configure: enable additional code cleanliness warnings
[nasm.git] / preproc.c
blobb43cd61769285ae895efafe9dd42db459d361c77
1 /* ----------------------------------------------------------------------- *
3 * Copyright 1996-2016 The NASM Authors - All Rights Reserved
4 * See the file AUTHORS included with the NASM distribution for
5 * the specific copyright holders.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following
9 * conditions are met:
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above
14 * copyright notice, this list of conditions and the following
15 * disclaimer in the documentation and/or other materials provided
16 * with the distribution.
18 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
19 * CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
20 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
21 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
22 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
23 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
25 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
26 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
29 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
30 * EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32 * ----------------------------------------------------------------------- */
35 * preproc.c macro preprocessor for the Netwide Assembler
38 /* Typical flow of text through preproc
40 * pp_getline gets tokenized lines, either
42 * from a macro expansion
44 * or
45 * {
46 * read_line gets raw text from stdmacpos, or predef, or current input file
47 * tokenize converts to tokens
48 * }
50 * expand_mmac_params is used to expand %1 etc., unless a macro is being
51 * defined or a false conditional is being processed
52 * (%0, %1, %+1, %-1, %%foo
54 * do_directive checks for directives
56 * expand_smacro is used to expand single line macros
58 * expand_mmacro is used to expand multi-line macros
60 * detoken is used to convert the line back to text
63 #include "compiler.h"
65 #include <stdio.h>
66 #include <stdarg.h>
67 #include <stdlib.h>
68 #include <stddef.h>
69 #include <string.h>
70 #include <ctype.h>
71 #include <limits.h>
72 #include <inttypes.h>
74 #include "nasm.h"
75 #include "nasmlib.h"
76 #include "preproc.h"
77 #include "hashtbl.h"
78 #include "quote.h"
79 #include "stdscan.h"
80 #include "eval.h"
81 #include "tokens.h"
82 #include "tables.h"
84 typedef struct SMacro SMacro;
85 typedef struct MMacro MMacro;
86 typedef struct MMacroInvocation MMacroInvocation;
87 typedef struct Context Context;
88 typedef struct Token Token;
89 typedef struct Blocks Blocks;
90 typedef struct Line Line;
91 typedef struct Include Include;
92 typedef struct Cond Cond;
93 typedef struct IncPath IncPath;
96 * Note on the storage of both SMacro and MMacros: the hash table
97 * indexes them case-insensitively, and we then have to go through a
98 * linked list of potential case aliases (and, for MMacros, parameter
99 * ranges); this is to preserve the matching semantics of the earlier
100 * code. If the number of case aliases for a specific macro is a
101 * performance issue, you may want to reconsider your coding style.
105 * Store the definition of a single-line macro.
107 struct SMacro {
108 SMacro *next;
109 char *name;
110 bool casesense;
111 bool in_progress;
112 unsigned int nparam;
113 Token *expansion;
117 * Store the definition of a multi-line macro. This is also used to
118 * store the interiors of `%rep...%endrep' blocks, which are
119 * effectively self-re-invoking multi-line macros which simply
120 * don't have a name or bother to appear in the hash tables. %rep
121 * blocks are signified by having a NULL `name' field.
123 * In a MMacro describing a `%rep' block, the `in_progress' field
124 * isn't merely boolean, but gives the number of repeats left to
125 * run.
127 * The `next' field is used for storing MMacros in hash tables; the
128 * `next_active' field is for stacking them on istk entries.
130 * When a MMacro is being expanded, `params', `iline', `nparam',
131 * `paramlen', `rotate' and `unique' are local to the invocation.
133 struct MMacro {
134 MMacro *next;
135 MMacroInvocation *prev; /* previous invocation */
136 char *name;
137 int nparam_min, nparam_max;
138 bool casesense;
139 bool plus; /* is the last parameter greedy? */
140 bool nolist; /* is this macro listing-inhibited? */
141 int64_t in_progress; /* is this macro currently being expanded? */
142 int32_t max_depth; /* maximum number of recursive expansions allowed */
143 Token *dlist; /* All defaults as one list */
144 Token **defaults; /* Parameter default pointers */
145 int ndefs; /* number of default parameters */
146 Line *expansion;
148 MMacro *next_active;
149 MMacro *rep_nest; /* used for nesting %rep */
150 Token **params; /* actual parameters */
151 Token *iline; /* invocation line */
152 unsigned int nparam, rotate;
153 int *paramlen;
154 uint64_t unique;
155 int lineno; /* Current line number on expansion */
156 uint64_t condcnt; /* number of if blocks... */
160 /* Store the definition of a multi-line macro, as defined in a
161 * previous recursive macro expansion.
163 struct MMacroInvocation {
164 MMacroInvocation *prev; /* previous invocation */
165 Token **params; /* actual parameters */
166 Token *iline; /* invocation line */
167 unsigned int nparam, rotate;
168 int *paramlen;
169 uint64_t unique;
170 uint64_t condcnt;
175 * The context stack is composed of a linked list of these.
177 struct Context {
178 Context *next;
179 char *name;
180 struct hash_table localmac;
181 uint32_t number;
185 * This is the internal form which we break input lines up into.
186 * Typically stored in linked lists.
188 * Note that `type' serves a double meaning: TOK_SMAC_PARAM is not
189 * necessarily used as-is, but is intended to denote the number of
190 * the substituted parameter. So in the definition
192 * %define a(x,y) ( (x) & ~(y) )
194 * the token representing `x' will have its type changed to
195 * TOK_SMAC_PARAM, but the one representing `y' will be
196 * TOK_SMAC_PARAM+1.
198 * TOK_INTERNAL_STRING is a dirty hack: it's a single string token
199 * which doesn't need quotes around it. Used in the pre-include
200 * mechanism as an alternative to trying to find a sensible type of
201 * quote to use on the filename we were passed.
203 enum pp_token_type {
204 TOK_NONE = 0, TOK_WHITESPACE, TOK_COMMENT, TOK_ID,
205 TOK_PREPROC_ID, TOK_STRING,
206 TOK_NUMBER, TOK_FLOAT, TOK_SMAC_END, TOK_OTHER,
207 TOK_INTERNAL_STRING,
208 TOK_PREPROC_Q, TOK_PREPROC_QQ,
209 TOK_PASTE, /* %+ */
210 TOK_INDIRECT, /* %[...] */
211 TOK_SMAC_PARAM, /* MUST BE LAST IN THE LIST!!! */
212 TOK_MAX = INT_MAX /* Keep compiler from reducing the range */
215 #define PP_CONCAT_MASK(x) (1 << (x))
216 #define PP_CONCAT_MATCH(t, mask) (PP_CONCAT_MASK((t)->type) & mask)
218 struct tokseq_match {
219 int mask_head;
220 int mask_tail;
223 struct Token {
224 Token *next;
225 char *text;
226 union {
227 SMacro *mac; /* associated macro for TOK_SMAC_END */
228 size_t len; /* scratch length field */
229 } a; /* Auxiliary data */
230 enum pp_token_type type;
234 * Multi-line macro definitions are stored as a linked list of
235 * these, which is essentially a container to allow several linked
236 * lists of Tokens.
238 * Note that in this module, linked lists are treated as stacks
239 * wherever possible. For this reason, Lines are _pushed_ on to the
240 * `expansion' field in MMacro structures, so that the linked list,
241 * if walked, would give the macro lines in reverse order; this
242 * means that we can walk the list when expanding a macro, and thus
243 * push the lines on to the `expansion' field in _istk_ in reverse
244 * order (so that when popped back off they are in the right
245 * order). It may seem cockeyed, and it relies on my design having
246 * an even number of steps in, but it works...
248 * Some of these structures, rather than being actual lines, are
249 * markers delimiting the end of the expansion of a given macro.
250 * This is for use in the cycle-tracking and %rep-handling code.
251 * Such structures have `finishes' non-NULL, and `first' NULL. All
252 * others have `finishes' NULL, but `first' may still be NULL if
253 * the line is blank.
255 struct Line {
256 Line *next;
257 MMacro *finishes;
258 Token *first;
262 * To handle an arbitrary level of file inclusion, we maintain a
263 * stack (ie linked list) of these things.
265 struct Include {
266 Include *next;
267 FILE *fp;
268 Cond *conds;
269 Line *expansion;
270 char *fname;
271 int lineno, lineinc;
272 MMacro *mstk; /* stack of active macros/reps */
276 * Include search path. This is simply a list of strings which get
277 * prepended, in turn, to the name of an include file, in an
278 * attempt to find the file if it's not in the current directory.
280 struct IncPath {
281 IncPath *next;
282 char *path;
286 * Conditional assembly: we maintain a separate stack of these for
287 * each level of file inclusion. (The only reason we keep the
288 * stacks separate is to ensure that a stray `%endif' in a file
289 * included from within the true branch of a `%if' won't terminate
290 * it and cause confusion: instead, rightly, it'll cause an error.)
292 struct Cond {
293 Cond *next;
294 int state;
296 enum {
298 * These states are for use just after %if or %elif: IF_TRUE
299 * means the condition has evaluated to truth so we are
300 * currently emitting, whereas IF_FALSE means we are not
301 * currently emitting but will start doing so if a %else comes
302 * up. In these states, all directives are admissible: %elif,
303 * %else and %endif. (And of course %if.)
305 COND_IF_TRUE, COND_IF_FALSE,
307 * These states come up after a %else: ELSE_TRUE means we're
308 * emitting, and ELSE_FALSE means we're not. In ELSE_* states,
309 * any %elif or %else will cause an error.
311 COND_ELSE_TRUE, COND_ELSE_FALSE,
313 * These states mean that we're not emitting now, and also that
314 * nothing until %endif will be emitted at all. COND_DONE is
315 * used when we've had our moment of emission
316 * and have now started seeing %elifs. COND_NEVER is used when
317 * the condition construct in question is contained within a
318 * non-emitting branch of a larger condition construct,
319 * or if there is an error.
321 COND_DONE, COND_NEVER
323 #define emitting(x) ( (x) == COND_IF_TRUE || (x) == COND_ELSE_TRUE )
326 * These defines are used as the possible return values for do_directive
328 #define NO_DIRECTIVE_FOUND 0
329 #define DIRECTIVE_FOUND 1
332 * This define sets the upper limit for smacro and recursive mmacro
333 * expansions
335 #define DEADMAN_LIMIT (1 << 20)
337 /* max reps */
338 #define REP_LIMIT ((INT64_C(1) << 62))
341 * Condition codes. Note that we use c_ prefix not C_ because C_ is
342 * used in nasm.h for the "real" condition codes. At _this_ level,
343 * we treat CXZ and ECXZ as condition codes, albeit non-invertible
344 * ones, so we need a different enum...
346 static const char * const conditions[] = {
347 "a", "ae", "b", "be", "c", "cxz", "e", "ecxz", "g", "ge", "l", "le",
348 "na", "nae", "nb", "nbe", "nc", "ne", "ng", "nge", "nl", "nle", "no",
349 "np", "ns", "nz", "o", "p", "pe", "po", "rcxz", "s", "z"
351 enum pp_conds {
352 c_A, c_AE, c_B, c_BE, c_C, c_CXZ, c_E, c_ECXZ, c_G, c_GE, c_L, c_LE,
353 c_NA, c_NAE, c_NB, c_NBE, c_NC, c_NE, c_NG, c_NGE, c_NL, c_NLE, c_NO,
354 c_NP, c_NS, c_NZ, c_O, c_P, c_PE, c_PO, c_RCXZ, c_S, c_Z,
355 c_none = -1
357 static const enum pp_conds inverse_ccs[] = {
358 c_NA, c_NAE, c_NB, c_NBE, c_NC, -1, c_NE, -1, c_NG, c_NGE, c_NL, c_NLE,
359 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,
360 c_Z, c_NO, c_NP, c_PO, c_PE, -1, c_NS, c_NZ
364 * Directive names.
366 /* If this is a an IF, ELIF, ELSE or ENDIF keyword */
367 static int is_condition(enum preproc_token arg)
369 return PP_IS_COND(arg) || (arg == PP_ELSE) || (arg == PP_ENDIF);
372 /* For TASM compatibility we need to be able to recognise TASM compatible
373 * conditional compilation directives. Using the NASM pre-processor does
374 * not work, so we look for them specifically from the following list and
375 * then jam in the equivalent NASM directive into the input stream.
378 enum {
379 TM_ARG, TM_ELIF, TM_ELSE, TM_ENDIF, TM_IF, TM_IFDEF, TM_IFDIFI,
380 TM_IFNDEF, TM_INCLUDE, TM_LOCAL
383 static const char * const tasm_directives[] = {
384 "arg", "elif", "else", "endif", "if", "ifdef", "ifdifi",
385 "ifndef", "include", "local"
388 static int StackSize = 4;
389 static char *StackPointer = "ebp";
390 static int ArgOffset = 8;
391 static int LocalOffset = 0;
393 static Context *cstk;
394 static Include *istk;
395 static IncPath *ipath = NULL;
397 static int pass; /* HACK: pass 0 = generate dependencies only */
398 static StrList **dephead, **deptail; /* Dependency list */
400 static uint64_t unique; /* unique identifier numbers */
402 static Line *predef = NULL;
403 static bool do_predef;
406 * The current set of multi-line macros we have defined.
408 static struct hash_table mmacros;
411 * The current set of single-line macros we have defined.
413 static struct hash_table smacros;
416 * The multi-line macro we are currently defining, or the %rep
417 * block we are currently reading, if any.
419 static MMacro *defining;
421 static uint64_t nested_mac_count;
422 static uint64_t nested_rep_count;
425 * The number of macro parameters to allocate space for at a time.
427 #define PARAM_DELTA 16
430 * The standard macro set: defined in macros.c in the array nasm_stdmac.
431 * This gives our position in the macro set, when we're processing it.
433 static macros_t *stdmacpos;
436 * The extra standard macros that come from the object format, if
437 * any.
439 static macros_t *extrastdmac = NULL;
440 static bool any_extrastdmac;
443 * Tokens are allocated in blocks to improve speed
445 #define TOKEN_BLOCKSIZE 4096
446 static Token *freeTokens = NULL;
447 struct Blocks {
448 Blocks *next;
449 void *chunk;
452 static Blocks blocks = { NULL, NULL };
455 * Forward declarations.
457 static Token *expand_mmac_params(Token * tline);
458 static Token *expand_smacro(Token * tline);
459 static Token *expand_id(Token * tline);
460 static Context *get_ctx(const char *name, const char **namep);
461 static void make_tok_num(Token * tok, int64_t val);
462 static void pp_verror(int severity, const char *fmt, va_list ap);
463 static vefunc real_verror;
464 static void *new_Block(size_t size);
465 static void delete_Blocks(void);
466 static Token *new_Token(Token * next, enum pp_token_type type,
467 const char *text, int txtlen);
468 static Token *delete_Token(Token * t);
471 * Macros for safe checking of token pointers, avoid *(NULL)
473 #define tok_type_(x,t) ((x) && (x)->type == (t))
474 #define skip_white_(x) if (tok_type_((x), TOK_WHITESPACE)) (x)=(x)->next
475 #define tok_is_(x,v) (tok_type_((x), TOK_OTHER) && !strcmp((x)->text,(v)))
476 #define tok_isnt_(x,v) ((x) && ((x)->type!=TOK_OTHER || strcmp((x)->text,(v))))
479 * nasm_unquote with error if the string contains NUL characters.
480 * If the string contains NUL characters, issue an error and return
481 * the C len, i.e. truncate at the NUL.
483 static size_t nasm_unquote_cstr(char *qstr, enum preproc_token directive)
485 size_t len = nasm_unquote(qstr, NULL);
486 size_t clen = strlen(qstr);
488 if (len != clen)
489 nasm_error(ERR_NONFATAL, "NUL character in `%s' directive",
490 pp_directives[directive]);
492 return clen;
496 * In-place reverse a list of tokens.
498 static Token *reverse_tokens(Token *t)
500 Token *prev = NULL;
501 Token *next;
503 while (t) {
504 next = t->next;
505 t->next = prev;
506 prev = t;
507 t = next;
510 return prev;
514 * Handle TASM specific directives, which do not contain a % in
515 * front of them. We do it here because I could not find any other
516 * place to do it for the moment, and it is a hack (ideally it would
517 * be nice to be able to use the NASM pre-processor to do it).
519 static char *check_tasm_directive(char *line)
521 int32_t i, j, k, m, len;
522 char *p, *q, *oldline, oldchar;
524 p = nasm_skip_spaces(line);
526 /* Binary search for the directive name */
527 i = -1;
528 j = ARRAY_SIZE(tasm_directives);
529 q = nasm_skip_word(p);
530 len = q - p;
531 if (len) {
532 oldchar = p[len];
533 p[len] = 0;
534 while (j - i > 1) {
535 k = (j + i) / 2;
536 m = nasm_stricmp(p, tasm_directives[k]);
537 if (m == 0) {
538 /* We have found a directive, so jam a % in front of it
539 * so that NASM will then recognise it as one if it's own.
541 p[len] = oldchar;
542 len = strlen(p);
543 oldline = line;
544 line = nasm_malloc(len + 2);
545 line[0] = '%';
546 if (k == TM_IFDIFI) {
548 * NASM does not recognise IFDIFI, so we convert
549 * it to %if 0. This is not used in NASM
550 * compatible code, but does need to parse for the
551 * TASM macro package.
553 strcpy(line + 1, "if 0");
554 } else {
555 memcpy(line + 1, p, len + 1);
557 nasm_free(oldline);
558 return line;
559 } else if (m < 0) {
560 j = k;
561 } else
562 i = k;
564 p[len] = oldchar;
566 return line;
570 * The pre-preprocessing stage... This function translates line
571 * number indications as they emerge from GNU cpp (`# lineno "file"
572 * flags') into NASM preprocessor line number indications (`%line
573 * lineno file').
575 static char *prepreproc(char *line)
577 int lineno, fnlen;
578 char *fname, *oldline;
580 if (line[0] == '#' && line[1] == ' ') {
581 oldline = line;
582 fname = oldline + 2;
583 lineno = atoi(fname);
584 fname += strspn(fname, "0123456789 ");
585 if (*fname == '"')
586 fname++;
587 fnlen = strcspn(fname, "\"");
588 line = nasm_malloc(20 + fnlen);
589 snprintf(line, 20 + fnlen, "%%line %d %.*s", lineno, fnlen, fname);
590 nasm_free(oldline);
592 if (tasm_compatible_mode)
593 return check_tasm_directive(line);
594 return line;
598 * Free a linked list of tokens.
600 static void free_tlist(Token * list)
602 while (list)
603 list = delete_Token(list);
607 * Free a linked list of lines.
609 static void free_llist(Line * list)
611 Line *l, *tmp;
612 list_for_each_safe(l, tmp, list) {
613 free_tlist(l->first);
614 nasm_free(l);
619 * Free an MMacro
621 static void free_mmacro(MMacro * m)
623 nasm_free(m->name);
624 free_tlist(m->dlist);
625 nasm_free(m->defaults);
626 free_llist(m->expansion);
627 nasm_free(m);
631 * Free all currently defined macros, and free the hash tables
633 static void free_smacro_table(struct hash_table *smt)
635 SMacro *s, *tmp;
636 const char *key;
637 struct hash_tbl_node *it = NULL;
639 while ((s = hash_iterate(smt, &it, &key)) != NULL) {
640 nasm_free((void *)key);
641 list_for_each_safe(s, tmp, s) {
642 nasm_free(s->name);
643 free_tlist(s->expansion);
644 nasm_free(s);
647 hash_free(smt);
650 static void free_mmacro_table(struct hash_table *mmt)
652 MMacro *m, *tmp;
653 const char *key;
654 struct hash_tbl_node *it = NULL;
656 it = NULL;
657 while ((m = hash_iterate(mmt, &it, &key)) != NULL) {
658 nasm_free((void *)key);
659 list_for_each_safe(m ,tmp, m)
660 free_mmacro(m);
662 hash_free(mmt);
665 static void free_macros(void)
667 free_smacro_table(&smacros);
668 free_mmacro_table(&mmacros);
672 * Initialize the hash tables
674 static void init_macros(void)
676 hash_init(&smacros, HASH_LARGE);
677 hash_init(&mmacros, HASH_LARGE);
681 * Pop the context stack.
683 static void ctx_pop(void)
685 Context *c = cstk;
687 cstk = cstk->next;
688 free_smacro_table(&c->localmac);
689 nasm_free(c->name);
690 nasm_free(c);
694 * Search for a key in the hash index; adding it if necessary
695 * (in which case we initialize the data pointer to NULL.)
697 static void **
698 hash_findi_add(struct hash_table *hash, const char *str)
700 struct hash_insert hi;
701 void **r;
702 char *strx;
704 r = hash_findi(hash, str, &hi);
705 if (r)
706 return r;
708 strx = nasm_strdup(str); /* Use a more efficient allocator here? */
709 return hash_add(&hi, strx, NULL);
713 * Like hash_findi, but returns the data element rather than a pointer
714 * to it. Used only when not adding a new element, hence no third
715 * argument.
717 static void *
718 hash_findix(struct hash_table *hash, const char *str)
720 void **p;
722 p = hash_findi(hash, str, NULL);
723 return p ? *p : NULL;
727 * read line from standart macros set,
728 * if there no more left -- return NULL
730 static char *line_from_stdmac(void)
732 unsigned char c;
733 const unsigned char *p = stdmacpos;
734 char *line, *q;
735 size_t len = 0;
737 if (!stdmacpos)
738 return NULL;
740 while ((c = *p++)) {
741 if (c >= 0x80)
742 len += pp_directives_len[c - 0x80] + 1;
743 else
744 len++;
747 line = nasm_malloc(len + 1);
748 q = line;
749 while ((c = *stdmacpos++)) {
750 if (c >= 0x80) {
751 memcpy(q, pp_directives[c - 0x80], pp_directives_len[c - 0x80]);
752 q += pp_directives_len[c - 0x80];
753 *q++ = ' ';
754 } else {
755 *q++ = c;
758 stdmacpos = p;
759 *q = '\0';
761 if (!*stdmacpos) {
762 /* This was the last of the standard macro chain... */
763 stdmacpos = NULL;
764 if (any_extrastdmac) {
765 stdmacpos = extrastdmac;
766 any_extrastdmac = false;
767 } else if (do_predef) {
768 Line *pd, *l;
769 Token *head, **tail, *t;
772 * Nasty hack: here we push the contents of
773 * `predef' on to the top-level expansion stack,
774 * since this is the most convenient way to
775 * implement the pre-include and pre-define
776 * features.
778 list_for_each(pd, predef) {
779 head = NULL;
780 tail = &head;
781 list_for_each(t, pd->first) {
782 *tail = new_Token(NULL, t->type, t->text, 0);
783 tail = &(*tail)->next;
786 l = nasm_malloc(sizeof(Line));
787 l->next = istk->expansion;
788 l->first = head;
789 l->finishes = NULL;
791 istk->expansion = l;
793 do_predef = false;
797 return line;
800 static char *read_line(void)
802 unsigned int size, c, next;
803 const unsigned int delta = 512;
804 const unsigned int pad = 8;
805 unsigned int nr_cont = 0;
806 bool cont = false;
807 char *buffer, *p;
809 /* Standart macros set (predefined) goes first */
810 p = line_from_stdmac();
811 if (p)
812 return p;
814 size = delta;
815 p = buffer = nasm_malloc(size);
817 for (;;) {
818 c = fgetc(istk->fp);
819 if ((int)(c) == EOF) {
820 p[0] = 0;
821 break;
824 switch (c) {
825 case '\r':
826 next = fgetc(istk->fp);
827 if (next != '\n')
828 ungetc(next, istk->fp);
829 if (cont) {
830 cont = false;
831 continue;
833 break;
835 case '\n':
836 if (cont) {
837 cont = false;
838 continue;
840 break;
842 case '\\':
843 next = fgetc(istk->fp);
844 ungetc(next, istk->fp);
845 if (next == '\r' || next == '\n') {
846 cont = true;
847 nr_cont++;
848 continue;
850 break;
853 if (c == '\r' || c == '\n') {
854 *p++ = 0;
855 break;
858 if (p >= (buffer + size - pad)) {
859 buffer = nasm_realloc(buffer, size + delta);
860 p = buffer + size - pad;
861 size += delta;
864 *p++ = (unsigned char)c;
867 if (p == buffer) {
868 nasm_free(buffer);
869 return NULL;
872 src_set_linnum(src_get_linnum() + istk->lineinc +
873 (nr_cont * istk->lineinc));
876 * Handle spurious ^Z, which may be inserted into source files
877 * by some file transfer utilities.
879 buffer[strcspn(buffer, "\032")] = '\0';
881 nasmlist->line(LIST_READ, buffer);
883 return buffer;
887 * Tokenize a line of text. This is a very simple process since we
888 * don't need to parse the value out of e.g. numeric tokens: we
889 * simply split one string into many.
891 static Token *tokenize(char *line)
893 char c, *p = line;
894 enum pp_token_type type;
895 Token *list = NULL;
896 Token *t, **tail = &list;
898 while (*line) {
899 p = line;
900 if (*p == '%') {
901 p++;
902 if (*p == '+' && !nasm_isdigit(p[1])) {
903 p++;
904 type = TOK_PASTE;
905 } else if (nasm_isdigit(*p) ||
906 ((*p == '-' || *p == '+') && nasm_isdigit(p[1]))) {
907 do {
908 p++;
910 while (nasm_isdigit(*p));
911 type = TOK_PREPROC_ID;
912 } else if (*p == '{') {
913 p++;
914 while (*p) {
915 if (*p == '}')
916 break;
917 p[-1] = *p;
918 p++;
920 if (*p != '}')
921 nasm_error(ERR_WARNING | ERR_PASS1,
922 "unterminated %%{ construct");
923 p[-1] = '\0';
924 if (*p)
925 p++;
926 type = TOK_PREPROC_ID;
927 } else if (*p == '[') {
928 int lvl = 1;
929 line += 2; /* Skip the leading %[ */
930 p++;
931 while (lvl && (c = *p++)) {
932 switch (c) {
933 case ']':
934 lvl--;
935 break;
936 case '%':
937 if (*p == '[')
938 lvl++;
939 break;
940 case '\'':
941 case '\"':
942 case '`':
943 p = nasm_skip_string(p - 1) + 1;
944 break;
945 default:
946 break;
949 p--;
950 if (*p)
951 *p++ = '\0';
952 if (lvl)
953 nasm_error(ERR_NONFATAL|ERR_PASS1,
954 "unterminated %%[ construct");
955 type = TOK_INDIRECT;
956 } else if (*p == '?') {
957 type = TOK_PREPROC_Q; /* %? */
958 p++;
959 if (*p == '?') {
960 type = TOK_PREPROC_QQ; /* %?? */
961 p++;
963 } else if (*p == '!') {
964 type = TOK_PREPROC_ID;
965 p++;
966 if (isidchar(*p)) {
967 do {
968 p++;
970 while (isidchar(*p));
971 } else if (*p == '\'' || *p == '\"' || *p == '`') {
972 p = nasm_skip_string(p);
973 if (*p)
974 p++;
975 else
976 nasm_error(ERR_NONFATAL|ERR_PASS1,
977 "unterminated %%! string");
978 } else {
979 /* %! without string or identifier */
980 type = TOK_OTHER; /* Legacy behavior... */
982 } else if (isidchar(*p) ||
983 ((*p == '!' || *p == '%' || *p == '$') &&
984 isidchar(p[1]))) {
985 do {
986 p++;
988 while (isidchar(*p));
989 type = TOK_PREPROC_ID;
990 } else {
991 type = TOK_OTHER;
992 if (*p == '%')
993 p++;
995 } else if (isidstart(*p) || (*p == '$' && isidstart(p[1]))) {
996 type = TOK_ID;
997 p++;
998 while (*p && isidchar(*p))
999 p++;
1000 } else if (*p == '\'' || *p == '"' || *p == '`') {
1002 * A string token.
1004 type = TOK_STRING;
1005 p = nasm_skip_string(p);
1007 if (*p) {
1008 p++;
1009 } else {
1010 nasm_error(ERR_WARNING|ERR_PASS1, "unterminated string");
1011 /* Handling unterminated strings by UNV */
1012 /* type = -1; */
1014 } else if (p[0] == '$' && p[1] == '$') {
1015 type = TOK_OTHER; /* TOKEN_BASE */
1016 p += 2;
1017 } else if (isnumstart(*p)) {
1018 bool is_hex = false;
1019 bool is_float = false;
1020 bool has_e = false;
1021 char c, *r;
1024 * A numeric token.
1027 if (*p == '$') {
1028 p++;
1029 is_hex = true;
1032 for (;;) {
1033 c = *p++;
1035 if (!is_hex && (c == 'e' || c == 'E')) {
1036 has_e = true;
1037 if (*p == '+' || *p == '-') {
1039 * e can only be followed by +/- if it is either a
1040 * prefixed hex number or a floating-point number
1042 p++;
1043 is_float = true;
1045 } else if (c == 'H' || c == 'h' || c == 'X' || c == 'x') {
1046 is_hex = true;
1047 } else if (c == 'P' || c == 'p') {
1048 is_float = true;
1049 if (*p == '+' || *p == '-')
1050 p++;
1051 } else if (isnumchar(c) || c == '_')
1052 ; /* just advance */
1053 else if (c == '.') {
1055 * we need to deal with consequences of the legacy
1056 * parser, like "1.nolist" being two tokens
1057 * (TOK_NUMBER, TOK_ID) here; at least give it
1058 * a shot for now. In the future, we probably need
1059 * a flex-based scanner with proper pattern matching
1060 * to do it as well as it can be done. Nothing in
1061 * the world is going to help the person who wants
1062 * 0x123.p16 interpreted as two tokens, though.
1064 r = p;
1065 while (*r == '_')
1066 r++;
1068 if (nasm_isdigit(*r) || (is_hex && nasm_isxdigit(*r)) ||
1069 (!is_hex && (*r == 'e' || *r == 'E')) ||
1070 (*r == 'p' || *r == 'P')) {
1071 p = r;
1072 is_float = true;
1073 } else
1074 break; /* Terminate the token */
1075 } else
1076 break;
1078 p--; /* Point to first character beyond number */
1080 if (p == line+1 && *line == '$') {
1081 type = TOK_OTHER; /* TOKEN_HERE */
1082 } else {
1083 if (has_e && !is_hex) {
1084 /* 1e13 is floating-point, but 1e13h is not */
1085 is_float = true;
1088 type = is_float ? TOK_FLOAT : TOK_NUMBER;
1090 } else if (nasm_isspace(*p)) {
1091 type = TOK_WHITESPACE;
1092 p = nasm_skip_spaces(p);
1094 * Whitespace just before end-of-line is discarded by
1095 * pretending it's a comment; whitespace just before a
1096 * comment gets lumped into the comment.
1098 if (!*p || *p == ';') {
1099 type = TOK_COMMENT;
1100 while (*p)
1101 p++;
1103 } else if (*p == ';') {
1104 type = TOK_COMMENT;
1105 while (*p)
1106 p++;
1107 } else {
1109 * Anything else is an operator of some kind. We check
1110 * for all the double-character operators (>>, <<, //,
1111 * %%, <=, >=, ==, !=, <>, &&, ||, ^^), but anything
1112 * else is a single-character operator.
1114 type = TOK_OTHER;
1115 if ((p[0] == '>' && p[1] == '>') ||
1116 (p[0] == '<' && p[1] == '<') ||
1117 (p[0] == '/' && p[1] == '/') ||
1118 (p[0] == '<' && p[1] == '=') ||
1119 (p[0] == '>' && p[1] == '=') ||
1120 (p[0] == '=' && p[1] == '=') ||
1121 (p[0] == '!' && p[1] == '=') ||
1122 (p[0] == '<' && p[1] == '>') ||
1123 (p[0] == '&' && p[1] == '&') ||
1124 (p[0] == '|' && p[1] == '|') ||
1125 (p[0] == '^' && p[1] == '^')) {
1126 p++;
1128 p++;
1131 /* Handling unterminated string by UNV */
1132 /*if (type == -1)
1134 *tail = t = new_Token(NULL, TOK_STRING, line, p-line+1);
1135 t->text[p-line] = *line;
1136 tail = &t->next;
1138 else */
1139 if (type != TOK_COMMENT) {
1140 *tail = t = new_Token(NULL, type, line, p - line);
1141 tail = &t->next;
1143 line = p;
1145 return list;
1149 * this function allocates a new managed block of memory and
1150 * returns a pointer to the block. The managed blocks are
1151 * deleted only all at once by the delete_Blocks function.
1153 static void *new_Block(size_t size)
1155 Blocks *b = &blocks;
1157 /* first, get to the end of the linked list */
1158 while (b->next)
1159 b = b->next;
1160 /* now allocate the requested chunk */
1161 b->chunk = nasm_malloc(size);
1163 /* now allocate a new block for the next request */
1164 b->next = nasm_zalloc(sizeof(Blocks));
1165 return b->chunk;
1169 * this function deletes all managed blocks of memory
1171 static void delete_Blocks(void)
1173 Blocks *a, *b = &blocks;
1176 * keep in mind that the first block, pointed to by blocks
1177 * is a static and not dynamically allocated, so we don't
1178 * free it.
1180 while (b) {
1181 if (b->chunk)
1182 nasm_free(b->chunk);
1183 a = b;
1184 b = b->next;
1185 if (a != &blocks)
1186 nasm_free(a);
1188 memset(&blocks, 0, sizeof(blocks));
1192 * this function creates a new Token and passes a pointer to it
1193 * back to the caller. It sets the type and text elements, and
1194 * also the a.mac and next elements to NULL.
1196 static Token *new_Token(Token * next, enum pp_token_type type,
1197 const char *text, int txtlen)
1199 Token *t;
1200 int i;
1202 if (!freeTokens) {
1203 freeTokens = (Token *) new_Block(TOKEN_BLOCKSIZE * sizeof(Token));
1204 for (i = 0; i < TOKEN_BLOCKSIZE - 1; i++)
1205 freeTokens[i].next = &freeTokens[i + 1];
1206 freeTokens[i].next = NULL;
1208 t = freeTokens;
1209 freeTokens = t->next;
1210 t->next = next;
1211 t->a.mac = NULL;
1212 t->type = type;
1213 if (type == TOK_WHITESPACE || !text) {
1214 t->text = NULL;
1215 } else {
1216 if (txtlen == 0)
1217 txtlen = strlen(text);
1218 t->text = nasm_malloc(txtlen+1);
1219 memcpy(t->text, text, txtlen);
1220 t->text[txtlen] = '\0';
1222 return t;
1225 static Token *delete_Token(Token * t)
1227 Token *next = t->next;
1228 nasm_free(t->text);
1229 t->next = freeTokens;
1230 freeTokens = t;
1231 return next;
1235 * Convert a line of tokens back into text.
1236 * If expand_locals is not zero, identifiers of the form "%$*xxx"
1237 * will be transformed into ..@ctxnum.xxx
1239 static char *detoken(Token * tlist, bool expand_locals)
1241 Token *t;
1242 char *line, *p;
1243 const char *q;
1244 int len = 0;
1246 list_for_each(t, tlist) {
1247 if (t->type == TOK_PREPROC_ID && t->text[1] == '!') {
1248 char *v;
1249 char *q = t->text;
1251 v = t->text + 2;
1252 if (*v == '\'' || *v == '\"' || *v == '`') {
1253 size_t len = nasm_unquote(v, NULL);
1254 size_t clen = strlen(v);
1256 if (len != clen) {
1257 nasm_error(ERR_NONFATAL | ERR_PASS1,
1258 "NUL character in %%! string");
1259 v = NULL;
1263 if (v) {
1264 char *p = getenv(v);
1265 if (!p) {
1266 nasm_error(ERR_NONFATAL | ERR_PASS1,
1267 "nonexistent environment variable `%s'", v);
1268 p = "";
1270 t->text = nasm_strdup(p);
1272 nasm_free(q);
1275 /* Expand local macros here and not during preprocessing */
1276 if (expand_locals &&
1277 t->type == TOK_PREPROC_ID && t->text &&
1278 t->text[0] == '%' && t->text[1] == '$') {
1279 const char *q;
1280 char *p;
1281 Context *ctx = get_ctx(t->text, &q);
1282 if (ctx) {
1283 char buffer[40];
1284 snprintf(buffer, sizeof(buffer), "..@%"PRIu32".", ctx->number);
1285 p = nasm_strcat(buffer, q);
1286 nasm_free(t->text);
1287 t->text = p;
1290 if (t->type == TOK_WHITESPACE)
1291 len++;
1292 else if (t->text)
1293 len += strlen(t->text);
1296 p = line = nasm_malloc(len + 1);
1298 list_for_each(t, tlist) {
1299 if (t->type == TOK_WHITESPACE) {
1300 *p++ = ' ';
1301 } else if (t->text) {
1302 q = t->text;
1303 while (*q)
1304 *p++ = *q++;
1307 *p = '\0';
1309 return line;
1313 * A scanner, suitable for use by the expression evaluator, which
1314 * operates on a line of Tokens. Expects a pointer to a pointer to
1315 * the first token in the line to be passed in as its private_data
1316 * field.
1318 * FIX: This really needs to be unified with stdscan.
1320 static int ppscan(void *private_data, struct tokenval *tokval)
1322 Token **tlineptr = private_data;
1323 Token *tline;
1324 char ourcopy[MAX_KEYWORD+1], *p, *r, *s;
1326 do {
1327 tline = *tlineptr;
1328 *tlineptr = tline ? tline->next : NULL;
1329 } while (tline && (tline->type == TOK_WHITESPACE ||
1330 tline->type == TOK_COMMENT));
1332 if (!tline)
1333 return tokval->t_type = TOKEN_EOS;
1335 tokval->t_charptr = tline->text;
1337 if (tline->text[0] == '$' && !tline->text[1])
1338 return tokval->t_type = TOKEN_HERE;
1339 if (tline->text[0] == '$' && tline->text[1] == '$' && !tline->text[2])
1340 return tokval->t_type = TOKEN_BASE;
1342 if (tline->type == TOK_ID) {
1343 p = tokval->t_charptr = tline->text;
1344 if (p[0] == '$') {
1345 tokval->t_charptr++;
1346 return tokval->t_type = TOKEN_ID;
1349 for (r = p, s = ourcopy; *r; r++) {
1350 if (r >= p+MAX_KEYWORD)
1351 return tokval->t_type = TOKEN_ID; /* Not a keyword */
1352 *s++ = nasm_tolower(*r);
1354 *s = '\0';
1355 /* right, so we have an identifier sitting in temp storage. now,
1356 * is it actually a register or instruction name, or what? */
1357 return nasm_token_hash(ourcopy, tokval);
1360 if (tline->type == TOK_NUMBER) {
1361 bool rn_error;
1362 tokval->t_integer = readnum(tline->text, &rn_error);
1363 tokval->t_charptr = tline->text;
1364 if (rn_error)
1365 return tokval->t_type = TOKEN_ERRNUM;
1366 else
1367 return tokval->t_type = TOKEN_NUM;
1370 if (tline->type == TOK_FLOAT) {
1371 return tokval->t_type = TOKEN_FLOAT;
1374 if (tline->type == TOK_STRING) {
1375 char bq, *ep;
1377 bq = tline->text[0];
1378 tokval->t_charptr = tline->text;
1379 tokval->t_inttwo = nasm_unquote(tline->text, &ep);
1381 if (ep[0] != bq || ep[1] != '\0')
1382 return tokval->t_type = TOKEN_ERRSTR;
1383 else
1384 return tokval->t_type = TOKEN_STR;
1387 if (tline->type == TOK_OTHER) {
1388 if (!strcmp(tline->text, "<<"))
1389 return tokval->t_type = TOKEN_SHL;
1390 if (!strcmp(tline->text, ">>"))
1391 return tokval->t_type = TOKEN_SHR;
1392 if (!strcmp(tline->text, "//"))
1393 return tokval->t_type = TOKEN_SDIV;
1394 if (!strcmp(tline->text, "%%"))
1395 return tokval->t_type = TOKEN_SMOD;
1396 if (!strcmp(tline->text, "=="))
1397 return tokval->t_type = TOKEN_EQ;
1398 if (!strcmp(tline->text, "<>"))
1399 return tokval->t_type = TOKEN_NE;
1400 if (!strcmp(tline->text, "!="))
1401 return tokval->t_type = TOKEN_NE;
1402 if (!strcmp(tline->text, "<="))
1403 return tokval->t_type = TOKEN_LE;
1404 if (!strcmp(tline->text, ">="))
1405 return tokval->t_type = TOKEN_GE;
1406 if (!strcmp(tline->text, "&&"))
1407 return tokval->t_type = TOKEN_DBL_AND;
1408 if (!strcmp(tline->text, "^^"))
1409 return tokval->t_type = TOKEN_DBL_XOR;
1410 if (!strcmp(tline->text, "||"))
1411 return tokval->t_type = TOKEN_DBL_OR;
1415 * We have no other options: just return the first character of
1416 * the token text.
1418 return tokval->t_type = tline->text[0];
1422 * Compare a string to the name of an existing macro; this is a
1423 * simple wrapper which calls either strcmp or nasm_stricmp
1424 * depending on the value of the `casesense' parameter.
1426 static int mstrcmp(const char *p, const char *q, bool casesense)
1428 return casesense ? strcmp(p, q) : nasm_stricmp(p, q);
1432 * Compare a string to the name of an existing macro; this is a
1433 * simple wrapper which calls either strcmp or nasm_stricmp
1434 * depending on the value of the `casesense' parameter.
1436 static int mmemcmp(const char *p, const char *q, size_t l, bool casesense)
1438 return casesense ? memcmp(p, q, l) : nasm_memicmp(p, q, l);
1442 * Return the Context structure associated with a %$ token. Return
1443 * NULL, having _already_ reported an error condition, if the
1444 * context stack isn't deep enough for the supplied number of $
1445 * signs.
1447 * If "namep" is non-NULL, set it to the pointer to the macro name
1448 * tail, i.e. the part beyond %$...
1450 static Context *get_ctx(const char *name, const char **namep)
1452 Context *ctx;
1453 int i;
1455 if (namep)
1456 *namep = name;
1458 if (!name || name[0] != '%' || name[1] != '$')
1459 return NULL;
1461 if (!cstk) {
1462 nasm_error(ERR_NONFATAL, "`%s': context stack is empty", name);
1463 return NULL;
1466 name += 2;
1467 ctx = cstk;
1468 i = 0;
1469 while (ctx && *name == '$') {
1470 name++;
1471 i++;
1472 ctx = ctx->next;
1474 if (!ctx) {
1475 nasm_error(ERR_NONFATAL, "`%s': context stack is only"
1476 " %d level%s deep", name, i, (i == 1 ? "" : "s"));
1477 return NULL;
1480 if (namep)
1481 *namep = name;
1483 return ctx;
1487 * Check to see if a file is already in a string list
1489 static bool in_list(const StrList *list, const char *str)
1491 while (list) {
1492 if (!strcmp(list->str, str))
1493 return true;
1494 list = list->next;
1496 return false;
1500 * Open an include file. This routine must always return a valid
1501 * file pointer if it returns - it's responsible for throwing an
1502 * ERR_FATAL and bombing out completely if not. It should also try
1503 * the include path one by one until it finds the file or reaches
1504 * the end of the path.
1506 static FILE *inc_fopen(const char *file, StrList **dhead, StrList ***dtail,
1507 bool missing_ok)
1509 FILE *fp;
1510 char *prefix = "";
1511 IncPath *ip = ipath;
1512 int len = strlen(file);
1513 size_t prefix_len = 0;
1514 StrList *sl;
1516 while (1) {
1517 sl = nasm_malloc(prefix_len+len+1+sizeof sl->next);
1518 memcpy(sl->str, prefix, prefix_len);
1519 memcpy(sl->str+prefix_len, file, len+1);
1520 fp = fopen(sl->str, "r");
1521 if (fp && dhead && !in_list(*dhead, sl->str)) {
1522 sl->next = NULL;
1523 **dtail = sl;
1524 *dtail = &sl->next;
1525 } else {
1526 nasm_free(sl);
1528 if (fp)
1529 return fp;
1530 if (!ip) {
1531 if (!missing_ok)
1532 break;
1533 prefix = NULL;
1534 } else {
1535 prefix = ip->path;
1536 ip = ip->next;
1538 if (prefix) {
1539 prefix_len = strlen(prefix);
1540 } else {
1541 /* -MG given and file not found */
1542 if (dhead && !in_list(*dhead, file)) {
1543 sl = nasm_malloc(len+1+sizeof sl->next);
1544 sl->next = NULL;
1545 strcpy(sl->str, file);
1546 **dtail = sl;
1547 *dtail = &sl->next;
1549 return NULL;
1553 nasm_error(ERR_FATAL, "unable to open include file `%s'", file);
1554 return NULL;
1558 * Determine if we should warn on defining a single-line macro of
1559 * name `name', with `nparam' parameters. If nparam is 0 or -1, will
1560 * return true if _any_ single-line macro of that name is defined.
1561 * Otherwise, will return true if a single-line macro with either
1562 * `nparam' or no parameters is defined.
1564 * If a macro with precisely the right number of parameters is
1565 * defined, or nparam is -1, the address of the definition structure
1566 * will be returned in `defn'; otherwise NULL will be returned. If `defn'
1567 * is NULL, no action will be taken regarding its contents, and no
1568 * error will occur.
1570 * Note that this is also called with nparam zero to resolve
1571 * `ifdef'.
1573 * If you already know which context macro belongs to, you can pass
1574 * the context pointer as first parameter; if you won't but name begins
1575 * with %$ the context will be automatically computed. If all_contexts
1576 * is true, macro will be searched in outer contexts as well.
1578 static bool
1579 smacro_defined(Context * ctx, const char *name, int nparam, SMacro ** defn,
1580 bool nocase)
1582 struct hash_table *smtbl;
1583 SMacro *m;
1585 if (ctx) {
1586 smtbl = &ctx->localmac;
1587 } else if (name[0] == '%' && name[1] == '$') {
1588 if (cstk)
1589 ctx = get_ctx(name, &name);
1590 if (!ctx)
1591 return false; /* got to return _something_ */
1592 smtbl = &ctx->localmac;
1593 } else {
1594 smtbl = &smacros;
1596 m = (SMacro *) hash_findix(smtbl, name);
1598 while (m) {
1599 if (!mstrcmp(m->name, name, m->casesense && nocase) &&
1600 (nparam <= 0 || m->nparam == 0 || nparam == (int) m->nparam)) {
1601 if (defn) {
1602 if (nparam == (int) m->nparam || nparam == -1)
1603 *defn = m;
1604 else
1605 *defn = NULL;
1607 return true;
1609 m = m->next;
1612 return false;
1616 * Count and mark off the parameters in a multi-line macro call.
1617 * This is called both from within the multi-line macro expansion
1618 * code, and also to mark off the default parameters when provided
1619 * in a %macro definition line.
1621 static void count_mmac_params(Token * t, int *nparam, Token *** params)
1623 int paramsize, brace;
1625 *nparam = paramsize = 0;
1626 *params = NULL;
1627 while (t) {
1628 /* +1: we need space for the final NULL */
1629 if (*nparam+1 >= paramsize) {
1630 paramsize += PARAM_DELTA;
1631 *params = nasm_realloc(*params, sizeof(**params) * paramsize);
1633 skip_white_(t);
1634 brace = 0;
1635 if (tok_is_(t, "{"))
1636 brace++;
1637 (*params)[(*nparam)++] = t;
1638 if (brace) {
1639 while (brace && (t = t->next) != NULL) {
1640 if (tok_is_(t, "{"))
1641 brace++;
1642 else if (tok_is_(t, "}"))
1643 brace--;
1646 if (t) {
1648 * Now we've found the closing brace, look further
1649 * for the comma.
1651 t = t->next;
1652 skip_white_(t);
1653 if (tok_isnt_(t, ",")) {
1654 nasm_error(ERR_NONFATAL,
1655 "braces do not enclose all of macro parameter");
1656 while (tok_isnt_(t, ","))
1657 t = t->next;
1660 } else {
1661 while (tok_isnt_(t, ","))
1662 t = t->next;
1664 if (t) { /* got a comma/brace */
1665 t = t->next; /* eat the comma */
1671 * Determine whether one of the various `if' conditions is true or
1672 * not.
1674 * We must free the tline we get passed.
1676 static bool if_condition(Token * tline, enum preproc_token ct)
1678 enum pp_conditional i = PP_COND(ct);
1679 bool j;
1680 Token *t, *tt, **tptr, *origline;
1681 struct tokenval tokval;
1682 expr *evalresult;
1683 enum pp_token_type needtype;
1684 char *p;
1686 origline = tline;
1688 switch (i) {
1689 case PPC_IFCTX:
1690 j = false; /* have we matched yet? */
1691 while (true) {
1692 skip_white_(tline);
1693 if (!tline)
1694 break;
1695 if (tline->type != TOK_ID) {
1696 nasm_error(ERR_NONFATAL,
1697 "`%s' expects context identifiers", pp_directives[ct]);
1698 free_tlist(origline);
1699 return -1;
1701 if (cstk && cstk->name && !nasm_stricmp(tline->text, cstk->name))
1702 j = true;
1703 tline = tline->next;
1705 break;
1707 case PPC_IFDEF:
1708 j = false; /* have we matched yet? */
1709 while (tline) {
1710 skip_white_(tline);
1711 if (!tline || (tline->type != TOK_ID &&
1712 (tline->type != TOK_PREPROC_ID ||
1713 tline->text[1] != '$'))) {
1714 nasm_error(ERR_NONFATAL,
1715 "`%s' expects macro identifiers", pp_directives[ct]);
1716 goto fail;
1718 if (smacro_defined(NULL, tline->text, 0, NULL, true))
1719 j = true;
1720 tline = tline->next;
1722 break;
1724 case PPC_IFENV:
1725 tline = expand_smacro(tline);
1726 j = false; /* have we matched yet? */
1727 while (tline) {
1728 skip_white_(tline);
1729 if (!tline || (tline->type != TOK_ID &&
1730 tline->type != TOK_STRING &&
1731 (tline->type != TOK_PREPROC_ID ||
1732 tline->text[1] != '!'))) {
1733 nasm_error(ERR_NONFATAL,
1734 "`%s' expects environment variable names",
1735 pp_directives[ct]);
1736 goto fail;
1738 p = tline->text;
1739 if (tline->type == TOK_PREPROC_ID)
1740 p += 2; /* Skip leading %! */
1741 if (*p == '\'' || *p == '\"' || *p == '`')
1742 nasm_unquote_cstr(p, ct);
1743 if (getenv(p))
1744 j = true;
1745 tline = tline->next;
1747 break;
1749 case PPC_IFIDN:
1750 case PPC_IFIDNI:
1751 tline = expand_smacro(tline);
1752 t = tt = tline;
1753 while (tok_isnt_(tt, ","))
1754 tt = tt->next;
1755 if (!tt) {
1756 nasm_error(ERR_NONFATAL,
1757 "`%s' expects two comma-separated arguments",
1758 pp_directives[ct]);
1759 goto fail;
1761 tt = tt->next;
1762 j = true; /* assume equality unless proved not */
1763 while ((t->type != TOK_OTHER || strcmp(t->text, ",")) && tt) {
1764 if (tt->type == TOK_OTHER && !strcmp(tt->text, ",")) {
1765 nasm_error(ERR_NONFATAL, "`%s': more than one comma on line",
1766 pp_directives[ct]);
1767 goto fail;
1769 if (t->type == TOK_WHITESPACE) {
1770 t = t->next;
1771 continue;
1773 if (tt->type == TOK_WHITESPACE) {
1774 tt = tt->next;
1775 continue;
1777 if (tt->type != t->type) {
1778 j = false; /* found mismatching tokens */
1779 break;
1781 /* When comparing strings, need to unquote them first */
1782 if (t->type == TOK_STRING) {
1783 size_t l1 = nasm_unquote(t->text, NULL);
1784 size_t l2 = nasm_unquote(tt->text, NULL);
1786 if (l1 != l2) {
1787 j = false;
1788 break;
1790 if (mmemcmp(t->text, tt->text, l1, i == PPC_IFIDN)) {
1791 j = false;
1792 break;
1794 } else if (mstrcmp(tt->text, t->text, i == PPC_IFIDN) != 0) {
1795 j = false; /* found mismatching tokens */
1796 break;
1799 t = t->next;
1800 tt = tt->next;
1802 if ((t->type != TOK_OTHER || strcmp(t->text, ",")) || tt)
1803 j = false; /* trailing gunk on one end or other */
1804 break;
1806 case PPC_IFMACRO:
1808 bool found = false;
1809 MMacro searching, *mmac;
1811 skip_white_(tline);
1812 tline = expand_id(tline);
1813 if (!tok_type_(tline, TOK_ID)) {
1814 nasm_error(ERR_NONFATAL,
1815 "`%s' expects a macro name", pp_directives[ct]);
1816 goto fail;
1818 searching.name = nasm_strdup(tline->text);
1819 searching.casesense = true;
1820 searching.plus = false;
1821 searching.nolist = false;
1822 searching.in_progress = 0;
1823 searching.max_depth = 0;
1824 searching.rep_nest = NULL;
1825 searching.nparam_min = 0;
1826 searching.nparam_max = INT_MAX;
1827 tline = expand_smacro(tline->next);
1828 skip_white_(tline);
1829 if (!tline) {
1830 } else if (!tok_type_(tline, TOK_NUMBER)) {
1831 nasm_error(ERR_NONFATAL,
1832 "`%s' expects a parameter count or nothing",
1833 pp_directives[ct]);
1834 } else {
1835 searching.nparam_min = searching.nparam_max =
1836 readnum(tline->text, &j);
1837 if (j)
1838 nasm_error(ERR_NONFATAL,
1839 "unable to parse parameter count `%s'",
1840 tline->text);
1842 if (tline && tok_is_(tline->next, "-")) {
1843 tline = tline->next->next;
1844 if (tok_is_(tline, "*"))
1845 searching.nparam_max = INT_MAX;
1846 else if (!tok_type_(tline, TOK_NUMBER))
1847 nasm_error(ERR_NONFATAL,
1848 "`%s' expects a parameter count after `-'",
1849 pp_directives[ct]);
1850 else {
1851 searching.nparam_max = readnum(tline->text, &j);
1852 if (j)
1853 nasm_error(ERR_NONFATAL,
1854 "unable to parse parameter count `%s'",
1855 tline->text);
1856 if (searching.nparam_min > searching.nparam_max)
1857 nasm_error(ERR_NONFATAL,
1858 "minimum parameter count exceeds maximum");
1861 if (tline && tok_is_(tline->next, "+")) {
1862 tline = tline->next;
1863 searching.plus = true;
1865 mmac = (MMacro *) hash_findix(&mmacros, searching.name);
1866 while (mmac) {
1867 if (!strcmp(mmac->name, searching.name) &&
1868 (mmac->nparam_min <= searching.nparam_max
1869 || searching.plus)
1870 && (searching.nparam_min <= mmac->nparam_max
1871 || mmac->plus)) {
1872 found = true;
1873 break;
1875 mmac = mmac->next;
1877 if (tline && tline->next)
1878 nasm_error(ERR_WARNING|ERR_PASS1,
1879 "trailing garbage after %%ifmacro ignored");
1880 nasm_free(searching.name);
1881 j = found;
1882 break;
1885 case PPC_IFID:
1886 needtype = TOK_ID;
1887 goto iftype;
1888 case PPC_IFNUM:
1889 needtype = TOK_NUMBER;
1890 goto iftype;
1891 case PPC_IFSTR:
1892 needtype = TOK_STRING;
1893 goto iftype;
1895 iftype:
1896 t = tline = expand_smacro(tline);
1898 while (tok_type_(t, TOK_WHITESPACE) ||
1899 (needtype == TOK_NUMBER &&
1900 tok_type_(t, TOK_OTHER) &&
1901 (t->text[0] == '-' || t->text[0] == '+') &&
1902 !t->text[1]))
1903 t = t->next;
1905 j = tok_type_(t, needtype);
1906 break;
1908 case PPC_IFTOKEN:
1909 t = tline = expand_smacro(tline);
1910 while (tok_type_(t, TOK_WHITESPACE))
1911 t = t->next;
1913 j = false;
1914 if (t) {
1915 t = t->next; /* Skip the actual token */
1916 while (tok_type_(t, TOK_WHITESPACE))
1917 t = t->next;
1918 j = !t; /* Should be nothing left */
1920 break;
1922 case PPC_IFEMPTY:
1923 t = tline = expand_smacro(tline);
1924 while (tok_type_(t, TOK_WHITESPACE))
1925 t = t->next;
1927 j = !t; /* Should be empty */
1928 break;
1930 case PPC_IF:
1931 t = tline = expand_smacro(tline);
1932 tptr = &t;
1933 tokval.t_type = TOKEN_INVALID;
1934 evalresult = evaluate(ppscan, tptr, &tokval,
1935 NULL, pass | CRITICAL, NULL);
1936 if (!evalresult)
1937 return -1;
1938 if (tokval.t_type)
1939 nasm_error(ERR_WARNING|ERR_PASS1,
1940 "trailing garbage after expression ignored");
1941 if (!is_simple(evalresult)) {
1942 nasm_error(ERR_NONFATAL,
1943 "non-constant value given to `%s'", pp_directives[ct]);
1944 goto fail;
1946 j = reloc_value(evalresult) != 0;
1947 break;
1949 default:
1950 nasm_error(ERR_FATAL,
1951 "preprocessor directive `%s' not yet implemented",
1952 pp_directives[ct]);
1953 goto fail;
1956 free_tlist(origline);
1957 return j ^ PP_NEGATIVE(ct);
1959 fail:
1960 free_tlist(origline);
1961 return -1;
1965 * Common code for defining an smacro
1967 static bool define_smacro(Context *ctx, const char *mname, bool casesense,
1968 int nparam, Token *expansion)
1970 SMacro *smac, **smhead;
1971 struct hash_table *smtbl;
1973 if (smacro_defined(ctx, mname, nparam, &smac, casesense)) {
1974 if (!smac) {
1975 nasm_error(ERR_WARNING|ERR_PASS1,
1976 "single-line macro `%s' defined both with and"
1977 " without parameters", mname);
1979 * Some instances of the old code considered this a failure,
1980 * some others didn't. What is the right thing to do here?
1982 free_tlist(expansion);
1983 return false; /* Failure */
1984 } else {
1986 * We're redefining, so we have to take over an
1987 * existing SMacro structure. This means freeing
1988 * what was already in it.
1990 nasm_free(smac->name);
1991 free_tlist(smac->expansion);
1993 } else {
1994 smtbl = ctx ? &ctx->localmac : &smacros;
1995 smhead = (SMacro **) hash_findi_add(smtbl, mname);
1996 smac = nasm_malloc(sizeof(SMacro));
1997 smac->next = *smhead;
1998 *smhead = smac;
2000 smac->name = nasm_strdup(mname);
2001 smac->casesense = casesense;
2002 smac->nparam = nparam;
2003 smac->expansion = expansion;
2004 smac->in_progress = false;
2005 return true; /* Success */
2009 * Undefine an smacro
2011 static void undef_smacro(Context *ctx, const char *mname)
2013 SMacro **smhead, *s, **sp;
2014 struct hash_table *smtbl;
2016 smtbl = ctx ? &ctx->localmac : &smacros;
2017 smhead = (SMacro **)hash_findi(smtbl, mname, NULL);
2019 if (smhead) {
2021 * We now have a macro name... go hunt for it.
2023 sp = smhead;
2024 while ((s = *sp) != NULL) {
2025 if (!mstrcmp(s->name, mname, s->casesense)) {
2026 *sp = s->next;
2027 nasm_free(s->name);
2028 free_tlist(s->expansion);
2029 nasm_free(s);
2030 } else {
2031 sp = &s->next;
2038 * Parse a mmacro specification.
2040 static bool parse_mmacro_spec(Token *tline, MMacro *def, const char *directive)
2042 bool err;
2044 tline = tline->next;
2045 skip_white_(tline);
2046 tline = expand_id(tline);
2047 if (!tok_type_(tline, TOK_ID)) {
2048 nasm_error(ERR_NONFATAL, "`%s' expects a macro name", directive);
2049 return false;
2052 def->prev = NULL;
2053 def->name = nasm_strdup(tline->text);
2054 def->plus = false;
2055 def->nolist = false;
2056 def->in_progress = 0;
2057 def->rep_nest = NULL;
2058 def->nparam_min = 0;
2059 def->nparam_max = 0;
2061 tline = expand_smacro(tline->next);
2062 skip_white_(tline);
2063 if (!tok_type_(tline, TOK_NUMBER)) {
2064 nasm_error(ERR_NONFATAL, "`%s' expects a parameter count", directive);
2065 } else {
2066 def->nparam_min = def->nparam_max =
2067 readnum(tline->text, &err);
2068 if (err)
2069 nasm_error(ERR_NONFATAL,
2070 "unable to parse parameter count `%s'", tline->text);
2072 if (tline && tok_is_(tline->next, "-")) {
2073 tline = tline->next->next;
2074 if (tok_is_(tline, "*")) {
2075 def->nparam_max = INT_MAX;
2076 } else if (!tok_type_(tline, TOK_NUMBER)) {
2077 nasm_error(ERR_NONFATAL,
2078 "`%s' expects a parameter count after `-'", directive);
2079 } else {
2080 def->nparam_max = readnum(tline->text, &err);
2081 if (err) {
2082 nasm_error(ERR_NONFATAL, "unable to parse parameter count `%s'",
2083 tline->text);
2085 if (def->nparam_min > def->nparam_max) {
2086 nasm_error(ERR_NONFATAL, "minimum parameter count exceeds maximum");
2090 if (tline && tok_is_(tline->next, "+")) {
2091 tline = tline->next;
2092 def->plus = true;
2094 if (tline && tok_type_(tline->next, TOK_ID) &&
2095 !nasm_stricmp(tline->next->text, ".nolist")) {
2096 tline = tline->next;
2097 def->nolist = true;
2101 * Handle default parameters.
2103 if (tline && tline->next) {
2104 def->dlist = tline->next;
2105 tline->next = NULL;
2106 count_mmac_params(def->dlist, &def->ndefs, &def->defaults);
2107 } else {
2108 def->dlist = NULL;
2109 def->defaults = NULL;
2111 def->expansion = NULL;
2113 if (def->defaults && def->ndefs > def->nparam_max - def->nparam_min &&
2114 !def->plus)
2115 nasm_error(ERR_WARNING|ERR_PASS1|ERR_WARN_MDP,
2116 "too many default macro parameters");
2118 return true;
2123 * Decode a size directive
2125 static int parse_size(const char *str) {
2126 static const char *size_names[] =
2127 { "byte", "dword", "oword", "qword", "tword", "word", "yword" };
2128 static const int sizes[] =
2129 { 0, 1, 4, 16, 8, 10, 2, 32 };
2131 return sizes[bsii(str, size_names, ARRAY_SIZE(size_names))+1];
2135 * find and process preprocessor directive in passed line
2136 * Find out if a line contains a preprocessor directive, and deal
2137 * with it if so.
2139 * If a directive _is_ found, it is the responsibility of this routine
2140 * (and not the caller) to free_tlist() the line.
2142 * @param tline a pointer to the current tokeninzed line linked list
2143 * @return DIRECTIVE_FOUND or NO_DIRECTIVE_FOUND
2146 static int do_directive(Token * tline)
2148 enum preproc_token i;
2149 int j;
2150 bool err;
2151 int nparam;
2152 bool nolist;
2153 bool casesense;
2154 int k, m;
2155 int offset;
2156 char *p, *pp;
2157 const char *mname;
2158 Include *inc;
2159 Context *ctx;
2160 Cond *cond;
2161 MMacro *mmac, **mmhead;
2162 Token *t = NULL, *tt, *param_start, *macro_start, *last, **tptr, *origline;
2163 Line *l;
2164 struct tokenval tokval;
2165 expr *evalresult;
2166 MMacro *tmp_defining; /* Used when manipulating rep_nest */
2167 int64_t count;
2168 size_t len;
2169 int severity;
2171 origline = tline;
2173 skip_white_(tline);
2174 if (!tline || !tok_type_(tline, TOK_PREPROC_ID) ||
2175 (tline->text[1] == '%' || tline->text[1] == '$'
2176 || tline->text[1] == '!'))
2177 return NO_DIRECTIVE_FOUND;
2179 i = pp_token_hash(tline->text);
2182 * FIXME: We zap execution of PP_RMACRO, PP_IRMACRO, PP_EXITMACRO
2183 * since they are known to be buggy at moment, we need to fix them
2184 * in future release (2.09-2.10)
2186 if (i == PP_RMACRO || i == PP_IRMACRO || i == PP_EXITMACRO) {
2187 nasm_error(ERR_NONFATAL, "unknown preprocessor directive `%s'",
2188 tline->text);
2189 return NO_DIRECTIVE_FOUND;
2193 * If we're in a non-emitting branch of a condition construct,
2194 * or walking to the end of an already terminated %rep block,
2195 * we should ignore all directives except for condition
2196 * directives.
2198 if (((istk->conds && !emitting(istk->conds->state)) ||
2199 (istk->mstk && !istk->mstk->in_progress)) && !is_condition(i)) {
2200 return NO_DIRECTIVE_FOUND;
2204 * If we're defining a macro or reading a %rep block, we should
2205 * ignore all directives except for %macro/%imacro (which nest),
2206 * %endm/%endmacro, and (only if we're in a %rep block) %endrep.
2207 * If we're in a %rep block, another %rep nests, so should be let through.
2209 if (defining && i != PP_MACRO && i != PP_IMACRO &&
2210 i != PP_RMACRO && i != PP_IRMACRO &&
2211 i != PP_ENDMACRO && i != PP_ENDM &&
2212 (defining->name || (i != PP_ENDREP && i != PP_REP))) {
2213 return NO_DIRECTIVE_FOUND;
2216 if (defining) {
2217 if (i == PP_MACRO || i == PP_IMACRO ||
2218 i == PP_RMACRO || i == PP_IRMACRO) {
2219 nested_mac_count++;
2220 return NO_DIRECTIVE_FOUND;
2221 } else if (nested_mac_count > 0) {
2222 if (i == PP_ENDMACRO) {
2223 nested_mac_count--;
2224 return NO_DIRECTIVE_FOUND;
2227 if (!defining->name) {
2228 if (i == PP_REP) {
2229 nested_rep_count++;
2230 return NO_DIRECTIVE_FOUND;
2231 } else if (nested_rep_count > 0) {
2232 if (i == PP_ENDREP) {
2233 nested_rep_count--;
2234 return NO_DIRECTIVE_FOUND;
2240 switch (i) {
2241 case PP_INVALID:
2242 nasm_error(ERR_NONFATAL, "unknown preprocessor directive `%s'",
2243 tline->text);
2244 return NO_DIRECTIVE_FOUND; /* didn't get it */
2246 case PP_STACKSIZE:
2247 /* Directive to tell NASM what the default stack size is. The
2248 * default is for a 16-bit stack, and this can be overriden with
2249 * %stacksize large.
2251 tline = tline->next;
2252 if (tline && tline->type == TOK_WHITESPACE)
2253 tline = tline->next;
2254 if (!tline || tline->type != TOK_ID) {
2255 nasm_error(ERR_NONFATAL, "`%%stacksize' missing size parameter");
2256 free_tlist(origline);
2257 return DIRECTIVE_FOUND;
2259 if (nasm_stricmp(tline->text, "flat") == 0) {
2260 /* All subsequent ARG directives are for a 32-bit stack */
2261 StackSize = 4;
2262 StackPointer = "ebp";
2263 ArgOffset = 8;
2264 LocalOffset = 0;
2265 } else if (nasm_stricmp(tline->text, "flat64") == 0) {
2266 /* All subsequent ARG directives are for a 64-bit stack */
2267 StackSize = 8;
2268 StackPointer = "rbp";
2269 ArgOffset = 16;
2270 LocalOffset = 0;
2271 } else if (nasm_stricmp(tline->text, "large") == 0) {
2272 /* All subsequent ARG directives are for a 16-bit stack,
2273 * far function call.
2275 StackSize = 2;
2276 StackPointer = "bp";
2277 ArgOffset = 4;
2278 LocalOffset = 0;
2279 } else if (nasm_stricmp(tline->text, "small") == 0) {
2280 /* All subsequent ARG directives are for a 16-bit stack,
2281 * far function call. We don't support near functions.
2283 StackSize = 2;
2284 StackPointer = "bp";
2285 ArgOffset = 6;
2286 LocalOffset = 0;
2287 } else {
2288 nasm_error(ERR_NONFATAL, "`%%stacksize' invalid size type");
2289 free_tlist(origline);
2290 return DIRECTIVE_FOUND;
2292 free_tlist(origline);
2293 return DIRECTIVE_FOUND;
2295 case PP_ARG:
2296 /* TASM like ARG directive to define arguments to functions, in
2297 * the following form:
2299 * ARG arg1:WORD, arg2:DWORD, arg4:QWORD
2301 offset = ArgOffset;
2302 do {
2303 char *arg, directive[256];
2304 int size = StackSize;
2306 /* Find the argument name */
2307 tline = tline->next;
2308 if (tline && tline->type == TOK_WHITESPACE)
2309 tline = tline->next;
2310 if (!tline || tline->type != TOK_ID) {
2311 nasm_error(ERR_NONFATAL, "`%%arg' missing argument parameter");
2312 free_tlist(origline);
2313 return DIRECTIVE_FOUND;
2315 arg = tline->text;
2317 /* Find the argument size type */
2318 tline = tline->next;
2319 if (!tline || tline->type != TOK_OTHER
2320 || tline->text[0] != ':') {
2321 nasm_error(ERR_NONFATAL,
2322 "Syntax error processing `%%arg' directive");
2323 free_tlist(origline);
2324 return DIRECTIVE_FOUND;
2326 tline = tline->next;
2327 if (!tline || tline->type != TOK_ID) {
2328 nasm_error(ERR_NONFATAL, "`%%arg' missing size type parameter");
2329 free_tlist(origline);
2330 return DIRECTIVE_FOUND;
2333 /* Allow macro expansion of type parameter */
2334 tt = tokenize(tline->text);
2335 tt = expand_smacro(tt);
2336 size = parse_size(tt->text);
2337 if (!size) {
2338 nasm_error(ERR_NONFATAL,
2339 "Invalid size type for `%%arg' missing directive");
2340 free_tlist(tt);
2341 free_tlist(origline);
2342 return DIRECTIVE_FOUND;
2344 free_tlist(tt);
2346 /* Round up to even stack slots */
2347 size = ALIGN(size, StackSize);
2349 /* Now define the macro for the argument */
2350 snprintf(directive, sizeof(directive), "%%define %s (%s+%d)",
2351 arg, StackPointer, offset);
2352 do_directive(tokenize(directive));
2353 offset += size;
2355 /* Move to the next argument in the list */
2356 tline = tline->next;
2357 if (tline && tline->type == TOK_WHITESPACE)
2358 tline = tline->next;
2359 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2360 ArgOffset = offset;
2361 free_tlist(origline);
2362 return DIRECTIVE_FOUND;
2364 case PP_LOCAL:
2365 /* TASM like LOCAL directive to define local variables for a
2366 * function, in the following form:
2368 * LOCAL local1:WORD, local2:DWORD, local4:QWORD = LocalSize
2370 * The '= LocalSize' at the end is ignored by NASM, but is
2371 * required by TASM to define the local parameter size (and used
2372 * by the TASM macro package).
2374 offset = LocalOffset;
2375 do {
2376 char *local, directive[256];
2377 int size = StackSize;
2379 /* Find the argument name */
2380 tline = tline->next;
2381 if (tline && tline->type == TOK_WHITESPACE)
2382 tline = tline->next;
2383 if (!tline || tline->type != TOK_ID) {
2384 nasm_error(ERR_NONFATAL,
2385 "`%%local' missing argument parameter");
2386 free_tlist(origline);
2387 return DIRECTIVE_FOUND;
2389 local = tline->text;
2391 /* Find the argument size type */
2392 tline = tline->next;
2393 if (!tline || tline->type != TOK_OTHER
2394 || tline->text[0] != ':') {
2395 nasm_error(ERR_NONFATAL,
2396 "Syntax error processing `%%local' directive");
2397 free_tlist(origline);
2398 return DIRECTIVE_FOUND;
2400 tline = tline->next;
2401 if (!tline || tline->type != TOK_ID) {
2402 nasm_error(ERR_NONFATAL,
2403 "`%%local' missing size type parameter");
2404 free_tlist(origline);
2405 return DIRECTIVE_FOUND;
2408 /* Allow macro expansion of type parameter */
2409 tt = tokenize(tline->text);
2410 tt = expand_smacro(tt);
2411 size = parse_size(tt->text);
2412 if (!size) {
2413 nasm_error(ERR_NONFATAL,
2414 "Invalid size type for `%%local' missing directive");
2415 free_tlist(tt);
2416 free_tlist(origline);
2417 return DIRECTIVE_FOUND;
2419 free_tlist(tt);
2421 /* Round up to even stack slots */
2422 size = ALIGN(size, StackSize);
2424 offset += size; /* Negative offset, increment before */
2426 /* Now define the macro for the argument */
2427 snprintf(directive, sizeof(directive), "%%define %s (%s-%d)",
2428 local, StackPointer, offset);
2429 do_directive(tokenize(directive));
2431 /* Now define the assign to setup the enter_c macro correctly */
2432 snprintf(directive, sizeof(directive),
2433 "%%assign %%$localsize %%$localsize+%d", size);
2434 do_directive(tokenize(directive));
2436 /* Move to the next argument in the list */
2437 tline = tline->next;
2438 if (tline && tline->type == TOK_WHITESPACE)
2439 tline = tline->next;
2440 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2441 LocalOffset = offset;
2442 free_tlist(origline);
2443 return DIRECTIVE_FOUND;
2445 case PP_CLEAR:
2446 if (tline->next)
2447 nasm_error(ERR_WARNING|ERR_PASS1,
2448 "trailing garbage after `%%clear' ignored");
2449 free_macros();
2450 init_macros();
2451 free_tlist(origline);
2452 return DIRECTIVE_FOUND;
2454 case PP_DEPEND:
2455 t = tline->next = expand_smacro(tline->next);
2456 skip_white_(t);
2457 if (!t || (t->type != TOK_STRING &&
2458 t->type != TOK_INTERNAL_STRING)) {
2459 nasm_error(ERR_NONFATAL, "`%%depend' expects a file name");
2460 free_tlist(origline);
2461 return DIRECTIVE_FOUND; /* but we did _something_ */
2463 if (t->next)
2464 nasm_error(ERR_WARNING|ERR_PASS1,
2465 "trailing garbage after `%%depend' ignored");
2466 p = t->text;
2467 if (t->type != TOK_INTERNAL_STRING)
2468 nasm_unquote_cstr(p, i);
2469 if (dephead && !in_list(*dephead, p)) {
2470 StrList *sl = nasm_malloc(strlen(p)+1+sizeof sl->next);
2471 sl->next = NULL;
2472 strcpy(sl->str, p);
2473 *deptail = sl;
2474 deptail = &sl->next;
2476 free_tlist(origline);
2477 return DIRECTIVE_FOUND;
2479 case PP_INCLUDE:
2480 t = tline->next = expand_smacro(tline->next);
2481 skip_white_(t);
2483 if (!t || (t->type != TOK_STRING &&
2484 t->type != TOK_INTERNAL_STRING)) {
2485 nasm_error(ERR_NONFATAL, "`%%include' expects a file name");
2486 free_tlist(origline);
2487 return DIRECTIVE_FOUND; /* but we did _something_ */
2489 if (t->next)
2490 nasm_error(ERR_WARNING|ERR_PASS1,
2491 "trailing garbage after `%%include' ignored");
2492 p = t->text;
2493 if (t->type != TOK_INTERNAL_STRING)
2494 nasm_unquote_cstr(p, i);
2495 inc = nasm_malloc(sizeof(Include));
2496 inc->next = istk;
2497 inc->conds = NULL;
2498 inc->fp = inc_fopen(p, dephead, &deptail, pass == 0);
2499 if (!inc->fp) {
2500 /* -MG given but file not found */
2501 nasm_free(inc);
2502 } else {
2503 inc->fname = src_set_fname(nasm_strdup(p));
2504 inc->lineno = src_set_linnum(0);
2505 inc->lineinc = 1;
2506 inc->expansion = NULL;
2507 inc->mstk = NULL;
2508 istk = inc;
2509 nasmlist->uplevel(LIST_INCLUDE);
2511 free_tlist(origline);
2512 return DIRECTIVE_FOUND;
2514 case PP_USE:
2516 static macros_t *use_pkg;
2517 const char *pkg_macro = NULL;
2519 tline = tline->next;
2520 skip_white_(tline);
2521 tline = expand_id(tline);
2523 if (!tline || (tline->type != TOK_STRING &&
2524 tline->type != TOK_INTERNAL_STRING &&
2525 tline->type != TOK_ID)) {
2526 nasm_error(ERR_NONFATAL, "`%%use' expects a package name");
2527 free_tlist(origline);
2528 return DIRECTIVE_FOUND; /* but we did _something_ */
2530 if (tline->next)
2531 nasm_error(ERR_WARNING|ERR_PASS1,
2532 "trailing garbage after `%%use' ignored");
2533 if (tline->type == TOK_STRING)
2534 nasm_unquote_cstr(tline->text, i);
2535 use_pkg = nasm_stdmac_find_package(tline->text);
2536 if (!use_pkg)
2537 nasm_error(ERR_NONFATAL, "unknown `%%use' package: %s", tline->text);
2538 else
2539 pkg_macro = (char *)use_pkg + 1; /* The first string will be <%define>__USE_*__ */
2540 if (use_pkg && ! smacro_defined(NULL, pkg_macro, 0, NULL, true)) {
2541 /* Not already included, go ahead and include it */
2542 stdmacpos = use_pkg;
2544 free_tlist(origline);
2545 return DIRECTIVE_FOUND;
2547 case PP_PUSH:
2548 case PP_REPL:
2549 case PP_POP:
2550 tline = tline->next;
2551 skip_white_(tline);
2552 tline = expand_id(tline);
2553 if (tline) {
2554 if (!tok_type_(tline, TOK_ID)) {
2555 nasm_error(ERR_NONFATAL, "`%s' expects a context identifier",
2556 pp_directives[i]);
2557 free_tlist(origline);
2558 return DIRECTIVE_FOUND; /* but we did _something_ */
2560 if (tline->next)
2561 nasm_error(ERR_WARNING|ERR_PASS1,
2562 "trailing garbage after `%s' ignored",
2563 pp_directives[i]);
2564 p = nasm_strdup(tline->text);
2565 } else {
2566 p = NULL; /* Anonymous */
2569 if (i == PP_PUSH) {
2570 ctx = nasm_malloc(sizeof(Context));
2571 ctx->next = cstk;
2572 hash_init(&ctx->localmac, HASH_SMALL);
2573 ctx->name = p;
2574 ctx->number = unique++;
2575 cstk = ctx;
2576 } else {
2577 /* %pop or %repl */
2578 if (!cstk) {
2579 nasm_error(ERR_NONFATAL, "`%s': context stack is empty",
2580 pp_directives[i]);
2581 } else if (i == PP_POP) {
2582 if (p && (!cstk->name || nasm_stricmp(p, cstk->name)))
2583 nasm_error(ERR_NONFATAL, "`%%pop' in wrong context: %s, "
2584 "expected %s",
2585 cstk->name ? cstk->name : "anonymous", p);
2586 else
2587 ctx_pop();
2588 } else {
2589 /* i == PP_REPL */
2590 nasm_free(cstk->name);
2591 cstk->name = p;
2592 p = NULL;
2594 nasm_free(p);
2596 free_tlist(origline);
2597 return DIRECTIVE_FOUND;
2598 case PP_FATAL:
2599 severity = ERR_FATAL;
2600 goto issue_error;
2601 case PP_ERROR:
2602 severity = ERR_NONFATAL;
2603 goto issue_error;
2604 case PP_WARNING:
2605 severity = ERR_WARNING|ERR_WARN_USER;
2606 goto issue_error;
2608 issue_error:
2610 /* Only error out if this is the final pass */
2611 if (pass != 2 && i != PP_FATAL)
2612 return DIRECTIVE_FOUND;
2614 tline->next = expand_smacro(tline->next);
2615 tline = tline->next;
2616 skip_white_(tline);
2617 t = tline ? tline->next : NULL;
2618 skip_white_(t);
2619 if (tok_type_(tline, TOK_STRING) && !t) {
2620 /* The line contains only a quoted string */
2621 p = tline->text;
2622 nasm_unquote(p, NULL); /* Ignore NUL character truncation */
2623 nasm_error(severity, "%s", p);
2624 } else {
2625 /* Not a quoted string, or more than a quoted string */
2626 p = detoken(tline, false);
2627 nasm_error(severity, "%s", p);
2628 nasm_free(p);
2630 free_tlist(origline);
2631 return DIRECTIVE_FOUND;
2634 CASE_PP_IF:
2635 if (istk->conds && !emitting(istk->conds->state))
2636 j = COND_NEVER;
2637 else {
2638 j = if_condition(tline->next, i);
2639 tline->next = NULL; /* it got freed */
2640 j = j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2642 cond = nasm_malloc(sizeof(Cond));
2643 cond->next = istk->conds;
2644 cond->state = j;
2645 istk->conds = cond;
2646 if(istk->mstk)
2647 istk->mstk->condcnt ++;
2648 free_tlist(origline);
2649 return DIRECTIVE_FOUND;
2651 CASE_PP_ELIF:
2652 if (!istk->conds)
2653 nasm_error(ERR_FATAL, "`%s': no matching `%%if'", pp_directives[i]);
2654 switch(istk->conds->state) {
2655 case COND_IF_TRUE:
2656 istk->conds->state = COND_DONE;
2657 break;
2659 case COND_DONE:
2660 case COND_NEVER:
2661 break;
2663 case COND_ELSE_TRUE:
2664 case COND_ELSE_FALSE:
2665 nasm_error(ERR_WARNING|ERR_PASS1|ERR_PP_PRECOND,
2666 "`%%elif' after `%%else' ignored");
2667 istk->conds->state = COND_NEVER;
2668 break;
2670 case COND_IF_FALSE:
2672 * IMPORTANT: In the case of %if, we will already have
2673 * called expand_mmac_params(); however, if we're
2674 * processing an %elif we must have been in a
2675 * non-emitting mode, which would have inhibited
2676 * the normal invocation of expand_mmac_params().
2677 * Therefore, we have to do it explicitly here.
2679 j = if_condition(expand_mmac_params(tline->next), i);
2680 tline->next = NULL; /* it got freed */
2681 istk->conds->state =
2682 j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2683 break;
2685 free_tlist(origline);
2686 return DIRECTIVE_FOUND;
2688 case PP_ELSE:
2689 if (tline->next)
2690 nasm_error(ERR_WARNING|ERR_PASS1|ERR_PP_PRECOND,
2691 "trailing garbage after `%%else' ignored");
2692 if (!istk->conds)
2693 nasm_fatal(0, "`%%else: no matching `%%if'");
2694 switch(istk->conds->state) {
2695 case COND_IF_TRUE:
2696 case COND_DONE:
2697 istk->conds->state = COND_ELSE_FALSE;
2698 break;
2700 case COND_NEVER:
2701 break;
2703 case COND_IF_FALSE:
2704 istk->conds->state = COND_ELSE_TRUE;
2705 break;
2707 case COND_ELSE_TRUE:
2708 case COND_ELSE_FALSE:
2709 nasm_error(ERR_WARNING|ERR_PASS1|ERR_PP_PRECOND,
2710 "`%%else' after `%%else' ignored.");
2711 istk->conds->state = COND_NEVER;
2712 break;
2714 free_tlist(origline);
2715 return DIRECTIVE_FOUND;
2717 case PP_ENDIF:
2718 if (tline->next)
2719 nasm_error(ERR_WARNING|ERR_PASS1|ERR_PP_PRECOND,
2720 "trailing garbage after `%%endif' ignored");
2721 if (!istk->conds)
2722 nasm_error(ERR_FATAL, "`%%endif': no matching `%%if'");
2723 cond = istk->conds;
2724 istk->conds = cond->next;
2725 nasm_free(cond);
2726 if(istk->mstk)
2727 istk->mstk->condcnt --;
2728 free_tlist(origline);
2729 return DIRECTIVE_FOUND;
2731 case PP_RMACRO:
2732 case PP_IRMACRO:
2733 case PP_MACRO:
2734 case PP_IMACRO:
2735 if (defining) {
2736 nasm_error(ERR_FATAL, "`%s': already defining a macro",
2737 pp_directives[i]);
2738 return DIRECTIVE_FOUND;
2740 defining = nasm_malloc(sizeof(MMacro));
2741 defining->max_depth =
2742 (i == PP_RMACRO) || (i == PP_IRMACRO) ? DEADMAN_LIMIT : 0;
2743 defining->casesense = (i == PP_MACRO) || (i == PP_RMACRO);
2744 if (!parse_mmacro_spec(tline, defining, pp_directives[i])) {
2745 nasm_free(defining);
2746 defining = NULL;
2747 return DIRECTIVE_FOUND;
2750 mmac = (MMacro *) hash_findix(&mmacros, defining->name);
2751 while (mmac) {
2752 if (!strcmp(mmac->name, defining->name) &&
2753 (mmac->nparam_min <= defining->nparam_max
2754 || defining->plus)
2755 && (defining->nparam_min <= mmac->nparam_max
2756 || mmac->plus)) {
2757 nasm_error(ERR_WARNING|ERR_PASS1,
2758 "redefining multi-line macro `%s'", defining->name);
2759 return DIRECTIVE_FOUND;
2761 mmac = mmac->next;
2763 free_tlist(origline);
2764 return DIRECTIVE_FOUND;
2766 case PP_ENDM:
2767 case PP_ENDMACRO:
2768 if (! (defining && defining->name)) {
2769 nasm_error(ERR_NONFATAL, "`%s': not defining a macro", tline->text);
2770 return DIRECTIVE_FOUND;
2772 mmhead = (MMacro **) hash_findi_add(&mmacros, defining->name);
2773 defining->next = *mmhead;
2774 *mmhead = defining;
2775 defining = NULL;
2776 free_tlist(origline);
2777 return DIRECTIVE_FOUND;
2779 case PP_EXITMACRO:
2781 * We must search along istk->expansion until we hit a
2782 * macro-end marker for a macro with a name. Then we
2783 * bypass all lines between exitmacro and endmacro.
2785 list_for_each(l, istk->expansion)
2786 if (l->finishes && l->finishes->name)
2787 break;
2789 if (l) {
2791 * Remove all conditional entries relative to this
2792 * macro invocation. (safe to do in this context)
2794 for ( ; l->finishes->condcnt > 0; l->finishes->condcnt --) {
2795 cond = istk->conds;
2796 istk->conds = cond->next;
2797 nasm_free(cond);
2799 istk->expansion = l;
2800 } else {
2801 nasm_error(ERR_NONFATAL, "`%%exitmacro' not within `%%macro' block");
2803 free_tlist(origline);
2804 return DIRECTIVE_FOUND;
2806 case PP_UNMACRO:
2807 case PP_UNIMACRO:
2809 MMacro **mmac_p;
2810 MMacro spec;
2812 spec.casesense = (i == PP_UNMACRO);
2813 if (!parse_mmacro_spec(tline, &spec, pp_directives[i])) {
2814 return DIRECTIVE_FOUND;
2816 mmac_p = (MMacro **) hash_findi(&mmacros, spec.name, NULL);
2817 while (mmac_p && *mmac_p) {
2818 mmac = *mmac_p;
2819 if (mmac->casesense == spec.casesense &&
2820 !mstrcmp(mmac->name, spec.name, spec.casesense) &&
2821 mmac->nparam_min == spec.nparam_min &&
2822 mmac->nparam_max == spec.nparam_max &&
2823 mmac->plus == spec.plus) {
2824 *mmac_p = mmac->next;
2825 free_mmacro(mmac);
2826 } else {
2827 mmac_p = &mmac->next;
2830 free_tlist(origline);
2831 free_tlist(spec.dlist);
2832 return DIRECTIVE_FOUND;
2835 case PP_ROTATE:
2836 if (tline->next && tline->next->type == TOK_WHITESPACE)
2837 tline = tline->next;
2838 if (!tline->next) {
2839 free_tlist(origline);
2840 nasm_error(ERR_NONFATAL, "`%%rotate' missing rotate count");
2841 return DIRECTIVE_FOUND;
2843 t = expand_smacro(tline->next);
2844 tline->next = NULL;
2845 free_tlist(origline);
2846 tline = t;
2847 tptr = &t;
2848 tokval.t_type = TOKEN_INVALID;
2849 evalresult =
2850 evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
2851 free_tlist(tline);
2852 if (!evalresult)
2853 return DIRECTIVE_FOUND;
2854 if (tokval.t_type)
2855 nasm_error(ERR_WARNING|ERR_PASS1,
2856 "trailing garbage after expression ignored");
2857 if (!is_simple(evalresult)) {
2858 nasm_error(ERR_NONFATAL, "non-constant value given to `%%rotate'");
2859 return DIRECTIVE_FOUND;
2861 mmac = istk->mstk;
2862 while (mmac && !mmac->name) /* avoid mistaking %reps for macros */
2863 mmac = mmac->next_active;
2864 if (!mmac) {
2865 nasm_error(ERR_NONFATAL, "`%%rotate' invoked outside a macro call");
2866 } else if (mmac->nparam == 0) {
2867 nasm_error(ERR_NONFATAL,
2868 "`%%rotate' invoked within macro without parameters");
2869 } else {
2870 int rotate = mmac->rotate + reloc_value(evalresult);
2872 rotate %= (int)mmac->nparam;
2873 if (rotate < 0)
2874 rotate += mmac->nparam;
2876 mmac->rotate = rotate;
2878 return DIRECTIVE_FOUND;
2880 case PP_REP:
2881 nolist = false;
2882 do {
2883 tline = tline->next;
2884 } while (tok_type_(tline, TOK_WHITESPACE));
2886 if (tok_type_(tline, TOK_ID) &&
2887 nasm_stricmp(tline->text, ".nolist") == 0) {
2888 nolist = true;
2889 do {
2890 tline = tline->next;
2891 } while (tok_type_(tline, TOK_WHITESPACE));
2894 if (tline) {
2895 t = expand_smacro(tline);
2896 tptr = &t;
2897 tokval.t_type = TOKEN_INVALID;
2898 evalresult =
2899 evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
2900 if (!evalresult) {
2901 free_tlist(origline);
2902 return DIRECTIVE_FOUND;
2904 if (tokval.t_type)
2905 nasm_error(ERR_WARNING|ERR_PASS1,
2906 "trailing garbage after expression ignored");
2907 if (!is_simple(evalresult)) {
2908 nasm_error(ERR_NONFATAL, "non-constant value given to `%%rep'");
2909 return DIRECTIVE_FOUND;
2911 count = reloc_value(evalresult);
2912 if (count >= REP_LIMIT) {
2913 nasm_error(ERR_NONFATAL, "`%%rep' value exceeds limit");
2914 count = 0;
2915 } else
2916 count++;
2917 } else {
2918 nasm_error(ERR_NONFATAL, "`%%rep' expects a repeat count");
2919 count = 0;
2921 free_tlist(origline);
2923 tmp_defining = defining;
2924 defining = nasm_malloc(sizeof(MMacro));
2925 defining->prev = NULL;
2926 defining->name = NULL; /* flags this macro as a %rep block */
2927 defining->casesense = false;
2928 defining->plus = false;
2929 defining->nolist = nolist;
2930 defining->in_progress = count;
2931 defining->max_depth = 0;
2932 defining->nparam_min = defining->nparam_max = 0;
2933 defining->defaults = NULL;
2934 defining->dlist = NULL;
2935 defining->expansion = NULL;
2936 defining->next_active = istk->mstk;
2937 defining->rep_nest = tmp_defining;
2938 return DIRECTIVE_FOUND;
2940 case PP_ENDREP:
2941 if (!defining || defining->name) {
2942 nasm_error(ERR_NONFATAL, "`%%endrep': no matching `%%rep'");
2943 return DIRECTIVE_FOUND;
2947 * Now we have a "macro" defined - although it has no name
2948 * and we won't be entering it in the hash tables - we must
2949 * push a macro-end marker for it on to istk->expansion.
2950 * After that, it will take care of propagating itself (a
2951 * macro-end marker line for a macro which is really a %rep
2952 * block will cause the macro to be re-expanded, complete
2953 * with another macro-end marker to ensure the process
2954 * continues) until the whole expansion is forcibly removed
2955 * from istk->expansion by a %exitrep.
2957 l = nasm_malloc(sizeof(Line));
2958 l->next = istk->expansion;
2959 l->finishes = defining;
2960 l->first = NULL;
2961 istk->expansion = l;
2963 istk->mstk = defining;
2965 nasmlist->uplevel(defining->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
2966 tmp_defining = defining;
2967 defining = defining->rep_nest;
2968 free_tlist(origline);
2969 return DIRECTIVE_FOUND;
2971 case PP_EXITREP:
2973 * We must search along istk->expansion until we hit a
2974 * macro-end marker for a macro with no name. Then we set
2975 * its `in_progress' flag to 0.
2977 list_for_each(l, istk->expansion)
2978 if (l->finishes && !l->finishes->name)
2979 break;
2981 if (l)
2982 l->finishes->in_progress = 1;
2983 else
2984 nasm_error(ERR_NONFATAL, "`%%exitrep' not within `%%rep' block");
2985 free_tlist(origline);
2986 return DIRECTIVE_FOUND;
2988 case PP_XDEFINE:
2989 case PP_IXDEFINE:
2990 case PP_DEFINE:
2991 case PP_IDEFINE:
2992 casesense = (i == PP_DEFINE || i == PP_XDEFINE);
2994 tline = tline->next;
2995 skip_white_(tline);
2996 tline = expand_id(tline);
2997 if (!tline || (tline->type != TOK_ID &&
2998 (tline->type != TOK_PREPROC_ID ||
2999 tline->text[1] != '$'))) {
3000 nasm_error(ERR_NONFATAL, "`%s' expects a macro identifier",
3001 pp_directives[i]);
3002 free_tlist(origline);
3003 return DIRECTIVE_FOUND;
3006 ctx = get_ctx(tline->text, &mname);
3007 last = tline;
3008 param_start = tline = tline->next;
3009 nparam = 0;
3011 /* Expand the macro definition now for %xdefine and %ixdefine */
3012 if ((i == PP_XDEFINE) || (i == PP_IXDEFINE))
3013 tline = expand_smacro(tline);
3015 if (tok_is_(tline, "(")) {
3017 * This macro has parameters.
3020 tline = tline->next;
3021 while (1) {
3022 skip_white_(tline);
3023 if (!tline) {
3024 nasm_error(ERR_NONFATAL, "parameter identifier expected");
3025 free_tlist(origline);
3026 return DIRECTIVE_FOUND;
3028 if (tline->type != TOK_ID) {
3029 nasm_error(ERR_NONFATAL,
3030 "`%s': parameter identifier expected",
3031 tline->text);
3032 free_tlist(origline);
3033 return DIRECTIVE_FOUND;
3035 tline->type = TOK_SMAC_PARAM + nparam++;
3036 tline = tline->next;
3037 skip_white_(tline);
3038 if (tok_is_(tline, ",")) {
3039 tline = tline->next;
3040 } else {
3041 if (!tok_is_(tline, ")")) {
3042 nasm_error(ERR_NONFATAL,
3043 "`)' expected to terminate macro template");
3044 free_tlist(origline);
3045 return DIRECTIVE_FOUND;
3047 break;
3050 last = tline;
3051 tline = tline->next;
3053 if (tok_type_(tline, TOK_WHITESPACE))
3054 last = tline, tline = tline->next;
3055 macro_start = NULL;
3056 last->next = NULL;
3057 t = tline;
3058 while (t) {
3059 if (t->type == TOK_ID) {
3060 list_for_each(tt, param_start)
3061 if (tt->type >= TOK_SMAC_PARAM &&
3062 !strcmp(tt->text, t->text))
3063 t->type = tt->type;
3065 tt = t->next;
3066 t->next = macro_start;
3067 macro_start = t;
3068 t = tt;
3071 * Good. We now have a macro name, a parameter count, and a
3072 * token list (in reverse order) for an expansion. We ought
3073 * to be OK just to create an SMacro, store it, and let
3074 * free_tlist have the rest of the line (which we have
3075 * carefully re-terminated after chopping off the expansion
3076 * from the end).
3078 define_smacro(ctx, mname, casesense, nparam, macro_start);
3079 free_tlist(origline);
3080 return DIRECTIVE_FOUND;
3082 case PP_UNDEF:
3083 tline = tline->next;
3084 skip_white_(tline);
3085 tline = expand_id(tline);
3086 if (!tline || (tline->type != TOK_ID &&
3087 (tline->type != TOK_PREPROC_ID ||
3088 tline->text[1] != '$'))) {
3089 nasm_error(ERR_NONFATAL, "`%%undef' expects a macro identifier");
3090 free_tlist(origline);
3091 return DIRECTIVE_FOUND;
3093 if (tline->next) {
3094 nasm_error(ERR_WARNING|ERR_PASS1,
3095 "trailing garbage after macro name ignored");
3098 /* Find the context that symbol belongs to */
3099 ctx = get_ctx(tline->text, &mname);
3100 undef_smacro(ctx, mname);
3101 free_tlist(origline);
3102 return DIRECTIVE_FOUND;
3104 case PP_DEFSTR:
3105 case PP_IDEFSTR:
3106 casesense = (i == PP_DEFSTR);
3108 tline = tline->next;
3109 skip_white_(tline);
3110 tline = expand_id(tline);
3111 if (!tline || (tline->type != TOK_ID &&
3112 (tline->type != TOK_PREPROC_ID ||
3113 tline->text[1] != '$'))) {
3114 nasm_error(ERR_NONFATAL, "`%s' expects a macro identifier",
3115 pp_directives[i]);
3116 free_tlist(origline);
3117 return DIRECTIVE_FOUND;
3120 ctx = get_ctx(tline->text, &mname);
3121 last = tline;
3122 tline = expand_smacro(tline->next);
3123 last->next = NULL;
3125 while (tok_type_(tline, TOK_WHITESPACE))
3126 tline = delete_Token(tline);
3128 p = detoken(tline, false);
3129 macro_start = nasm_malloc(sizeof(*macro_start));
3130 macro_start->next = NULL;
3131 macro_start->text = nasm_quote(p, strlen(p));
3132 macro_start->type = TOK_STRING;
3133 macro_start->a.mac = NULL;
3134 nasm_free(p);
3137 * We now have a macro name, an implicit parameter count of
3138 * zero, and a string token to use as an expansion. Create
3139 * and store an SMacro.
3141 define_smacro(ctx, mname, casesense, 0, macro_start);
3142 free_tlist(origline);
3143 return DIRECTIVE_FOUND;
3145 case PP_DEFTOK:
3146 case PP_IDEFTOK:
3147 casesense = (i == PP_DEFTOK);
3149 tline = tline->next;
3150 skip_white_(tline);
3151 tline = expand_id(tline);
3152 if (!tline || (tline->type != TOK_ID &&
3153 (tline->type != TOK_PREPROC_ID ||
3154 tline->text[1] != '$'))) {
3155 nasm_error(ERR_NONFATAL,
3156 "`%s' expects a macro identifier as first parameter",
3157 pp_directives[i]);
3158 free_tlist(origline);
3159 return DIRECTIVE_FOUND;
3161 ctx = get_ctx(tline->text, &mname);
3162 last = tline;
3163 tline = expand_smacro(tline->next);
3164 last->next = NULL;
3166 t = tline;
3167 while (tok_type_(t, TOK_WHITESPACE))
3168 t = t->next;
3169 /* t should now point to the string */
3170 if (!tok_type_(t, TOK_STRING)) {
3171 nasm_error(ERR_NONFATAL,
3172 "`%s` requires string as second parameter",
3173 pp_directives[i]);
3174 free_tlist(tline);
3175 free_tlist(origline);
3176 return DIRECTIVE_FOUND;
3180 * Convert the string to a token stream. Note that smacros
3181 * are stored with the token stream reversed, so we have to
3182 * reverse the output of tokenize().
3184 nasm_unquote_cstr(t->text, i);
3185 macro_start = reverse_tokens(tokenize(t->text));
3188 * We now have a macro name, an implicit parameter count of
3189 * zero, and a numeric token to use as an expansion. Create
3190 * and store an SMacro.
3192 define_smacro(ctx, mname, casesense, 0, macro_start);
3193 free_tlist(tline);
3194 free_tlist(origline);
3195 return DIRECTIVE_FOUND;
3197 case PP_PATHSEARCH:
3199 FILE *fp;
3200 StrList *xsl = NULL;
3201 StrList **xst = &xsl;
3203 casesense = true;
3205 tline = tline->next;
3206 skip_white_(tline);
3207 tline = expand_id(tline);
3208 if (!tline || (tline->type != TOK_ID &&
3209 (tline->type != TOK_PREPROC_ID ||
3210 tline->text[1] != '$'))) {
3211 nasm_error(ERR_NONFATAL,
3212 "`%%pathsearch' expects a macro identifier as first parameter");
3213 free_tlist(origline);
3214 return DIRECTIVE_FOUND;
3216 ctx = get_ctx(tline->text, &mname);
3217 last = tline;
3218 tline = expand_smacro(tline->next);
3219 last->next = NULL;
3221 t = tline;
3222 while (tok_type_(t, TOK_WHITESPACE))
3223 t = t->next;
3225 if (!t || (t->type != TOK_STRING &&
3226 t->type != TOK_INTERNAL_STRING)) {
3227 nasm_error(ERR_NONFATAL, "`%%pathsearch' expects a file name");
3228 free_tlist(tline);
3229 free_tlist(origline);
3230 return DIRECTIVE_FOUND; /* but we did _something_ */
3232 if (t->next)
3233 nasm_error(ERR_WARNING|ERR_PASS1,
3234 "trailing garbage after `%%pathsearch' ignored");
3235 p = t->text;
3236 if (t->type != TOK_INTERNAL_STRING)
3237 nasm_unquote(p, NULL);
3239 fp = inc_fopen(p, &xsl, &xst, true);
3240 if (fp) {
3241 p = xsl->str;
3242 fclose(fp); /* Don't actually care about the file */
3244 macro_start = nasm_malloc(sizeof(*macro_start));
3245 macro_start->next = NULL;
3246 macro_start->text = nasm_quote(p, strlen(p));
3247 macro_start->type = TOK_STRING;
3248 macro_start->a.mac = NULL;
3249 if (xsl)
3250 nasm_free(xsl);
3253 * We now have a macro name, an implicit parameter count of
3254 * zero, and a string token to use as an expansion. Create
3255 * and store an SMacro.
3257 define_smacro(ctx, mname, casesense, 0, macro_start);
3258 free_tlist(tline);
3259 free_tlist(origline);
3260 return DIRECTIVE_FOUND;
3263 case PP_STRLEN:
3264 casesense = true;
3266 tline = tline->next;
3267 skip_white_(tline);
3268 tline = expand_id(tline);
3269 if (!tline || (tline->type != TOK_ID &&
3270 (tline->type != TOK_PREPROC_ID ||
3271 tline->text[1] != '$'))) {
3272 nasm_error(ERR_NONFATAL,
3273 "`%%strlen' expects a macro identifier as first parameter");
3274 free_tlist(origline);
3275 return DIRECTIVE_FOUND;
3277 ctx = get_ctx(tline->text, &mname);
3278 last = tline;
3279 tline = expand_smacro(tline->next);
3280 last->next = NULL;
3282 t = tline;
3283 while (tok_type_(t, TOK_WHITESPACE))
3284 t = t->next;
3285 /* t should now point to the string */
3286 if (!tok_type_(t, TOK_STRING)) {
3287 nasm_error(ERR_NONFATAL,
3288 "`%%strlen` requires string as second parameter");
3289 free_tlist(tline);
3290 free_tlist(origline);
3291 return DIRECTIVE_FOUND;
3294 macro_start = nasm_malloc(sizeof(*macro_start));
3295 macro_start->next = NULL;
3296 make_tok_num(macro_start, nasm_unquote(t->text, NULL));
3297 macro_start->a.mac = NULL;
3300 * We now have a macro name, an implicit parameter count of
3301 * zero, and a numeric token to use as an expansion. Create
3302 * and store an SMacro.
3304 define_smacro(ctx, mname, casesense, 0, macro_start);
3305 free_tlist(tline);
3306 free_tlist(origline);
3307 return DIRECTIVE_FOUND;
3309 case PP_STRCAT:
3310 casesense = true;
3312 tline = tline->next;
3313 skip_white_(tline);
3314 tline = expand_id(tline);
3315 if (!tline || (tline->type != TOK_ID &&
3316 (tline->type != TOK_PREPROC_ID ||
3317 tline->text[1] != '$'))) {
3318 nasm_error(ERR_NONFATAL,
3319 "`%%strcat' expects a macro identifier as first parameter");
3320 free_tlist(origline);
3321 return DIRECTIVE_FOUND;
3323 ctx = get_ctx(tline->text, &mname);
3324 last = tline;
3325 tline = expand_smacro(tline->next);
3326 last->next = NULL;
3328 len = 0;
3329 list_for_each(t, tline) {
3330 switch (t->type) {
3331 case TOK_WHITESPACE:
3332 break;
3333 case TOK_STRING:
3334 len += t->a.len = nasm_unquote(t->text, NULL);
3335 break;
3336 case TOK_OTHER:
3337 if (!strcmp(t->text, ",")) /* permit comma separators */
3338 break;
3339 /* else fall through */
3340 default:
3341 nasm_error(ERR_NONFATAL,
3342 "non-string passed to `%%strcat' (%d)", t->type);
3343 free_tlist(tline);
3344 free_tlist(origline);
3345 return DIRECTIVE_FOUND;
3349 p = pp = nasm_malloc(len);
3350 list_for_each(t, tline) {
3351 if (t->type == TOK_STRING) {
3352 memcpy(p, t->text, t->a.len);
3353 p += t->a.len;
3358 * We now have a macro name, an implicit parameter count of
3359 * zero, and a numeric token to use as an expansion. Create
3360 * and store an SMacro.
3362 macro_start = new_Token(NULL, TOK_STRING, NULL, 0);
3363 macro_start->text = nasm_quote(pp, len);
3364 nasm_free(pp);
3365 define_smacro(ctx, mname, casesense, 0, macro_start);
3366 free_tlist(tline);
3367 free_tlist(origline);
3368 return DIRECTIVE_FOUND;
3370 case PP_SUBSTR:
3372 int64_t start, count;
3373 size_t len;
3375 casesense = true;
3377 tline = tline->next;
3378 skip_white_(tline);
3379 tline = expand_id(tline);
3380 if (!tline || (tline->type != TOK_ID &&
3381 (tline->type != TOK_PREPROC_ID ||
3382 tline->text[1] != '$'))) {
3383 nasm_error(ERR_NONFATAL,
3384 "`%%substr' expects a macro identifier as first parameter");
3385 free_tlist(origline);
3386 return DIRECTIVE_FOUND;
3388 ctx = get_ctx(tline->text, &mname);
3389 last = tline;
3390 tline = expand_smacro(tline->next);
3391 last->next = NULL;
3393 if (tline) /* skip expanded id */
3394 t = tline->next;
3395 while (tok_type_(t, TOK_WHITESPACE))
3396 t = t->next;
3398 /* t should now point to the string */
3399 if (!tok_type_(t, TOK_STRING)) {
3400 nasm_error(ERR_NONFATAL,
3401 "`%%substr` requires string as second parameter");
3402 free_tlist(tline);
3403 free_tlist(origline);
3404 return DIRECTIVE_FOUND;
3407 tt = t->next;
3408 tptr = &tt;
3409 tokval.t_type = TOKEN_INVALID;
3410 evalresult = evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
3411 if (!evalresult) {
3412 free_tlist(tline);
3413 free_tlist(origline);
3414 return DIRECTIVE_FOUND;
3415 } else if (!is_simple(evalresult)) {
3416 nasm_error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3417 free_tlist(tline);
3418 free_tlist(origline);
3419 return DIRECTIVE_FOUND;
3421 start = evalresult->value - 1;
3423 while (tok_type_(tt, TOK_WHITESPACE))
3424 tt = tt->next;
3425 if (!tt) {
3426 count = 1; /* Backwards compatibility: one character */
3427 } else {
3428 tokval.t_type = TOKEN_INVALID;
3429 evalresult = evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
3430 if (!evalresult) {
3431 free_tlist(tline);
3432 free_tlist(origline);
3433 return DIRECTIVE_FOUND;
3434 } else if (!is_simple(evalresult)) {
3435 nasm_error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3436 free_tlist(tline);
3437 free_tlist(origline);
3438 return DIRECTIVE_FOUND;
3440 count = evalresult->value;
3443 len = nasm_unquote(t->text, NULL);
3445 /* make start and count being in range */
3446 if (start < 0)
3447 start = 0;
3448 if (count < 0)
3449 count = len + count + 1 - start;
3450 if (start + count > (int64_t)len)
3451 count = len - start;
3452 if (!len || count < 0 || start >=(int64_t)len)
3453 start = -1, count = 0; /* empty string */
3455 macro_start = nasm_malloc(sizeof(*macro_start));
3456 macro_start->next = NULL;
3457 macro_start->text = nasm_quote((start < 0) ? "" : t->text + start, count);
3458 macro_start->type = TOK_STRING;
3459 macro_start->a.mac = NULL;
3462 * We now have a macro name, an implicit parameter count of
3463 * zero, and a numeric token to use as an expansion. Create
3464 * and store an SMacro.
3466 define_smacro(ctx, mname, casesense, 0, macro_start);
3467 free_tlist(tline);
3468 free_tlist(origline);
3469 return DIRECTIVE_FOUND;
3472 case PP_ASSIGN:
3473 case PP_IASSIGN:
3474 casesense = (i == PP_ASSIGN);
3476 tline = tline->next;
3477 skip_white_(tline);
3478 tline = expand_id(tline);
3479 if (!tline || (tline->type != TOK_ID &&
3480 (tline->type != TOK_PREPROC_ID ||
3481 tline->text[1] != '$'))) {
3482 nasm_error(ERR_NONFATAL,
3483 "`%%%sassign' expects a macro identifier",
3484 (i == PP_IASSIGN ? "i" : ""));
3485 free_tlist(origline);
3486 return DIRECTIVE_FOUND;
3488 ctx = get_ctx(tline->text, &mname);
3489 last = tline;
3490 tline = expand_smacro(tline->next);
3491 last->next = NULL;
3493 t = tline;
3494 tptr = &t;
3495 tokval.t_type = TOKEN_INVALID;
3496 evalresult = evaluate(ppscan, tptr, &tokval, NULL, pass, NULL);
3497 free_tlist(tline);
3498 if (!evalresult) {
3499 free_tlist(origline);
3500 return DIRECTIVE_FOUND;
3503 if (tokval.t_type)
3504 nasm_error(ERR_WARNING|ERR_PASS1,
3505 "trailing garbage after expression ignored");
3507 if (!is_simple(evalresult)) {
3508 nasm_error(ERR_NONFATAL,
3509 "non-constant value given to `%%%sassign'",
3510 (i == PP_IASSIGN ? "i" : ""));
3511 free_tlist(origline);
3512 return DIRECTIVE_FOUND;
3515 macro_start = nasm_malloc(sizeof(*macro_start));
3516 macro_start->next = NULL;
3517 make_tok_num(macro_start, reloc_value(evalresult));
3518 macro_start->a.mac = NULL;
3521 * We now have a macro name, an implicit parameter count of
3522 * zero, and a numeric token to use as an expansion. Create
3523 * and store an SMacro.
3525 define_smacro(ctx, mname, casesense, 0, macro_start);
3526 free_tlist(origline);
3527 return DIRECTIVE_FOUND;
3529 case PP_LINE:
3531 * Syntax is `%line nnn[+mmm] [filename]'
3533 tline = tline->next;
3534 skip_white_(tline);
3535 if (!tok_type_(tline, TOK_NUMBER)) {
3536 nasm_error(ERR_NONFATAL, "`%%line' expects line number");
3537 free_tlist(origline);
3538 return DIRECTIVE_FOUND;
3540 k = readnum(tline->text, &err);
3541 m = 1;
3542 tline = tline->next;
3543 if (tok_is_(tline, "+")) {
3544 tline = tline->next;
3545 if (!tok_type_(tline, TOK_NUMBER)) {
3546 nasm_error(ERR_NONFATAL, "`%%line' expects line increment");
3547 free_tlist(origline);
3548 return DIRECTIVE_FOUND;
3550 m = readnum(tline->text, &err);
3551 tline = tline->next;
3553 skip_white_(tline);
3554 src_set_linnum(k);
3555 istk->lineinc = m;
3556 if (tline) {
3557 nasm_free(src_set_fname(detoken(tline, false)));
3559 free_tlist(origline);
3560 return DIRECTIVE_FOUND;
3562 default:
3563 nasm_error(ERR_FATAL,
3564 "preprocessor directive `%s' not yet implemented",
3565 pp_directives[i]);
3566 return DIRECTIVE_FOUND;
3571 * Ensure that a macro parameter contains a condition code and
3572 * nothing else. Return the condition code index if so, or -1
3573 * otherwise.
3575 static int find_cc(Token * t)
3577 Token *tt;
3579 if (!t)
3580 return -1; /* Probably a %+ without a space */
3582 skip_white_(t);
3583 if (t->type != TOK_ID)
3584 return -1;
3585 tt = t->next;
3586 skip_white_(tt);
3587 if (tt && (tt->type != TOK_OTHER || strcmp(tt->text, ",")))
3588 return -1;
3590 return bsii(t->text, (const char **)conditions, ARRAY_SIZE(conditions));
3594 * This routines walks over tokens strem and hadnles tokens
3595 * pasting, if @handle_explicit passed then explicit pasting
3596 * term is handled, otherwise -- implicit pastings only.
3598 static bool paste_tokens(Token **head, const struct tokseq_match *m,
3599 size_t mnum, bool handle_explicit)
3601 Token *tok, *next, **prev_next, **prev_nonspace;
3602 bool pasted = false;
3603 char *buf, *p;
3604 size_t len, i;
3607 * The last token before pasting. We need it
3608 * to be able to connect new handled tokens.
3609 * In other words if there were a tokens stream
3611 * A -> B -> C -> D
3613 * and we've joined tokens B and C, the resulting
3614 * stream should be
3616 * A -> BC -> D
3618 tok = *head;
3619 prev_next = NULL;
3621 if (!tok_type_(tok, TOK_WHITESPACE) && !tok_type_(tok, TOK_PASTE))
3622 prev_nonspace = head;
3623 else
3624 prev_nonspace = NULL;
3626 while (tok && (next = tok->next)) {
3628 switch (tok->type) {
3629 case TOK_WHITESPACE:
3630 /* Zap redundant whitespaces */
3631 while (tok_type_(next, TOK_WHITESPACE))
3632 next = delete_Token(next);
3633 tok->next = next;
3634 break;
3636 case TOK_PASTE:
3637 /* Explicit pasting */
3638 if (!handle_explicit)
3639 break;
3640 next = delete_Token(tok);
3642 while (tok_type_(next, TOK_WHITESPACE))
3643 next = delete_Token(next);
3645 if (!pasted)
3646 pasted = true;
3648 /* Left pasting token is start of line */
3649 if (!prev_nonspace)
3650 nasm_error(ERR_FATAL, "No lvalue found on pasting");
3653 * No ending token, this might happen in two
3654 * cases
3656 * 1) There indeed no right token at all
3657 * 2) There is a bare "%define ID" statement,
3658 * and @ID does expand to whitespace.
3660 * So technically we need to do a grammar analysis
3661 * in another stage of parsing, but for now lets don't
3662 * change the behaviour people used to. Simply allow
3663 * whitespace after paste token.
3665 if (!next) {
3667 * Zap ending space tokens and that's all.
3669 tok = (*prev_nonspace)->next;
3670 while (tok_type_(tok, TOK_WHITESPACE))
3671 tok = delete_Token(tok);
3672 tok = *prev_nonspace;
3673 tok->next = NULL;
3674 break;
3677 tok = *prev_nonspace;
3678 while (tok_type_(tok, TOK_WHITESPACE))
3679 tok = delete_Token(tok);
3680 len = strlen(tok->text);
3681 len += strlen(next->text);
3683 p = buf = nasm_malloc(len + 1);
3684 strcpy(p, tok->text);
3685 p = strchr(p, '\0');
3686 strcpy(p, next->text);
3688 delete_Token(tok);
3690 tok = tokenize(buf);
3691 nasm_free(buf);
3693 *prev_nonspace = tok;
3694 while (tok && tok->next)
3695 tok = tok->next;
3697 tok->next = delete_Token(next);
3699 /* Restart from pasted tokens head */
3700 tok = *prev_nonspace;
3701 break;
3703 default:
3704 /* implicit pasting */
3705 for (i = 0; i < mnum; i++) {
3706 if (!(PP_CONCAT_MATCH(tok, m[i].mask_head)))
3707 continue;
3709 len = 0;
3710 while (next && PP_CONCAT_MATCH(next, m[i].mask_tail)) {
3711 len += strlen(next->text);
3712 next = next->next;
3715 /* No match */
3716 if (tok == next)
3717 break;
3719 len += strlen(tok->text);
3720 p = buf = nasm_malloc(len + 1);
3722 while (tok != next) {
3723 strcpy(p, tok->text);
3724 p = strchr(p, '\0');
3725 tok = delete_Token(tok);
3728 tok = tokenize(buf);
3729 nasm_free(buf);
3731 if (prev_next)
3732 *prev_next = tok;
3733 else
3734 *head = tok;
3737 * Connect pasted into original stream,
3738 * ie A -> new-tokens -> B
3740 while (tok && tok->next)
3741 tok = tok->next;
3742 tok->next = next;
3744 if (!pasted)
3745 pasted = true;
3747 /* Restart from pasted tokens head */
3748 tok = prev_next ? *prev_next : *head;
3751 break;
3754 prev_next = &tok->next;
3756 if (tok->next &&
3757 !tok_type_(tok->next, TOK_WHITESPACE) &&
3758 !tok_type_(tok->next, TOK_PASTE))
3759 prev_nonspace = prev_next;
3761 tok = tok->next;
3764 return pasted;
3768 * expands to a list of tokens from %{x:y}
3770 static Token *expand_mmac_params_range(MMacro *mac, Token *tline, Token ***last)
3772 Token *t = tline, **tt, *tm, *head;
3773 char *pos;
3774 int fst, lst, j, i;
3776 pos = strchr(tline->text, ':');
3777 nasm_assert(pos);
3779 lst = atoi(pos + 1);
3780 fst = atoi(tline->text + 1);
3783 * only macros params are accounted so
3784 * if someone passes %0 -- we reject such
3785 * value(s)
3787 if (lst == 0 || fst == 0)
3788 goto err;
3790 /* the values should be sane */
3791 if ((fst > (int)mac->nparam || fst < (-(int)mac->nparam)) ||
3792 (lst > (int)mac->nparam || lst < (-(int)mac->nparam)))
3793 goto err;
3795 fst = fst < 0 ? fst + (int)mac->nparam + 1: fst;
3796 lst = lst < 0 ? lst + (int)mac->nparam + 1: lst;
3798 /* counted from zero */
3799 fst--, lst--;
3802 * It will be at least one token. Note we
3803 * need to scan params until separator, otherwise
3804 * only first token will be passed.
3806 tm = mac->params[(fst + mac->rotate) % mac->nparam];
3807 head = new_Token(NULL, tm->type, tm->text, 0);
3808 tt = &head->next, tm = tm->next;
3809 while (tok_isnt_(tm, ",")) {
3810 t = new_Token(NULL, tm->type, tm->text, 0);
3811 *tt = t, tt = &t->next, tm = tm->next;
3814 if (fst < lst) {
3815 for (i = fst + 1; i <= lst; i++) {
3816 t = new_Token(NULL, TOK_OTHER, ",", 0);
3817 *tt = t, tt = &t->next;
3818 j = (i + mac->rotate) % mac->nparam;
3819 tm = mac->params[j];
3820 while (tok_isnt_(tm, ",")) {
3821 t = new_Token(NULL, tm->type, tm->text, 0);
3822 *tt = t, tt = &t->next, tm = tm->next;
3825 } else {
3826 for (i = fst - 1; i >= lst; i--) {
3827 t = new_Token(NULL, TOK_OTHER, ",", 0);
3828 *tt = t, tt = &t->next;
3829 j = (i + mac->rotate) % mac->nparam;
3830 tm = mac->params[j];
3831 while (tok_isnt_(tm, ",")) {
3832 t = new_Token(NULL, tm->type, tm->text, 0);
3833 *tt = t, tt = &t->next, tm = tm->next;
3838 *last = tt;
3839 return head;
3841 err:
3842 nasm_error(ERR_NONFATAL, "`%%{%s}': macro parameters out of range",
3843 &tline->text[1]);
3844 return tline;
3848 * Expand MMacro-local things: parameter references (%0, %n, %+n,
3849 * %-n) and MMacro-local identifiers (%%foo) as well as
3850 * macro indirection (%[...]) and range (%{..:..}).
3852 static Token *expand_mmac_params(Token * tline)
3854 Token *t, *tt, **tail, *thead;
3855 bool changed = false;
3856 char *pos;
3858 tail = &thead;
3859 thead = NULL;
3861 while (tline) {
3862 if (tline->type == TOK_PREPROC_ID &&
3863 (((tline->text[1] == '+' || tline->text[1] == '-') && tline->text[2]) ||
3864 (tline->text[1] >= '0' && tline->text[1] <= '9') ||
3865 tline->text[1] == '%')) {
3866 char *text = NULL;
3867 int type = 0, cc; /* type = 0 to placate optimisers */
3868 char tmpbuf[30];
3869 unsigned int n;
3870 int i;
3871 MMacro *mac;
3873 t = tline;
3874 tline = tline->next;
3876 mac = istk->mstk;
3877 while (mac && !mac->name) /* avoid mistaking %reps for macros */
3878 mac = mac->next_active;
3879 if (!mac) {
3880 nasm_error(ERR_NONFATAL, "`%s': not in a macro call", t->text);
3881 } else {
3882 pos = strchr(t->text, ':');
3883 if (!pos) {
3884 switch (t->text[1]) {
3886 * We have to make a substitution of one of the
3887 * forms %1, %-1, %+1, %%foo, %0.
3889 case '0':
3890 type = TOK_NUMBER;
3891 snprintf(tmpbuf, sizeof(tmpbuf), "%d", mac->nparam);
3892 text = nasm_strdup(tmpbuf);
3893 break;
3894 case '%':
3895 type = TOK_ID;
3896 snprintf(tmpbuf, sizeof(tmpbuf), "..@%"PRIu64".",
3897 mac->unique);
3898 text = nasm_strcat(tmpbuf, t->text + 2);
3899 break;
3900 case '-':
3901 n = atoi(t->text + 2) - 1;
3902 if (n >= mac->nparam)
3903 tt = NULL;
3904 else {
3905 if (mac->nparam > 1)
3906 n = (n + mac->rotate) % mac->nparam;
3907 tt = mac->params[n];
3909 cc = find_cc(tt);
3910 if (cc == -1) {
3911 nasm_error(ERR_NONFATAL,
3912 "macro parameter %d is not a condition code",
3913 n + 1);
3914 text = NULL;
3915 } else {
3916 type = TOK_ID;
3917 if (inverse_ccs[cc] == -1) {
3918 nasm_error(ERR_NONFATAL,
3919 "condition code `%s' is not invertible",
3920 conditions[cc]);
3921 text = NULL;
3922 } else
3923 text = nasm_strdup(conditions[inverse_ccs[cc]]);
3925 break;
3926 case '+':
3927 n = atoi(t->text + 2) - 1;
3928 if (n >= mac->nparam)
3929 tt = NULL;
3930 else {
3931 if (mac->nparam > 1)
3932 n = (n + mac->rotate) % mac->nparam;
3933 tt = mac->params[n];
3935 cc = find_cc(tt);
3936 if (cc == -1) {
3937 nasm_error(ERR_NONFATAL,
3938 "macro parameter %d is not a condition code",
3939 n + 1);
3940 text = NULL;
3941 } else {
3942 type = TOK_ID;
3943 text = nasm_strdup(conditions[cc]);
3945 break;
3946 default:
3947 n = atoi(t->text + 1) - 1;
3948 if (n >= mac->nparam)
3949 tt = NULL;
3950 else {
3951 if (mac->nparam > 1)
3952 n = (n + mac->rotate) % mac->nparam;
3953 tt = mac->params[n];
3955 if (tt) {
3956 for (i = 0; i < mac->paramlen[n]; i++) {
3957 *tail = new_Token(NULL, tt->type, tt->text, 0);
3958 tail = &(*tail)->next;
3959 tt = tt->next;
3962 text = NULL; /* we've done it here */
3963 break;
3965 } else {
3967 * seems we have a parameters range here
3969 Token *head, **last;
3970 head = expand_mmac_params_range(mac, t, &last);
3971 if (head != t) {
3972 *tail = head;
3973 *last = tline;
3974 tline = head;
3975 text = NULL;
3979 if (!text) {
3980 delete_Token(t);
3981 } else {
3982 *tail = t;
3983 tail = &t->next;
3984 t->type = type;
3985 nasm_free(t->text);
3986 t->text = text;
3987 t->a.mac = NULL;
3989 changed = true;
3990 continue;
3991 } else if (tline->type == TOK_INDIRECT) {
3992 t = tline;
3993 tline = tline->next;
3994 tt = tokenize(t->text);
3995 tt = expand_mmac_params(tt);
3996 tt = expand_smacro(tt);
3997 *tail = tt;
3998 while (tt) {
3999 tt->a.mac = NULL; /* Necessary? */
4000 tail = &tt->next;
4001 tt = tt->next;
4003 delete_Token(t);
4004 changed = true;
4005 } else {
4006 t = *tail = tline;
4007 tline = tline->next;
4008 t->a.mac = NULL;
4009 tail = &t->next;
4012 *tail = NULL;
4014 if (changed) {
4015 const struct tokseq_match t[] = {
4017 PP_CONCAT_MASK(TOK_ID) |
4018 PP_CONCAT_MASK(TOK_FLOAT), /* head */
4019 PP_CONCAT_MASK(TOK_ID) |
4020 PP_CONCAT_MASK(TOK_NUMBER) |
4021 PP_CONCAT_MASK(TOK_FLOAT) |
4022 PP_CONCAT_MASK(TOK_OTHER) /* tail */
4025 PP_CONCAT_MASK(TOK_NUMBER), /* head */
4026 PP_CONCAT_MASK(TOK_NUMBER) /* tail */
4029 paste_tokens(&thead, t, ARRAY_SIZE(t), false);
4032 return thead;
4036 * Expand all single-line macro calls made in the given line.
4037 * Return the expanded version of the line. The original is deemed
4038 * to be destroyed in the process. (In reality we'll just move
4039 * Tokens from input to output a lot of the time, rather than
4040 * actually bothering to destroy and replicate.)
4043 static Token *expand_smacro(Token * tline)
4045 Token *t, *tt, *mstart, **tail, *thead;
4046 SMacro *head = NULL, *m;
4047 Token **params;
4048 int *paramsize;
4049 unsigned int nparam, sparam;
4050 int brackets;
4051 Token *org_tline = tline;
4052 Context *ctx;
4053 const char *mname;
4054 int deadman = DEADMAN_LIMIT;
4055 bool expanded;
4058 * Trick: we should avoid changing the start token pointer since it can
4059 * be contained in "next" field of other token. Because of this
4060 * we allocate a copy of first token and work with it; at the end of
4061 * routine we copy it back
4063 if (org_tline) {
4064 tline = new_Token(org_tline->next, org_tline->type,
4065 org_tline->text, 0);
4066 tline->a.mac = org_tline->a.mac;
4067 nasm_free(org_tline->text);
4068 org_tline->text = NULL;
4071 expanded = true; /* Always expand %+ at least once */
4073 again:
4074 thead = NULL;
4075 tail = &thead;
4077 while (tline) { /* main token loop */
4078 if (!--deadman) {
4079 nasm_error(ERR_NONFATAL, "interminable macro recursion");
4080 goto err;
4083 if ((mname = tline->text)) {
4084 /* if this token is a local macro, look in local context */
4085 if (tline->type == TOK_ID) {
4086 head = (SMacro *)hash_findix(&smacros, mname);
4087 } else if (tline->type == TOK_PREPROC_ID) {
4088 ctx = get_ctx(mname, &mname);
4089 head = ctx ? (SMacro *)hash_findix(&ctx->localmac, mname) : NULL;
4090 } else
4091 head = NULL;
4094 * We've hit an identifier. As in is_mmacro below, we first
4095 * check whether the identifier is a single-line macro at
4096 * all, then think about checking for parameters if
4097 * necessary.
4099 list_for_each(m, head)
4100 if (!mstrcmp(m->name, mname, m->casesense))
4101 break;
4102 if (m) {
4103 mstart = tline;
4104 params = NULL;
4105 paramsize = NULL;
4106 if (m->nparam == 0) {
4108 * Simple case: the macro is parameterless. Discard the
4109 * one token that the macro call took, and push the
4110 * expansion back on the to-do stack.
4112 if (!m->expansion) {
4113 if (!strcmp("__FILE__", m->name)) {
4114 int32_t num = 0;
4115 char *file = NULL;
4116 src_get(&num, &file);
4117 tline->text = nasm_quote(file, strlen(file));
4118 tline->type = TOK_STRING;
4119 nasm_free(file);
4120 continue;
4122 if (!strcmp("__LINE__", m->name)) {
4123 nasm_free(tline->text);
4124 make_tok_num(tline, src_get_linnum());
4125 continue;
4127 if (!strcmp("__BITS__", m->name)) {
4128 nasm_free(tline->text);
4129 make_tok_num(tline, globalbits);
4130 continue;
4132 tline = delete_Token(tline);
4133 continue;
4135 } else {
4137 * Complicated case: at least one macro with this name
4138 * exists and takes parameters. We must find the
4139 * parameters in the call, count them, find the SMacro
4140 * that corresponds to that form of the macro call, and
4141 * substitute for the parameters when we expand. What a
4142 * pain.
4144 /*tline = tline->next;
4145 skip_white_(tline); */
4146 do {
4147 t = tline->next;
4148 while (tok_type_(t, TOK_SMAC_END)) {
4149 t->a.mac->in_progress = false;
4150 t->text = NULL;
4151 t = tline->next = delete_Token(t);
4153 tline = t;
4154 } while (tok_type_(tline, TOK_WHITESPACE));
4155 if (!tok_is_(tline, "(")) {
4157 * This macro wasn't called with parameters: ignore
4158 * the call. (Behaviour borrowed from gnu cpp.)
4160 tline = mstart;
4161 m = NULL;
4162 } else {
4163 int paren = 0;
4164 int white = 0;
4165 brackets = 0;
4166 nparam = 0;
4167 sparam = PARAM_DELTA;
4168 params = nasm_malloc(sparam * sizeof(Token *));
4169 params[0] = tline->next;
4170 paramsize = nasm_malloc(sparam * sizeof(int));
4171 paramsize[0] = 0;
4172 while (true) { /* parameter loop */
4174 * For some unusual expansions
4175 * which concatenates function call
4177 t = tline->next;
4178 while (tok_type_(t, TOK_SMAC_END)) {
4179 t->a.mac->in_progress = false;
4180 t->text = NULL;
4181 t = tline->next = delete_Token(t);
4183 tline = t;
4185 if (!tline) {
4186 nasm_error(ERR_NONFATAL,
4187 "macro call expects terminating `)'");
4188 break;
4190 if (tline->type == TOK_WHITESPACE
4191 && brackets <= 0) {
4192 if (paramsize[nparam])
4193 white++;
4194 else
4195 params[nparam] = tline->next;
4196 continue; /* parameter loop */
4198 if (tline->type == TOK_OTHER
4199 && tline->text[1] == 0) {
4200 char ch = tline->text[0];
4201 if (ch == ',' && !paren && brackets <= 0) {
4202 if (++nparam >= sparam) {
4203 sparam += PARAM_DELTA;
4204 params = nasm_realloc(params,
4205 sparam * sizeof(Token *));
4206 paramsize = nasm_realloc(paramsize,
4207 sparam * sizeof(int));
4209 params[nparam] = tline->next;
4210 paramsize[nparam] = 0;
4211 white = 0;
4212 continue; /* parameter loop */
4214 if (ch == '{' &&
4215 (brackets > 0 || (brackets == 0 &&
4216 !paramsize[nparam])))
4218 if (!(brackets++)) {
4219 params[nparam] = tline->next;
4220 continue; /* parameter loop */
4223 if (ch == '}' && brackets > 0)
4224 if (--brackets == 0) {
4225 brackets = -1;
4226 continue; /* parameter loop */
4228 if (ch == '(' && !brackets)
4229 paren++;
4230 if (ch == ')' && brackets <= 0)
4231 if (--paren < 0)
4232 break;
4234 if (brackets < 0) {
4235 brackets = 0;
4236 nasm_error(ERR_NONFATAL, "braces do not "
4237 "enclose all of macro parameter");
4239 paramsize[nparam] += white + 1;
4240 white = 0;
4241 } /* parameter loop */
4242 nparam++;
4243 while (m && (m->nparam != nparam ||
4244 mstrcmp(m->name, mname,
4245 m->casesense)))
4246 m = m->next;
4247 if (!m)
4248 nasm_error(ERR_WARNING|ERR_PASS1|ERR_WARN_MNP,
4249 "macro `%s' exists, "
4250 "but not taking %d parameters",
4251 mstart->text, nparam);
4254 if (m && m->in_progress)
4255 m = NULL;
4256 if (!m) { /* in progess or didn't find '(' or wrong nparam */
4258 * Design question: should we handle !tline, which
4259 * indicates missing ')' here, or expand those
4260 * macros anyway, which requires the (t) test a few
4261 * lines down?
4263 nasm_free(params);
4264 nasm_free(paramsize);
4265 tline = mstart;
4266 } else {
4268 * Expand the macro: we are placed on the last token of the
4269 * call, so that we can easily split the call from the
4270 * following tokens. We also start by pushing an SMAC_END
4271 * token for the cycle removal.
4273 t = tline;
4274 if (t) {
4275 tline = t->next;
4276 t->next = NULL;
4278 tt = new_Token(tline, TOK_SMAC_END, NULL, 0);
4279 tt->a.mac = m;
4280 m->in_progress = true;
4281 tline = tt;
4282 list_for_each(t, m->expansion) {
4283 if (t->type >= TOK_SMAC_PARAM) {
4284 Token *pcopy = tline, **ptail = &pcopy;
4285 Token *ttt, *pt;
4286 int i;
4288 ttt = params[t->type - TOK_SMAC_PARAM];
4289 i = paramsize[t->type - TOK_SMAC_PARAM];
4290 while (--i >= 0) {
4291 pt = *ptail = new_Token(tline, ttt->type,
4292 ttt->text, 0);
4293 ptail = &pt->next;
4294 ttt = ttt->next;
4296 tline = pcopy;
4297 } else if (t->type == TOK_PREPROC_Q) {
4298 tt = new_Token(tline, TOK_ID, mname, 0);
4299 tline = tt;
4300 } else if (t->type == TOK_PREPROC_QQ) {
4301 tt = new_Token(tline, TOK_ID, m->name, 0);
4302 tline = tt;
4303 } else {
4304 tt = new_Token(tline, t->type, t->text, 0);
4305 tline = tt;
4310 * Having done that, get rid of the macro call, and clean
4311 * up the parameters.
4313 nasm_free(params);
4314 nasm_free(paramsize);
4315 free_tlist(mstart);
4316 expanded = true;
4317 continue; /* main token loop */
4322 if (tline->type == TOK_SMAC_END) {
4323 tline->a.mac->in_progress = false;
4324 tline = delete_Token(tline);
4325 } else {
4326 t = *tail = tline;
4327 tline = tline->next;
4328 t->a.mac = NULL;
4329 t->next = NULL;
4330 tail = &t->next;
4335 * Now scan the entire line and look for successive TOK_IDs that resulted
4336 * after expansion (they can't be produced by tokenize()). The successive
4337 * TOK_IDs should be concatenated.
4338 * Also we look for %+ tokens and concatenate the tokens before and after
4339 * them (without white spaces in between).
4341 if (expanded) {
4342 const struct tokseq_match t[] = {
4344 PP_CONCAT_MASK(TOK_ID) |
4345 PP_CONCAT_MASK(TOK_PREPROC_ID), /* head */
4346 PP_CONCAT_MASK(TOK_ID) |
4347 PP_CONCAT_MASK(TOK_PREPROC_ID) |
4348 PP_CONCAT_MASK(TOK_NUMBER) /* tail */
4351 if (paste_tokens(&thead, t, ARRAY_SIZE(t), true)) {
4353 * If we concatenated something, *and* we had previously expanded
4354 * an actual macro, scan the lines again for macros...
4356 tline = thead;
4357 expanded = false;
4358 goto again;
4362 err:
4363 if (org_tline) {
4364 if (thead) {
4365 *org_tline = *thead;
4366 /* since we just gave text to org_line, don't free it */
4367 thead->text = NULL;
4368 delete_Token(thead);
4369 } else {
4370 /* the expression expanded to empty line;
4371 we can't return NULL for some reasons
4372 we just set the line to a single WHITESPACE token. */
4373 memset(org_tline, 0, sizeof(*org_tline));
4374 org_tline->text = NULL;
4375 org_tline->type = TOK_WHITESPACE;
4377 thead = org_tline;
4380 return thead;
4384 * Similar to expand_smacro but used exclusively with macro identifiers
4385 * right before they are fetched in. The reason is that there can be
4386 * identifiers consisting of several subparts. We consider that if there
4387 * are more than one element forming the name, user wants a expansion,
4388 * otherwise it will be left as-is. Example:
4390 * %define %$abc cde
4392 * the identifier %$abc will be left as-is so that the handler for %define
4393 * will suck it and define the corresponding value. Other case:
4395 * %define _%$abc cde
4397 * In this case user wants name to be expanded *before* %define starts
4398 * working, so we'll expand %$abc into something (if it has a value;
4399 * otherwise it will be left as-is) then concatenate all successive
4400 * PP_IDs into one.
4402 static Token *expand_id(Token * tline)
4404 Token *cur, *oldnext = NULL;
4406 if (!tline || !tline->next)
4407 return tline;
4409 cur = tline;
4410 while (cur->next &&
4411 (cur->next->type == TOK_ID ||
4412 cur->next->type == TOK_PREPROC_ID
4413 || cur->next->type == TOK_NUMBER))
4414 cur = cur->next;
4416 /* If identifier consists of just one token, don't expand */
4417 if (cur == tline)
4418 return tline;
4420 if (cur) {
4421 oldnext = cur->next; /* Detach the tail past identifier */
4422 cur->next = NULL; /* so that expand_smacro stops here */
4425 tline = expand_smacro(tline);
4427 if (cur) {
4428 /* expand_smacro possibly changhed tline; re-scan for EOL */
4429 cur = tline;
4430 while (cur && cur->next)
4431 cur = cur->next;
4432 if (cur)
4433 cur->next = oldnext;
4436 return tline;
4440 * Determine whether the given line constitutes a multi-line macro
4441 * call, and return the MMacro structure called if so. Doesn't have
4442 * to check for an initial label - that's taken care of in
4443 * expand_mmacro - but must check numbers of parameters. Guaranteed
4444 * to be called with tline->type == TOK_ID, so the putative macro
4445 * name is easy to find.
4447 static MMacro *is_mmacro(Token * tline, Token *** params_array)
4449 MMacro *head, *m;
4450 Token **params;
4451 int nparam;
4453 head = (MMacro *) hash_findix(&mmacros, tline->text);
4456 * Efficiency: first we see if any macro exists with the given
4457 * name. If not, we can return NULL immediately. _Then_ we
4458 * count the parameters, and then we look further along the
4459 * list if necessary to find the proper MMacro.
4461 list_for_each(m, head)
4462 if (!mstrcmp(m->name, tline->text, m->casesense))
4463 break;
4464 if (!m)
4465 return NULL;
4468 * OK, we have a potential macro. Count and demarcate the
4469 * parameters.
4471 count_mmac_params(tline->next, &nparam, &params);
4474 * So we know how many parameters we've got. Find the MMacro
4475 * structure that handles this number.
4477 while (m) {
4478 if (m->nparam_min <= nparam
4479 && (m->plus || nparam <= m->nparam_max)) {
4481 * This one is right. Just check if cycle removal
4482 * prohibits us using it before we actually celebrate...
4484 if (m->in_progress > m->max_depth) {
4485 if (m->max_depth > 0) {
4486 nasm_error(ERR_WARNING,
4487 "reached maximum recursion depth of %i",
4488 m->max_depth);
4490 nasm_free(params);
4491 return NULL;
4494 * It's right, and we can use it. Add its default
4495 * parameters to the end of our list if necessary.
4497 if (m->defaults && nparam < m->nparam_min + m->ndefs) {
4498 params =
4499 nasm_realloc(params,
4500 ((m->nparam_min + m->ndefs +
4501 1) * sizeof(*params)));
4502 while (nparam < m->nparam_min + m->ndefs) {
4503 params[nparam] = m->defaults[nparam - m->nparam_min];
4504 nparam++;
4508 * If we've gone over the maximum parameter count (and
4509 * we're in Plus mode), ignore parameters beyond
4510 * nparam_max.
4512 if (m->plus && nparam > m->nparam_max)
4513 nparam = m->nparam_max;
4515 * Then terminate the parameter list, and leave.
4517 if (!params) { /* need this special case */
4518 params = nasm_malloc(sizeof(*params));
4519 nparam = 0;
4521 params[nparam] = NULL;
4522 *params_array = params;
4523 return m;
4526 * This one wasn't right: look for the next one with the
4527 * same name.
4529 list_for_each(m, m->next)
4530 if (!mstrcmp(m->name, tline->text, m->casesense))
4531 break;
4535 * After all that, we didn't find one with the right number of
4536 * parameters. Issue a warning, and fail to expand the macro.
4538 nasm_error(ERR_WARNING|ERR_PASS1|ERR_WARN_MNP,
4539 "macro `%s' exists, but not taking %d parameters",
4540 tline->text, nparam);
4541 nasm_free(params);
4542 return NULL;
4547 * Save MMacro invocation specific fields in
4548 * preparation for a recursive macro expansion
4550 static void push_mmacro(MMacro *m)
4552 MMacroInvocation *i;
4554 i = nasm_malloc(sizeof(MMacroInvocation));
4555 i->prev = m->prev;
4556 i->params = m->params;
4557 i->iline = m->iline;
4558 i->nparam = m->nparam;
4559 i->rotate = m->rotate;
4560 i->paramlen = m->paramlen;
4561 i->unique = m->unique;
4562 i->condcnt = m->condcnt;
4563 m->prev = i;
4568 * Restore MMacro invocation specific fields that were
4569 * saved during a previous recursive macro expansion
4571 static void pop_mmacro(MMacro *m)
4573 MMacroInvocation *i;
4575 if (m->prev) {
4576 i = m->prev;
4577 m->prev = i->prev;
4578 m->params = i->params;
4579 m->iline = i->iline;
4580 m->nparam = i->nparam;
4581 m->rotate = i->rotate;
4582 m->paramlen = i->paramlen;
4583 m->unique = i->unique;
4584 m->condcnt = i->condcnt;
4585 nasm_free(i);
4591 * Expand the multi-line macro call made by the given line, if
4592 * there is one to be expanded. If there is, push the expansion on
4593 * istk->expansion and return 1. Otherwise return 0.
4595 static int expand_mmacro(Token * tline)
4597 Token *startline = tline;
4598 Token *label = NULL;
4599 int dont_prepend = 0;
4600 Token **params, *t, *tt;
4601 MMacro *m;
4602 Line *l, *ll;
4603 int i, nparam, *paramlen;
4604 const char *mname;
4606 t = tline;
4607 skip_white_(t);
4608 /* if (!tok_type_(t, TOK_ID)) Lino 02/25/02 */
4609 if (!tok_type_(t, TOK_ID) && !tok_type_(t, TOK_PREPROC_ID))
4610 return 0;
4611 m = is_mmacro(t, &params);
4612 if (m) {
4613 mname = t->text;
4614 } else {
4615 Token *last;
4617 * We have an id which isn't a macro call. We'll assume
4618 * it might be a label; we'll also check to see if a
4619 * colon follows it. Then, if there's another id after
4620 * that lot, we'll check it again for macro-hood.
4622 label = last = t;
4623 t = t->next;
4624 if (tok_type_(t, TOK_WHITESPACE))
4625 last = t, t = t->next;
4626 if (tok_is_(t, ":")) {
4627 dont_prepend = 1;
4628 last = t, t = t->next;
4629 if (tok_type_(t, TOK_WHITESPACE))
4630 last = t, t = t->next;
4632 if (!tok_type_(t, TOK_ID) || !(m = is_mmacro(t, &params)))
4633 return 0;
4634 last->next = NULL;
4635 mname = t->text;
4636 tline = t;
4640 * Fix up the parameters: this involves stripping leading and
4641 * trailing whitespace, then stripping braces if they are
4642 * present.
4644 for (nparam = 0; params[nparam]; nparam++) ;
4645 paramlen = nparam ? nasm_malloc(nparam * sizeof(*paramlen)) : NULL;
4647 for (i = 0; params[i]; i++) {
4648 int brace = 0;
4649 int comma = (!m->plus || i < nparam - 1);
4651 t = params[i];
4652 skip_white_(t);
4653 if (tok_is_(t, "{"))
4654 t = t->next, brace++, comma = false;
4655 params[i] = t;
4656 paramlen[i] = 0;
4657 while (t) {
4658 if (comma && t->type == TOK_OTHER && !strcmp(t->text, ","))
4659 break; /* ... because we have hit a comma */
4660 if (comma && t->type == TOK_WHITESPACE
4661 && tok_is_(t->next, ","))
4662 break; /* ... or a space then a comma */
4663 if (brace && t->type == TOK_OTHER) {
4664 if (t->text[0] == '{')
4665 brace++; /* ... or a nested opening brace */
4666 else if (t->text[0] == '}')
4667 if (!--brace)
4668 break; /* ... or a brace */
4670 t = t->next;
4671 paramlen[i]++;
4673 if (brace)
4674 nasm_error(ERR_NONFATAL, "macro params should be enclosed in braces");
4678 * OK, we have a MMacro structure together with a set of
4679 * parameters. We must now go through the expansion and push
4680 * copies of each Line on to istk->expansion. Substitution of
4681 * parameter tokens and macro-local tokens doesn't get done
4682 * until the single-line macro substitution process; this is
4683 * because delaying them allows us to change the semantics
4684 * later through %rotate.
4686 * First, push an end marker on to istk->expansion, mark this
4687 * macro as in progress, and set up its invocation-specific
4688 * variables.
4690 ll = nasm_malloc(sizeof(Line));
4691 ll->next = istk->expansion;
4692 ll->finishes = m;
4693 ll->first = NULL;
4694 istk->expansion = ll;
4697 * Save the previous MMacro expansion in the case of
4698 * macro recursion
4700 if (m->max_depth && m->in_progress)
4701 push_mmacro(m);
4703 m->in_progress ++;
4704 m->params = params;
4705 m->iline = tline;
4706 m->nparam = nparam;
4707 m->rotate = 0;
4708 m->paramlen = paramlen;
4709 m->unique = unique++;
4710 m->lineno = 0;
4711 m->condcnt = 0;
4713 m->next_active = istk->mstk;
4714 istk->mstk = m;
4716 list_for_each(l, m->expansion) {
4717 Token **tail;
4719 ll = nasm_malloc(sizeof(Line));
4720 ll->finishes = NULL;
4721 ll->next = istk->expansion;
4722 istk->expansion = ll;
4723 tail = &ll->first;
4725 list_for_each(t, l->first) {
4726 Token *x = t;
4727 switch (t->type) {
4728 case TOK_PREPROC_Q:
4729 tt = *tail = new_Token(NULL, TOK_ID, mname, 0);
4730 break;
4731 case TOK_PREPROC_QQ:
4732 tt = *tail = new_Token(NULL, TOK_ID, m->name, 0);
4733 break;
4734 case TOK_PREPROC_ID:
4735 if (t->text[1] == '0' && t->text[2] == '0') {
4736 dont_prepend = -1;
4737 x = label;
4738 if (!x)
4739 continue;
4741 /* fall through */
4742 default:
4743 tt = *tail = new_Token(NULL, x->type, x->text, 0);
4744 break;
4746 tail = &tt->next;
4748 *tail = NULL;
4752 * If we had a label, push it on as the first line of
4753 * the macro expansion.
4755 if (label) {
4756 if (dont_prepend < 0)
4757 free_tlist(startline);
4758 else {
4759 ll = nasm_malloc(sizeof(Line));
4760 ll->finishes = NULL;
4761 ll->next = istk->expansion;
4762 istk->expansion = ll;
4763 ll->first = startline;
4764 if (!dont_prepend) {
4765 while (label->next)
4766 label = label->next;
4767 label->next = tt = new_Token(NULL, TOK_OTHER, ":", 0);
4772 nasmlist->uplevel(m->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
4774 return 1;
4778 * This function adds macro names to error messages, and suppresses
4779 * them if necessary.
4781 static void pp_verror(int severity, const char *fmt, va_list arg)
4783 char buff[BUFSIZ];
4784 MMacro *mmac = NULL;
4785 int delta = 0;
4788 * If we're in a dead branch of IF or something like it, ignore the error.
4789 * However, because %else etc are evaluated in the state context
4790 * of the previous branch, errors might get lost:
4791 * %if 0 ... %else trailing garbage ... %endif
4792 * So %else etc should set the ERR_PP_PRECOND flag.
4794 if ((severity & ERR_MASK) < ERR_FATAL &&
4795 istk && istk->conds &&
4796 ((severity & ERR_PP_PRECOND) ?
4797 istk->conds->state == COND_NEVER :
4798 emitting(istk->conds->state)))
4799 return;
4801 /* get %macro name */
4802 if (!(severity & ERR_NOFILE) && istk && istk->mstk) {
4803 mmac = istk->mstk;
4804 /* but %rep blocks should be skipped */
4805 while (mmac && !mmac->name)
4806 mmac = mmac->next_active, delta++;
4809 if (mmac) {
4810 vsnprintf(buff, sizeof(buff), fmt, arg);
4812 nasm_set_verror(real_verror);
4813 nasm_error(severity, "(%s:%d) %s",
4814 mmac->name, mmac->lineno - delta, buff);
4815 nasm_set_verror(pp_verror);
4816 } else {
4817 real_verror(severity, fmt, arg);
4821 static void
4822 pp_reset(char *file, int apass, StrList **deplist)
4824 Token *t;
4826 cstk = NULL;
4827 istk = nasm_malloc(sizeof(Include));
4828 istk->next = NULL;
4829 istk->conds = NULL;
4830 istk->expansion = NULL;
4831 istk->mstk = NULL;
4832 istk->fp = fopen(file, "r");
4833 istk->fname = NULL;
4834 src_set_fname(nasm_strdup(file));
4835 src_set_linnum(0);
4836 istk->lineinc = 1;
4837 if (!istk->fp)
4838 nasm_fatal(ERR_NOFILE, "unable to open input file `%s'", file);
4839 defining = NULL;
4840 nested_mac_count = 0;
4841 nested_rep_count = 0;
4842 init_macros();
4843 unique = 0;
4844 if (tasm_compatible_mode) {
4845 stdmacpos = nasm_stdmac;
4846 } else {
4847 stdmacpos = nasm_stdmac_after_tasm;
4849 any_extrastdmac = extrastdmac && *extrastdmac;
4850 do_predef = true;
4853 * 0 for dependencies, 1 for preparatory passes, 2 for final pass.
4854 * The caller, however, will also pass in 3 for preprocess-only so
4855 * we can set __PASS__ accordingly.
4857 pass = apass > 2 ? 2 : apass;
4859 dephead = deptail = deplist;
4860 if (deplist) {
4861 StrList *sl = nasm_malloc(strlen(file)+1+sizeof sl->next);
4862 sl->next = NULL;
4863 strcpy(sl->str, file);
4864 *deptail = sl;
4865 deptail = &sl->next;
4869 * Define the __PASS__ macro. This is defined here unlike
4870 * all the other builtins, because it is special -- it varies between
4871 * passes.
4873 t = nasm_malloc(sizeof(*t));
4874 t->next = NULL;
4875 make_tok_num(t, apass);
4876 t->a.mac = NULL;
4877 define_smacro(NULL, "__PASS__", true, 0, t);
4880 static char *pp_getline(void)
4882 char *line;
4883 Token *tline;
4885 real_verror = nasm_set_verror(pp_verror);
4887 while (1) {
4889 * Fetch a tokenized line, either from the macro-expansion
4890 * buffer or from the input file.
4892 tline = NULL;
4893 while (istk->expansion && istk->expansion->finishes) {
4894 Line *l = istk->expansion;
4895 if (!l->finishes->name && l->finishes->in_progress > 1) {
4896 Line *ll;
4899 * This is a macro-end marker for a macro with no
4900 * name, which means it's not really a macro at all
4901 * but a %rep block, and the `in_progress' field is
4902 * more than 1, meaning that we still need to
4903 * repeat. (1 means the natural last repetition; 0
4904 * means termination by %exitrep.) We have
4905 * therefore expanded up to the %endrep, and must
4906 * push the whole block on to the expansion buffer
4907 * again. We don't bother to remove the macro-end
4908 * marker: we'd only have to generate another one
4909 * if we did.
4911 l->finishes->in_progress--;
4912 list_for_each(l, l->finishes->expansion) {
4913 Token *t, *tt, **tail;
4915 ll = nasm_malloc(sizeof(Line));
4916 ll->next = istk->expansion;
4917 ll->finishes = NULL;
4918 ll->first = NULL;
4919 tail = &ll->first;
4921 list_for_each(t, l->first) {
4922 if (t->text || t->type == TOK_WHITESPACE) {
4923 tt = *tail = new_Token(NULL, t->type, t->text, 0);
4924 tail = &tt->next;
4928 istk->expansion = ll;
4930 } else {
4932 * Check whether a `%rep' was started and not ended
4933 * within this macro expansion. This can happen and
4934 * should be detected. It's a fatal error because
4935 * I'm too confused to work out how to recover
4936 * sensibly from it.
4938 if (defining) {
4939 if (defining->name)
4940 nasm_panic(0, "defining with name in expansion");
4941 else if (istk->mstk->name)
4942 nasm_fatal(0, "`%%rep' without `%%endrep' within"
4943 " expansion of macro `%s'",
4944 istk->mstk->name);
4948 * FIXME: investigate the relationship at this point between
4949 * istk->mstk and l->finishes
4952 MMacro *m = istk->mstk;
4953 istk->mstk = m->next_active;
4954 if (m->name) {
4956 * This was a real macro call, not a %rep, and
4957 * therefore the parameter information needs to
4958 * be freed.
4960 if (m->prev) {
4961 pop_mmacro(m);
4962 l->finishes->in_progress --;
4963 } else {
4964 nasm_free(m->params);
4965 free_tlist(m->iline);
4966 nasm_free(m->paramlen);
4967 l->finishes->in_progress = 0;
4969 } else
4970 free_mmacro(m);
4972 istk->expansion = l->next;
4973 nasm_free(l);
4974 nasmlist->downlevel(LIST_MACRO);
4977 while (1) { /* until we get a line we can use */
4979 if (istk->expansion) { /* from a macro expansion */
4980 char *p;
4981 Line *l = istk->expansion;
4982 if (istk->mstk)
4983 istk->mstk->lineno++;
4984 tline = l->first;
4985 istk->expansion = l->next;
4986 nasm_free(l);
4987 p = detoken(tline, false);
4988 nasmlist->line(LIST_MACRO, p);
4989 nasm_free(p);
4990 break;
4992 line = read_line();
4993 if (line) { /* from the current input file */
4994 line = prepreproc(line);
4995 tline = tokenize(line);
4996 nasm_free(line);
4997 break;
5000 * The current file has ended; work down the istk
5003 Include *i = istk;
5004 fclose(i->fp);
5005 if (i->conds) {
5006 /* nasm_error can't be conditionally suppressed */
5007 nasm_fatal(0,
5008 "expected `%%endif' before end of file");
5010 /* only set line and file name if there's a next node */
5011 if (i->next) {
5012 src_set_linnum(i->lineno);
5013 nasm_free(src_set_fname(nasm_strdup(i->fname)));
5015 istk = i->next;
5016 nasmlist->downlevel(LIST_INCLUDE);
5017 nasm_free(i);
5018 if (!istk) {
5019 line = NULL;
5020 goto done;
5022 if (istk->expansion && istk->expansion->finishes)
5023 break;
5028 * We must expand MMacro parameters and MMacro-local labels
5029 * _before_ we plunge into directive processing, to cope
5030 * with things like `%define something %1' such as STRUC
5031 * uses. Unless we're _defining_ a MMacro, in which case
5032 * those tokens should be left alone to go into the
5033 * definition; and unless we're in a non-emitting
5034 * condition, in which case we don't want to meddle with
5035 * anything.
5037 if (!defining && !(istk->conds && !emitting(istk->conds->state))
5038 && !(istk->mstk && !istk->mstk->in_progress)) {
5039 tline = expand_mmac_params(tline);
5043 * Check the line to see if it's a preprocessor directive.
5045 if (do_directive(tline) == DIRECTIVE_FOUND) {
5046 continue;
5047 } else if (defining) {
5049 * We're defining a multi-line macro. We emit nothing
5050 * at all, and just
5051 * shove the tokenized line on to the macro definition.
5053 Line *l = nasm_malloc(sizeof(Line));
5054 l->next = defining->expansion;
5055 l->first = tline;
5056 l->finishes = NULL;
5057 defining->expansion = l;
5058 continue;
5059 } else if (istk->conds && !emitting(istk->conds->state)) {
5061 * We're in a non-emitting branch of a condition block.
5062 * Emit nothing at all, not even a blank line: when we
5063 * emerge from the condition we'll give a line-number
5064 * directive so we keep our place correctly.
5066 free_tlist(tline);
5067 continue;
5068 } else if (istk->mstk && !istk->mstk->in_progress) {
5070 * We're in a %rep block which has been terminated, so
5071 * we're walking through to the %endrep without
5072 * emitting anything. Emit nothing at all, not even a
5073 * blank line: when we emerge from the %rep block we'll
5074 * give a line-number directive so we keep our place
5075 * correctly.
5077 free_tlist(tline);
5078 continue;
5079 } else {
5080 tline = expand_smacro(tline);
5081 if (!expand_mmacro(tline)) {
5083 * De-tokenize the line again, and emit it.
5085 line = detoken(tline, true);
5086 free_tlist(tline);
5087 break;
5088 } else {
5089 continue; /* expand_mmacro calls free_tlist */
5094 done:
5095 nasm_set_verror(real_verror);
5096 return line;
5099 static void pp_cleanup(int pass)
5101 real_verror = nasm_set_verror(pp_verror);
5103 if (defining) {
5104 if (defining->name) {
5105 nasm_error(ERR_NONFATAL,
5106 "end of file while still defining macro `%s'",
5107 defining->name);
5108 } else {
5109 nasm_error(ERR_NONFATAL, "end of file while still in %%rep");
5112 free_mmacro(defining);
5113 defining = NULL;
5116 nasm_set_verror(real_verror);
5118 while (cstk)
5119 ctx_pop();
5120 free_macros();
5121 while (istk) {
5122 Include *i = istk;
5123 istk = istk->next;
5124 fclose(i->fp);
5125 nasm_free(i->fname);
5126 nasm_free(i);
5128 while (cstk)
5129 ctx_pop();
5130 nasm_free(src_set_fname(NULL));
5131 if (pass == 0) {
5132 IncPath *i;
5133 free_llist(predef);
5134 predef = NULL;
5135 delete_Blocks();
5136 freeTokens = NULL;
5137 while ((i = ipath)) {
5138 ipath = i->next;
5139 if (i->path)
5140 nasm_free(i->path);
5141 nasm_free(i);
5146 static void pp_include_path(char *path)
5148 IncPath *i;
5150 i = nasm_malloc(sizeof(IncPath));
5151 i->path = path ? nasm_strdup(path) : NULL;
5152 i->next = NULL;
5154 if (ipath) {
5155 IncPath *j = ipath;
5156 while (j->next)
5157 j = j->next;
5158 j->next = i;
5159 } else {
5160 ipath = i;
5164 static void pp_pre_include(char *fname)
5166 Token *inc, *space, *name;
5167 Line *l;
5169 name = new_Token(NULL, TOK_INTERNAL_STRING, fname, 0);
5170 space = new_Token(name, TOK_WHITESPACE, NULL, 0);
5171 inc = new_Token(space, TOK_PREPROC_ID, "%include", 0);
5173 l = nasm_malloc(sizeof(Line));
5174 l->next = predef;
5175 l->first = inc;
5176 l->finishes = NULL;
5177 predef = l;
5180 static void pp_pre_define(char *definition)
5182 Token *def, *space;
5183 Line *l;
5184 char *equals;
5186 real_verror = nasm_set_verror(pp_verror);
5188 equals = strchr(definition, '=');
5189 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
5190 def = new_Token(space, TOK_PREPROC_ID, "%define", 0);
5191 if (equals)
5192 *equals = ' ';
5193 space->next = tokenize(definition);
5194 if (equals)
5195 *equals = '=';
5197 if (space->next->type != TOK_PREPROC_ID &&
5198 space->next->type != TOK_ID)
5199 nasm_error(ERR_WARNING, "pre-defining non ID `%s\'\n", definition);
5201 l = nasm_malloc(sizeof(Line));
5202 l->next = predef;
5203 l->first = def;
5204 l->finishes = NULL;
5205 predef = l;
5207 nasm_set_verror(real_verror);
5210 static void pp_pre_undefine(char *definition)
5212 Token *def, *space;
5213 Line *l;
5215 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
5216 def = new_Token(space, TOK_PREPROC_ID, "%undef", 0);
5217 space->next = tokenize(definition);
5219 l = nasm_malloc(sizeof(Line));
5220 l->next = predef;
5221 l->first = def;
5222 l->finishes = NULL;
5223 predef = l;
5226 static void pp_extra_stdmac(macros_t *macros)
5228 extrastdmac = macros;
5231 static void make_tok_num(Token * tok, int64_t val)
5233 char numbuf[32];
5234 snprintf(numbuf, sizeof(numbuf), "%"PRId64"", val);
5235 tok->text = nasm_strdup(numbuf);
5236 tok->type = TOK_NUMBER;
5239 struct preproc_ops nasmpp = {
5240 pp_reset,
5241 pp_getline,
5242 pp_cleanup,
5243 pp_extra_stdmac,
5244 pp_pre_define,
5245 pp_pre_undefine,
5246 pp_pre_include,
5247 pp_include_path