preproc: Don't access out of bound data on malformed input
[nasm.git] / asm / preproc.c
blobecf89f1b8474404db8692866fad14ba865909032
1 /* ----------------------------------------------------------------------- *
3 * Copyright 1996-2018 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>
73 #include "nasm.h"
74 #include "nasmlib.h"
75 #include "error.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... */
159 const char *fname; /* File where defined */
160 int32_t xline; /* First line in macro */
164 /* Store the definition of a multi-line macro, as defined in a
165 * previous recursive macro expansion.
167 struct MMacroInvocation {
168 MMacroInvocation *prev; /* previous invocation */
169 Token **params; /* actual parameters */
170 Token *iline; /* invocation line */
171 unsigned int nparam, rotate;
172 int *paramlen;
173 uint64_t unique;
174 uint64_t condcnt;
179 * The context stack is composed of a linked list of these.
181 struct Context {
182 Context *next;
183 char *name;
184 struct hash_table localmac;
185 uint32_t number;
189 * This is the internal form which we break input lines up into.
190 * Typically stored in linked lists.
192 * Note that `type' serves a double meaning: TOK_SMAC_PARAM is not
193 * necessarily used as-is, but is intended to denote the number of
194 * the substituted parameter. So in the definition
196 * %define a(x,y) ( (x) & ~(y) )
198 * the token representing `x' will have its type changed to
199 * TOK_SMAC_PARAM, but the one representing `y' will be
200 * TOK_SMAC_PARAM+1.
202 * TOK_INTERNAL_STRING is a dirty hack: it's a single string token
203 * which doesn't need quotes around it. Used in the pre-include
204 * mechanism as an alternative to trying to find a sensible type of
205 * quote to use on the filename we were passed.
207 enum pp_token_type {
208 TOK_NONE = 0, TOK_WHITESPACE, TOK_COMMENT, TOK_ID,
209 TOK_PREPROC_ID, TOK_STRING,
210 TOK_NUMBER, TOK_FLOAT, TOK_SMAC_END, TOK_OTHER,
211 TOK_INTERNAL_STRING,
212 TOK_PREPROC_Q, TOK_PREPROC_QQ,
213 TOK_PASTE, /* %+ */
214 TOK_INDIRECT, /* %[...] */
215 TOK_SMAC_PARAM, /* MUST BE LAST IN THE LIST!!! */
216 TOK_MAX = INT_MAX /* Keep compiler from reducing the range */
219 #define PP_CONCAT_MASK(x) (1 << (x))
220 #define PP_CONCAT_MATCH(t, mask) (PP_CONCAT_MASK((t)->type) & mask)
222 struct tokseq_match {
223 int mask_head;
224 int mask_tail;
227 struct Token {
228 Token *next;
229 char *text;
230 union {
231 SMacro *mac; /* associated macro for TOK_SMAC_END */
232 size_t len; /* scratch length field */
233 } a; /* Auxiliary data */
234 enum pp_token_type type;
238 * Multi-line macro definitions are stored as a linked list of
239 * these, which is essentially a container to allow several linked
240 * lists of Tokens.
242 * Note that in this module, linked lists are treated as stacks
243 * wherever possible. For this reason, Lines are _pushed_ on to the
244 * `expansion' field in MMacro structures, so that the linked list,
245 * if walked, would give the macro lines in reverse order; this
246 * means that we can walk the list when expanding a macro, and thus
247 * push the lines on to the `expansion' field in _istk_ in reverse
248 * order (so that when popped back off they are in the right
249 * order). It may seem cockeyed, and it relies on my design having
250 * an even number of steps in, but it works...
252 * Some of these structures, rather than being actual lines, are
253 * markers delimiting the end of the expansion of a given macro.
254 * This is for use in the cycle-tracking and %rep-handling code.
255 * Such structures have `finishes' non-NULL, and `first' NULL. All
256 * others have `finishes' NULL, but `first' may still be NULL if
257 * the line is blank.
259 struct Line {
260 Line *next;
261 MMacro *finishes;
262 Token *first;
266 * To handle an arbitrary level of file inclusion, we maintain a
267 * stack (ie linked list) of these things.
269 struct Include {
270 Include *next;
271 FILE *fp;
272 Cond *conds;
273 Line *expansion;
274 const char *fname;
275 int lineno, lineinc;
276 MMacro *mstk; /* stack of active macros/reps */
280 * Include search path. This is simply a list of strings which get
281 * prepended, in turn, to the name of an include file, in an
282 * attempt to find the file if it's not in the current directory.
284 struct IncPath {
285 IncPath *next;
286 char *path;
290 * File real name hash, so we don't have to re-search the include
291 * path for every pass (and potentially more than that if a file
292 * is used more than once.)
294 struct hash_table FileHash;
297 * Conditional assembly: we maintain a separate stack of these for
298 * each level of file inclusion. (The only reason we keep the
299 * stacks separate is to ensure that a stray `%endif' in a file
300 * included from within the true branch of a `%if' won't terminate
301 * it and cause confusion: instead, rightly, it'll cause an error.)
303 struct Cond {
304 Cond *next;
305 int state;
307 enum {
309 * These states are for use just after %if or %elif: IF_TRUE
310 * means the condition has evaluated to truth so we are
311 * currently emitting, whereas IF_FALSE means we are not
312 * currently emitting but will start doing so if a %else comes
313 * up. In these states, all directives are admissible: %elif,
314 * %else and %endif. (And of course %if.)
316 COND_IF_TRUE, COND_IF_FALSE,
318 * These states come up after a %else: ELSE_TRUE means we're
319 * emitting, and ELSE_FALSE means we're not. In ELSE_* states,
320 * any %elif or %else will cause an error.
322 COND_ELSE_TRUE, COND_ELSE_FALSE,
324 * These states mean that we're not emitting now, and also that
325 * nothing until %endif will be emitted at all. COND_DONE is
326 * used when we've had our moment of emission
327 * and have now started seeing %elifs. COND_NEVER is used when
328 * the condition construct in question is contained within a
329 * non-emitting branch of a larger condition construct,
330 * or if there is an error.
332 COND_DONE, COND_NEVER
334 #define emitting(x) ( (x) == COND_IF_TRUE || (x) == COND_ELSE_TRUE )
337 * These defines are used as the possible return values for do_directive
339 #define NO_DIRECTIVE_FOUND 0
340 #define DIRECTIVE_FOUND 1
342 /* max reps */
343 #define REP_LIMIT ((INT64_C(1) << 62))
346 * Condition codes. Note that we use c_ prefix not C_ because C_ is
347 * used in nasm.h for the "real" condition codes. At _this_ level,
348 * we treat CXZ and ECXZ as condition codes, albeit non-invertible
349 * ones, so we need a different enum...
351 static const char * const conditions[] = {
352 "a", "ae", "b", "be", "c", "cxz", "e", "ecxz", "g", "ge", "l", "le",
353 "na", "nae", "nb", "nbe", "nc", "ne", "ng", "nge", "nl", "nle", "no",
354 "np", "ns", "nz", "o", "p", "pe", "po", "rcxz", "s", "z"
356 enum pp_conds {
357 c_A, c_AE, c_B, c_BE, c_C, c_CXZ, c_E, c_ECXZ, c_G, c_GE, c_L, c_LE,
358 c_NA, c_NAE, c_NB, c_NBE, c_NC, c_NE, c_NG, c_NGE, c_NL, c_NLE, c_NO,
359 c_NP, c_NS, c_NZ, c_O, c_P, c_PE, c_PO, c_RCXZ, c_S, c_Z,
360 c_none = -1
362 static const enum pp_conds inverse_ccs[] = {
363 c_NA, c_NAE, c_NB, c_NBE, c_NC, -1, c_NE, -1, c_NG, c_NGE, c_NL, c_NLE,
364 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,
365 c_Z, c_NO, c_NP, c_PO, c_PE, -1, c_NS, c_NZ
369 * Directive names.
371 /* If this is a an IF, ELIF, ELSE or ENDIF keyword */
372 static int is_condition(enum preproc_token arg)
374 return PP_IS_COND(arg) || (arg == PP_ELSE) || (arg == PP_ENDIF);
377 /* For TASM compatibility we need to be able to recognise TASM compatible
378 * conditional compilation directives. Using the NASM pre-processor does
379 * not work, so we look for them specifically from the following list and
380 * then jam in the equivalent NASM directive into the input stream.
383 enum {
384 TM_ARG, TM_ELIF, TM_ELSE, TM_ENDIF, TM_IF, TM_IFDEF, TM_IFDIFI,
385 TM_IFNDEF, TM_INCLUDE, TM_LOCAL
388 static const char * const tasm_directives[] = {
389 "arg", "elif", "else", "endif", "if", "ifdef", "ifdifi",
390 "ifndef", "include", "local"
393 static int StackSize = 4;
394 static const char *StackPointer = "ebp";
395 static int ArgOffset = 8;
396 static int LocalOffset = 0;
398 static Context *cstk;
399 static Include *istk;
400 static IncPath *ipath = NULL;
402 static int pass; /* HACK: pass 0 = generate dependencies only */
403 static StrList **dephead;
405 static uint64_t unique; /* unique identifier numbers */
407 static Line *predef = NULL;
408 static bool do_predef;
411 * The current set of multi-line macros we have defined.
413 static struct hash_table mmacros;
416 * The current set of single-line macros we have defined.
418 static struct hash_table smacros;
421 * The multi-line macro we are currently defining, or the %rep
422 * block we are currently reading, if any.
424 static MMacro *defining;
426 static uint64_t nested_mac_count;
427 static uint64_t nested_rep_count;
430 * The number of macro parameters to allocate space for at a time.
432 #define PARAM_DELTA 16
435 * The standard macro set: defined in macros.c in a set of arrays.
436 * This gives our position in any macro set, while we are processing it.
437 * The stdmacset is an array of such macro sets.
439 static macros_t *stdmacpos;
440 static macros_t **stdmacnext;
441 static macros_t *stdmacros[8];
442 static macros_t *extrastdmac;
445 * Tokens are allocated in blocks to improve speed
447 #define TOKEN_BLOCKSIZE 4096
448 static Token *freeTokens = NULL;
449 struct Blocks {
450 Blocks *next;
451 void *chunk;
454 static Blocks blocks = { NULL, NULL };
457 * Forward declarations.
459 static void pp_add_stdmac(macros_t *macros);
460 static Token *expand_mmac_params(Token * tline);
461 static Token *expand_smacro(Token * tline);
462 static Token *expand_id(Token * tline);
463 static Context *get_ctx(const char *name, const char **namep);
464 static void make_tok_num(Token * tok, int64_t val);
465 static void pp_verror(int severity, const char *fmt, va_list ap);
466 static vefunc real_verror;
467 static void *new_Block(size_t size);
468 static void delete_Blocks(void);
469 static Token *new_Token(Token * next, enum pp_token_type type,
470 const char *text, int txtlen);
471 static Token *delete_Token(Token * t);
474 * Macros for safe checking of token pointers, avoid *(NULL)
476 #define tok_type_(x,t) ((x) && (x)->type == (t))
477 #define skip_white_(x) if (tok_type_((x), TOK_WHITESPACE)) (x)=(x)->next
478 #define tok_is_(x,v) (tok_type_((x), TOK_OTHER) && !strcmp((x)->text,(v)))
479 #define tok_isnt_(x,v) ((x) && ((x)->type!=TOK_OTHER || strcmp((x)->text,(v))))
482 * nasm_unquote with error if the string contains NUL characters.
483 * If the string contains NUL characters, issue an error and return
484 * the C len, i.e. truncate at the NUL.
486 static size_t nasm_unquote_cstr(char *qstr, enum preproc_token directive)
488 size_t len = nasm_unquote(qstr, NULL);
489 size_t clen = strlen(qstr);
491 if (len != clen)
492 nasm_error(ERR_NONFATAL, "NUL character in `%s' directive",
493 pp_directives[directive]);
495 return clen;
499 * In-place reverse a list of tokens.
501 static Token *reverse_tokens(Token *t)
503 Token *prev = NULL;
504 Token *next;
506 while (t) {
507 next = t->next;
508 t->next = prev;
509 prev = t;
510 t = next;
513 return prev;
517 * Handle TASM specific directives, which do not contain a % in
518 * front of them. We do it here because I could not find any other
519 * place to do it for the moment, and it is a hack (ideally it would
520 * be nice to be able to use the NASM pre-processor to do it).
522 static char *check_tasm_directive(char *line)
524 int32_t i, j, k, m, len;
525 char *p, *q, *oldline, oldchar;
527 p = nasm_skip_spaces(line);
529 /* Binary search for the directive name */
530 i = -1;
531 j = ARRAY_SIZE(tasm_directives);
532 q = nasm_skip_word(p);
533 len = q - p;
534 if (len) {
535 oldchar = p[len];
536 p[len] = 0;
537 while (j - i > 1) {
538 k = (j + i) / 2;
539 m = nasm_stricmp(p, tasm_directives[k]);
540 if (m == 0) {
541 /* We have found a directive, so jam a % in front of it
542 * so that NASM will then recognise it as one if it's own.
544 p[len] = oldchar;
545 len = strlen(p);
546 oldline = line;
547 line = nasm_malloc(len + 2);
548 line[0] = '%';
549 if (k == TM_IFDIFI) {
551 * NASM does not recognise IFDIFI, so we convert
552 * it to %if 0. This is not used in NASM
553 * compatible code, but does need to parse for the
554 * TASM macro package.
556 strcpy(line + 1, "if 0");
557 } else {
558 memcpy(line + 1, p, len + 1);
560 nasm_free(oldline);
561 return line;
562 } else if (m < 0) {
563 j = k;
564 } else
565 i = k;
567 p[len] = oldchar;
569 return line;
573 * The pre-preprocessing stage... This function translates line
574 * number indications as they emerge from GNU cpp (`# lineno "file"
575 * flags') into NASM preprocessor line number indications (`%line
576 * lineno file').
578 static char *prepreproc(char *line)
580 int lineno, fnlen;
581 char *fname, *oldline;
583 if (line[0] == '#' && line[1] == ' ') {
584 oldline = line;
585 fname = oldline + 2;
586 lineno = atoi(fname);
587 fname += strspn(fname, "0123456789 ");
588 if (*fname == '"')
589 fname++;
590 fnlen = strcspn(fname, "\"");
591 line = nasm_malloc(20 + fnlen);
592 snprintf(line, 20 + fnlen, "%%line %d %.*s", lineno, fnlen, fname);
593 nasm_free(oldline);
595 if (tasm_compatible_mode)
596 return check_tasm_directive(line);
597 return line;
601 * Free a linked list of tokens.
603 static void free_tlist(Token * list)
605 while (list)
606 list = delete_Token(list);
610 * Free a linked list of lines.
612 static void free_llist(Line * list)
614 Line *l, *tmp;
615 list_for_each_safe(l, tmp, list) {
616 free_tlist(l->first);
617 nasm_free(l);
622 * Free an MMacro
624 static void free_mmacro(MMacro * m)
626 nasm_free(m->name);
627 free_tlist(m->dlist);
628 nasm_free(m->defaults);
629 free_llist(m->expansion);
630 nasm_free(m);
634 * Free all currently defined macros, and free the hash tables
636 static void free_smacro_table(struct hash_table *smt)
638 SMacro *s, *tmp;
639 const char *key;
640 struct hash_tbl_node *it = NULL;
642 while ((s = hash_iterate(smt, &it, &key)) != NULL) {
643 nasm_free((void *)key);
644 list_for_each_safe(s, tmp, s) {
645 nasm_free(s->name);
646 free_tlist(s->expansion);
647 nasm_free(s);
650 hash_free(smt);
653 static void free_mmacro_table(struct hash_table *mmt)
655 MMacro *m, *tmp;
656 const char *key;
657 struct hash_tbl_node *it = NULL;
659 it = NULL;
660 while ((m = hash_iterate(mmt, &it, &key)) != NULL) {
661 nasm_free((void *)key);
662 list_for_each_safe(m ,tmp, m)
663 free_mmacro(m);
665 hash_free(mmt);
668 static void free_macros(void)
670 free_smacro_table(&smacros);
671 free_mmacro_table(&mmacros);
675 * Initialize the hash tables
677 static void init_macros(void)
679 hash_init(&smacros, HASH_LARGE);
680 hash_init(&mmacros, HASH_LARGE);
684 * Pop the context stack.
686 static void ctx_pop(void)
688 Context *c = cstk;
690 cstk = cstk->next;
691 free_smacro_table(&c->localmac);
692 nasm_free(c->name);
693 nasm_free(c);
697 * Search for a key in the hash index; adding it if necessary
698 * (in which case we initialize the data pointer to NULL.)
700 static void **
701 hash_findi_add(struct hash_table *hash, const char *str)
703 struct hash_insert hi;
704 void **r;
705 char *strx;
707 r = hash_findi(hash, str, &hi);
708 if (r)
709 return r;
711 strx = nasm_strdup(str); /* Use a more efficient allocator here? */
712 return hash_add(&hi, strx, NULL);
716 * Like hash_findi, but returns the data element rather than a pointer
717 * to it. Used only when not adding a new element, hence no third
718 * argument.
720 static void *
721 hash_findix(struct hash_table *hash, const char *str)
723 void **p;
725 p = hash_findi(hash, str, NULL);
726 return p ? *p : NULL;
730 * read line from standart macros set,
731 * if there no more left -- return NULL
733 static char *line_from_stdmac(void)
735 unsigned char c;
736 const unsigned char *p = stdmacpos;
737 char *line, *q;
738 size_t len = 0;
740 if (!stdmacpos)
741 return NULL;
743 while ((c = *p++)) {
744 if (c >= 0x80)
745 len += pp_directives_len[c - 0x80] + 1;
746 else
747 len++;
750 line = nasm_malloc(len + 1);
751 q = line;
752 while ((c = *stdmacpos++)) {
753 if (c >= 0x80) {
754 memcpy(q, pp_directives[c - 0x80], pp_directives_len[c - 0x80]);
755 q += pp_directives_len[c - 0x80];
756 *q++ = ' ';
757 } else {
758 *q++ = c;
761 stdmacpos = p;
762 *q = '\0';
764 if (!*stdmacpos) {
765 /* This was the last of this particular macro set */
766 stdmacpos = NULL;
767 if (*stdmacnext) {
768 stdmacpos = *stdmacnext++;
769 } else if (do_predef) {
770 Line *pd, *l;
771 Token *head, **tail, *t;
774 * Nasty hack: here we push the contents of
775 * `predef' on to the top-level expansion stack,
776 * since this is the most convenient way to
777 * implement the pre-include and pre-define
778 * features.
780 list_for_each(pd, predef) {
781 head = NULL;
782 tail = &head;
783 list_for_each(t, pd->first) {
784 *tail = new_Token(NULL, t->type, t->text, 0);
785 tail = &(*tail)->next;
788 l = nasm_malloc(sizeof(Line));
789 l->next = istk->expansion;
790 l->first = head;
791 l->finishes = NULL;
793 istk->expansion = l;
795 do_predef = false;
799 return line;
802 static char *read_line(void)
804 unsigned int size, c, next;
805 const unsigned int delta = 512;
806 const unsigned int pad = 8;
807 unsigned int nr_cont = 0;
808 bool cont = false;
809 char *buffer, *p;
811 /* Standart macros set (predefined) goes first */
812 p = line_from_stdmac();
813 if (p)
814 return p;
816 size = delta;
817 p = buffer = nasm_malloc(size);
819 for (;;) {
820 c = fgetc(istk->fp);
821 if ((int)(c) == EOF) {
822 p[0] = 0;
823 break;
826 switch (c) {
827 case '\r':
828 next = fgetc(istk->fp);
829 if (next != '\n')
830 ungetc(next, istk->fp);
831 if (cont) {
832 cont = false;
833 continue;
835 break;
837 case '\n':
838 if (cont) {
839 cont = false;
840 continue;
842 break;
844 case '\\':
845 next = fgetc(istk->fp);
846 ungetc(next, istk->fp);
847 if (next == '\r' || next == '\n') {
848 cont = true;
849 nr_cont++;
850 continue;
852 break;
855 if (c == '\r' || c == '\n') {
856 *p++ = 0;
857 break;
860 if (p >= (buffer + size - pad)) {
861 buffer = nasm_realloc(buffer, size + delta);
862 p = buffer + size - pad;
863 size += delta;
866 *p++ = (unsigned char)c;
869 if (p == buffer) {
870 nasm_free(buffer);
871 return NULL;
874 src_set_linnum(src_get_linnum() + istk->lineinc +
875 (nr_cont * istk->lineinc));
878 * Handle spurious ^Z, which may be inserted into source files
879 * by some file transfer utilities.
881 buffer[strcspn(buffer, "\032")] = '\0';
883 lfmt->line(LIST_READ, buffer);
885 return buffer;
889 * Tokenize a line of text. This is a very simple process since we
890 * don't need to parse the value out of e.g. numeric tokens: we
891 * simply split one string into many.
893 static Token *tokenize(char *line)
895 char c, *p = line;
896 enum pp_token_type type;
897 Token *list = NULL;
898 Token *t, **tail = &list;
900 while (*line) {
901 p = line;
902 if (*p == '%') {
903 p++;
904 if (*p == '+' && !nasm_isdigit(p[1])) {
905 p++;
906 type = TOK_PASTE;
907 } else if (nasm_isdigit(*p) ||
908 ((*p == '-' || *p == '+') && nasm_isdigit(p[1]))) {
909 do {
910 p++;
912 while (nasm_isdigit(*p));
913 type = TOK_PREPROC_ID;
914 } else if (*p == '{') {
915 p++;
916 while (*p) {
917 if (*p == '}')
918 break;
919 p[-1] = *p;
920 p++;
922 if (*p != '}')
923 nasm_error(ERR_WARNING | ERR_PASS1,
924 "unterminated %%{ construct");
925 p[-1] = '\0';
926 if (*p)
927 p++;
928 type = TOK_PREPROC_ID;
929 } else if (*p == '[') {
930 int lvl = 1;
931 line += 2; /* Skip the leading %[ */
932 p++;
933 while (lvl && (c = *p++)) {
934 switch (c) {
935 case ']':
936 lvl--;
937 break;
938 case '%':
939 if (*p == '[')
940 lvl++;
941 break;
942 case '\'':
943 case '\"':
944 case '`':
945 p = nasm_skip_string(p - 1);
946 if (*p)
947 p++;
948 break;
949 default:
950 break;
953 p--;
954 if (*p)
955 *p++ = '\0';
956 if (lvl)
957 nasm_error(ERR_NONFATAL|ERR_PASS1,
958 "unterminated %%[ construct");
959 type = TOK_INDIRECT;
960 } else if (*p == '?') {
961 type = TOK_PREPROC_Q; /* %? */
962 p++;
963 if (*p == '?') {
964 type = TOK_PREPROC_QQ; /* %?? */
965 p++;
967 } else if (*p == '!') {
968 type = TOK_PREPROC_ID;
969 p++;
970 if (isidchar(*p)) {
971 do {
972 p++;
974 while (isidchar(*p));
975 } else if (*p == '\'' || *p == '\"' || *p == '`') {
976 p = nasm_skip_string(p);
977 if (*p)
978 p++;
979 else
980 nasm_error(ERR_NONFATAL|ERR_PASS1,
981 "unterminated %%! string");
982 } else {
983 /* %! without string or identifier */
984 type = TOK_OTHER; /* Legacy behavior... */
986 } else if (isidchar(*p) ||
987 ((*p == '!' || *p == '%' || *p == '$') &&
988 isidchar(p[1]))) {
989 do {
990 p++;
992 while (isidchar(*p));
993 type = TOK_PREPROC_ID;
994 } else {
995 type = TOK_OTHER;
996 if (*p == '%')
997 p++;
999 } else if (isidstart(*p) || (*p == '$' && isidstart(p[1]))) {
1000 type = TOK_ID;
1001 p++;
1002 while (*p && isidchar(*p))
1003 p++;
1004 } else if (*p == '\'' || *p == '"' || *p == '`') {
1006 * A string token.
1008 type = TOK_STRING;
1009 p = nasm_skip_string(p);
1011 if (*p) {
1012 p++;
1013 } else {
1014 nasm_error(ERR_WARNING|ERR_PASS1, "unterminated string");
1015 /* Handling unterminated strings by UNV */
1016 /* type = -1; */
1018 } else if (p[0] == '$' && p[1] == '$') {
1019 type = TOK_OTHER; /* TOKEN_BASE */
1020 p += 2;
1021 } else if (isnumstart(*p)) {
1022 bool is_hex = false;
1023 bool is_float = false;
1024 bool has_e = false;
1025 char c, *r;
1028 * A numeric token.
1031 if (*p == '$') {
1032 p++;
1033 is_hex = true;
1036 for (;;) {
1037 c = *p++;
1039 if (!is_hex && (c == 'e' || c == 'E')) {
1040 has_e = true;
1041 if (*p == '+' || *p == '-') {
1043 * e can only be followed by +/- if it is either a
1044 * prefixed hex number or a floating-point number
1046 p++;
1047 is_float = true;
1049 } else if (c == 'H' || c == 'h' || c == 'X' || c == 'x') {
1050 is_hex = true;
1051 } else if (c == 'P' || c == 'p') {
1052 is_float = true;
1053 if (*p == '+' || *p == '-')
1054 p++;
1055 } else if (isnumchar(c))
1056 ; /* just advance */
1057 else if (c == '.') {
1059 * we need to deal with consequences of the legacy
1060 * parser, like "1.nolist" being two tokens
1061 * (TOK_NUMBER, TOK_ID) here; at least give it
1062 * a shot for now. In the future, we probably need
1063 * a flex-based scanner with proper pattern matching
1064 * to do it as well as it can be done. Nothing in
1065 * the world is going to help the person who wants
1066 * 0x123.p16 interpreted as two tokens, though.
1068 r = p;
1069 while (*r == '_')
1070 r++;
1072 if (nasm_isdigit(*r) || (is_hex && nasm_isxdigit(*r)) ||
1073 (!is_hex && (*r == 'e' || *r == 'E')) ||
1074 (*r == 'p' || *r == 'P')) {
1075 p = r;
1076 is_float = true;
1077 } else
1078 break; /* Terminate the token */
1079 } else
1080 break;
1082 p--; /* Point to first character beyond number */
1084 if (p == line+1 && *line == '$') {
1085 type = TOK_OTHER; /* TOKEN_HERE */
1086 } else {
1087 if (has_e && !is_hex) {
1088 /* 1e13 is floating-point, but 1e13h is not */
1089 is_float = true;
1092 type = is_float ? TOK_FLOAT : TOK_NUMBER;
1094 } else if (nasm_isspace(*p)) {
1095 type = TOK_WHITESPACE;
1096 p = nasm_skip_spaces(p);
1098 * Whitespace just before end-of-line is discarded by
1099 * pretending it's a comment; whitespace just before a
1100 * comment gets lumped into the comment.
1102 if (!*p || *p == ';') {
1103 type = TOK_COMMENT;
1104 while (*p)
1105 p++;
1107 } else if (*p == ';') {
1108 type = TOK_COMMENT;
1109 while (*p)
1110 p++;
1111 } else {
1113 * Anything else is an operator of some kind. We check
1114 * for all the double-character operators (>>, <<, //,
1115 * %%, <=, >=, ==, !=, <>, &&, ||, ^^), but anything
1116 * else is a single-character operator.
1118 type = TOK_OTHER;
1119 if ((p[0] == '>' && p[1] == '>') ||
1120 (p[0] == '<' && p[1] == '<') ||
1121 (p[0] == '/' && p[1] == '/') ||
1122 (p[0] == '<' && p[1] == '=') ||
1123 (p[0] == '>' && p[1] == '=') ||
1124 (p[0] == '=' && p[1] == '=') ||
1125 (p[0] == '!' && p[1] == '=') ||
1126 (p[0] == '<' && p[1] == '>') ||
1127 (p[0] == '&' && p[1] == '&') ||
1128 (p[0] == '|' && p[1] == '|') ||
1129 (p[0] == '^' && p[1] == '^')) {
1130 p++;
1132 p++;
1135 /* Handling unterminated string by UNV */
1136 /*if (type == -1)
1138 *tail = t = new_Token(NULL, TOK_STRING, line, p-line+1);
1139 t->text[p-line] = *line;
1140 tail = &t->next;
1142 else */
1143 if (type != TOK_COMMENT) {
1144 *tail = t = new_Token(NULL, type, line, p - line);
1145 tail = &t->next;
1147 line = p;
1149 return list;
1153 * this function allocates a new managed block of memory and
1154 * returns a pointer to the block. The managed blocks are
1155 * deleted only all at once by the delete_Blocks function.
1157 static void *new_Block(size_t size)
1159 Blocks *b = &blocks;
1161 /* first, get to the end of the linked list */
1162 while (b->next)
1163 b = b->next;
1164 /* now allocate the requested chunk */
1165 b->chunk = nasm_malloc(size);
1167 /* now allocate a new block for the next request */
1168 b->next = nasm_zalloc(sizeof(Blocks));
1169 return b->chunk;
1173 * this function deletes all managed blocks of memory
1175 static void delete_Blocks(void)
1177 Blocks *a, *b = &blocks;
1180 * keep in mind that the first block, pointed to by blocks
1181 * is a static and not dynamically allocated, so we don't
1182 * free it.
1184 while (b) {
1185 if (b->chunk)
1186 nasm_free(b->chunk);
1187 a = b;
1188 b = b->next;
1189 if (a != &blocks)
1190 nasm_free(a);
1192 memset(&blocks, 0, sizeof(blocks));
1196 * this function creates a new Token and passes a pointer to it
1197 * back to the caller. It sets the type and text elements, and
1198 * also the a.mac and next elements to NULL.
1200 static Token *new_Token(Token * next, enum pp_token_type type,
1201 const char *text, int txtlen)
1203 Token *t;
1204 int i;
1206 if (!freeTokens) {
1207 freeTokens = (Token *) new_Block(TOKEN_BLOCKSIZE * sizeof(Token));
1208 for (i = 0; i < TOKEN_BLOCKSIZE - 1; i++)
1209 freeTokens[i].next = &freeTokens[i + 1];
1210 freeTokens[i].next = NULL;
1212 t = freeTokens;
1213 freeTokens = t->next;
1214 t->next = next;
1215 t->a.mac = NULL;
1216 t->type = type;
1217 if (type == TOK_WHITESPACE || !text) {
1218 t->text = NULL;
1219 } else {
1220 if (txtlen == 0)
1221 txtlen = strlen(text);
1222 t->text = nasm_malloc(txtlen+1);
1223 memcpy(t->text, text, txtlen);
1224 t->text[txtlen] = '\0';
1226 return t;
1229 static Token *delete_Token(Token * t)
1231 Token *next = t->next;
1232 nasm_free(t->text);
1233 t->next = freeTokens;
1234 freeTokens = t;
1235 return next;
1239 * Convert a line of tokens back into text.
1240 * If expand_locals is not zero, identifiers of the form "%$*xxx"
1241 * will be transformed into ..@ctxnum.xxx
1243 static char *detoken(Token * tlist, bool expand_locals)
1245 Token *t;
1246 char *line, *p;
1247 const char *q;
1248 int len = 0;
1250 list_for_each(t, tlist) {
1251 if (t->type == TOK_PREPROC_ID && t->text &&
1252 t->text[0] && t->text[1] == '!') {
1253 char *v;
1254 char *q = t->text;
1256 v = t->text + 2;
1257 if (*v == '\'' || *v == '\"' || *v == '`') {
1258 size_t len = nasm_unquote(v, NULL);
1259 size_t clen = strlen(v);
1261 if (len != clen) {
1262 nasm_error(ERR_NONFATAL | ERR_PASS1,
1263 "NUL character in %%! string");
1264 v = NULL;
1268 if (v) {
1269 char *p = getenv(v);
1270 if (!p) {
1271 nasm_error(ERR_NONFATAL | ERR_PASS1,
1272 "nonexistent environment variable `%s'", v);
1274 * FIXME We better should investigate if accessing
1275 * ->text[1] without ->text[0] is safe enough.
1277 t->text = nasm_zalloc(2);
1278 } else
1279 t->text = nasm_strdup(p);
1280 nasm_free(q);
1284 /* Expand local macros here and not during preprocessing */
1285 if (expand_locals &&
1286 t->type == TOK_PREPROC_ID && t->text &&
1287 t->text[0] == '%' && t->text[1] == '$') {
1288 const char *q;
1289 char *p;
1290 Context *ctx = get_ctx(t->text, &q);
1291 if (ctx) {
1292 char buffer[40];
1293 snprintf(buffer, sizeof(buffer), "..@%"PRIu32".", ctx->number);
1294 p = nasm_strcat(buffer, q);
1295 nasm_free(t->text);
1296 t->text = p;
1299 if (t->type == TOK_WHITESPACE)
1300 len++;
1301 else if (t->text)
1302 len += strlen(t->text);
1305 p = line = nasm_malloc(len + 1);
1307 list_for_each(t, tlist) {
1308 if (t->type == TOK_WHITESPACE) {
1309 *p++ = ' ';
1310 } else if (t->text) {
1311 q = t->text;
1312 while (*q)
1313 *p++ = *q++;
1316 *p = '\0';
1318 return line;
1322 * A scanner, suitable for use by the expression evaluator, which
1323 * operates on a line of Tokens. Expects a pointer to a pointer to
1324 * the first token in the line to be passed in as its private_data
1325 * field.
1327 * FIX: This really needs to be unified with stdscan.
1329 static int ppscan(void *private_data, struct tokenval *tokval)
1331 Token **tlineptr = private_data;
1332 Token *tline;
1333 char ourcopy[MAX_KEYWORD+1], *p, *r, *s;
1335 do {
1336 tline = *tlineptr;
1337 *tlineptr = tline ? tline->next : NULL;
1338 } while (tline && (tline->type == TOK_WHITESPACE ||
1339 tline->type == TOK_COMMENT));
1341 if (!tline)
1342 return tokval->t_type = TOKEN_EOS;
1344 tokval->t_charptr = tline->text;
1346 if (tline->text[0] == '$' && !tline->text[1])
1347 return tokval->t_type = TOKEN_HERE;
1348 if (tline->text[0] == '$' && tline->text[1] == '$' && !tline->text[2])
1349 return tokval->t_type = TOKEN_BASE;
1351 if (tline->type == TOK_ID) {
1352 p = tokval->t_charptr = tline->text;
1353 if (p[0] == '$') {
1354 tokval->t_charptr++;
1355 return tokval->t_type = TOKEN_ID;
1358 for (r = p, s = ourcopy; *r; r++) {
1359 if (r >= p+MAX_KEYWORD)
1360 return tokval->t_type = TOKEN_ID; /* Not a keyword */
1361 *s++ = nasm_tolower(*r);
1363 *s = '\0';
1364 /* right, so we have an identifier sitting in temp storage. now,
1365 * is it actually a register or instruction name, or what? */
1366 return nasm_token_hash(ourcopy, tokval);
1369 if (tline->type == TOK_NUMBER) {
1370 bool rn_error;
1371 tokval->t_integer = readnum(tline->text, &rn_error);
1372 tokval->t_charptr = tline->text;
1373 if (rn_error)
1374 return tokval->t_type = TOKEN_ERRNUM;
1375 else
1376 return tokval->t_type = TOKEN_NUM;
1379 if (tline->type == TOK_FLOAT) {
1380 return tokval->t_type = TOKEN_FLOAT;
1383 if (tline->type == TOK_STRING) {
1384 char bq, *ep;
1386 bq = tline->text[0];
1387 tokval->t_charptr = tline->text;
1388 tokval->t_inttwo = nasm_unquote(tline->text, &ep);
1390 if (ep[0] != bq || ep[1] != '\0')
1391 return tokval->t_type = TOKEN_ERRSTR;
1392 else
1393 return tokval->t_type = TOKEN_STR;
1396 if (tline->type == TOK_OTHER) {
1397 if (!strcmp(tline->text, "<<"))
1398 return tokval->t_type = TOKEN_SHL;
1399 if (!strcmp(tline->text, ">>"))
1400 return tokval->t_type = TOKEN_SHR;
1401 if (!strcmp(tline->text, "//"))
1402 return tokval->t_type = TOKEN_SDIV;
1403 if (!strcmp(tline->text, "%%"))
1404 return tokval->t_type = TOKEN_SMOD;
1405 if (!strcmp(tline->text, "=="))
1406 return tokval->t_type = TOKEN_EQ;
1407 if (!strcmp(tline->text, "<>"))
1408 return tokval->t_type = TOKEN_NE;
1409 if (!strcmp(tline->text, "!="))
1410 return tokval->t_type = TOKEN_NE;
1411 if (!strcmp(tline->text, "<="))
1412 return tokval->t_type = TOKEN_LE;
1413 if (!strcmp(tline->text, ">="))
1414 return tokval->t_type = TOKEN_GE;
1415 if (!strcmp(tline->text, "&&"))
1416 return tokval->t_type = TOKEN_DBL_AND;
1417 if (!strcmp(tline->text, "^^"))
1418 return tokval->t_type = TOKEN_DBL_XOR;
1419 if (!strcmp(tline->text, "||"))
1420 return tokval->t_type = TOKEN_DBL_OR;
1424 * We have no other options: just return the first character of
1425 * the token text.
1427 return tokval->t_type = tline->text[0];
1431 * Compare a string to the name of an existing macro; this is a
1432 * simple wrapper which calls either strcmp or nasm_stricmp
1433 * depending on the value of the `casesense' parameter.
1435 static int mstrcmp(const char *p, const char *q, bool casesense)
1437 return casesense ? strcmp(p, q) : nasm_stricmp(p, q);
1441 * Compare a string to the name of an existing macro; this is a
1442 * simple wrapper which calls either strcmp or nasm_stricmp
1443 * depending on the value of the `casesense' parameter.
1445 static int mmemcmp(const char *p, const char *q, size_t l, bool casesense)
1447 return casesense ? memcmp(p, q, l) : nasm_memicmp(p, q, l);
1451 * Return the Context structure associated with a %$ token. Return
1452 * NULL, having _already_ reported an error condition, if the
1453 * context stack isn't deep enough for the supplied number of $
1454 * signs.
1456 * If "namep" is non-NULL, set it to the pointer to the macro name
1457 * tail, i.e. the part beyond %$...
1459 static Context *get_ctx(const char *name, const char **namep)
1461 Context *ctx;
1462 int i;
1464 if (namep)
1465 *namep = name;
1467 if (!name || name[0] != '%' || name[1] != '$')
1468 return NULL;
1470 if (!cstk) {
1471 nasm_error(ERR_NONFATAL, "`%s': context stack is empty", name);
1472 return NULL;
1475 name += 2;
1476 ctx = cstk;
1477 i = 0;
1478 while (ctx && *name == '$') {
1479 name++;
1480 i++;
1481 ctx = ctx->next;
1483 if (!ctx) {
1484 nasm_error(ERR_NONFATAL, "`%s': context stack is only"
1485 " %d level%s deep", name, i, (i == 1 ? "" : "s"));
1486 return NULL;
1489 if (namep)
1490 *namep = name;
1492 return ctx;
1496 * Open an include file. This routine must always return a valid
1497 * file pointer if it returns - it's responsible for throwing an
1498 * ERR_FATAL and bombing out completely if not. It should also try
1499 * the include path one by one until it finds the file or reaches
1500 * the end of the path.
1502 * Note: for INC_PROBE the function returns NULL at all times;
1503 * instead look for the
1505 enum incopen_mode {
1506 INC_NEEDED, /* File must exist */
1507 INC_OPTIONAL, /* Missing is OK */
1508 INC_PROBE /* Only an existence probe */
1511 /* This is conducts a full pathname search */
1512 static FILE *inc_fopen_search(const char *file, StrList **slpath,
1513 enum incopen_mode omode, enum file_flags fmode)
1515 FILE *fp;
1516 char *prefix = "";
1517 const IncPath *ip = ipath;
1518 int len;
1519 StrList *sl;
1520 char *sp;
1521 bool found;
1523 while (1) {
1524 sp = nasm_catfile(prefix, file);
1525 len = strlen(sp) + 1;
1526 sl = nasm_malloc(len + sizeof sl->next);
1527 memcpy(sl->str, sp, len);
1528 sl->next = NULL;
1529 nasm_free(sp);
1531 if (omode == INC_PROBE) {
1532 fp = NULL;
1533 found = nasm_file_exists(sl->str);
1534 } else {
1535 fp = nasm_open_read(sl->str, fmode);
1536 found = (fp != NULL);
1538 if (found) {
1539 *slpath = sl;
1540 return fp;
1543 nasm_free(sl);
1545 if (!ip)
1546 return NULL;
1548 prefix = ip->path;
1549 ip = ip->next;
1554 * Open a file, or test for the presence of one (depending on omode),
1555 * considering the include path.
1557 static FILE *inc_fopen(const char *file,
1558 StrList **dhead,
1559 const char **found_path,
1560 enum incopen_mode omode,
1561 enum file_flags fmode)
1563 StrList *sl;
1564 struct hash_insert hi;
1565 void **hp;
1566 char *path;
1567 FILE *fp = NULL;
1569 hp = hash_find(&FileHash, file, &hi);
1570 if (hp) {
1571 path = *hp;
1572 if (path || omode != INC_NEEDED) {
1573 nasm_add_string_to_strlist(dhead, path ? path : file);
1575 } else {
1576 /* Need to do the actual path search */
1577 size_t file_len;
1579 sl = NULL;
1580 fp = inc_fopen_search(file, &sl, omode, fmode);
1582 file_len = strlen(file);
1584 if (!sl) {
1585 /* Store negative result for this file */
1586 sl = nasm_malloc(file_len + 1 + sizeof sl->next);
1587 memcpy(sl->str, file, file_len+1);
1588 sl->next = NULL;
1589 file = sl->str;
1590 path = NULL;
1591 } else {
1592 path = sl->str;
1593 file = strchr(path, '\0') - file_len;
1596 hash_add(&hi, file, path); /* Positive or negative result */
1599 * Add file to dependency path. The in_list() is needed
1600 * in case the file was already added with %depend.
1602 if (path || omode != INC_NEEDED)
1603 nasm_add_to_strlist(dhead, sl);
1606 if (!path) {
1607 if (omode == INC_NEEDED)
1608 nasm_fatal(0, "unable to open include file `%s'", file);
1610 if (found_path)
1611 *found_path = NULL;
1613 return NULL;
1616 if (!fp && omode != INC_PROBE)
1617 fp = nasm_open_read(path, fmode);
1619 if (found_path)
1620 *found_path = path;
1622 return fp;
1626 * Opens an include or input file. Public version, for use by modules
1627 * that get a file:lineno pair and need to look at the file again
1628 * (e.g. the CodeView debug backend). Returns NULL on failure.
1630 FILE *pp_input_fopen(const char *filename, enum file_flags mode)
1632 return inc_fopen(filename, NULL, NULL, INC_OPTIONAL, mode);
1636 * Determine if we should warn on defining a single-line macro of
1637 * name `name', with `nparam' parameters. If nparam is 0 or -1, will
1638 * return true if _any_ single-line macro of that name is defined.
1639 * Otherwise, will return true if a single-line macro with either
1640 * `nparam' or no parameters is defined.
1642 * If a macro with precisely the right number of parameters is
1643 * defined, or nparam is -1, the address of the definition structure
1644 * will be returned in `defn'; otherwise NULL will be returned. If `defn'
1645 * is NULL, no action will be taken regarding its contents, and no
1646 * error will occur.
1648 * Note that this is also called with nparam zero to resolve
1649 * `ifdef'.
1651 * If you already know which context macro belongs to, you can pass
1652 * the context pointer as first parameter; if you won't but name begins
1653 * with %$ the context will be automatically computed. If all_contexts
1654 * is true, macro will be searched in outer contexts as well.
1656 static bool
1657 smacro_defined(Context * ctx, const char *name, int nparam, SMacro ** defn,
1658 bool nocase)
1660 struct hash_table *smtbl;
1661 SMacro *m;
1663 if (ctx) {
1664 smtbl = &ctx->localmac;
1665 } else if (name[0] == '%' && name[1] == '$') {
1666 if (cstk)
1667 ctx = get_ctx(name, &name);
1668 if (!ctx)
1669 return false; /* got to return _something_ */
1670 smtbl = &ctx->localmac;
1671 } else {
1672 smtbl = &smacros;
1674 m = (SMacro *) hash_findix(smtbl, name);
1676 while (m) {
1677 if (!mstrcmp(m->name, name, m->casesense && nocase) &&
1678 (nparam <= 0 || m->nparam == 0 || nparam == (int) m->nparam)) {
1679 if (defn) {
1680 if (nparam == (int) m->nparam || nparam == -1)
1681 *defn = m;
1682 else
1683 *defn = NULL;
1685 return true;
1687 m = m->next;
1690 return false;
1694 * Count and mark off the parameters in a multi-line macro call.
1695 * This is called both from within the multi-line macro expansion
1696 * code, and also to mark off the default parameters when provided
1697 * in a %macro definition line.
1699 static void count_mmac_params(Token * t, int *nparam, Token *** params)
1701 int paramsize, brace;
1703 *nparam = paramsize = 0;
1704 *params = NULL;
1705 while (t) {
1706 /* +1: we need space for the final NULL */
1707 if (*nparam+1 >= paramsize) {
1708 paramsize += PARAM_DELTA;
1709 *params = nasm_realloc(*params, sizeof(**params) * paramsize);
1711 skip_white_(t);
1712 brace = 0;
1713 if (tok_is_(t, "{"))
1714 brace++;
1715 (*params)[(*nparam)++] = t;
1716 if (brace) {
1717 while (brace && (t = t->next) != NULL) {
1718 if (tok_is_(t, "{"))
1719 brace++;
1720 else if (tok_is_(t, "}"))
1721 brace--;
1724 if (t) {
1726 * Now we've found the closing brace, look further
1727 * for the comma.
1729 t = t->next;
1730 skip_white_(t);
1731 if (tok_isnt_(t, ",")) {
1732 nasm_error(ERR_NONFATAL,
1733 "braces do not enclose all of macro parameter");
1734 while (tok_isnt_(t, ","))
1735 t = t->next;
1738 } else {
1739 while (tok_isnt_(t, ","))
1740 t = t->next;
1742 if (t) { /* got a comma/brace */
1743 t = t->next; /* eat the comma */
1749 * Determine whether one of the various `if' conditions is true or
1750 * not.
1752 * We must free the tline we get passed.
1754 static bool if_condition(Token * tline, enum preproc_token ct)
1756 enum pp_conditional i = PP_COND(ct);
1757 bool j;
1758 Token *t, *tt, **tptr, *origline;
1759 struct tokenval tokval;
1760 expr *evalresult;
1761 enum pp_token_type needtype;
1762 char *p;
1764 origline = tline;
1766 switch (i) {
1767 case PPC_IFCTX:
1768 j = false; /* have we matched yet? */
1769 while (true) {
1770 skip_white_(tline);
1771 if (!tline)
1772 break;
1773 if (tline->type != TOK_ID) {
1774 nasm_error(ERR_NONFATAL,
1775 "`%s' expects context identifiers", pp_directives[ct]);
1776 free_tlist(origline);
1777 return -1;
1779 if (cstk && cstk->name && !nasm_stricmp(tline->text, cstk->name))
1780 j = true;
1781 tline = tline->next;
1783 break;
1785 case PPC_IFDEF:
1786 j = false; /* have we matched yet? */
1787 while (tline) {
1788 skip_white_(tline);
1789 if (!tline || (tline->type != TOK_ID &&
1790 (tline->type != TOK_PREPROC_ID ||
1791 tline->text[1] != '$'))) {
1792 nasm_error(ERR_NONFATAL,
1793 "`%s' expects macro identifiers", pp_directives[ct]);
1794 goto fail;
1796 if (smacro_defined(NULL, tline->text, 0, NULL, true))
1797 j = true;
1798 tline = tline->next;
1800 break;
1802 case PPC_IFENV:
1803 tline = expand_smacro(tline);
1804 j = false; /* have we matched yet? */
1805 while (tline) {
1806 skip_white_(tline);
1807 if (!tline || (tline->type != TOK_ID &&
1808 tline->type != TOK_STRING &&
1809 (tline->type != TOK_PREPROC_ID ||
1810 tline->text[1] != '!'))) {
1811 nasm_error(ERR_NONFATAL,
1812 "`%s' expects environment variable names",
1813 pp_directives[ct]);
1814 goto fail;
1816 p = tline->text;
1817 if (tline->type == TOK_PREPROC_ID)
1818 p += 2; /* Skip leading %! */
1819 if (*p == '\'' || *p == '\"' || *p == '`')
1820 nasm_unquote_cstr(p, ct);
1821 if (getenv(p))
1822 j = true;
1823 tline = tline->next;
1825 break;
1827 case PPC_IFIDN:
1828 case PPC_IFIDNI:
1829 tline = expand_smacro(tline);
1830 t = tt = tline;
1831 while (tok_isnt_(tt, ","))
1832 tt = tt->next;
1833 if (!tt) {
1834 nasm_error(ERR_NONFATAL,
1835 "`%s' expects two comma-separated arguments",
1836 pp_directives[ct]);
1837 goto fail;
1839 tt = tt->next;
1840 j = true; /* assume equality unless proved not */
1841 while ((t->type != TOK_OTHER || strcmp(t->text, ",")) && tt) {
1842 if (tt->type == TOK_OTHER && !strcmp(tt->text, ",")) {
1843 nasm_error(ERR_NONFATAL, "`%s': more than one comma on line",
1844 pp_directives[ct]);
1845 goto fail;
1847 if (t->type == TOK_WHITESPACE) {
1848 t = t->next;
1849 continue;
1851 if (tt->type == TOK_WHITESPACE) {
1852 tt = tt->next;
1853 continue;
1855 if (tt->type != t->type) {
1856 j = false; /* found mismatching tokens */
1857 break;
1859 /* When comparing strings, need to unquote them first */
1860 if (t->type == TOK_STRING) {
1861 size_t l1 = nasm_unquote(t->text, NULL);
1862 size_t l2 = nasm_unquote(tt->text, NULL);
1864 if (l1 != l2) {
1865 j = false;
1866 break;
1868 if (mmemcmp(t->text, tt->text, l1, i == PPC_IFIDN)) {
1869 j = false;
1870 break;
1872 } else if (mstrcmp(tt->text, t->text, i == PPC_IFIDN) != 0) {
1873 j = false; /* found mismatching tokens */
1874 break;
1877 t = t->next;
1878 tt = tt->next;
1880 if ((t->type != TOK_OTHER || strcmp(t->text, ",")) || tt)
1881 j = false; /* trailing gunk on one end or other */
1882 break;
1884 case PPC_IFMACRO:
1886 bool found = false;
1887 MMacro searching, *mmac;
1889 skip_white_(tline);
1890 tline = expand_id(tline);
1891 if (!tok_type_(tline, TOK_ID)) {
1892 nasm_error(ERR_NONFATAL,
1893 "`%s' expects a macro name", pp_directives[ct]);
1894 goto fail;
1896 searching.name = nasm_strdup(tline->text);
1897 searching.casesense = true;
1898 searching.plus = false;
1899 searching.nolist = false;
1900 searching.in_progress = 0;
1901 searching.max_depth = 0;
1902 searching.rep_nest = NULL;
1903 searching.nparam_min = 0;
1904 searching.nparam_max = INT_MAX;
1905 tline = expand_smacro(tline->next);
1906 skip_white_(tline);
1907 if (!tline) {
1908 } else if (!tok_type_(tline, TOK_NUMBER)) {
1909 nasm_error(ERR_NONFATAL,
1910 "`%s' expects a parameter count or nothing",
1911 pp_directives[ct]);
1912 } else {
1913 searching.nparam_min = searching.nparam_max =
1914 readnum(tline->text, &j);
1915 if (j)
1916 nasm_error(ERR_NONFATAL,
1917 "unable to parse parameter count `%s'",
1918 tline->text);
1920 if (tline && tok_is_(tline->next, "-")) {
1921 tline = tline->next->next;
1922 if (tok_is_(tline, "*"))
1923 searching.nparam_max = INT_MAX;
1924 else if (!tok_type_(tline, TOK_NUMBER))
1925 nasm_error(ERR_NONFATAL,
1926 "`%s' expects a parameter count after `-'",
1927 pp_directives[ct]);
1928 else {
1929 searching.nparam_max = readnum(tline->text, &j);
1930 if (j)
1931 nasm_error(ERR_NONFATAL,
1932 "unable to parse parameter count `%s'",
1933 tline->text);
1934 if (searching.nparam_min > searching.nparam_max) {
1935 nasm_error(ERR_NONFATAL,
1936 "minimum parameter count exceeds maximum");
1937 searching.nparam_max = searching.nparam_min;
1941 if (tline && tok_is_(tline->next, "+")) {
1942 tline = tline->next;
1943 searching.plus = true;
1945 mmac = (MMacro *) hash_findix(&mmacros, searching.name);
1946 while (mmac) {
1947 if (!strcmp(mmac->name, searching.name) &&
1948 (mmac->nparam_min <= searching.nparam_max
1949 || searching.plus)
1950 && (searching.nparam_min <= mmac->nparam_max
1951 || mmac->plus)) {
1952 found = true;
1953 break;
1955 mmac = mmac->next;
1957 if (tline && tline->next)
1958 nasm_error(ERR_WARNING|ERR_PASS1,
1959 "trailing garbage after %%ifmacro ignored");
1960 nasm_free(searching.name);
1961 j = found;
1962 break;
1965 case PPC_IFID:
1966 needtype = TOK_ID;
1967 goto iftype;
1968 case PPC_IFNUM:
1969 needtype = TOK_NUMBER;
1970 goto iftype;
1971 case PPC_IFSTR:
1972 needtype = TOK_STRING;
1973 goto iftype;
1975 iftype:
1976 t = tline = expand_smacro(tline);
1978 while (tok_type_(t, TOK_WHITESPACE) ||
1979 (needtype == TOK_NUMBER &&
1980 tok_type_(t, TOK_OTHER) &&
1981 (t->text[0] == '-' || t->text[0] == '+') &&
1982 !t->text[1]))
1983 t = t->next;
1985 j = tok_type_(t, needtype);
1986 break;
1988 case PPC_IFTOKEN:
1989 t = tline = expand_smacro(tline);
1990 while (tok_type_(t, TOK_WHITESPACE))
1991 t = t->next;
1993 j = false;
1994 if (t) {
1995 t = t->next; /* Skip the actual token */
1996 while (tok_type_(t, TOK_WHITESPACE))
1997 t = t->next;
1998 j = !t; /* Should be nothing left */
2000 break;
2002 case PPC_IFEMPTY:
2003 t = tline = expand_smacro(tline);
2004 while (tok_type_(t, TOK_WHITESPACE))
2005 t = t->next;
2007 j = !t; /* Should be empty */
2008 break;
2010 case PPC_IF:
2011 t = tline = expand_smacro(tline);
2012 tptr = &t;
2013 tokval.t_type = TOKEN_INVALID;
2014 evalresult = evaluate(ppscan, tptr, &tokval,
2015 NULL, pass | CRITICAL, NULL);
2016 if (!evalresult)
2017 return -1;
2018 if (tokval.t_type)
2019 nasm_error(ERR_WARNING|ERR_PASS1,
2020 "trailing garbage after expression ignored");
2021 if (!is_simple(evalresult)) {
2022 nasm_error(ERR_NONFATAL,
2023 "non-constant value given to `%s'", pp_directives[ct]);
2024 goto fail;
2026 j = reloc_value(evalresult) != 0;
2027 break;
2029 default:
2030 nasm_error(ERR_FATAL,
2031 "preprocessor directive `%s' not yet implemented",
2032 pp_directives[ct]);
2033 goto fail;
2036 free_tlist(origline);
2037 return j ^ PP_NEGATIVE(ct);
2039 fail:
2040 free_tlist(origline);
2041 return -1;
2045 * Common code for defining an smacro
2047 static bool define_smacro(Context *ctx, const char *mname, bool casesense,
2048 int nparam, Token *expansion)
2050 SMacro *smac, **smhead;
2051 struct hash_table *smtbl;
2053 if (smacro_defined(ctx, mname, nparam, &smac, casesense)) {
2054 if (!smac) {
2055 nasm_error(ERR_WARNING|ERR_PASS1,
2056 "single-line macro `%s' defined both with and"
2057 " without parameters", mname);
2059 * Some instances of the old code considered this a failure,
2060 * some others didn't. What is the right thing to do here?
2062 free_tlist(expansion);
2063 return false; /* Failure */
2064 } else {
2066 * We're redefining, so we have to take over an
2067 * existing SMacro structure. This means freeing
2068 * what was already in it.
2070 nasm_free(smac->name);
2071 free_tlist(smac->expansion);
2073 } else {
2074 smtbl = ctx ? &ctx->localmac : &smacros;
2075 smhead = (SMacro **) hash_findi_add(smtbl, mname);
2076 smac = nasm_malloc(sizeof(SMacro));
2077 smac->next = *smhead;
2078 *smhead = smac;
2080 smac->name = nasm_strdup(mname);
2081 smac->casesense = casesense;
2082 smac->nparam = nparam;
2083 smac->expansion = expansion;
2084 smac->in_progress = false;
2085 return true; /* Success */
2089 * Undefine an smacro
2091 static void undef_smacro(Context *ctx, const char *mname)
2093 SMacro **smhead, *s, **sp;
2094 struct hash_table *smtbl;
2096 smtbl = ctx ? &ctx->localmac : &smacros;
2097 smhead = (SMacro **)hash_findi(smtbl, mname, NULL);
2099 if (smhead) {
2101 * We now have a macro name... go hunt for it.
2103 sp = smhead;
2104 while ((s = *sp) != NULL) {
2105 if (!mstrcmp(s->name, mname, s->casesense)) {
2106 *sp = s->next;
2107 nasm_free(s->name);
2108 free_tlist(s->expansion);
2109 nasm_free(s);
2110 } else {
2111 sp = &s->next;
2118 * Parse a mmacro specification.
2120 static bool parse_mmacro_spec(Token *tline, MMacro *def, const char *directive)
2122 bool err;
2124 tline = tline->next;
2125 skip_white_(tline);
2126 tline = expand_id(tline);
2127 if (!tok_type_(tline, TOK_ID)) {
2128 nasm_error(ERR_NONFATAL, "`%s' expects a macro name", directive);
2129 return false;
2132 def->prev = NULL;
2133 def->name = nasm_strdup(tline->text);
2134 def->plus = false;
2135 def->nolist = false;
2136 def->in_progress = 0;
2137 def->rep_nest = NULL;
2138 def->nparam_min = 0;
2139 def->nparam_max = 0;
2141 tline = expand_smacro(tline->next);
2142 skip_white_(tline);
2143 if (!tok_type_(tline, TOK_NUMBER)) {
2144 nasm_error(ERR_NONFATAL, "`%s' expects a parameter count", directive);
2145 } else {
2146 def->nparam_min = def->nparam_max =
2147 readnum(tline->text, &err);
2148 if (err)
2149 nasm_error(ERR_NONFATAL,
2150 "unable to parse parameter count `%s'", tline->text);
2152 if (tline && tok_is_(tline->next, "-")) {
2153 tline = tline->next->next;
2154 if (tok_is_(tline, "*")) {
2155 def->nparam_max = INT_MAX;
2156 } else if (!tok_type_(tline, TOK_NUMBER)) {
2157 nasm_error(ERR_NONFATAL,
2158 "`%s' expects a parameter count after `-'", directive);
2159 } else {
2160 def->nparam_max = readnum(tline->text, &err);
2161 if (err) {
2162 nasm_error(ERR_NONFATAL, "unable to parse parameter count `%s'",
2163 tline->text);
2165 if (def->nparam_min > def->nparam_max) {
2166 nasm_error(ERR_NONFATAL, "minimum parameter count exceeds maximum");
2167 def->nparam_max = def->nparam_min;
2171 if (tline && tok_is_(tline->next, "+")) {
2172 tline = tline->next;
2173 def->plus = true;
2175 if (tline && tok_type_(tline->next, TOK_ID) &&
2176 !nasm_stricmp(tline->next->text, ".nolist")) {
2177 tline = tline->next;
2178 def->nolist = true;
2182 * Handle default parameters.
2184 if (tline && tline->next) {
2185 def->dlist = tline->next;
2186 tline->next = NULL;
2187 count_mmac_params(def->dlist, &def->ndefs, &def->defaults);
2188 } else {
2189 def->dlist = NULL;
2190 def->defaults = NULL;
2192 def->expansion = NULL;
2194 if (def->defaults && def->ndefs > def->nparam_max - def->nparam_min &&
2195 !def->plus)
2196 nasm_error(ERR_WARNING|ERR_PASS1|ERR_WARN_MDP,
2197 "too many default macro parameters");
2199 return true;
2204 * Decode a size directive
2206 static int parse_size(const char *str) {
2207 static const char *size_names[] =
2208 { "byte", "dword", "oword", "qword", "tword", "word", "yword" };
2209 static const int sizes[] =
2210 { 0, 1, 4, 16, 8, 10, 2, 32 };
2211 return str ? sizes[bsii(str, size_names, ARRAY_SIZE(size_names))+1] : 0;
2215 * Process a preprocessor %pragma directive. Currently there are none.
2216 * Gets passed the token list starting with the "preproc" token from
2217 * "%pragma preproc".
2219 static void do_pragma_preproc(Token *tline)
2221 /* Skip to the real stuff */
2222 tline = tline->next;
2223 skip_white_(tline);
2224 if (!tline)
2225 return;
2227 (void)tline; /* Nothing else to do at present */
2231 * find and process preprocessor directive in passed line
2232 * Find out if a line contains a preprocessor directive, and deal
2233 * with it if so.
2235 * If a directive _is_ found, it is the responsibility of this routine
2236 * (and not the caller) to free_tlist() the line.
2238 * @param tline a pointer to the current tokeninzed line linked list
2239 * @param output if this directive generated output
2240 * @return DIRECTIVE_FOUND or NO_DIRECTIVE_FOUND
2243 static int do_directive(Token *tline, char **output)
2245 enum preproc_token i;
2246 int j;
2247 bool err;
2248 int nparam;
2249 bool nolist;
2250 bool casesense;
2251 int k, m;
2252 int offset;
2253 char *p, *pp;
2254 const char *found_path;
2255 const char *mname;
2256 Include *inc;
2257 Context *ctx;
2258 Cond *cond;
2259 MMacro *mmac, **mmhead;
2260 Token *t = NULL, *tt, *param_start, *macro_start, *last, **tptr, *origline;
2261 Line *l;
2262 struct tokenval tokval;
2263 expr *evalresult;
2264 MMacro *tmp_defining; /* Used when manipulating rep_nest */
2265 int64_t count;
2266 size_t len;
2267 int severity;
2269 *output = NULL; /* No output generated */
2270 origline = tline;
2272 skip_white_(tline);
2273 if (!tline || !tok_type_(tline, TOK_PREPROC_ID) ||
2274 (tline->text[0] && (tline->text[1] == '%' ||
2275 tline->text[1] == '$' ||
2276 tline->text[1] == '!')))
2277 return NO_DIRECTIVE_FOUND;
2279 i = pp_token_hash(tline->text);
2282 * FIXME: We zap execution of PP_RMACRO, PP_IRMACRO, PP_EXITMACRO
2283 * since they are known to be buggy at moment, we need to fix them
2284 * in future release (2.09-2.10)
2286 if (i == PP_RMACRO || i == PP_IRMACRO || i == PP_EXITMACRO) {
2287 nasm_error(ERR_NONFATAL, "unknown preprocessor directive `%s'",
2288 tline->text);
2289 return NO_DIRECTIVE_FOUND;
2293 * If we're in a non-emitting branch of a condition construct,
2294 * or walking to the end of an already terminated %rep block,
2295 * we should ignore all directives except for condition
2296 * directives.
2298 if (((istk->conds && !emitting(istk->conds->state)) ||
2299 (istk->mstk && !istk->mstk->in_progress)) && !is_condition(i)) {
2300 return NO_DIRECTIVE_FOUND;
2304 * If we're defining a macro or reading a %rep block, we should
2305 * ignore all directives except for %macro/%imacro (which nest),
2306 * %endm/%endmacro, and (only if we're in a %rep block) %endrep.
2307 * If we're in a %rep block, another %rep nests, so should be let through.
2309 if (defining && i != PP_MACRO && i != PP_IMACRO &&
2310 i != PP_RMACRO && i != PP_IRMACRO &&
2311 i != PP_ENDMACRO && i != PP_ENDM &&
2312 (defining->name || (i != PP_ENDREP && i != PP_REP))) {
2313 return NO_DIRECTIVE_FOUND;
2316 if (defining) {
2317 if (i == PP_MACRO || i == PP_IMACRO ||
2318 i == PP_RMACRO || i == PP_IRMACRO) {
2319 nested_mac_count++;
2320 return NO_DIRECTIVE_FOUND;
2321 } else if (nested_mac_count > 0) {
2322 if (i == PP_ENDMACRO) {
2323 nested_mac_count--;
2324 return NO_DIRECTIVE_FOUND;
2327 if (!defining->name) {
2328 if (i == PP_REP) {
2329 nested_rep_count++;
2330 return NO_DIRECTIVE_FOUND;
2331 } else if (nested_rep_count > 0) {
2332 if (i == PP_ENDREP) {
2333 nested_rep_count--;
2334 return NO_DIRECTIVE_FOUND;
2340 switch (i) {
2341 case PP_INVALID:
2342 nasm_error(ERR_NONFATAL, "unknown preprocessor directive `%s'",
2343 tline->text);
2344 return NO_DIRECTIVE_FOUND; /* didn't get it */
2346 case PP_PRAGMA:
2348 * %pragma namespace options...
2350 * The namespace "preproc" is reserved for the preprocessor;
2351 * all other namespaces generate a [pragma] assembly directive.
2353 * Invalid %pragmas are ignored and may have different
2354 * meaning in future versions of NASM.
2356 tline = tline->next;
2357 skip_white_(tline);
2358 tline = expand_smacro(tline);
2359 if (tok_type_(tline, TOK_ID)) {
2360 if (!nasm_stricmp(tline->text, "preproc")) {
2361 /* Preprocessor pragma */
2362 do_pragma_preproc(tline);
2363 } else {
2364 /* Build the assembler directive */
2365 t = new_Token(NULL, TOK_OTHER, "[", 1);
2366 t->next = new_Token(NULL, TOK_ID, "pragma", 6);
2367 t->next->next = new_Token(tline, TOK_WHITESPACE, NULL, 0);
2368 tline = t;
2369 for (t = tline; t->next; t = t->next)
2371 t->next = new_Token(NULL, TOK_OTHER, "]", 1);
2372 /* true here can be revisited in the future */
2373 *output = detoken(tline, true);
2376 free_tlist(origline);
2377 return DIRECTIVE_FOUND;
2379 case PP_STACKSIZE:
2380 /* Directive to tell NASM what the default stack size is. The
2381 * default is for a 16-bit stack, and this can be overriden with
2382 * %stacksize large.
2384 tline = tline->next;
2385 if (tline && tline->type == TOK_WHITESPACE)
2386 tline = tline->next;
2387 if (!tline || tline->type != TOK_ID) {
2388 nasm_error(ERR_NONFATAL, "`%%stacksize' missing size parameter");
2389 free_tlist(origline);
2390 return DIRECTIVE_FOUND;
2392 if (nasm_stricmp(tline->text, "flat") == 0) {
2393 /* All subsequent ARG directives are for a 32-bit stack */
2394 StackSize = 4;
2395 StackPointer = "ebp";
2396 ArgOffset = 8;
2397 LocalOffset = 0;
2398 } else if (nasm_stricmp(tline->text, "flat64") == 0) {
2399 /* All subsequent ARG directives are for a 64-bit stack */
2400 StackSize = 8;
2401 StackPointer = "rbp";
2402 ArgOffset = 16;
2403 LocalOffset = 0;
2404 } else if (nasm_stricmp(tline->text, "large") == 0) {
2405 /* All subsequent ARG directives are for a 16-bit stack,
2406 * far function call.
2408 StackSize = 2;
2409 StackPointer = "bp";
2410 ArgOffset = 4;
2411 LocalOffset = 0;
2412 } else if (nasm_stricmp(tline->text, "small") == 0) {
2413 /* All subsequent ARG directives are for a 16-bit stack,
2414 * far function call. We don't support near functions.
2416 StackSize = 2;
2417 StackPointer = "bp";
2418 ArgOffset = 6;
2419 LocalOffset = 0;
2420 } else {
2421 nasm_error(ERR_NONFATAL, "`%%stacksize' invalid size type");
2422 free_tlist(origline);
2423 return DIRECTIVE_FOUND;
2425 free_tlist(origline);
2426 return DIRECTIVE_FOUND;
2428 case PP_ARG:
2429 /* TASM like ARG directive to define arguments to functions, in
2430 * the following form:
2432 * ARG arg1:WORD, arg2:DWORD, arg4:QWORD
2434 offset = ArgOffset;
2435 do {
2436 char *arg, directive[256];
2437 int size = StackSize;
2439 /* Find the argument name */
2440 tline = tline->next;
2441 if (tline && tline->type == TOK_WHITESPACE)
2442 tline = tline->next;
2443 if (!tline || tline->type != TOK_ID) {
2444 nasm_error(ERR_NONFATAL, "`%%arg' missing argument parameter");
2445 free_tlist(origline);
2446 return DIRECTIVE_FOUND;
2448 arg = tline->text;
2450 /* Find the argument size type */
2451 tline = tline->next;
2452 if (!tline || tline->type != TOK_OTHER
2453 || tline->text[0] != ':') {
2454 nasm_error(ERR_NONFATAL,
2455 "Syntax error processing `%%arg' directive");
2456 free_tlist(origline);
2457 return DIRECTIVE_FOUND;
2459 tline = tline->next;
2460 if (!tline || tline->type != TOK_ID) {
2461 nasm_error(ERR_NONFATAL, "`%%arg' missing size type parameter");
2462 free_tlist(origline);
2463 return DIRECTIVE_FOUND;
2466 /* Allow macro expansion of type parameter */
2467 tt = tokenize(tline->text);
2468 tt = expand_smacro(tt);
2469 size = parse_size(tt->text);
2470 if (!size) {
2471 nasm_error(ERR_NONFATAL,
2472 "Invalid size type for `%%arg' missing directive");
2473 free_tlist(tt);
2474 free_tlist(origline);
2475 return DIRECTIVE_FOUND;
2477 free_tlist(tt);
2479 /* Round up to even stack slots */
2480 size = ALIGN(size, StackSize);
2482 /* Now define the macro for the argument */
2483 snprintf(directive, sizeof(directive), "%%define %s (%s+%d)",
2484 arg, StackPointer, offset);
2485 do_directive(tokenize(directive), output);
2486 offset += size;
2488 /* Move to the next argument in the list */
2489 tline = tline->next;
2490 if (tline && tline->type == TOK_WHITESPACE)
2491 tline = tline->next;
2492 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2493 ArgOffset = offset;
2494 free_tlist(origline);
2495 return DIRECTIVE_FOUND;
2497 case PP_LOCAL:
2498 /* TASM like LOCAL directive to define local variables for a
2499 * function, in the following form:
2501 * LOCAL local1:WORD, local2:DWORD, local4:QWORD = LocalSize
2503 * The '= LocalSize' at the end is ignored by NASM, but is
2504 * required by TASM to define the local parameter size (and used
2505 * by the TASM macro package).
2507 offset = LocalOffset;
2508 do {
2509 char *local, directive[256];
2510 int size = StackSize;
2512 /* Find the argument name */
2513 tline = tline->next;
2514 if (tline && tline->type == TOK_WHITESPACE)
2515 tline = tline->next;
2516 if (!tline || tline->type != TOK_ID) {
2517 nasm_error(ERR_NONFATAL,
2518 "`%%local' missing argument parameter");
2519 free_tlist(origline);
2520 return DIRECTIVE_FOUND;
2522 local = tline->text;
2524 /* Find the argument size type */
2525 tline = tline->next;
2526 if (!tline || tline->type != TOK_OTHER
2527 || tline->text[0] != ':') {
2528 nasm_error(ERR_NONFATAL,
2529 "Syntax error processing `%%local' directive");
2530 free_tlist(origline);
2531 return DIRECTIVE_FOUND;
2533 tline = tline->next;
2534 if (!tline || tline->type != TOK_ID) {
2535 nasm_error(ERR_NONFATAL,
2536 "`%%local' missing size type parameter");
2537 free_tlist(origline);
2538 return DIRECTIVE_FOUND;
2541 /* Allow macro expansion of type parameter */
2542 tt = tokenize(tline->text);
2543 tt = expand_smacro(tt);
2544 size = parse_size(tt->text);
2545 if (!size) {
2546 nasm_error(ERR_NONFATAL,
2547 "Invalid size type for `%%local' missing directive");
2548 free_tlist(tt);
2549 free_tlist(origline);
2550 return DIRECTIVE_FOUND;
2552 free_tlist(tt);
2554 /* Round up to even stack slots */
2555 size = ALIGN(size, StackSize);
2557 offset += size; /* Negative offset, increment before */
2559 /* Now define the macro for the argument */
2560 snprintf(directive, sizeof(directive), "%%define %s (%s-%d)",
2561 local, StackPointer, offset);
2562 do_directive(tokenize(directive), output);
2564 /* Now define the assign to setup the enter_c macro correctly */
2565 snprintf(directive, sizeof(directive),
2566 "%%assign %%$localsize %%$localsize+%d", size);
2567 do_directive(tokenize(directive), output);
2569 /* Move to the next argument in the list */
2570 tline = tline->next;
2571 if (tline && tline->type == TOK_WHITESPACE)
2572 tline = tline->next;
2573 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2574 LocalOffset = offset;
2575 free_tlist(origline);
2576 return DIRECTIVE_FOUND;
2578 case PP_CLEAR:
2579 if (tline->next)
2580 nasm_error(ERR_WARNING|ERR_PASS1,
2581 "trailing garbage after `%%clear' ignored");
2582 free_macros();
2583 init_macros();
2584 free_tlist(origline);
2585 return DIRECTIVE_FOUND;
2587 case PP_DEPEND:
2588 t = tline->next = expand_smacro(tline->next);
2589 skip_white_(t);
2590 if (!t || (t->type != TOK_STRING &&
2591 t->type != TOK_INTERNAL_STRING)) {
2592 nasm_error(ERR_NONFATAL, "`%%depend' expects a file name");
2593 free_tlist(origline);
2594 return DIRECTIVE_FOUND; /* but we did _something_ */
2596 if (t->next)
2597 nasm_error(ERR_WARNING|ERR_PASS1,
2598 "trailing garbage after `%%depend' ignored");
2599 p = t->text;
2600 if (t->type != TOK_INTERNAL_STRING)
2601 nasm_unquote_cstr(p, i);
2602 nasm_add_string_to_strlist(dephead, p);
2603 free_tlist(origline);
2604 return DIRECTIVE_FOUND;
2606 case PP_INCLUDE:
2607 t = tline->next = expand_smacro(tline->next);
2608 skip_white_(t);
2610 if (!t || (t->type != TOK_STRING &&
2611 t->type != TOK_INTERNAL_STRING)) {
2612 nasm_error(ERR_NONFATAL, "`%%include' expects a file name");
2613 free_tlist(origline);
2614 return DIRECTIVE_FOUND; /* but we did _something_ */
2616 if (t->next)
2617 nasm_error(ERR_WARNING|ERR_PASS1,
2618 "trailing garbage after `%%include' ignored");
2619 p = t->text;
2620 if (t->type != TOK_INTERNAL_STRING)
2621 nasm_unquote_cstr(p, i);
2622 inc = nasm_malloc(sizeof(Include));
2623 inc->next = istk;
2624 inc->conds = NULL;
2625 found_path = NULL;
2626 inc->fp = inc_fopen(p, dephead, &found_path,
2627 pass == 0 ? INC_OPTIONAL : INC_NEEDED, NF_TEXT);
2628 if (!inc->fp) {
2629 /* -MG given but file not found */
2630 nasm_free(inc);
2631 } else {
2632 inc->fname = src_set_fname(found_path ? found_path : p);
2633 inc->lineno = src_set_linnum(0);
2634 inc->lineinc = 1;
2635 inc->expansion = NULL;
2636 inc->mstk = NULL;
2637 istk = inc;
2638 lfmt->uplevel(LIST_INCLUDE);
2640 free_tlist(origline);
2641 return DIRECTIVE_FOUND;
2643 case PP_USE:
2645 static macros_t *use_pkg;
2646 const char *pkg_macro = NULL;
2648 tline = tline->next;
2649 skip_white_(tline);
2650 tline = expand_id(tline);
2652 if (!tline || (tline->type != TOK_STRING &&
2653 tline->type != TOK_INTERNAL_STRING &&
2654 tline->type != TOK_ID)) {
2655 nasm_error(ERR_NONFATAL, "`%%use' expects a package name");
2656 free_tlist(origline);
2657 return DIRECTIVE_FOUND; /* but we did _something_ */
2659 if (tline->next)
2660 nasm_error(ERR_WARNING|ERR_PASS1,
2661 "trailing garbage after `%%use' ignored");
2662 if (tline->type == TOK_STRING)
2663 nasm_unquote_cstr(tline->text, i);
2664 use_pkg = nasm_stdmac_find_package(tline->text);
2665 if (!use_pkg)
2666 nasm_error(ERR_NONFATAL, "unknown `%%use' package: %s", tline->text);
2667 else
2668 pkg_macro = (char *)use_pkg + 1; /* The first string will be <%define>__USE_*__ */
2669 if (use_pkg && ! smacro_defined(NULL, pkg_macro, 0, NULL, true)) {
2670 /* Not already included, go ahead and include it */
2671 stdmacpos = use_pkg;
2673 free_tlist(origline);
2674 return DIRECTIVE_FOUND;
2676 case PP_PUSH:
2677 case PP_REPL:
2678 case PP_POP:
2679 tline = tline->next;
2680 skip_white_(tline);
2681 tline = expand_id(tline);
2682 if (tline) {
2683 if (!tok_type_(tline, TOK_ID)) {
2684 nasm_error(ERR_NONFATAL, "`%s' expects a context identifier",
2685 pp_directives[i]);
2686 free_tlist(origline);
2687 return DIRECTIVE_FOUND; /* but we did _something_ */
2689 if (tline->next)
2690 nasm_error(ERR_WARNING|ERR_PASS1,
2691 "trailing garbage after `%s' ignored",
2692 pp_directives[i]);
2693 p = nasm_strdup(tline->text);
2694 } else {
2695 p = NULL; /* Anonymous */
2698 if (i == PP_PUSH) {
2699 ctx = nasm_malloc(sizeof(Context));
2700 ctx->next = cstk;
2701 hash_init(&ctx->localmac, HASH_SMALL);
2702 ctx->name = p;
2703 ctx->number = unique++;
2704 cstk = ctx;
2705 } else {
2706 /* %pop or %repl */
2707 if (!cstk) {
2708 nasm_error(ERR_NONFATAL, "`%s': context stack is empty",
2709 pp_directives[i]);
2710 } else if (i == PP_POP) {
2711 if (p && (!cstk->name || nasm_stricmp(p, cstk->name)))
2712 nasm_error(ERR_NONFATAL, "`%%pop' in wrong context: %s, "
2713 "expected %s",
2714 cstk->name ? cstk->name : "anonymous", p);
2715 else
2716 ctx_pop();
2717 } else {
2718 /* i == PP_REPL */
2719 nasm_free(cstk->name);
2720 cstk->name = p;
2721 p = NULL;
2723 nasm_free(p);
2725 free_tlist(origline);
2726 return DIRECTIVE_FOUND;
2727 case PP_FATAL:
2728 severity = ERR_FATAL;
2729 goto issue_error;
2730 case PP_ERROR:
2731 severity = ERR_NONFATAL;
2732 goto issue_error;
2733 case PP_WARNING:
2734 severity = ERR_WARNING|ERR_WARN_USER;
2735 goto issue_error;
2737 issue_error:
2739 /* Only error out if this is the final pass */
2740 if (pass != 2 && i != PP_FATAL)
2741 return DIRECTIVE_FOUND;
2743 tline->next = expand_smacro(tline->next);
2744 tline = tline->next;
2745 skip_white_(tline);
2746 t = tline ? tline->next : NULL;
2747 skip_white_(t);
2748 if (tok_type_(tline, TOK_STRING) && !t) {
2749 /* The line contains only a quoted string */
2750 p = tline->text;
2751 nasm_unquote(p, NULL); /* Ignore NUL character truncation */
2752 nasm_error(severity, "%s", p);
2753 } else {
2754 /* Not a quoted string, or more than a quoted string */
2755 p = detoken(tline, false);
2756 nasm_error(severity, "%s", p);
2757 nasm_free(p);
2759 free_tlist(origline);
2760 return DIRECTIVE_FOUND;
2763 CASE_PP_IF:
2764 if (istk->conds && !emitting(istk->conds->state))
2765 j = COND_NEVER;
2766 else {
2767 j = if_condition(tline->next, i);
2768 tline->next = NULL; /* it got freed */
2769 j = j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2771 cond = nasm_malloc(sizeof(Cond));
2772 cond->next = istk->conds;
2773 cond->state = j;
2774 istk->conds = cond;
2775 if(istk->mstk)
2776 istk->mstk->condcnt ++;
2777 free_tlist(origline);
2778 return DIRECTIVE_FOUND;
2780 CASE_PP_ELIF:
2781 if (!istk->conds)
2782 nasm_error(ERR_FATAL, "`%s': no matching `%%if'", pp_directives[i]);
2783 switch(istk->conds->state) {
2784 case COND_IF_TRUE:
2785 istk->conds->state = COND_DONE;
2786 break;
2788 case COND_DONE:
2789 case COND_NEVER:
2790 break;
2792 case COND_ELSE_TRUE:
2793 case COND_ELSE_FALSE:
2794 nasm_error(ERR_WARNING|ERR_PASS1|ERR_PP_PRECOND,
2795 "`%%elif' after `%%else' ignored");
2796 istk->conds->state = COND_NEVER;
2797 break;
2799 case COND_IF_FALSE:
2801 * IMPORTANT: In the case of %if, we will already have
2802 * called expand_mmac_params(); however, if we're
2803 * processing an %elif we must have been in a
2804 * non-emitting mode, which would have inhibited
2805 * the normal invocation of expand_mmac_params().
2806 * Therefore, we have to do it explicitly here.
2808 j = if_condition(expand_mmac_params(tline->next), i);
2809 tline->next = NULL; /* it got freed */
2810 istk->conds->state =
2811 j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2812 break;
2814 free_tlist(origline);
2815 return DIRECTIVE_FOUND;
2817 case PP_ELSE:
2818 if (tline->next)
2819 nasm_error(ERR_WARNING|ERR_PASS1|ERR_PP_PRECOND,
2820 "trailing garbage after `%%else' ignored");
2821 if (!istk->conds)
2822 nasm_fatal(0, "`%%else: no matching `%%if'");
2823 switch(istk->conds->state) {
2824 case COND_IF_TRUE:
2825 case COND_DONE:
2826 istk->conds->state = COND_ELSE_FALSE;
2827 break;
2829 case COND_NEVER:
2830 break;
2832 case COND_IF_FALSE:
2833 istk->conds->state = COND_ELSE_TRUE;
2834 break;
2836 case COND_ELSE_TRUE:
2837 case COND_ELSE_FALSE:
2838 nasm_error(ERR_WARNING|ERR_PASS1|ERR_PP_PRECOND,
2839 "`%%else' after `%%else' ignored.");
2840 istk->conds->state = COND_NEVER;
2841 break;
2843 free_tlist(origline);
2844 return DIRECTIVE_FOUND;
2846 case PP_ENDIF:
2847 if (tline->next)
2848 nasm_error(ERR_WARNING|ERR_PASS1|ERR_PP_PRECOND,
2849 "trailing garbage after `%%endif' ignored");
2850 if (!istk->conds)
2851 nasm_error(ERR_FATAL, "`%%endif': no matching `%%if'");
2852 cond = istk->conds;
2853 istk->conds = cond->next;
2854 nasm_free(cond);
2855 if(istk->mstk)
2856 istk->mstk->condcnt --;
2857 free_tlist(origline);
2858 return DIRECTIVE_FOUND;
2860 case PP_RMACRO:
2861 case PP_IRMACRO:
2862 case PP_MACRO:
2863 case PP_IMACRO:
2864 if (defining) {
2865 nasm_error(ERR_FATAL, "`%s': already defining a macro",
2866 pp_directives[i]);
2867 return DIRECTIVE_FOUND;
2869 defining = nasm_zalloc(sizeof(MMacro));
2870 defining->max_depth = ((i == PP_RMACRO) || (i == PP_IRMACRO))
2871 ? nasm_limit[LIMIT_MACROS] : 0;
2872 defining->casesense = (i == PP_MACRO) || (i == PP_RMACRO);
2873 if (!parse_mmacro_spec(tline, defining, pp_directives[i])) {
2874 nasm_free(defining);
2875 defining = NULL;
2876 return DIRECTIVE_FOUND;
2879 src_get(&defining->xline, &defining->fname);
2881 mmac = (MMacro *) hash_findix(&mmacros, defining->name);
2882 while (mmac) {
2883 if (!strcmp(mmac->name, defining->name) &&
2884 (mmac->nparam_min <= defining->nparam_max
2885 || defining->plus)
2886 && (defining->nparam_min <= mmac->nparam_max
2887 || mmac->plus)) {
2888 nasm_error(ERR_WARNING|ERR_PASS1,
2889 "redefining multi-line macro `%s'", defining->name);
2890 return DIRECTIVE_FOUND;
2892 mmac = mmac->next;
2894 free_tlist(origline);
2895 return DIRECTIVE_FOUND;
2897 case PP_ENDM:
2898 case PP_ENDMACRO:
2899 if (! (defining && defining->name)) {
2900 nasm_error(ERR_NONFATAL, "`%s': not defining a macro", tline->text);
2901 return DIRECTIVE_FOUND;
2903 mmhead = (MMacro **) hash_findi_add(&mmacros, defining->name);
2904 defining->next = *mmhead;
2905 *mmhead = defining;
2906 defining = NULL;
2907 free_tlist(origline);
2908 return DIRECTIVE_FOUND;
2910 case PP_EXITMACRO:
2912 * We must search along istk->expansion until we hit a
2913 * macro-end marker for a macro with a name. Then we
2914 * bypass all lines between exitmacro and endmacro.
2916 list_for_each(l, istk->expansion)
2917 if (l->finishes && l->finishes->name)
2918 break;
2920 if (l) {
2922 * Remove all conditional entries relative to this
2923 * macro invocation. (safe to do in this context)
2925 for ( ; l->finishes->condcnt > 0; l->finishes->condcnt --) {
2926 cond = istk->conds;
2927 istk->conds = cond->next;
2928 nasm_free(cond);
2930 istk->expansion = l;
2931 } else {
2932 nasm_error(ERR_NONFATAL, "`%%exitmacro' not within `%%macro' block");
2934 free_tlist(origline);
2935 return DIRECTIVE_FOUND;
2937 case PP_UNMACRO:
2938 case PP_UNIMACRO:
2940 MMacro **mmac_p;
2941 MMacro spec;
2943 spec.casesense = (i == PP_UNMACRO);
2944 if (!parse_mmacro_spec(tline, &spec, pp_directives[i])) {
2945 return DIRECTIVE_FOUND;
2947 mmac_p = (MMacro **) hash_findi(&mmacros, spec.name, NULL);
2948 while (mmac_p && *mmac_p) {
2949 mmac = *mmac_p;
2950 if (mmac->casesense == spec.casesense &&
2951 !mstrcmp(mmac->name, spec.name, spec.casesense) &&
2952 mmac->nparam_min == spec.nparam_min &&
2953 mmac->nparam_max == spec.nparam_max &&
2954 mmac->plus == spec.plus) {
2955 *mmac_p = mmac->next;
2956 free_mmacro(mmac);
2957 } else {
2958 mmac_p = &mmac->next;
2961 free_tlist(origline);
2962 free_tlist(spec.dlist);
2963 return DIRECTIVE_FOUND;
2966 case PP_ROTATE:
2967 if (tline->next && tline->next->type == TOK_WHITESPACE)
2968 tline = tline->next;
2969 if (!tline->next) {
2970 free_tlist(origline);
2971 nasm_error(ERR_NONFATAL, "`%%rotate' missing rotate count");
2972 return DIRECTIVE_FOUND;
2974 t = expand_smacro(tline->next);
2975 tline->next = NULL;
2976 free_tlist(origline);
2977 tline = t;
2978 tptr = &t;
2979 tokval.t_type = TOKEN_INVALID;
2980 evalresult =
2981 evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
2982 free_tlist(tline);
2983 if (!evalresult)
2984 return DIRECTIVE_FOUND;
2985 if (tokval.t_type)
2986 nasm_error(ERR_WARNING|ERR_PASS1,
2987 "trailing garbage after expression ignored");
2988 if (!is_simple(evalresult)) {
2989 nasm_error(ERR_NONFATAL, "non-constant value given to `%%rotate'");
2990 return DIRECTIVE_FOUND;
2992 mmac = istk->mstk;
2993 while (mmac && !mmac->name) /* avoid mistaking %reps for macros */
2994 mmac = mmac->next_active;
2995 if (!mmac) {
2996 nasm_error(ERR_NONFATAL, "`%%rotate' invoked outside a macro call");
2997 } else if (mmac->nparam == 0) {
2998 nasm_error(ERR_NONFATAL,
2999 "`%%rotate' invoked within macro without parameters");
3000 } else {
3001 int rotate = mmac->rotate + reloc_value(evalresult);
3003 rotate %= (int)mmac->nparam;
3004 if (rotate < 0)
3005 rotate += mmac->nparam;
3007 mmac->rotate = rotate;
3009 return DIRECTIVE_FOUND;
3011 case PP_REP:
3012 nolist = false;
3013 do {
3014 tline = tline->next;
3015 } while (tok_type_(tline, TOK_WHITESPACE));
3017 if (tok_type_(tline, TOK_ID) &&
3018 nasm_stricmp(tline->text, ".nolist") == 0) {
3019 nolist = true;
3020 do {
3021 tline = tline->next;
3022 } while (tok_type_(tline, TOK_WHITESPACE));
3025 if (tline) {
3026 t = expand_smacro(tline);
3027 tptr = &t;
3028 tokval.t_type = TOKEN_INVALID;
3029 evalresult =
3030 evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
3031 if (!evalresult) {
3032 free_tlist(origline);
3033 return DIRECTIVE_FOUND;
3035 if (tokval.t_type)
3036 nasm_error(ERR_WARNING|ERR_PASS1,
3037 "trailing garbage after expression ignored");
3038 if (!is_simple(evalresult)) {
3039 nasm_error(ERR_NONFATAL, "non-constant value given to `%%rep'");
3040 return DIRECTIVE_FOUND;
3042 count = reloc_value(evalresult);
3043 if (count > nasm_limit[LIMIT_REP]) {
3044 nasm_error(ERR_NONFATAL,
3045 "`%%rep' count %"PRId64" exceeds limit (currently %"PRId64")",
3046 count, nasm_limit[LIMIT_REP]);
3047 count = 0;
3048 } else if (count < 0) {
3049 nasm_error(ERR_WARNING|ERR_PASS2|ERR_WARN_NEG_REP,
3050 "negative `%%rep' count: %"PRId64, count);
3051 count = 0;
3052 } else {
3053 count++;
3055 } else {
3056 nasm_error(ERR_NONFATAL, "`%%rep' expects a repeat count");
3057 count = 0;
3059 free_tlist(origline);
3061 tmp_defining = defining;
3062 defining = nasm_malloc(sizeof(MMacro));
3063 defining->prev = NULL;
3064 defining->name = NULL; /* flags this macro as a %rep block */
3065 defining->casesense = false;
3066 defining->plus = false;
3067 defining->nolist = nolist;
3068 defining->in_progress = count;
3069 defining->max_depth = 0;
3070 defining->nparam_min = defining->nparam_max = 0;
3071 defining->defaults = NULL;
3072 defining->dlist = NULL;
3073 defining->expansion = NULL;
3074 defining->next_active = istk->mstk;
3075 defining->rep_nest = tmp_defining;
3076 return DIRECTIVE_FOUND;
3078 case PP_ENDREP:
3079 if (!defining || defining->name) {
3080 nasm_error(ERR_NONFATAL, "`%%endrep': no matching `%%rep'");
3081 return DIRECTIVE_FOUND;
3085 * Now we have a "macro" defined - although it has no name
3086 * and we won't be entering it in the hash tables - we must
3087 * push a macro-end marker for it on to istk->expansion.
3088 * After that, it will take care of propagating itself (a
3089 * macro-end marker line for a macro which is really a %rep
3090 * block will cause the macro to be re-expanded, complete
3091 * with another macro-end marker to ensure the process
3092 * continues) until the whole expansion is forcibly removed
3093 * from istk->expansion by a %exitrep.
3095 l = nasm_malloc(sizeof(Line));
3096 l->next = istk->expansion;
3097 l->finishes = defining;
3098 l->first = NULL;
3099 istk->expansion = l;
3101 istk->mstk = defining;
3103 lfmt->uplevel(defining->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
3104 tmp_defining = defining;
3105 defining = defining->rep_nest;
3106 free_tlist(origline);
3107 return DIRECTIVE_FOUND;
3109 case PP_EXITREP:
3111 * We must search along istk->expansion until we hit a
3112 * macro-end marker for a macro with no name. Then we set
3113 * its `in_progress' flag to 0.
3115 list_for_each(l, istk->expansion)
3116 if (l->finishes && !l->finishes->name)
3117 break;
3119 if (l)
3120 l->finishes->in_progress = 1;
3121 else
3122 nasm_error(ERR_NONFATAL, "`%%exitrep' not within `%%rep' block");
3123 free_tlist(origline);
3124 return DIRECTIVE_FOUND;
3126 case PP_XDEFINE:
3127 case PP_IXDEFINE:
3128 case PP_DEFINE:
3129 case PP_IDEFINE:
3130 casesense = (i == PP_DEFINE || i == PP_XDEFINE);
3132 tline = tline->next;
3133 skip_white_(tline);
3134 tline = expand_id(tline);
3135 if (!tline || (tline->type != TOK_ID &&
3136 (tline->type != TOK_PREPROC_ID ||
3137 tline->text[1] != '$'))) {
3138 nasm_error(ERR_NONFATAL, "`%s' expects a macro identifier",
3139 pp_directives[i]);
3140 free_tlist(origline);
3141 return DIRECTIVE_FOUND;
3144 ctx = get_ctx(tline->text, &mname);
3145 last = tline;
3146 param_start = tline = tline->next;
3147 nparam = 0;
3149 /* Expand the macro definition now for %xdefine and %ixdefine */
3150 if ((i == PP_XDEFINE) || (i == PP_IXDEFINE))
3151 tline = expand_smacro(tline);
3153 if (tok_is_(tline, "(")) {
3155 * This macro has parameters.
3158 tline = tline->next;
3159 while (1) {
3160 skip_white_(tline);
3161 if (!tline) {
3162 nasm_error(ERR_NONFATAL, "parameter identifier expected");
3163 free_tlist(origline);
3164 return DIRECTIVE_FOUND;
3166 if (tline->type != TOK_ID) {
3167 nasm_error(ERR_NONFATAL,
3168 "`%s': parameter identifier expected",
3169 tline->text);
3170 free_tlist(origline);
3171 return DIRECTIVE_FOUND;
3173 tline->type = TOK_SMAC_PARAM + nparam++;
3174 tline = tline->next;
3175 skip_white_(tline);
3176 if (tok_is_(tline, ",")) {
3177 tline = tline->next;
3178 } else {
3179 if (!tok_is_(tline, ")")) {
3180 nasm_error(ERR_NONFATAL,
3181 "`)' expected to terminate macro template");
3182 free_tlist(origline);
3183 return DIRECTIVE_FOUND;
3185 break;
3188 last = tline;
3189 tline = tline->next;
3191 if (tok_type_(tline, TOK_WHITESPACE))
3192 last = tline, tline = tline->next;
3193 macro_start = NULL;
3194 last->next = NULL;
3195 t = tline;
3196 while (t) {
3197 if (t->type == TOK_ID) {
3198 list_for_each(tt, param_start)
3199 if (tt->type >= TOK_SMAC_PARAM &&
3200 !strcmp(tt->text, t->text))
3201 t->type = tt->type;
3203 tt = t->next;
3204 t->next = macro_start;
3205 macro_start = t;
3206 t = tt;
3209 * Good. We now have a macro name, a parameter count, and a
3210 * token list (in reverse order) for an expansion. We ought
3211 * to be OK just to create an SMacro, store it, and let
3212 * free_tlist have the rest of the line (which we have
3213 * carefully re-terminated after chopping off the expansion
3214 * from the end).
3216 define_smacro(ctx, mname, casesense, nparam, macro_start);
3217 free_tlist(origline);
3218 return DIRECTIVE_FOUND;
3220 case PP_UNDEF:
3221 tline = tline->next;
3222 skip_white_(tline);
3223 tline = expand_id(tline);
3224 if (!tline || (tline->type != TOK_ID &&
3225 (tline->type != TOK_PREPROC_ID ||
3226 tline->text[1] != '$'))) {
3227 nasm_error(ERR_NONFATAL, "`%%undef' expects a macro identifier");
3228 free_tlist(origline);
3229 return DIRECTIVE_FOUND;
3231 if (tline->next) {
3232 nasm_error(ERR_WARNING|ERR_PASS1,
3233 "trailing garbage after macro name ignored");
3236 /* Find the context that symbol belongs to */
3237 ctx = get_ctx(tline->text, &mname);
3238 undef_smacro(ctx, mname);
3239 free_tlist(origline);
3240 return DIRECTIVE_FOUND;
3242 case PP_DEFSTR:
3243 case PP_IDEFSTR:
3244 casesense = (i == PP_DEFSTR);
3246 tline = tline->next;
3247 skip_white_(tline);
3248 tline = expand_id(tline);
3249 if (!tline || (tline->type != TOK_ID &&
3250 (tline->type != TOK_PREPROC_ID ||
3251 tline->text[1] != '$'))) {
3252 nasm_error(ERR_NONFATAL, "`%s' expects a macro identifier",
3253 pp_directives[i]);
3254 free_tlist(origline);
3255 return DIRECTIVE_FOUND;
3258 ctx = get_ctx(tline->text, &mname);
3259 last = tline;
3260 tline = expand_smacro(tline->next);
3261 last->next = NULL;
3263 while (tok_type_(tline, TOK_WHITESPACE))
3264 tline = delete_Token(tline);
3266 p = detoken(tline, false);
3267 macro_start = nasm_malloc(sizeof(*macro_start));
3268 macro_start->next = NULL;
3269 macro_start->text = nasm_quote(p, strlen(p));
3270 macro_start->type = TOK_STRING;
3271 macro_start->a.mac = NULL;
3272 nasm_free(p);
3275 * We now have a macro name, an implicit parameter count of
3276 * zero, and a string token to use as an expansion. Create
3277 * and store an SMacro.
3279 define_smacro(ctx, mname, casesense, 0, macro_start);
3280 free_tlist(origline);
3281 return DIRECTIVE_FOUND;
3283 case PP_DEFTOK:
3284 case PP_IDEFTOK:
3285 casesense = (i == PP_DEFTOK);
3287 tline = tline->next;
3288 skip_white_(tline);
3289 tline = expand_id(tline);
3290 if (!tline || (tline->type != TOK_ID &&
3291 (tline->type != TOK_PREPROC_ID ||
3292 tline->text[1] != '$'))) {
3293 nasm_error(ERR_NONFATAL,
3294 "`%s' expects a macro identifier as first parameter",
3295 pp_directives[i]);
3296 free_tlist(origline);
3297 return DIRECTIVE_FOUND;
3299 ctx = get_ctx(tline->text, &mname);
3300 last = tline;
3301 tline = expand_smacro(tline->next);
3302 last->next = NULL;
3304 t = tline;
3305 while (tok_type_(t, TOK_WHITESPACE))
3306 t = t->next;
3307 /* t should now point to the string */
3308 if (!tok_type_(t, TOK_STRING)) {
3309 nasm_error(ERR_NONFATAL,
3310 "`%s` requires string as second parameter",
3311 pp_directives[i]);
3312 free_tlist(tline);
3313 free_tlist(origline);
3314 return DIRECTIVE_FOUND;
3318 * Convert the string to a token stream. Note that smacros
3319 * are stored with the token stream reversed, so we have to
3320 * reverse the output of tokenize().
3322 nasm_unquote_cstr(t->text, i);
3323 macro_start = reverse_tokens(tokenize(t->text));
3326 * We now have a macro name, an implicit parameter count of
3327 * zero, and a numeric token to use as an expansion. Create
3328 * and store an SMacro.
3330 define_smacro(ctx, mname, casesense, 0, macro_start);
3331 free_tlist(tline);
3332 free_tlist(origline);
3333 return DIRECTIVE_FOUND;
3335 case PP_PATHSEARCH:
3337 const char *found_path;
3339 casesense = true;
3341 tline = tline->next;
3342 skip_white_(tline);
3343 tline = expand_id(tline);
3344 if (!tline || (tline->type != TOK_ID &&
3345 (tline->type != TOK_PREPROC_ID ||
3346 tline->text[1] != '$'))) {
3347 nasm_error(ERR_NONFATAL,
3348 "`%%pathsearch' expects a macro identifier as first parameter");
3349 free_tlist(origline);
3350 return DIRECTIVE_FOUND;
3352 ctx = get_ctx(tline->text, &mname);
3353 last = tline;
3354 tline = expand_smacro(tline->next);
3355 last->next = NULL;
3357 t = tline;
3358 while (tok_type_(t, TOK_WHITESPACE))
3359 t = t->next;
3361 if (!t || (t->type != TOK_STRING &&
3362 t->type != TOK_INTERNAL_STRING)) {
3363 nasm_error(ERR_NONFATAL, "`%%pathsearch' expects a file name");
3364 free_tlist(tline);
3365 free_tlist(origline);
3366 return DIRECTIVE_FOUND; /* but we did _something_ */
3368 if (t->next)
3369 nasm_error(ERR_WARNING|ERR_PASS1,
3370 "trailing garbage after `%%pathsearch' ignored");
3371 p = t->text;
3372 if (t->type != TOK_INTERNAL_STRING)
3373 nasm_unquote(p, NULL);
3375 inc_fopen(p, NULL, &found_path, INC_PROBE, NF_BINARY);
3376 if (!found_path)
3377 found_path = p;
3378 macro_start = nasm_malloc(sizeof(*macro_start));
3379 macro_start->next = NULL;
3380 macro_start->text = nasm_quote(found_path, strlen(found_path));
3381 macro_start->type = TOK_STRING;
3382 macro_start->a.mac = NULL;
3385 * We now have a macro name, an implicit parameter count of
3386 * zero, and a string token to use as an expansion. Create
3387 * and store an SMacro.
3389 define_smacro(ctx, mname, casesense, 0, macro_start);
3390 free_tlist(tline);
3391 free_tlist(origline);
3392 return DIRECTIVE_FOUND;
3395 case PP_STRLEN:
3396 casesense = true;
3398 tline = tline->next;
3399 skip_white_(tline);
3400 tline = expand_id(tline);
3401 if (!tline || (tline->type != TOK_ID &&
3402 (tline->type != TOK_PREPROC_ID ||
3403 tline->text[1] != '$'))) {
3404 nasm_error(ERR_NONFATAL,
3405 "`%%strlen' expects a macro identifier as first parameter");
3406 free_tlist(origline);
3407 return DIRECTIVE_FOUND;
3409 ctx = get_ctx(tline->text, &mname);
3410 last = tline;
3411 tline = expand_smacro(tline->next);
3412 last->next = NULL;
3414 t = tline;
3415 while (tok_type_(t, TOK_WHITESPACE))
3416 t = t->next;
3417 /* t should now point to the string */
3418 if (!tok_type_(t, TOK_STRING)) {
3419 nasm_error(ERR_NONFATAL,
3420 "`%%strlen` requires string as second parameter");
3421 free_tlist(tline);
3422 free_tlist(origline);
3423 return DIRECTIVE_FOUND;
3426 macro_start = nasm_malloc(sizeof(*macro_start));
3427 macro_start->next = NULL;
3428 make_tok_num(macro_start, nasm_unquote(t->text, NULL));
3429 macro_start->a.mac = NULL;
3432 * We now have a macro name, an implicit parameter count of
3433 * zero, and a numeric token to use as an expansion. Create
3434 * and store an SMacro.
3436 define_smacro(ctx, mname, casesense, 0, macro_start);
3437 free_tlist(tline);
3438 free_tlist(origline);
3439 return DIRECTIVE_FOUND;
3441 case PP_STRCAT:
3442 casesense = true;
3444 tline = tline->next;
3445 skip_white_(tline);
3446 tline = expand_id(tline);
3447 if (!tline || (tline->type != TOK_ID &&
3448 (tline->type != TOK_PREPROC_ID ||
3449 tline->text[1] != '$'))) {
3450 nasm_error(ERR_NONFATAL,
3451 "`%%strcat' expects a macro identifier as first parameter");
3452 free_tlist(origline);
3453 return DIRECTIVE_FOUND;
3455 ctx = get_ctx(tline->text, &mname);
3456 last = tline;
3457 tline = expand_smacro(tline->next);
3458 last->next = NULL;
3460 len = 0;
3461 list_for_each(t, tline) {
3462 switch (t->type) {
3463 case TOK_WHITESPACE:
3464 break;
3465 case TOK_STRING:
3466 len += t->a.len = nasm_unquote(t->text, NULL);
3467 break;
3468 case TOK_OTHER:
3469 if (!strcmp(t->text, ",")) /* permit comma separators */
3470 break;
3471 /* else fall through */
3472 default:
3473 nasm_error(ERR_NONFATAL,
3474 "non-string passed to `%%strcat' (%d)", t->type);
3475 free_tlist(tline);
3476 free_tlist(origline);
3477 return DIRECTIVE_FOUND;
3481 p = pp = nasm_malloc(len);
3482 list_for_each(t, tline) {
3483 if (t->type == TOK_STRING) {
3484 memcpy(p, t->text, t->a.len);
3485 p += t->a.len;
3490 * We now have a macro name, an implicit parameter count of
3491 * zero, and a numeric token to use as an expansion. Create
3492 * and store an SMacro.
3494 macro_start = new_Token(NULL, TOK_STRING, NULL, 0);
3495 macro_start->text = nasm_quote(pp, len);
3496 nasm_free(pp);
3497 define_smacro(ctx, mname, casesense, 0, macro_start);
3498 free_tlist(tline);
3499 free_tlist(origline);
3500 return DIRECTIVE_FOUND;
3502 case PP_SUBSTR:
3504 int64_t start, count;
3505 size_t len;
3507 casesense = true;
3509 tline = tline->next;
3510 skip_white_(tline);
3511 tline = expand_id(tline);
3512 if (!tline || (tline->type != TOK_ID &&
3513 (tline->type != TOK_PREPROC_ID ||
3514 tline->text[1] != '$'))) {
3515 nasm_error(ERR_NONFATAL,
3516 "`%%substr' expects a macro identifier as first parameter");
3517 free_tlist(origline);
3518 return DIRECTIVE_FOUND;
3520 ctx = get_ctx(tline->text, &mname);
3521 last = tline;
3522 tline = expand_smacro(tline->next);
3523 last->next = NULL;
3525 if (tline) /* skip expanded id */
3526 t = tline->next;
3527 while (tok_type_(t, TOK_WHITESPACE))
3528 t = t->next;
3530 /* t should now point to the string */
3531 if (!tok_type_(t, TOK_STRING)) {
3532 nasm_error(ERR_NONFATAL,
3533 "`%%substr` requires string as second parameter");
3534 free_tlist(tline);
3535 free_tlist(origline);
3536 return DIRECTIVE_FOUND;
3539 tt = t->next;
3540 tptr = &tt;
3541 tokval.t_type = TOKEN_INVALID;
3542 evalresult = evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
3543 if (!evalresult) {
3544 free_tlist(tline);
3545 free_tlist(origline);
3546 return DIRECTIVE_FOUND;
3547 } else if (!is_simple(evalresult)) {
3548 nasm_error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3549 free_tlist(tline);
3550 free_tlist(origline);
3551 return DIRECTIVE_FOUND;
3553 start = evalresult->value - 1;
3555 while (tok_type_(tt, TOK_WHITESPACE))
3556 tt = tt->next;
3557 if (!tt) {
3558 count = 1; /* Backwards compatibility: one character */
3559 } else {
3560 tokval.t_type = TOKEN_INVALID;
3561 evalresult = evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
3562 if (!evalresult) {
3563 free_tlist(tline);
3564 free_tlist(origline);
3565 return DIRECTIVE_FOUND;
3566 } else if (!is_simple(evalresult)) {
3567 nasm_error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3568 free_tlist(tline);
3569 free_tlist(origline);
3570 return DIRECTIVE_FOUND;
3572 count = evalresult->value;
3575 len = nasm_unquote(t->text, NULL);
3577 /* make start and count being in range */
3578 if (start < 0)
3579 start = 0;
3580 if (count < 0)
3581 count = len + count + 1 - start;
3582 if (start + count > (int64_t)len)
3583 count = len - start;
3584 if (!len || count < 0 || start >=(int64_t)len)
3585 start = -1, count = 0; /* empty string */
3587 macro_start = nasm_malloc(sizeof(*macro_start));
3588 macro_start->next = NULL;
3589 macro_start->text = nasm_quote((start < 0) ? "" : t->text + start, count);
3590 macro_start->type = TOK_STRING;
3591 macro_start->a.mac = NULL;
3594 * We now have a macro name, an implicit parameter count of
3595 * zero, and a numeric token to use as an expansion. Create
3596 * and store an SMacro.
3598 define_smacro(ctx, mname, casesense, 0, macro_start);
3599 free_tlist(tline);
3600 free_tlist(origline);
3601 return DIRECTIVE_FOUND;
3604 case PP_ASSIGN:
3605 case PP_IASSIGN:
3606 casesense = (i == PP_ASSIGN);
3608 tline = tline->next;
3609 skip_white_(tline);
3610 tline = expand_id(tline);
3611 if (!tline || (tline->type != TOK_ID &&
3612 (tline->type != TOK_PREPROC_ID ||
3613 tline->text[1] != '$'))) {
3614 nasm_error(ERR_NONFATAL,
3615 "`%%%sassign' expects a macro identifier",
3616 (i == PP_IASSIGN ? "i" : ""));
3617 free_tlist(origline);
3618 return DIRECTIVE_FOUND;
3620 ctx = get_ctx(tline->text, &mname);
3621 last = tline;
3622 tline = expand_smacro(tline->next);
3623 last->next = NULL;
3625 t = tline;
3626 tptr = &t;
3627 tokval.t_type = TOKEN_INVALID;
3628 evalresult = evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
3629 free_tlist(tline);
3630 if (!evalresult) {
3631 free_tlist(origline);
3632 return DIRECTIVE_FOUND;
3635 if (tokval.t_type)
3636 nasm_error(ERR_WARNING|ERR_PASS1,
3637 "trailing garbage after expression ignored");
3639 if (!is_simple(evalresult)) {
3640 nasm_error(ERR_NONFATAL,
3641 "non-constant value given to `%%%sassign'",
3642 (i == PP_IASSIGN ? "i" : ""));
3643 free_tlist(origline);
3644 return DIRECTIVE_FOUND;
3647 macro_start = nasm_malloc(sizeof(*macro_start));
3648 macro_start->next = NULL;
3649 make_tok_num(macro_start, reloc_value(evalresult));
3650 macro_start->a.mac = NULL;
3653 * We now have a macro name, an implicit parameter count of
3654 * zero, and a numeric token to use as an expansion. Create
3655 * and store an SMacro.
3657 define_smacro(ctx, mname, casesense, 0, macro_start);
3658 free_tlist(origline);
3659 return DIRECTIVE_FOUND;
3661 case PP_LINE:
3663 * Syntax is `%line nnn[+mmm] [filename]'
3665 tline = tline->next;
3666 skip_white_(tline);
3667 if (!tok_type_(tline, TOK_NUMBER)) {
3668 nasm_error(ERR_NONFATAL, "`%%line' expects line number");
3669 free_tlist(origline);
3670 return DIRECTIVE_FOUND;
3672 k = readnum(tline->text, &err);
3673 m = 1;
3674 tline = tline->next;
3675 if (tok_is_(tline, "+")) {
3676 tline = tline->next;
3677 if (!tok_type_(tline, TOK_NUMBER)) {
3678 nasm_error(ERR_NONFATAL, "`%%line' expects line increment");
3679 free_tlist(origline);
3680 return DIRECTIVE_FOUND;
3682 m = readnum(tline->text, &err);
3683 tline = tline->next;
3685 skip_white_(tline);
3686 src_set_linnum(k);
3687 istk->lineinc = m;
3688 if (tline) {
3689 char *fname = detoken(tline, false);
3690 src_set_fname(fname);
3691 nasm_free(fname);
3693 free_tlist(origline);
3694 return DIRECTIVE_FOUND;
3696 default:
3697 nasm_error(ERR_FATAL,
3698 "preprocessor directive `%s' not yet implemented",
3699 pp_directives[i]);
3700 return DIRECTIVE_FOUND;
3705 * Ensure that a macro parameter contains a condition code and
3706 * nothing else. Return the condition code index if so, or -1
3707 * otherwise.
3709 static int find_cc(Token * t)
3711 Token *tt;
3713 if (!t)
3714 return -1; /* Probably a %+ without a space */
3716 skip_white_(t);
3717 if (!t)
3718 return -1;
3719 if (t->type != TOK_ID)
3720 return -1;
3721 tt = t->next;
3722 skip_white_(tt);
3723 if (tt && (tt->type != TOK_OTHER || strcmp(tt->text, ",")))
3724 return -1;
3726 return bsii(t->text, (const char **)conditions, ARRAY_SIZE(conditions));
3730 * This routines walks over tokens strem and hadnles tokens
3731 * pasting, if @handle_explicit passed then explicit pasting
3732 * term is handled, otherwise -- implicit pastings only.
3734 static bool paste_tokens(Token **head, const struct tokseq_match *m,
3735 size_t mnum, bool handle_explicit)
3737 Token *tok, *next, **prev_next, **prev_nonspace;
3738 bool pasted = false;
3739 char *buf, *p;
3740 size_t len, i;
3743 * The last token before pasting. We need it
3744 * to be able to connect new handled tokens.
3745 * In other words if there were a tokens stream
3747 * A -> B -> C -> D
3749 * and we've joined tokens B and C, the resulting
3750 * stream should be
3752 * A -> BC -> D
3754 tok = *head;
3755 prev_next = NULL;
3757 if (!tok_type_(tok, TOK_WHITESPACE) && !tok_type_(tok, TOK_PASTE))
3758 prev_nonspace = head;
3759 else
3760 prev_nonspace = NULL;
3762 while (tok && (next = tok->next)) {
3764 switch (tok->type) {
3765 case TOK_WHITESPACE:
3766 /* Zap redundant whitespaces */
3767 while (tok_type_(next, TOK_WHITESPACE))
3768 next = delete_Token(next);
3769 tok->next = next;
3770 break;
3772 case TOK_PASTE:
3773 /* Explicit pasting */
3774 if (!handle_explicit)
3775 break;
3776 next = delete_Token(tok);
3778 while (tok_type_(next, TOK_WHITESPACE))
3779 next = delete_Token(next);
3781 if (!pasted)
3782 pasted = true;
3784 /* Left pasting token is start of line */
3785 if (!prev_nonspace)
3786 nasm_error(ERR_FATAL, "No lvalue found on pasting");
3789 * No ending token, this might happen in two
3790 * cases
3792 * 1) There indeed no right token at all
3793 * 2) There is a bare "%define ID" statement,
3794 * and @ID does expand to whitespace.
3796 * So technically we need to do a grammar analysis
3797 * in another stage of parsing, but for now lets don't
3798 * change the behaviour people used to. Simply allow
3799 * whitespace after paste token.
3801 if (!next) {
3803 * Zap ending space tokens and that's all.
3805 tok = (*prev_nonspace)->next;
3806 while (tok_type_(tok, TOK_WHITESPACE))
3807 tok = delete_Token(tok);
3808 tok = *prev_nonspace;
3809 tok->next = NULL;
3810 break;
3813 tok = *prev_nonspace;
3814 while (tok_type_(tok, TOK_WHITESPACE))
3815 tok = delete_Token(tok);
3816 len = strlen(tok->text);
3817 len += strlen(next->text);
3819 p = buf = nasm_malloc(len + 1);
3820 strcpy(p, tok->text);
3821 p = strchr(p, '\0');
3822 strcpy(p, next->text);
3824 delete_Token(tok);
3826 tok = tokenize(buf);
3827 nasm_free(buf);
3829 *prev_nonspace = tok;
3830 while (tok && tok->next)
3831 tok = tok->next;
3833 tok->next = delete_Token(next);
3835 /* Restart from pasted tokens head */
3836 tok = *prev_nonspace;
3837 break;
3839 default:
3840 /* implicit pasting */
3841 for (i = 0; i < mnum; i++) {
3842 if (!(PP_CONCAT_MATCH(tok, m[i].mask_head)))
3843 continue;
3845 len = 0;
3846 while (next && PP_CONCAT_MATCH(next, m[i].mask_tail)) {
3847 len += strlen(next->text);
3848 next = next->next;
3851 /* No match or no text to process */
3852 if (tok == next || len == 0)
3853 break;
3855 len += strlen(tok->text);
3856 p = buf = nasm_malloc(len + 1);
3858 strcpy(p, tok->text);
3859 p = strchr(p, '\0');
3860 tok = delete_Token(tok);
3862 while (tok != next) {
3863 if (PP_CONCAT_MATCH(tok, m[i].mask_tail)) {
3864 strcpy(p, tok->text);
3865 p = strchr(p, '\0');
3867 tok = delete_Token(tok);
3870 tok = tokenize(buf);
3871 nasm_free(buf);
3873 if (prev_next)
3874 *prev_next = tok;
3875 else
3876 *head = tok;
3879 * Connect pasted into original stream,
3880 * ie A -> new-tokens -> B
3882 while (tok && tok->next)
3883 tok = tok->next;
3884 tok->next = next;
3886 if (!pasted)
3887 pasted = true;
3889 /* Restart from pasted tokens head */
3890 tok = prev_next ? *prev_next : *head;
3893 break;
3896 prev_next = &tok->next;
3898 if (tok->next &&
3899 !tok_type_(tok->next, TOK_WHITESPACE) &&
3900 !tok_type_(tok->next, TOK_PASTE))
3901 prev_nonspace = prev_next;
3903 tok = tok->next;
3906 return pasted;
3910 * expands to a list of tokens from %{x:y}
3912 static Token *expand_mmac_params_range(MMacro *mac, Token *tline, Token ***last)
3914 Token *t = tline, **tt, *tm, *head;
3915 char *pos;
3916 int fst, lst, j, i;
3918 pos = strchr(tline->text, ':');
3919 nasm_assert(pos);
3921 lst = atoi(pos + 1);
3922 fst = atoi(tline->text + 1);
3925 * only macros params are accounted so
3926 * if someone passes %0 -- we reject such
3927 * value(s)
3929 if (lst == 0 || fst == 0)
3930 goto err;
3932 /* the values should be sane */
3933 if ((fst > (int)mac->nparam || fst < (-(int)mac->nparam)) ||
3934 (lst > (int)mac->nparam || lst < (-(int)mac->nparam)))
3935 goto err;
3937 fst = fst < 0 ? fst + (int)mac->nparam + 1: fst;
3938 lst = lst < 0 ? lst + (int)mac->nparam + 1: lst;
3940 /* counted from zero */
3941 fst--, lst--;
3944 * It will be at least one token. Note we
3945 * need to scan params until separator, otherwise
3946 * only first token will be passed.
3948 tm = mac->params[(fst + mac->rotate) % mac->nparam];
3949 if (!tm)
3950 goto err;
3951 head = new_Token(NULL, tm->type, tm->text, 0);
3952 tt = &head->next, tm = tm->next;
3953 while (tok_isnt_(tm, ",")) {
3954 t = new_Token(NULL, tm->type, tm->text, 0);
3955 *tt = t, tt = &t->next, tm = tm->next;
3958 if (fst < lst) {
3959 for (i = fst + 1; i <= lst; i++) {
3960 t = new_Token(NULL, TOK_OTHER, ",", 0);
3961 *tt = t, tt = &t->next;
3962 j = (i + mac->rotate) % mac->nparam;
3963 tm = mac->params[j];
3964 while (tok_isnt_(tm, ",")) {
3965 t = new_Token(NULL, tm->type, tm->text, 0);
3966 *tt = t, tt = &t->next, tm = tm->next;
3969 } else {
3970 for (i = fst - 1; i >= lst; i--) {
3971 t = new_Token(NULL, TOK_OTHER, ",", 0);
3972 *tt = t, tt = &t->next;
3973 j = (i + mac->rotate) % mac->nparam;
3974 tm = mac->params[j];
3975 while (tok_isnt_(tm, ",")) {
3976 t = new_Token(NULL, tm->type, tm->text, 0);
3977 *tt = t, tt = &t->next, tm = tm->next;
3982 *last = tt;
3983 return head;
3985 err:
3986 nasm_error(ERR_NONFATAL, "`%%{%s}': macro parameters out of range",
3987 &tline->text[1]);
3988 return tline;
3992 * Expand MMacro-local things: parameter references (%0, %n, %+n,
3993 * %-n) and MMacro-local identifiers (%%foo) as well as
3994 * macro indirection (%[...]) and range (%{..:..}).
3996 static Token *expand_mmac_params(Token * tline)
3998 Token *t, *tt, **tail, *thead;
3999 bool changed = false;
4000 char *pos;
4002 tail = &thead;
4003 thead = NULL;
4005 while (tline) {
4006 if (tline->type == TOK_PREPROC_ID && tline->text && tline->text[0] &&
4007 (((tline->text[1] == '+' || tline->text[1] == '-') && tline->text[2]) ||
4008 (tline->text[1] >= '0' && tline->text[1] <= '9') ||
4009 tline->text[1] == '%')) {
4010 char *text = NULL;
4011 int type = 0, cc; /* type = 0 to placate optimisers */
4012 char tmpbuf[30];
4013 unsigned int n;
4014 int i;
4015 MMacro *mac;
4017 t = tline;
4018 tline = tline->next;
4020 mac = istk->mstk;
4021 while (mac && !mac->name) /* avoid mistaking %reps for macros */
4022 mac = mac->next_active;
4023 if (!mac) {
4024 nasm_error(ERR_NONFATAL, "`%s': not in a macro call", t->text);
4025 } else {
4026 pos = strchr(t->text, ':');
4027 if (!pos) {
4028 switch (t->text[1]) {
4030 * We have to make a substitution of one of the
4031 * forms %1, %-1, %+1, %%foo, %0.
4033 case '0':
4034 type = TOK_NUMBER;
4035 snprintf(tmpbuf, sizeof(tmpbuf), "%d", mac->nparam);
4036 text = nasm_strdup(tmpbuf);
4037 break;
4038 case '%':
4039 type = TOK_ID;
4040 snprintf(tmpbuf, sizeof(tmpbuf), "..@%"PRIu64".",
4041 mac->unique);
4042 text = nasm_strcat(tmpbuf, t->text + 2);
4043 break;
4044 case '-':
4045 n = atoi(t->text + 2) - 1;
4046 if (n >= mac->nparam)
4047 tt = NULL;
4048 else {
4049 if (mac->nparam > 1)
4050 n = (n + mac->rotate) % mac->nparam;
4051 tt = mac->params[n];
4053 cc = find_cc(tt);
4054 if (cc == -1) {
4055 nasm_error(ERR_NONFATAL,
4056 "macro parameter %d is not a condition code",
4057 n + 1);
4058 text = NULL;
4059 } else {
4060 type = TOK_ID;
4061 if (inverse_ccs[cc] == -1) {
4062 nasm_error(ERR_NONFATAL,
4063 "condition code `%s' is not invertible",
4064 conditions[cc]);
4065 text = NULL;
4066 } else
4067 text = nasm_strdup(conditions[inverse_ccs[cc]]);
4069 break;
4070 case '+':
4071 n = atoi(t->text + 2) - 1;
4072 if (n >= mac->nparam)
4073 tt = NULL;
4074 else {
4075 if (mac->nparam > 1)
4076 n = (n + mac->rotate) % mac->nparam;
4077 tt = mac->params[n];
4079 cc = find_cc(tt);
4080 if (cc == -1) {
4081 nasm_error(ERR_NONFATAL,
4082 "macro parameter %d is not a condition code",
4083 n + 1);
4084 text = NULL;
4085 } else {
4086 type = TOK_ID;
4087 text = nasm_strdup(conditions[cc]);
4089 break;
4090 default:
4091 n = atoi(t->text + 1) - 1;
4092 if (n >= mac->nparam)
4093 tt = NULL;
4094 else {
4095 if (mac->nparam > 1)
4096 n = (n + mac->rotate) % mac->nparam;
4097 tt = mac->params[n];
4099 if (tt) {
4100 for (i = 0; i < mac->paramlen[n]; i++) {
4101 *tail = new_Token(NULL, tt->type, tt->text, 0);
4102 tail = &(*tail)->next;
4103 tt = tt->next;
4106 text = NULL; /* we've done it here */
4107 break;
4109 } else {
4111 * seems we have a parameters range here
4113 Token *head, **last;
4114 head = expand_mmac_params_range(mac, t, &last);
4115 if (head != t) {
4116 *tail = head;
4117 *last = tline;
4118 tline = head;
4119 text = NULL;
4123 if (!text) {
4124 delete_Token(t);
4125 } else {
4126 *tail = t;
4127 tail = &t->next;
4128 t->type = type;
4129 nasm_free(t->text);
4130 t->text = text;
4131 t->a.mac = NULL;
4133 changed = true;
4134 continue;
4135 } else if (tline->type == TOK_INDIRECT) {
4136 t = tline;
4137 tline = tline->next;
4138 tt = tokenize(t->text);
4139 tt = expand_mmac_params(tt);
4140 tt = expand_smacro(tt);
4141 *tail = tt;
4142 while (tt) {
4143 tt->a.mac = NULL; /* Necessary? */
4144 tail = &tt->next;
4145 tt = tt->next;
4147 delete_Token(t);
4148 changed = true;
4149 } else {
4150 t = *tail = tline;
4151 tline = tline->next;
4152 t->a.mac = NULL;
4153 tail = &t->next;
4156 *tail = NULL;
4158 if (changed) {
4159 const struct tokseq_match t[] = {
4161 PP_CONCAT_MASK(TOK_ID) |
4162 PP_CONCAT_MASK(TOK_FLOAT), /* head */
4163 PP_CONCAT_MASK(TOK_ID) |
4164 PP_CONCAT_MASK(TOK_NUMBER) |
4165 PP_CONCAT_MASK(TOK_FLOAT) |
4166 PP_CONCAT_MASK(TOK_OTHER) /* tail */
4169 PP_CONCAT_MASK(TOK_NUMBER), /* head */
4170 PP_CONCAT_MASK(TOK_NUMBER) /* tail */
4173 paste_tokens(&thead, t, ARRAY_SIZE(t), false);
4176 return thead;
4180 * Expand all single-line macro calls made in the given line.
4181 * Return the expanded version of the line. The original is deemed
4182 * to be destroyed in the process. (In reality we'll just move
4183 * Tokens from input to output a lot of the time, rather than
4184 * actually bothering to destroy and replicate.)
4187 static Token *expand_smacro(Token * tline)
4189 Token *t, *tt, *mstart, **tail, *thead;
4190 SMacro *head = NULL, *m;
4191 Token **params;
4192 int *paramsize;
4193 unsigned int nparam, sparam;
4194 int brackets;
4195 Token *org_tline = tline;
4196 Context *ctx;
4197 const char *mname;
4198 int64_t deadman = nasm_limit[LIMIT_MACROS];
4199 bool expanded;
4202 * Trick: we should avoid changing the start token pointer since it can
4203 * be contained in "next" field of other token. Because of this
4204 * we allocate a copy of first token and work with it; at the end of
4205 * routine we copy it back
4207 if (org_tline) {
4208 tline = new_Token(org_tline->next, org_tline->type,
4209 org_tline->text, 0);
4210 tline->a.mac = org_tline->a.mac;
4211 nasm_free(org_tline->text);
4212 org_tline->text = NULL;
4215 expanded = true; /* Always expand %+ at least once */
4217 again:
4218 thead = NULL;
4219 tail = &thead;
4221 while (tline) { /* main token loop */
4222 if (!--deadman) {
4223 nasm_error(ERR_NONFATAL, "interminable macro recursion");
4224 goto err;
4227 if ((mname = tline->text)) {
4228 /* if this token is a local macro, look in local context */
4229 if (tline->type == TOK_ID) {
4230 head = (SMacro *)hash_findix(&smacros, mname);
4231 } else if (tline->type == TOK_PREPROC_ID) {
4232 ctx = get_ctx(mname, &mname);
4233 head = ctx ? (SMacro *)hash_findix(&ctx->localmac, mname) : NULL;
4234 } else
4235 head = NULL;
4238 * We've hit an identifier. As in is_mmacro below, we first
4239 * check whether the identifier is a single-line macro at
4240 * all, then think about checking for parameters if
4241 * necessary.
4243 list_for_each(m, head)
4244 if (!mstrcmp(m->name, mname, m->casesense))
4245 break;
4246 if (m) {
4247 mstart = tline;
4248 params = NULL;
4249 paramsize = NULL;
4250 if (m->nparam == 0) {
4252 * Simple case: the macro is parameterless. Discard the
4253 * one token that the macro call took, and push the
4254 * expansion back on the to-do stack.
4256 if (!m->expansion) {
4257 if (!strcmp("__FILE__", m->name)) {
4258 const char *file = src_get_fname();
4259 /* nasm_free(tline->text); here? */
4260 tline->text = nasm_quote(file, strlen(file));
4261 tline->type = TOK_STRING;
4262 continue;
4264 if (!strcmp("__LINE__", m->name)) {
4265 nasm_free(tline->text);
4266 make_tok_num(tline, src_get_linnum());
4267 continue;
4269 if (!strcmp("__BITS__", m->name)) {
4270 nasm_free(tline->text);
4271 make_tok_num(tline, globalbits);
4272 continue;
4274 tline = delete_Token(tline);
4275 continue;
4277 } else {
4279 * Complicated case: at least one macro with this name
4280 * exists and takes parameters. We must find the
4281 * parameters in the call, count them, find the SMacro
4282 * that corresponds to that form of the macro call, and
4283 * substitute for the parameters when we expand. What a
4284 * pain.
4286 /*tline = tline->next;
4287 skip_white_(tline); */
4288 do {
4289 t = tline->next;
4290 while (tok_type_(t, TOK_SMAC_END)) {
4291 t->a.mac->in_progress = false;
4292 t->text = NULL;
4293 t = tline->next = delete_Token(t);
4295 tline = t;
4296 } while (tok_type_(tline, TOK_WHITESPACE));
4297 if (!tok_is_(tline, "(")) {
4299 * This macro wasn't called with parameters: ignore
4300 * the call. (Behaviour borrowed from gnu cpp.)
4302 tline = mstart;
4303 m = NULL;
4304 } else {
4305 int paren = 0;
4306 int white = 0;
4307 brackets = 0;
4308 nparam = 0;
4309 sparam = PARAM_DELTA;
4310 params = nasm_malloc(sparam * sizeof(Token *));
4311 params[0] = tline->next;
4312 paramsize = nasm_malloc(sparam * sizeof(int));
4313 paramsize[0] = 0;
4314 while (true) { /* parameter loop */
4316 * For some unusual expansions
4317 * which concatenates function call
4319 t = tline->next;
4320 while (tok_type_(t, TOK_SMAC_END)) {
4321 t->a.mac->in_progress = false;
4322 t->text = NULL;
4323 t = tline->next = delete_Token(t);
4325 tline = t;
4327 if (!tline) {
4328 nasm_error(ERR_NONFATAL,
4329 "macro call expects terminating `)'");
4330 break;
4332 if (tline->type == TOK_WHITESPACE
4333 && brackets <= 0) {
4334 if (paramsize[nparam])
4335 white++;
4336 else
4337 params[nparam] = tline->next;
4338 continue; /* parameter loop */
4340 if (tline->type == TOK_OTHER
4341 && tline->text[1] == 0) {
4342 char ch = tline->text[0];
4343 if (ch == ',' && !paren && brackets <= 0) {
4344 if (++nparam >= sparam) {
4345 sparam += PARAM_DELTA;
4346 params = nasm_realloc(params,
4347 sparam * sizeof(Token *));
4348 paramsize = nasm_realloc(paramsize,
4349 sparam * sizeof(int));
4351 params[nparam] = tline->next;
4352 paramsize[nparam] = 0;
4353 white = 0;
4354 continue; /* parameter loop */
4356 if (ch == '{' &&
4357 (brackets > 0 || (brackets == 0 &&
4358 !paramsize[nparam])))
4360 if (!(brackets++)) {
4361 params[nparam] = tline->next;
4362 continue; /* parameter loop */
4365 if (ch == '}' && brackets > 0)
4366 if (--brackets == 0) {
4367 brackets = -1;
4368 continue; /* parameter loop */
4370 if (ch == '(' && !brackets)
4371 paren++;
4372 if (ch == ')' && brackets <= 0)
4373 if (--paren < 0)
4374 break;
4376 if (brackets < 0) {
4377 brackets = 0;
4378 nasm_error(ERR_NONFATAL, "braces do not "
4379 "enclose all of macro parameter");
4381 paramsize[nparam] += white + 1;
4382 white = 0;
4383 } /* parameter loop */
4384 nparam++;
4385 while (m && (m->nparam != nparam ||
4386 mstrcmp(m->name, mname,
4387 m->casesense)))
4388 m = m->next;
4389 if (!m)
4390 nasm_error(ERR_WARNING|ERR_PASS1|ERR_WARN_MNP,
4391 "macro `%s' exists, "
4392 "but not taking %d parameters",
4393 mstart->text, nparam);
4396 if (m && m->in_progress)
4397 m = NULL;
4398 if (!m) { /* in progess or didn't find '(' or wrong nparam */
4400 * Design question: should we handle !tline, which
4401 * indicates missing ')' here, or expand those
4402 * macros anyway, which requires the (t) test a few
4403 * lines down?
4405 nasm_free(params);
4406 nasm_free(paramsize);
4407 tline = mstart;
4408 } else {
4410 * Expand the macro: we are placed on the last token of the
4411 * call, so that we can easily split the call from the
4412 * following tokens. We also start by pushing an SMAC_END
4413 * token for the cycle removal.
4415 t = tline;
4416 if (t) {
4417 tline = t->next;
4418 t->next = NULL;
4420 tt = new_Token(tline, TOK_SMAC_END, NULL, 0);
4421 tt->a.mac = m;
4422 m->in_progress = true;
4423 tline = tt;
4424 list_for_each(t, m->expansion) {
4425 if (t->type >= TOK_SMAC_PARAM) {
4426 Token *pcopy = tline, **ptail = &pcopy;
4427 Token *ttt, *pt;
4428 int i;
4430 ttt = params[t->type - TOK_SMAC_PARAM];
4431 i = paramsize[t->type - TOK_SMAC_PARAM];
4432 while (--i >= 0) {
4433 pt = *ptail = new_Token(tline, ttt->type,
4434 ttt->text, 0);
4435 ptail = &pt->next;
4436 ttt = ttt->next;
4437 if (!ttt && i > 0) {
4439 * FIXME: Need to handle more gracefully,
4440 * exiting early on agruments analysis.
4442 nasm_error(ERR_FATAL,
4443 "macro `%s' expects %d args",
4444 mstart->text,
4445 (int)paramsize[t->type - TOK_SMAC_PARAM]);
4448 tline = pcopy;
4449 } else if (t->type == TOK_PREPROC_Q) {
4450 tt = new_Token(tline, TOK_ID, mname, 0);
4451 tline = tt;
4452 } else if (t->type == TOK_PREPROC_QQ) {
4453 tt = new_Token(tline, TOK_ID, m->name, 0);
4454 tline = tt;
4455 } else {
4456 tt = new_Token(tline, t->type, t->text, 0);
4457 tline = tt;
4462 * Having done that, get rid of the macro call, and clean
4463 * up the parameters.
4465 nasm_free(params);
4466 nasm_free(paramsize);
4467 free_tlist(mstart);
4468 expanded = true;
4469 continue; /* main token loop */
4474 if (tline->type == TOK_SMAC_END) {
4475 /* On error path it might already be dropped */
4476 if (tline->a.mac)
4477 tline->a.mac->in_progress = false;
4478 tline = delete_Token(tline);
4479 } else {
4480 t = *tail = tline;
4481 tline = tline->next;
4482 t->a.mac = NULL;
4483 t->next = NULL;
4484 tail = &t->next;
4489 * Now scan the entire line and look for successive TOK_IDs that resulted
4490 * after expansion (they can't be produced by tokenize()). The successive
4491 * TOK_IDs should be concatenated.
4492 * Also we look for %+ tokens and concatenate the tokens before and after
4493 * them (without white spaces in between).
4495 if (expanded) {
4496 const struct tokseq_match t[] = {
4498 PP_CONCAT_MASK(TOK_ID) |
4499 PP_CONCAT_MASK(TOK_PREPROC_ID), /* head */
4500 PP_CONCAT_MASK(TOK_ID) |
4501 PP_CONCAT_MASK(TOK_PREPROC_ID) |
4502 PP_CONCAT_MASK(TOK_NUMBER) /* tail */
4505 if (paste_tokens(&thead, t, ARRAY_SIZE(t), true)) {
4507 * If we concatenated something, *and* we had previously expanded
4508 * an actual macro, scan the lines again for macros...
4510 tline = thead;
4511 expanded = false;
4512 goto again;
4516 err:
4517 if (org_tline) {
4518 if (thead) {
4519 *org_tline = *thead;
4520 /* since we just gave text to org_line, don't free it */
4521 thead->text = NULL;
4522 delete_Token(thead);
4523 } else {
4524 /* the expression expanded to empty line;
4525 we can't return NULL for some reasons
4526 we just set the line to a single WHITESPACE token. */
4527 memset(org_tline, 0, sizeof(*org_tline));
4528 org_tline->text = NULL;
4529 org_tline->type = TOK_WHITESPACE;
4531 thead = org_tline;
4534 return thead;
4538 * Similar to expand_smacro but used exclusively with macro identifiers
4539 * right before they are fetched in. The reason is that there can be
4540 * identifiers consisting of several subparts. We consider that if there
4541 * are more than one element forming the name, user wants a expansion,
4542 * otherwise it will be left as-is. Example:
4544 * %define %$abc cde
4546 * the identifier %$abc will be left as-is so that the handler for %define
4547 * will suck it and define the corresponding value. Other case:
4549 * %define _%$abc cde
4551 * In this case user wants name to be expanded *before* %define starts
4552 * working, so we'll expand %$abc into something (if it has a value;
4553 * otherwise it will be left as-is) then concatenate all successive
4554 * PP_IDs into one.
4556 static Token *expand_id(Token * tline)
4558 Token *cur, *oldnext = NULL;
4560 if (!tline || !tline->next)
4561 return tline;
4563 cur = tline;
4564 while (cur->next &&
4565 (cur->next->type == TOK_ID ||
4566 cur->next->type == TOK_PREPROC_ID
4567 || cur->next->type == TOK_NUMBER))
4568 cur = cur->next;
4570 /* If identifier consists of just one token, don't expand */
4571 if (cur == tline)
4572 return tline;
4574 if (cur) {
4575 oldnext = cur->next; /* Detach the tail past identifier */
4576 cur->next = NULL; /* so that expand_smacro stops here */
4579 tline = expand_smacro(tline);
4581 if (cur) {
4582 /* expand_smacro possibly changhed tline; re-scan for EOL */
4583 cur = tline;
4584 while (cur && cur->next)
4585 cur = cur->next;
4586 if (cur)
4587 cur->next = oldnext;
4590 return tline;
4594 * Determine whether the given line constitutes a multi-line macro
4595 * call, and return the MMacro structure called if so. Doesn't have
4596 * to check for an initial label - that's taken care of in
4597 * expand_mmacro - but must check numbers of parameters. Guaranteed
4598 * to be called with tline->type == TOK_ID, so the putative macro
4599 * name is easy to find.
4601 static MMacro *is_mmacro(Token * tline, Token *** params_array)
4603 MMacro *head, *m;
4604 Token **params;
4605 int nparam;
4607 head = (MMacro *) hash_findix(&mmacros, tline->text);
4610 * Efficiency: first we see if any macro exists with the given
4611 * name. If not, we can return NULL immediately. _Then_ we
4612 * count the parameters, and then we look further along the
4613 * list if necessary to find the proper MMacro.
4615 list_for_each(m, head)
4616 if (!mstrcmp(m->name, tline->text, m->casesense))
4617 break;
4618 if (!m)
4619 return NULL;
4622 * OK, we have a potential macro. Count and demarcate the
4623 * parameters.
4625 count_mmac_params(tline->next, &nparam, &params);
4628 * So we know how many parameters we've got. Find the MMacro
4629 * structure that handles this number.
4631 while (m) {
4632 if (m->nparam_min <= nparam
4633 && (m->plus || nparam <= m->nparam_max)) {
4635 * This one is right. Just check if cycle removal
4636 * prohibits us using it before we actually celebrate...
4638 if (m->in_progress > m->max_depth) {
4639 if (m->max_depth > 0) {
4640 nasm_error(ERR_WARNING,
4641 "reached maximum recursion depth of %i",
4642 m->max_depth);
4644 nasm_free(params);
4645 return NULL;
4648 * It's right, and we can use it. Add its default
4649 * parameters to the end of our list if necessary.
4651 if (m->defaults && nparam < m->nparam_min + m->ndefs) {
4652 params =
4653 nasm_realloc(params,
4654 ((m->nparam_min + m->ndefs +
4655 1) * sizeof(*params)));
4656 while (nparam < m->nparam_min + m->ndefs) {
4657 params[nparam] = m->defaults[nparam - m->nparam_min];
4658 nparam++;
4662 * If we've gone over the maximum parameter count (and
4663 * we're in Plus mode), ignore parameters beyond
4664 * nparam_max.
4666 if (m->plus && nparam > m->nparam_max)
4667 nparam = m->nparam_max;
4669 * Then terminate the parameter list, and leave.
4671 if (!params) { /* need this special case */
4672 params = nasm_malloc(sizeof(*params));
4673 nparam = 0;
4675 params[nparam] = NULL;
4676 *params_array = params;
4677 return m;
4680 * This one wasn't right: look for the next one with the
4681 * same name.
4683 list_for_each(m, m->next)
4684 if (!mstrcmp(m->name, tline->text, m->casesense))
4685 break;
4689 * After all that, we didn't find one with the right number of
4690 * parameters. Issue a warning, and fail to expand the macro.
4692 nasm_error(ERR_WARNING|ERR_PASS1|ERR_WARN_MNP,
4693 "macro `%s' exists, but not taking %d parameters",
4694 tline->text, nparam);
4695 nasm_free(params);
4696 return NULL;
4701 * Save MMacro invocation specific fields in
4702 * preparation for a recursive macro expansion
4704 static void push_mmacro(MMacro *m)
4706 MMacroInvocation *i;
4708 i = nasm_malloc(sizeof(MMacroInvocation));
4709 i->prev = m->prev;
4710 i->params = m->params;
4711 i->iline = m->iline;
4712 i->nparam = m->nparam;
4713 i->rotate = m->rotate;
4714 i->paramlen = m->paramlen;
4715 i->unique = m->unique;
4716 i->condcnt = m->condcnt;
4717 m->prev = i;
4722 * Restore MMacro invocation specific fields that were
4723 * saved during a previous recursive macro expansion
4725 static void pop_mmacro(MMacro *m)
4727 MMacroInvocation *i;
4729 if (m->prev) {
4730 i = m->prev;
4731 m->prev = i->prev;
4732 m->params = i->params;
4733 m->iline = i->iline;
4734 m->nparam = i->nparam;
4735 m->rotate = i->rotate;
4736 m->paramlen = i->paramlen;
4737 m->unique = i->unique;
4738 m->condcnt = i->condcnt;
4739 nasm_free(i);
4745 * Expand the multi-line macro call made by the given line, if
4746 * there is one to be expanded. If there is, push the expansion on
4747 * istk->expansion and return 1. Otherwise return 0.
4749 static int expand_mmacro(Token * tline)
4751 Token *startline = tline;
4752 Token *label = NULL;
4753 int dont_prepend = 0;
4754 Token **params, *t, *tt;
4755 MMacro *m;
4756 Line *l, *ll;
4757 int i, nparam, *paramlen;
4758 const char *mname;
4760 t = tline;
4761 skip_white_(t);
4762 /* if (!tok_type_(t, TOK_ID)) Lino 02/25/02 */
4763 if (!tok_type_(t, TOK_ID) && !tok_type_(t, TOK_PREPROC_ID))
4764 return 0;
4765 m = is_mmacro(t, &params);
4766 if (m) {
4767 mname = t->text;
4768 } else {
4769 Token *last;
4771 * We have an id which isn't a macro call. We'll assume
4772 * it might be a label; we'll also check to see if a
4773 * colon follows it. Then, if there's another id after
4774 * that lot, we'll check it again for macro-hood.
4776 label = last = t;
4777 t = t->next;
4778 if (tok_type_(t, TOK_WHITESPACE))
4779 last = t, t = t->next;
4780 if (tok_is_(t, ":")) {
4781 dont_prepend = 1;
4782 last = t, t = t->next;
4783 if (tok_type_(t, TOK_WHITESPACE))
4784 last = t, t = t->next;
4786 if (!tok_type_(t, TOK_ID) || !(m = is_mmacro(t, &params)))
4787 return 0;
4788 last->next = NULL;
4789 mname = t->text;
4790 tline = t;
4794 * Fix up the parameters: this involves stripping leading and
4795 * trailing whitespace, then stripping braces if they are
4796 * present.
4798 for (nparam = 0; params[nparam]; nparam++) ;
4799 paramlen = nparam ? nasm_malloc(nparam * sizeof(*paramlen)) : NULL;
4801 for (i = 0; params[i]; i++) {
4802 int brace = 0;
4803 int comma = (!m->plus || i < nparam - 1);
4805 t = params[i];
4806 skip_white_(t);
4807 if (tok_is_(t, "{"))
4808 t = t->next, brace++, comma = false;
4809 params[i] = t;
4810 paramlen[i] = 0;
4811 while (t) {
4812 if (comma && t->type == TOK_OTHER && !strcmp(t->text, ","))
4813 break; /* ... because we have hit a comma */
4814 if (comma && t->type == TOK_WHITESPACE
4815 && tok_is_(t->next, ","))
4816 break; /* ... or a space then a comma */
4817 if (brace && t->type == TOK_OTHER) {
4818 if (t->text[0] == '{')
4819 brace++; /* ... or a nested opening brace */
4820 else if (t->text[0] == '}')
4821 if (!--brace)
4822 break; /* ... or a brace */
4824 t = t->next;
4825 paramlen[i]++;
4827 if (brace)
4828 nasm_error(ERR_NONFATAL, "macro params should be enclosed in braces");
4832 * OK, we have a MMacro structure together with a set of
4833 * parameters. We must now go through the expansion and push
4834 * copies of each Line on to istk->expansion. Substitution of
4835 * parameter tokens and macro-local tokens doesn't get done
4836 * until the single-line macro substitution process; this is
4837 * because delaying them allows us to change the semantics
4838 * later through %rotate.
4840 * First, push an end marker on to istk->expansion, mark this
4841 * macro as in progress, and set up its invocation-specific
4842 * variables.
4844 ll = nasm_malloc(sizeof(Line));
4845 ll->next = istk->expansion;
4846 ll->finishes = m;
4847 ll->first = NULL;
4848 istk->expansion = ll;
4851 * Save the previous MMacro expansion in the case of
4852 * macro recursion
4854 if (m->max_depth && m->in_progress)
4855 push_mmacro(m);
4857 m->in_progress ++;
4858 m->params = params;
4859 m->iline = tline;
4860 m->nparam = nparam;
4861 m->rotate = 0;
4862 m->paramlen = paramlen;
4863 m->unique = unique++;
4864 m->lineno = 0;
4865 m->condcnt = 0;
4867 m->next_active = istk->mstk;
4868 istk->mstk = m;
4870 list_for_each(l, m->expansion) {
4871 Token **tail;
4873 ll = nasm_malloc(sizeof(Line));
4874 ll->finishes = NULL;
4875 ll->next = istk->expansion;
4876 istk->expansion = ll;
4877 tail = &ll->first;
4879 list_for_each(t, l->first) {
4880 Token *x = t;
4881 switch (t->type) {
4882 case TOK_PREPROC_Q:
4883 tt = *tail = new_Token(NULL, TOK_ID, mname, 0);
4884 break;
4885 case TOK_PREPROC_QQ:
4886 tt = *tail = new_Token(NULL, TOK_ID, m->name, 0);
4887 break;
4888 case TOK_PREPROC_ID:
4889 if (t->text[1] == '0' && t->text[2] == '0') {
4890 dont_prepend = -1;
4891 x = label;
4892 if (!x)
4893 continue;
4895 /* fall through */
4896 default:
4897 tt = *tail = new_Token(NULL, x->type, x->text, 0);
4898 break;
4900 tail = &tt->next;
4902 *tail = NULL;
4906 * If we had a label, push it on as the first line of
4907 * the macro expansion.
4909 if (label) {
4910 if (dont_prepend < 0)
4911 free_tlist(startline);
4912 else {
4913 ll = nasm_malloc(sizeof(Line));
4914 ll->finishes = NULL;
4915 ll->next = istk->expansion;
4916 istk->expansion = ll;
4917 ll->first = startline;
4918 if (!dont_prepend) {
4919 while (label->next)
4920 label = label->next;
4921 label->next = tt = new_Token(NULL, TOK_OTHER, ":", 0);
4926 lfmt->uplevel(m->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
4928 return 1;
4932 * This function adds macro names to error messages, and suppresses
4933 * them if necessary.
4935 static void pp_verror(int severity, const char *fmt, va_list arg)
4937 char buff[BUFSIZ];
4938 MMacro *mmac = NULL;
4939 int delta = 0;
4942 * If we're in a dead branch of IF or something like it, ignore the error.
4943 * However, because %else etc are evaluated in the state context
4944 * of the previous branch, errors might get lost:
4945 * %if 0 ... %else trailing garbage ... %endif
4946 * So %else etc should set the ERR_PP_PRECOND flag.
4948 if ((severity & ERR_MASK) < ERR_FATAL &&
4949 istk && istk->conds &&
4950 ((severity & ERR_PP_PRECOND) ?
4951 istk->conds->state == COND_NEVER :
4952 !emitting(istk->conds->state)))
4953 return;
4955 /* get %macro name */
4956 if (!(severity & ERR_NOFILE) && istk && istk->mstk) {
4957 mmac = istk->mstk;
4958 /* but %rep blocks should be skipped */
4959 while (mmac && !mmac->name)
4960 mmac = mmac->next_active, delta++;
4963 if (mmac) {
4964 vsnprintf(buff, sizeof(buff), fmt, arg);
4966 nasm_set_verror(real_verror);
4967 nasm_error(severity, "(%s:%d) %s",
4968 mmac->name, mmac->lineno - delta, buff);
4969 nasm_set_verror(pp_verror);
4970 } else {
4971 real_verror(severity, fmt, arg);
4975 static void
4976 pp_reset(const char *file, int apass, StrList **deplist)
4978 Token *t;
4980 cstk = NULL;
4981 istk = nasm_malloc(sizeof(Include));
4982 istk->next = NULL;
4983 istk->conds = NULL;
4984 istk->expansion = NULL;
4985 istk->mstk = NULL;
4986 istk->fp = nasm_open_read(file, NF_TEXT);
4987 istk->fname = NULL;
4988 src_set(0, file);
4989 istk->lineinc = 1;
4990 if (!istk->fp)
4991 nasm_fatal(ERR_NOFILE, "unable to open input file `%s'", file);
4992 defining = NULL;
4993 nested_mac_count = 0;
4994 nested_rep_count = 0;
4995 init_macros();
4996 unique = 0;
4998 if (tasm_compatible_mode)
4999 pp_add_stdmac(nasm_stdmac_tasm);
5001 pp_add_stdmac(nasm_stdmac_nasm);
5002 pp_add_stdmac(nasm_stdmac_version);
5004 if (extrastdmac)
5005 pp_add_stdmac(extrastdmac);
5007 stdmacpos = stdmacros[0];
5008 stdmacnext = &stdmacros[1];
5010 do_predef = true;
5013 * 0 for dependencies, 1 for preparatory passes, 2 for final pass.
5014 * The caller, however, will also pass in 3 for preprocess-only so
5015 * we can set __PASS__ accordingly.
5017 pass = apass > 2 ? 2 : apass;
5019 dephead = deplist;
5020 nasm_add_string_to_strlist(dephead, file);
5023 * Define the __PASS__ macro. This is defined here unlike
5024 * all the other builtins, because it is special -- it varies between
5025 * passes.
5027 t = nasm_malloc(sizeof(*t));
5028 t->next = NULL;
5029 make_tok_num(t, apass);
5030 t->a.mac = NULL;
5031 define_smacro(NULL, "__PASS__", true, 0, t);
5034 static void pp_init(void)
5036 hash_init(&FileHash, HASH_MEDIUM);
5039 static char *pp_getline(void)
5041 char *line;
5042 Token *tline;
5044 real_verror = nasm_set_verror(pp_verror);
5046 while (1) {
5048 * Fetch a tokenized line, either from the macro-expansion
5049 * buffer or from the input file.
5051 tline = NULL;
5052 while (istk->expansion && istk->expansion->finishes) {
5053 Line *l = istk->expansion;
5054 if (!l->finishes->name && l->finishes->in_progress > 1) {
5055 Line *ll;
5058 * This is a macro-end marker for a macro with no
5059 * name, which means it's not really a macro at all
5060 * but a %rep block, and the `in_progress' field is
5061 * more than 1, meaning that we still need to
5062 * repeat. (1 means the natural last repetition; 0
5063 * means termination by %exitrep.) We have
5064 * therefore expanded up to the %endrep, and must
5065 * push the whole block on to the expansion buffer
5066 * again. We don't bother to remove the macro-end
5067 * marker: we'd only have to generate another one
5068 * if we did.
5070 l->finishes->in_progress--;
5071 list_for_each(l, l->finishes->expansion) {
5072 Token *t, *tt, **tail;
5074 ll = nasm_malloc(sizeof(Line));
5075 ll->next = istk->expansion;
5076 ll->finishes = NULL;
5077 ll->first = NULL;
5078 tail = &ll->first;
5080 list_for_each(t, l->first) {
5081 if (t->text || t->type == TOK_WHITESPACE) {
5082 tt = *tail = new_Token(NULL, t->type, t->text, 0);
5083 tail = &tt->next;
5087 istk->expansion = ll;
5089 } else {
5091 * Check whether a `%rep' was started and not ended
5092 * within this macro expansion. This can happen and
5093 * should be detected. It's a fatal error because
5094 * I'm too confused to work out how to recover
5095 * sensibly from it.
5097 if (defining) {
5098 if (defining->name)
5099 nasm_panic(0, "defining with name in expansion");
5100 else if (istk->mstk->name)
5101 nasm_fatal(0, "`%%rep' without `%%endrep' within"
5102 " expansion of macro `%s'",
5103 istk->mstk->name);
5107 * FIXME: investigate the relationship at this point between
5108 * istk->mstk and l->finishes
5111 MMacro *m = istk->mstk;
5112 istk->mstk = m->next_active;
5113 if (m->name) {
5115 * This was a real macro call, not a %rep, and
5116 * therefore the parameter information needs to
5117 * be freed.
5119 if (m->prev) {
5120 pop_mmacro(m);
5121 l->finishes->in_progress --;
5122 } else {
5123 nasm_free(m->params);
5124 free_tlist(m->iline);
5125 nasm_free(m->paramlen);
5126 l->finishes->in_progress = 0;
5131 * FIXME It is incorrect to always free_mmacro here.
5132 * It leads to usage-after-free.
5134 * https://bugzilla.nasm.us/show_bug.cgi?id=3392414
5136 #if 0
5137 else
5138 free_mmacro(m);
5139 #endif
5141 istk->expansion = l->next;
5142 nasm_free(l);
5143 lfmt->downlevel(LIST_MACRO);
5146 while (1) { /* until we get a line we can use */
5148 if (istk->expansion) { /* from a macro expansion */
5149 char *p;
5150 Line *l = istk->expansion;
5151 if (istk->mstk)
5152 istk->mstk->lineno++;
5153 tline = l->first;
5154 istk->expansion = l->next;
5155 nasm_free(l);
5156 p = detoken(tline, false);
5157 lfmt->line(LIST_MACRO, p);
5158 nasm_free(p);
5159 break;
5161 line = read_line();
5162 if (line) { /* from the current input file */
5163 line = prepreproc(line);
5164 tline = tokenize(line);
5165 nasm_free(line);
5166 break;
5169 * The current file has ended; work down the istk
5172 Include *i = istk;
5173 fclose(i->fp);
5174 if (i->conds) {
5175 /* nasm_error can't be conditionally suppressed */
5176 nasm_fatal(0,
5177 "expected `%%endif' before end of file");
5179 /* only set line and file name if there's a next node */
5180 if (i->next)
5181 src_set(i->lineno, i->fname);
5182 istk = i->next;
5183 lfmt->downlevel(LIST_INCLUDE);
5184 nasm_free(i);
5185 if (!istk) {
5186 line = NULL;
5187 goto done;
5189 if (istk->expansion && istk->expansion->finishes)
5190 break;
5195 * We must expand MMacro parameters and MMacro-local labels
5196 * _before_ we plunge into directive processing, to cope
5197 * with things like `%define something %1' such as STRUC
5198 * uses. Unless we're _defining_ a MMacro, in which case
5199 * those tokens should be left alone to go into the
5200 * definition; and unless we're in a non-emitting
5201 * condition, in which case we don't want to meddle with
5202 * anything.
5204 if (!defining && !(istk->conds && !emitting(istk->conds->state))
5205 && !(istk->mstk && !istk->mstk->in_progress)) {
5206 tline = expand_mmac_params(tline);
5210 * Check the line to see if it's a preprocessor directive.
5212 if (do_directive(tline, &line) == DIRECTIVE_FOUND) {
5213 if (line)
5214 break; /* Directive generated output */
5215 else
5216 continue;
5217 } else if (defining) {
5219 * We're defining a multi-line macro. We emit nothing
5220 * at all, and just
5221 * shove the tokenized line on to the macro definition.
5223 Line *l = nasm_malloc(sizeof(Line));
5224 l->next = defining->expansion;
5225 l->first = tline;
5226 l->finishes = NULL;
5227 defining->expansion = l;
5228 continue;
5229 } else if (istk->conds && !emitting(istk->conds->state)) {
5231 * We're in a non-emitting branch of a condition block.
5232 * Emit nothing at all, not even a blank line: when we
5233 * emerge from the condition we'll give a line-number
5234 * directive so we keep our place correctly.
5236 free_tlist(tline);
5237 continue;
5238 } else if (istk->mstk && !istk->mstk->in_progress) {
5240 * We're in a %rep block which has been terminated, so
5241 * we're walking through to the %endrep without
5242 * emitting anything. Emit nothing at all, not even a
5243 * blank line: when we emerge from the %rep block we'll
5244 * give a line-number directive so we keep our place
5245 * correctly.
5247 free_tlist(tline);
5248 continue;
5249 } else {
5250 tline = expand_smacro(tline);
5251 if (!expand_mmacro(tline)) {
5253 * De-tokenize the line again, and emit it.
5255 line = detoken(tline, true);
5256 free_tlist(tline);
5257 break;
5258 } else {
5259 continue; /* expand_mmacro calls free_tlist */
5264 done:
5265 nasm_set_verror(real_verror);
5266 return line;
5269 static void pp_cleanup(int pass)
5271 real_verror = nasm_set_verror(pp_verror);
5273 if (defining) {
5274 if (defining->name) {
5275 nasm_error(ERR_NONFATAL,
5276 "end of file while still defining macro `%s'",
5277 defining->name);
5278 } else {
5279 nasm_error(ERR_NONFATAL, "end of file while still in %%rep");
5282 free_mmacro(defining);
5283 defining = NULL;
5286 nasm_set_verror(real_verror);
5288 while (cstk)
5289 ctx_pop();
5290 free_macros();
5291 while (istk) {
5292 Include *i = istk;
5293 istk = istk->next;
5294 fclose(i->fp);
5295 nasm_free(i);
5297 while (cstk)
5298 ctx_pop();
5299 src_set_fname(NULL);
5300 if (pass == 0) {
5301 IncPath *i;
5302 free_llist(predef);
5303 predef = NULL;
5304 delete_Blocks();
5305 freeTokens = NULL;
5306 while ((i = ipath)) {
5307 ipath = i->next;
5308 if (i->path)
5309 nasm_free(i->path);
5310 nasm_free(i);
5315 static void pp_include_path(char *path)
5317 IncPath *i;
5319 i = nasm_malloc(sizeof(IncPath));
5320 i->path = path ? nasm_strdup(path) : NULL;
5321 i->next = NULL;
5323 if (ipath) {
5324 IncPath *j = ipath;
5325 while (j->next)
5326 j = j->next;
5327 j->next = i;
5328 } else {
5329 ipath = i;
5333 static void pp_pre_include(char *fname)
5335 Token *inc, *space, *name;
5336 Line *l;
5338 name = new_Token(NULL, TOK_INTERNAL_STRING, fname, 0);
5339 space = new_Token(name, TOK_WHITESPACE, NULL, 0);
5340 inc = new_Token(space, TOK_PREPROC_ID, "%include", 0);
5342 l = nasm_malloc(sizeof(Line));
5343 l->next = predef;
5344 l->first = inc;
5345 l->finishes = NULL;
5346 predef = l;
5349 static void pp_pre_define(char *definition)
5351 Token *def, *space;
5352 Line *l;
5353 char *equals;
5355 real_verror = nasm_set_verror(pp_verror);
5357 equals = strchr(definition, '=');
5358 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
5359 def = new_Token(space, TOK_PREPROC_ID, "%define", 0);
5360 if (equals)
5361 *equals = ' ';
5362 space->next = tokenize(definition);
5363 if (equals)
5364 *equals = '=';
5366 if (space->next->type != TOK_PREPROC_ID &&
5367 space->next->type != TOK_ID)
5368 nasm_error(ERR_WARNING, "pre-defining non ID `%s\'\n", definition);
5370 l = nasm_malloc(sizeof(Line));
5371 l->next = predef;
5372 l->first = def;
5373 l->finishes = NULL;
5374 predef = l;
5376 nasm_set_verror(real_verror);
5379 static void pp_pre_undefine(char *definition)
5381 Token *def, *space;
5382 Line *l;
5384 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
5385 def = new_Token(space, TOK_PREPROC_ID, "%undef", 0);
5386 space->next = tokenize(definition);
5388 l = nasm_malloc(sizeof(Line));
5389 l->next = predef;
5390 l->first = def;
5391 l->finishes = NULL;
5392 predef = l;
5395 /* Insert an early preprocessor command that doesn't need special handling */
5396 static void pp_pre_command(const char *what, char *string)
5398 char *cmd;
5399 Token *def, *space;
5400 Line *l;
5402 def = tokenize(string);
5403 if (what) {
5404 cmd = nasm_strcat(what[0] == '%' ? "" : "%", what);
5405 space = new_Token(def, TOK_WHITESPACE, NULL, 0);
5406 def = new_Token(space, TOK_PREPROC_ID, cmd, 0);
5409 l = nasm_malloc(sizeof(Line));
5410 l->next = predef;
5411 l->first = def;
5412 l->finishes = NULL;
5413 predef = l;
5416 static void pp_add_stdmac(macros_t *macros)
5418 macros_t **mp;
5420 /* Find the end of the list and avoid duplicates */
5421 for (mp = stdmacros; *mp; mp++) {
5422 if (*mp == macros)
5423 return; /* Nothing to do */
5426 nasm_assert(mp < &stdmacros[ARRAY_SIZE(stdmacros)-1]);
5428 *mp = macros;
5431 static void pp_extra_stdmac(macros_t *macros)
5433 extrastdmac = macros;
5436 static void make_tok_num(Token * tok, int64_t val)
5438 char numbuf[32];
5439 snprintf(numbuf, sizeof(numbuf), "%"PRId64"", val);
5440 tok->text = nasm_strdup(numbuf);
5441 tok->type = TOK_NUMBER;
5444 static void pp_list_one_macro(MMacro *m, int severity)
5446 if (!m)
5447 return;
5449 /* We need to print the next_active list in reverse order */
5450 pp_list_one_macro(m->next_active, severity);
5452 if (m->name && !m->nolist) {
5453 src_set(m->xline + m->lineno, m->fname);
5454 nasm_error(severity, "... from macro `%s' defined here", m->name);
5458 static void pp_error_list_macros(int severity)
5460 int32_t saved_line;
5461 const char *saved_fname = NULL;
5463 severity |= ERR_PP_LISTMACRO | ERR_NO_SEVERITY;
5464 src_get(&saved_line, &saved_fname);
5466 if (istk)
5467 pp_list_one_macro(istk->mstk, severity);
5469 src_set(saved_line, saved_fname);
5472 const struct preproc_ops nasmpp = {
5473 pp_init,
5474 pp_reset,
5475 pp_getline,
5476 pp_cleanup,
5477 pp_extra_stdmac,
5478 pp_pre_define,
5479 pp_pre_undefine,
5480 pp_pre_include,
5481 pp_pre_command,
5482 pp_include_path,
5483 pp_error_list_macros,