= 0) {
4292 pt = *ptail = new_Token(tline, ttt->type,
4293 ttt->text, 0);
4294 ptail = &pt->next;
4295 ttt = ttt->next;
4297 tline = pcopy;
4298 } else if (t->type == TOK_PREPROC_Q) {
4299 tt = new_Token(tline, TOK_ID, mname, 0);
4300 tline = tt;
4301 } else if (t->type == TOK_PREPROC_QQ) {
4302 tt = new_Token(tline, TOK_ID, m->name, 0);
4303 tline = tt;
4304 } else {
4305 tt = new_Token(tline, t->type, t->text, 0);
4306 tline = tt;
4311 * Having done that, get rid of the macro call, and clean
4312 * up the parameters.
4314 nasm_free(params);
4315 nasm_free(paramsize);
4316 free_tlist(mstart);
4317 expanded = true;
4318 continue; /* main token loop */
4323 if (tline->type == TOK_SMAC_END) {
4324 tline->a.mac->in_progress = false;
4325 tline = delete_Token(tline);
4326 } else {
4327 t = *tail = tline;
4328 tline = tline->next;
4329 t->a.mac = NULL;
4330 t->next = NULL;
4331 tail = &t->next;
4336 * Now scan the entire line and look for successive TOK_IDs that resulted
4337 * after expansion (they can't be produced by tokenize()). The successive
4338 * TOK_IDs should be concatenated.
4339 * Also we look for %+ tokens and concatenate the tokens before and after
4340 * them (without white spaces in between).
4342 if (expanded) {
4343 const struct tokseq_match t[] = {
4345 PP_CONCAT_MASK(TOK_ID) |
4346 PP_CONCAT_MASK(TOK_PREPROC_ID), /* head */
4347 PP_CONCAT_MASK(TOK_ID) |
4348 PP_CONCAT_MASK(TOK_PREPROC_ID) |
4349 PP_CONCAT_MASK(TOK_NUMBER) /* tail */
4352 if (paste_tokens(&thead, t, ARRAY_SIZE(t), true)) {
4354 * If we concatenated something, *and* we had previously expanded
4355 * an actual macro, scan the lines again for macros...
4357 tline = thead;
4358 expanded = false;
4359 goto again;
4363 err:
4364 if (org_tline) {
4365 if (thead) {
4366 *org_tline = *thead;
4367 /* since we just gave text to org_line, don't free it */
4368 thead->text = NULL;
4369 delete_Token(thead);
4370 } else {
4371 /* the expression expanded to empty line;
4372 we can't return NULL for some reasons
4373 we just set the line to a single WHITESPACE token. */
4374 memset(org_tline, 0, sizeof(*org_tline));
4375 org_tline->text = NULL;
4376 org_tline->type = TOK_WHITESPACE;
4378 thead = org_tline;
4381 return thead;
4385 * Similar to expand_smacro but used exclusively with macro identifiers
4386 * right before they are fetched in. The reason is that there can be
4387 * identifiers consisting of several subparts. We consider that if there
4388 * are more than one element forming the name, user wants a expansion,
4389 * otherwise it will be left as-is. Example:
4391 * %define %$abc cde
4393 * the identifier %$abc will be left as-is so that the handler for %define
4394 * will suck it and define the corresponding value. Other case:
4396 * %define _%$abc cde
4398 * In this case user wants name to be expanded *before* %define starts
4399 * working, so we'll expand %$abc into something (if it has a value;
4400 * otherwise it will be left as-is) then concatenate all successive
4401 * PP_IDs into one.
4403 static Token *expand_id(Token * tline)
4405 Token *cur, *oldnext = NULL;
4407 if (!tline || !tline->next)
4408 return tline;
4410 cur = tline;
4411 while (cur->next &&
4412 (cur->next->type == TOK_ID ||
4413 cur->next->type == TOK_PREPROC_ID
4414 || cur->next->type == TOK_NUMBER))
4415 cur = cur->next;
4417 /* If identifier consists of just one token, don't expand */
4418 if (cur == tline)
4419 return tline;
4421 if (cur) {
4422 oldnext = cur->next; /* Detach the tail past identifier */
4423 cur->next = NULL; /* so that expand_smacro stops here */
4426 tline = expand_smacro(tline);
4428 if (cur) {
4429 /* expand_smacro possibly changhed tline; re-scan for EOL */
4430 cur = tline;
4431 while (cur && cur->next)
4432 cur = cur->next;
4433 if (cur)
4434 cur->next = oldnext;
4437 return tline;
4441 * Determine whether the given line constitutes a multi-line macro
4442 * call, and return the MMacro structure called if so. Doesn't have
4443 * to check for an initial label - that's taken care of in
4444 * expand_mmacro - but must check numbers of parameters. Guaranteed
4445 * to be called with tline->type == TOK_ID, so the putative macro
4446 * name is easy to find.
4448 static MMacro *is_mmacro(Token * tline, Token *** params_array)
4450 MMacro *head, *m;
4451 Token **params;
4452 int nparam;
4454 head = (MMacro *) hash_findix(&mmacros, tline->text);
4457 * Efficiency: first we see if any macro exists with the given
4458 * name. If not, we can return NULL immediately. _Then_ we
4459 * count the parameters, and then we look further along the
4460 * list if necessary to find the proper MMacro.
4462 list_for_each(m, head)
4463 if (!mstrcmp(m->name, tline->text, m->casesense))
4464 break;
4465 if (!m)
4466 return NULL;
4469 * OK, we have a potential macro. Count and demarcate the
4470 * parameters.
4472 count_mmac_params(tline->next, &nparam, &params);
4475 * So we know how many parameters we've got. Find the MMacro
4476 * structure that handles this number.
4478 while (m) {
4479 if (m->nparam_min <= nparam
4480 && (m->plus || nparam <= m->nparam_max)) {
4482 * This one is right. Just check if cycle removal
4483 * prohibits us using it before we actually celebrate...
4485 if (m->in_progress > m->max_depth) {
4486 if (m->max_depth > 0) {
4487 nasm_error(ERR_WARNING,
4488 "reached maximum recursion depth of %i",
4489 m->max_depth);
4491 nasm_free(params);
4492 return NULL;
4495 * It's right, and we can use it. Add its default
4496 * parameters to the end of our list if necessary.
4498 if (m->defaults && nparam < m->nparam_min + m->ndefs) {
4499 params =
4500 nasm_realloc(params,
4501 ((m->nparam_min + m->ndefs +
4502 1) * sizeof(*params)));
4503 while (nparam < m->nparam_min + m->ndefs) {
4504 params[nparam] = m->defaults[nparam - m->nparam_min];
4505 nparam++;
4509 * If we've gone over the maximum parameter count (and
4510 * we're in Plus mode), ignore parameters beyond
4511 * nparam_max.
4513 if (m->plus && nparam > m->nparam_max)
4514 nparam = m->nparam_max;
4516 * Then terminate the parameter list, and leave.
4518 if (!params) { /* need this special case */
4519 params = nasm_malloc(sizeof(*params));
4520 nparam = 0;
4522 params[nparam] = NULL;
4523 *params_array = params;
4524 return m;
4527 * This one wasn't right: look for the next one with the
4528 * same name.
4530 list_for_each(m, m->next)
4531 if (!mstrcmp(m->name, tline->text, m->casesense))
4532 break;
4536 * After all that, we didn't find one with the right number of
4537 * parameters. Issue a warning, and fail to expand the macro.
4539 nasm_error(ERR_WARNING|ERR_PASS1|ERR_WARN_MNP,
4540 "macro `%s' exists, but not taking %d parameters",
4541 tline->text, nparam);
4542 nasm_free(params);
4543 return NULL;
4548 * Save MMacro invocation specific fields in
4549 * preparation for a recursive macro expansion
4551 static void push_mmacro(MMacro *m)
4553 MMacroInvocation *i;
4555 i = nasm_malloc(sizeof(MMacroInvocation));
4556 i->prev = m->prev;
4557 i->params = m->params;
4558 i->iline = m->iline;
4559 i->nparam = m->nparam;
4560 i->rotate = m->rotate;
4561 i->paramlen = m->paramlen;
4562 i->unique = m->unique;
4563 i->condcnt = m->condcnt;
4564 m->prev = i;
4569 * Restore MMacro invocation specific fields that were
4570 * saved during a previous recursive macro expansion
4572 static void pop_mmacro(MMacro *m)
4574 MMacroInvocation *i;
4576 if (m->prev) {
4577 i = m->prev;
4578 m->prev = i->prev;
4579 m->params = i->params;
4580 m->iline = i->iline;
4581 m->nparam = i->nparam;
4582 m->rotate = i->rotate;
4583 m->paramlen = i->paramlen;
4584 m->unique = i->unique;
4585 m->condcnt = i->condcnt;
4586 nasm_free(i);
4592 * Expand the multi-line macro call made by the given line, if
4593 * there is one to be expanded. If there is, push the expansion on
4594 * istk->expansion and return 1. Otherwise return 0.
4596 static int expand_mmacro(Token * tline)
4598 Token *startline = tline;
4599 Token *label = NULL;
4600 int dont_prepend = 0;
4601 Token **params, *t, *tt;
4602 MMacro *m;
4603 Line *l, *ll;
4604 int i, nparam, *paramlen;
4605 const char *mname;
4607 t = tline;
4608 skip_white_(t);
4609 /* if (!tok_type_(t, TOK_ID)) Lino 02/25/02 */
4610 if (!tok_type_(t, TOK_ID) && !tok_type_(t, TOK_PREPROC_ID))
4611 return 0;
4612 m = is_mmacro(t, &params);
4613 if (m) {
4614 mname = t->text;
4615 } else {
4616 Token *last;
4618 * We have an id which isn't a macro call. We'll assume
4619 * it might be a label; we'll also check to see if a
4620 * colon follows it. Then, if there's another id after
4621 * that lot, we'll check it again for macro-hood.
4623 label = last = t;
4624 t = t->next;
4625 if (tok_type_(t, TOK_WHITESPACE))
4626 last = t, t = t->next;
4627 if (tok_is_(t, ":")) {
4628 dont_prepend = 1;
4629 last = t, t = t->next;
4630 if (tok_type_(t, TOK_WHITESPACE))
4631 last = t, t = t->next;
4633 if (!tok_type_(t, TOK_ID) || !(m = is_mmacro(t, &params)))
4634 return 0;
4635 last->next = NULL;
4636 mname = t->text;
4637 tline = t;
4641 * Fix up the parameters: this involves stripping leading and
4642 * trailing whitespace, then stripping braces if they are
4643 * present.
4645 for (nparam = 0; params[nparam]; nparam++) ;
4646 paramlen = nparam ? nasm_malloc(nparam * sizeof(*paramlen)) : NULL;
4648 for (i = 0; params[i]; i++) {
4649 int brace = 0;
4650 int comma = (!m->plus || i < nparam - 1);
4652 t = params[i];
4653 skip_white_(t);
4654 if (tok_is_(t, "{"))
4655 t = t->next, brace++, comma = false;
4656 params[i] = t;
4657 paramlen[i] = 0;
4658 while (t) {
4659 if (comma && t->type == TOK_OTHER && !strcmp(t->text, ","))
4660 break; /* ... because we have hit a comma */
4661 if (comma && t->type == TOK_WHITESPACE
4662 && tok_is_(t->next, ","))
4663 break; /* ... or a space then a comma */
4664 if (brace && t->type == TOK_OTHER) {
4665 if (t->text[0] == '{')
4666 brace++; /* ... or a nested opening brace */
4667 else if (t->text[0] == '}')
4668 if (!--brace)
4669 break; /* ... or a brace */
4671 t = t->next;
4672 paramlen[i]++;
4674 if (brace)
4675 nasm_error(ERR_NONFATAL, "macro params should be enclosed in braces");
4679 * OK, we have a MMacro structure together with a set of
4680 * parameters. We must now go through the expansion and push
4681 * copies of each Line on to istk->expansion. Substitution of
4682 * parameter tokens and macro-local tokens doesn't get done
4683 * until the single-line macro substitution process; this is
4684 * because delaying them allows us to change the semantics
4685 * later through %rotate.
4687 * First, push an end marker on to istk->expansion, mark this
4688 * macro as in progress, and set up its invocation-specific
4689 * variables.
4691 ll = nasm_malloc(sizeof(Line));
4692 ll->next = istk->expansion;
4693 ll->finishes = m;
4694 ll->first = NULL;
4695 istk->expansion = ll;
4698 * Save the previous MMacro expansion in the case of
4699 * macro recursion
4701 if (m->max_depth && m->in_progress)
4702 push_mmacro(m);
4704 m->in_progress ++;
4705 m->params = params;
4706 m->iline = tline;
4707 m->nparam = nparam;
4708 m->rotate = 0;
4709 m->paramlen = paramlen;
4710 m->unique = unique++;
4711 m->lineno = 0;
4712 m->condcnt = 0;
4714 m->next_active = istk->mstk;
4715 istk->mstk = m;
4717 list_for_each(l, m->expansion) {
4718 Token **tail;
4720 ll = nasm_malloc(sizeof(Line));
4721 ll->finishes = NULL;
4722 ll->next = istk->expansion;
4723 istk->expansion = ll;
4724 tail = &ll->first;
4726 list_for_each(t, l->first) {
4727 Token *x = t;
4728 switch (t->type) {
4729 case TOK_PREPROC_Q:
4730 tt = *tail = new_Token(NULL, TOK_ID, mname, 0);
4731 break;
4732 case TOK_PREPROC_QQ:
4733 tt = *tail = new_Token(NULL, TOK_ID, m->name, 0);
4734 break;
4735 case TOK_PREPROC_ID:
4736 if (t->text[1] == '0' && t->text[2] == '0') {
4737 dont_prepend = -1;
4738 x = label;
4739 if (!x)
4740 continue;
4742 /* fall through */
4743 default:
4744 tt = *tail = new_Token(NULL, x->type, x->text, 0);
4745 break;
4747 tail = &tt->next;
4749 *tail = NULL;
4753 * If we had a label, push it on as the first line of
4754 * the macro expansion.
4756 if (label) {
4757 if (dont_prepend < 0)
4758 free_tlist(startline);
4759 else {
4760 ll = nasm_malloc(sizeof(Line));
4761 ll->finishes = NULL;
4762 ll->next = istk->expansion;
4763 istk->expansion = ll;
4764 ll->first = startline;
4765 if (!dont_prepend) {
4766 while (label->next)
4767 label = label->next;
4768 label->next = tt = new_Token(NULL, TOK_OTHER, ":", 0);
4773 lfmt->uplevel(m->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
4775 return 1;
4779 * This function adds macro names to error messages, and suppresses
4780 * them if necessary.
4782 static void pp_verror(int severity, const char *fmt, va_list arg)
4784 char buff[BUFSIZ];
4785 MMacro *mmac = NULL;
4786 int delta = 0;
4789 * If we're in a dead branch of IF or something like it, ignore the error.
4790 * However, because %else etc are evaluated in the state context
4791 * of the previous branch, errors might get lost:
4792 * %if 0 ... %else trailing garbage ... %endif
4793 * So %else etc should set the ERR_PP_PRECOND flag.
4795 if ((severity & ERR_MASK) < ERR_FATAL &&
4796 istk && istk->conds &&
4797 ((severity & ERR_PP_PRECOND) ?
4798 istk->conds->state == COND_NEVER :
4799 emitting(istk->conds->state)))
4800 return;
4802 /* get %macro name */
4803 if (!(severity & ERR_NOFILE) && istk && istk->mstk) {
4804 mmac = istk->mstk;
4805 /* but %rep blocks should be skipped */
4806 while (mmac && !mmac->name)
4807 mmac = mmac->next_active, delta++;
4810 if (mmac) {
4811 vsnprintf(buff, sizeof(buff), fmt, arg);
4813 nasm_set_verror(real_verror);
4814 nasm_error(severity, "(%s:%d) %s",
4815 mmac->name, mmac->lineno - delta, buff);
4816 nasm_set_verror(pp_verror);
4817 } else {
4818 real_verror(severity, fmt, arg);
4822 static void
4823 pp_reset(char *file, int apass, StrList **deplist)
4825 Token *t;
4827 cstk = NULL;
4828 istk = nasm_malloc(sizeof(Include));
4829 istk->next = NULL;
4830 istk->conds = NULL;
4831 istk->expansion = NULL;
4832 istk->mstk = NULL;
4833 istk->fp = fopen(file, "r");
4834 istk->fname = NULL;
4835 src_set_fname(nasm_strdup(file));
4836 src_set_linnum(0);
4837 istk->lineinc = 1;
4838 if (!istk->fp)
4839 nasm_fatal(ERR_NOFILE, "unable to open input file `%s'", file);
4840 defining = NULL;
4841 nested_mac_count = 0;
4842 nested_rep_count = 0;
4843 init_macros();
4844 unique = 0;
4845 if (tasm_compatible_mode) {
4846 stdmacpos = nasm_stdmac;
4847 } else {
4848 stdmacpos = nasm_stdmac_after_tasm;
4850 any_extrastdmac = extrastdmac && *extrastdmac;
4851 do_predef = true;
4854 * 0 for dependencies, 1 for preparatory passes, 2 for final pass.
4855 * The caller, however, will also pass in 3 for preprocess-only so
4856 * we can set __PASS__ accordingly.
4858 pass = apass > 2 ? 2 : apass;
4860 dephead = deptail = deplist;
4861 if (deplist) {
4862 StrList *sl = nasm_malloc(strlen(file)+1+sizeof sl->next);
4863 sl->next = NULL;
4864 strcpy(sl->str, file);
4865 *deptail = sl;
4866 deptail = &sl->next;
4870 * Define the __PASS__ macro. This is defined here unlike
4871 * all the other builtins, because it is special -- it varies between
4872 * passes.
4874 t = nasm_malloc(sizeof(*t));
4875 t->next = NULL;
4876 make_tok_num(t, apass);
4877 t->a.mac = NULL;
4878 define_smacro(NULL, "__PASS__", true, 0, t);
4881 static char *pp_getline(void)
4883 char *line;
4884 Token *tline;
4886 real_verror = nasm_set_verror(pp_verror);
4888 while (1) {
4890 * Fetch a tokenized line, either from the macro-expansion
4891 * buffer or from the input file.
4893 tline = NULL;
4894 while (istk->expansion && istk->expansion->finishes) {
4895 Line *l = istk->expansion;
4896 if (!l->finishes->name && l->finishes->in_progress > 1) {
4897 Line *ll;
4900 * This is a macro-end marker for a macro with no
4901 * name, which means it's not really a macro at all
4902 * but a %rep block, and the `in_progress' field is
4903 * more than 1, meaning that we still need to
4904 * repeat. (1 means the natural last repetition; 0
4905 * means termination by %exitrep.) We have
4906 * therefore expanded up to the %endrep, and must
4907 * push the whole block on to the expansion buffer
4908 * again. We don't bother to remove the macro-end
4909 * marker: we'd only have to generate another one
4910 * if we did.
4912 l->finishes->in_progress--;
4913 list_for_each(l, l->finishes->expansion) {
4914 Token *t, *tt, **tail;
4916 ll = nasm_malloc(sizeof(Line));
4917 ll->next = istk->expansion;
4918 ll->finishes = NULL;
4919 ll->first = NULL;
4920 tail = &ll->first;
4922 list_for_each(t, l->first) {
4923 if (t->text || t->type == TOK_WHITESPACE) {
4924 tt = *tail = new_Token(NULL, t->type, t->text, 0);
4925 tail = &tt->next;
4929 istk->expansion = ll;
4931 } else {
4933 * Check whether a `%rep' was started and not ended
4934 * within this macro expansion. This can happen and
4935 * should be detected. It's a fatal error because
4936 * I'm too confused to work out how to recover
4937 * sensibly from it.
4939 if (defining) {
4940 if (defining->name)
4941 nasm_panic(0, "defining with name in expansion");
4942 else if (istk->mstk->name)
4943 nasm_fatal(0, "`%%rep' without `%%endrep' within"
4944 " expansion of macro `%s'",
4945 istk->mstk->name);
4949 * FIXME: investigate the relationship at this point between
4950 * istk->mstk and l->finishes
4953 MMacro *m = istk->mstk;
4954 istk->mstk = m->next_active;
4955 if (m->name) {
4957 * This was a real macro call, not a %rep, and
4958 * therefore the parameter information needs to
4959 * be freed.
4961 if (m->prev) {
4962 pop_mmacro(m);
4963 l->finishes->in_progress --;
4964 } else {
4965 nasm_free(m->params);
4966 free_tlist(m->iline);
4967 nasm_free(m->paramlen);
4968 l->finishes->in_progress = 0;
4970 } else
4971 free_mmacro(m);
4973 istk->expansion = l->next;
4974 nasm_free(l);
4975 lfmt->downlevel(LIST_MACRO);
4978 while (1) { /* until we get a line we can use */
4980 if (istk->expansion) { /* from a macro expansion */
4981 char *p;
4982 Line *l = istk->expansion;
4983 if (istk->mstk)
4984 istk->mstk->lineno++;
4985 tline = l->first;
4986 istk->expansion = l->next;
4987 nasm_free(l);
4988 p = detoken(tline, false);
4989 lfmt->line(LIST_MACRO, p);
4990 nasm_free(p);
4991 break;
4993 line = read_line();
4994 if (line) { /* from the current input file */
4995 line = prepreproc(line);
4996 tline = tokenize(line);
4997 nasm_free(line);
4998 break;
5001 * The current file has ended; work down the istk
5004 Include *i = istk;
5005 fclose(i->fp);
5006 if (i->conds) {
5007 /* nasm_error can't be conditionally suppressed */
5008 nasm_error(ERR_FATAL,
5009 "expected `%%endif' before end of file");
5011 /* only set line and file name if there's a next node */
5012 if (i->next) {
5013 src_set_linnum(i->lineno);
5014 nasm_free(src_set_fname(nasm_strdup(i->fname)));
5016 istk = i->next;
5017 lfmt->downlevel(LIST_INCLUDE);
5018 nasm_free(i);
5019 if (!istk) {
5020 line = NULL;
5021 goto done;
5023 if (istk->expansion && istk->expansion->finishes)
5024 break;
5029 * We must expand MMacro parameters and MMacro-local labels
5030 * _before_ we plunge into directive processing, to cope
5031 * with things like `%define something %1' such as STRUC
5032 * uses. Unless we're _defining_ a MMacro, in which case
5033 * those tokens should be left alone to go into the
5034 * definition; and unless we're in a non-emitting
5035 * condition, in which case we don't want to meddle with
5036 * anything.
5038 if (!defining && !(istk->conds && !emitting(istk->conds->state))
5039 && !(istk->mstk && !istk->mstk->in_progress)) {
5040 tline = expand_mmac_params(tline);
5044 * Check the line to see if it's a preprocessor directive.
5046 if (do_directive(tline) == DIRECTIVE_FOUND) {
5047 continue;
5048 } else if (defining) {
5050 * We're defining a multi-line macro. We emit nothing
5051 * at all, and just
5052 * shove the tokenized line on to the macro definition.
5054 Line *l = nasm_malloc(sizeof(Line));
5055 l->next = defining->expansion;
5056 l->first = tline;
5057 l->finishes = NULL;
5058 defining->expansion = l;
5059 continue;
5060 } else if (istk->conds && !emitting(istk->conds->state)) {
5062 * We're in a non-emitting branch of a condition block.
5063 * Emit nothing at all, not even a blank line: when we
5064 * emerge from the condition we'll give a line-number
5065 * directive so we keep our place correctly.
5067 free_tlist(tline);
5068 continue;
5069 } else if (istk->mstk && !istk->mstk->in_progress) {
5071 * We're in a %rep block which has been terminated, so
5072 * we're walking through to the %endrep without
5073 * emitting anything. Emit nothing at all, not even a
5074 * blank line: when we emerge from the %rep block we'll
5075 * give a line-number directive so we keep our place
5076 * correctly.
5078 free_tlist(tline);
5079 continue;
5080 } else {
5081 tline = expand_smacro(tline);
5082 if (!expand_mmacro(tline)) {
5084 * De-tokenize the line again, and emit it.
5086 line = detoken(tline, true);
5087 free_tlist(tline);
5088 break;
5089 } else {
5090 continue; /* expand_mmacro calls free_tlist */
5095 done:
5096 nasm_set_verror(real_verror);
5097 return line;
5100 static void pp_cleanup(int pass)
5102 real_verror = nasm_set_verror(pp_verror);
5104 if (defining) {
5105 if (defining->name) {
5106 nasm_error(ERR_NONFATAL,
5107 "end of file while still defining macro `%s'",
5108 defining->name);
5109 } else {
5110 nasm_error(ERR_NONFATAL, "end of file while still in %%rep");
5113 free_mmacro(defining);
5114 defining = NULL;
5117 nasm_set_verror(real_verror);
5119 while (cstk)
5120 ctx_pop();
5121 free_macros();
5122 while (istk) {
5123 Include *i = istk;
5124 istk = istk->next;
5125 fclose(i->fp);
5126 nasm_free(i->fname);
5127 nasm_free(i);
5129 while (cstk)
5130 ctx_pop();
5131 nasm_free(src_set_fname(NULL));
5132 if (pass == 0) {
5133 IncPath *i;
5134 free_llist(predef);
5135 predef = NULL;
5136 delete_Blocks();
5137 freeTokens = NULL;
5138 while ((i = ipath)) {
5139 ipath = i->next;
5140 if (i->path)
5141 nasm_free(i->path);
5142 nasm_free(i);
5147 static void pp_include_path(char *path)
5149 IncPath *i;
5151 i = nasm_malloc(sizeof(IncPath));
5152 i->path = path ? nasm_strdup(path) : NULL;
5153 i->next = NULL;
5155 if (ipath) {
5156 IncPath *j = ipath;
5157 while (j->next)
5158 j = j->next;
5159 j->next = i;
5160 } else {
5161 ipath = i;
5165 static void pp_pre_include(char *fname)
5167 Token *inc, *space, *name;
5168 Line *l;
5170 name = new_Token(NULL, TOK_INTERNAL_STRING, fname, 0);
5171 space = new_Token(name, TOK_WHITESPACE, NULL, 0);
5172 inc = new_Token(space, TOK_PREPROC_ID, "%include", 0);
5174 l = nasm_malloc(sizeof(Line));
5175 l->next = predef;
5176 l->first = inc;
5177 l->finishes = NULL;
5178 predef = l;
5181 static void pp_pre_define(char *definition)
5183 Token *def, *space;
5184 Line *l;
5185 char *equals;
5187 real_verror = nasm_set_verror(pp_verror);
5189 equals = strchr(definition, '=');
5190 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
5191 def = new_Token(space, TOK_PREPROC_ID, "%define", 0);
5192 if (equals)
5193 *equals = ' ';
5194 space->next = tokenize(definition);
5195 if (equals)
5196 *equals = '=';
5198 if (space->next->type != TOK_PREPROC_ID &&
5199 space->next->type != TOK_ID)
5200 nasm_error(ERR_WARNING, "pre-defining non ID `%s\'\n", definition);
5202 l = nasm_malloc(sizeof(Line));
5203 l->next = predef;
5204 l->first = def;
5205 l->finishes = NULL;
5206 predef = l;
5208 nasm_set_verror(real_verror);
5211 static void pp_pre_undefine(char *definition)
5213 Token *def, *space;
5214 Line *l;
5216 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
5217 def = new_Token(space, TOK_PREPROC_ID, "%undef", 0);
5218 space->next = tokenize(definition);
5220 l = nasm_malloc(sizeof(Line));
5221 l->next = predef;
5222 l->first = def;
5223 l->finishes = NULL;
5224 predef = l;
5227 static void pp_extra_stdmac(macros_t *macros)
5229 extrastdmac = macros;
5232 static void make_tok_num(Token * tok, int64_t val)
5234 char numbuf[32];
5235 snprintf(numbuf, sizeof(numbuf), "%"PRId64"", val);
5236 tok->text = nasm_strdup(numbuf);
5237 tok->type = TOK_NUMBER;
5240 struct preproc_ops nasmpp = {
5241 pp_reset,
5242 pp_getline,
5243 pp_cleanup,
5244 pp_extra_stdmac,
5245 pp_pre_define,
5246 pp_pre_undefine,
5247 pp_pre_include,
5248 pp_include_path