nasmdoc: we still miss Chuck, but...
[nasm.git] / preproc.c
blobba1acd55507179da7987757dd18955fe34510cc4
1 /* ----------------------------------------------------------------------- *
3 * Copyright 1996-2010 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 struct Token {
216 Token *next;
217 char *text;
218 union {
219 SMacro *mac; /* associated macro for TOK_SMAC_END */
220 size_t len; /* scratch length field */
221 } a; /* Auxiliary data */
222 enum pp_token_type type;
226 * Multi-line macro definitions are stored as a linked list of
227 * these, which is essentially a container to allow several linked
228 * lists of Tokens.
230 * Note that in this module, linked lists are treated as stacks
231 * wherever possible. For this reason, Lines are _pushed_ on to the
232 * `expansion' field in MMacro structures, so that the linked list,
233 * if walked, would give the macro lines in reverse order; this
234 * means that we can walk the list when expanding a macro, and thus
235 * push the lines on to the `expansion' field in _istk_ in reverse
236 * order (so that when popped back off they are in the right
237 * order). It may seem cockeyed, and it relies on my design having
238 * an even number of steps in, but it works...
240 * Some of these structures, rather than being actual lines, are
241 * markers delimiting the end of the expansion of a given macro.
242 * This is for use in the cycle-tracking and %rep-handling code.
243 * Such structures have `finishes' non-NULL, and `first' NULL. All
244 * others have `finishes' NULL, but `first' may still be NULL if
245 * the line is blank.
247 struct Line {
248 Line *next;
249 MMacro *finishes;
250 Token *first;
254 * To handle an arbitrary level of file inclusion, we maintain a
255 * stack (ie linked list) of these things.
257 struct Include {
258 Include *next;
259 FILE *fp;
260 Cond *conds;
261 Line *expansion;
262 char *fname;
263 int lineno, lineinc;
264 MMacro *mstk; /* stack of active macros/reps */
268 * Include search path. This is simply a list of strings which get
269 * prepended, in turn, to the name of an include file, in an
270 * attempt to find the file if it's not in the current directory.
272 struct IncPath {
273 IncPath *next;
274 char *path;
278 * Conditional assembly: we maintain a separate stack of these for
279 * each level of file inclusion. (The only reason we keep the
280 * stacks separate is to ensure that a stray `%endif' in a file
281 * included from within the true branch of a `%if' won't terminate
282 * it and cause confusion: instead, rightly, it'll cause an error.)
284 struct Cond {
285 Cond *next;
286 int state;
288 enum {
290 * These states are for use just after %if or %elif: IF_TRUE
291 * means the condition has evaluated to truth so we are
292 * currently emitting, whereas IF_FALSE means we are not
293 * currently emitting but will start doing so if a %else comes
294 * up. In these states, all directives are admissible: %elif,
295 * %else and %endif. (And of course %if.)
297 COND_IF_TRUE, COND_IF_FALSE,
299 * These states come up after a %else: ELSE_TRUE means we're
300 * emitting, and ELSE_FALSE means we're not. In ELSE_* states,
301 * any %elif or %else will cause an error.
303 COND_ELSE_TRUE, COND_ELSE_FALSE,
305 * These states mean that we're not emitting now, and also that
306 * nothing until %endif will be emitted at all. COND_DONE is
307 * used when we've had our moment of emission
308 * and have now started seeing %elifs. COND_NEVER is used when
309 * the condition construct in question is contained within a
310 * non-emitting branch of a larger condition construct,
311 * or if there is an error.
313 COND_DONE, COND_NEVER
315 #define emitting(x) ( (x) == COND_IF_TRUE || (x) == COND_ELSE_TRUE )
318 * These defines are used as the possible return values for do_directive
320 #define NO_DIRECTIVE_FOUND 0
321 #define DIRECTIVE_FOUND 1
324 * This define sets the upper limit for smacro and recursive mmacro
325 * expansions
327 #define DEADMAN_LIMIT (1 << 20)
330 * Condition codes. Note that we use c_ prefix not C_ because C_ is
331 * used in nasm.h for the "real" condition codes. At _this_ level,
332 * we treat CXZ and ECXZ as condition codes, albeit non-invertible
333 * ones, so we need a different enum...
335 static const char * const conditions[] = {
336 "a", "ae", "b", "be", "c", "cxz", "e", "ecxz", "g", "ge", "l", "le",
337 "na", "nae", "nb", "nbe", "nc", "ne", "ng", "nge", "nl", "nle", "no",
338 "np", "ns", "nz", "o", "p", "pe", "po", "rcxz", "s", "z"
340 enum pp_conds {
341 c_A, c_AE, c_B, c_BE, c_C, c_CXZ, c_E, c_ECXZ, c_G, c_GE, c_L, c_LE,
342 c_NA, c_NAE, c_NB, c_NBE, c_NC, c_NE, c_NG, c_NGE, c_NL, c_NLE, c_NO,
343 c_NP, c_NS, c_NZ, c_O, c_P, c_PE, c_PO, c_RCXZ, c_S, c_Z,
344 c_none = -1
346 static const enum pp_conds inverse_ccs[] = {
347 c_NA, c_NAE, c_NB, c_NBE, c_NC, -1, c_NE, -1, c_NG, c_NGE, c_NL, c_NLE,
348 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,
349 c_Z, c_NO, c_NP, c_PO, c_PE, -1, c_NS, c_NZ
353 * Directive names.
355 /* If this is a an IF, ELIF, ELSE or ENDIF keyword */
356 static int is_condition(enum preproc_token arg)
358 return PP_IS_COND(arg) || (arg == PP_ELSE) || (arg == PP_ENDIF);
361 /* For TASM compatibility we need to be able to recognise TASM compatible
362 * conditional compilation directives. Using the NASM pre-processor does
363 * not work, so we look for them specifically from the following list and
364 * then jam in the equivalent NASM directive into the input stream.
367 enum {
368 TM_ARG, TM_ELIF, TM_ELSE, TM_ENDIF, TM_IF, TM_IFDEF, TM_IFDIFI,
369 TM_IFNDEF, TM_INCLUDE, TM_LOCAL
372 static const char * const tasm_directives[] = {
373 "arg", "elif", "else", "endif", "if", "ifdef", "ifdifi",
374 "ifndef", "include", "local"
377 static int StackSize = 4;
378 static char *StackPointer = "ebp";
379 static int ArgOffset = 8;
380 static int LocalOffset = 0;
382 static Context *cstk;
383 static Include *istk;
384 static IncPath *ipath = NULL;
386 static int pass; /* HACK: pass 0 = generate dependencies only */
387 static StrList **dephead, **deptail; /* Dependency list */
389 static uint64_t unique; /* unique identifier numbers */
391 static Line *predef = NULL;
392 static bool do_predef;
394 static ListGen *list;
397 * The current set of multi-line macros we have defined.
399 static struct hash_table mmacros;
402 * The current set of single-line macros we have defined.
404 static struct hash_table smacros;
407 * The multi-line macro we are currently defining, or the %rep
408 * block we are currently reading, if any.
410 static MMacro *defining;
412 static uint64_t nested_mac_count;
413 static uint64_t nested_rep_count;
416 * The number of macro parameters to allocate space for at a time.
418 #define PARAM_DELTA 16
421 * The standard macro set: defined in macros.c in the array nasm_stdmac.
422 * This gives our position in the macro set, when we're processing it.
424 static macros_t *stdmacpos;
427 * The extra standard macros that come from the object format, if
428 * any.
430 static macros_t *extrastdmac = NULL;
431 static bool any_extrastdmac;
434 * Tokens are allocated in blocks to improve speed
436 #define TOKEN_BLOCKSIZE 4096
437 static Token *freeTokens = NULL;
438 struct Blocks {
439 Blocks *next;
440 void *chunk;
443 static Blocks blocks = { NULL, NULL };
446 * Forward declarations.
448 static Token *expand_mmac_params(Token * tline);
449 static Token *expand_smacro(Token * tline);
450 static Token *expand_id(Token * tline);
451 static Context *get_ctx(const char *name, const char **namep,
452 bool all_contexts);
453 static void make_tok_num(Token * tok, int64_t val);
454 static void error(int severity, const char *fmt, ...);
455 static void error_precond(int severity, const char *fmt, ...);
456 static void *new_Block(size_t size);
457 static void delete_Blocks(void);
458 static Token *new_Token(Token * next, enum pp_token_type type,
459 const char *text, int txtlen);
460 static Token *delete_Token(Token * t);
463 * Macros for safe checking of token pointers, avoid *(NULL)
465 #define tok_type_(x,t) ((x) && (x)->type == (t))
466 #define skip_white_(x) if (tok_type_((x), TOK_WHITESPACE)) (x)=(x)->next
467 #define tok_is_(x,v) (tok_type_((x), TOK_OTHER) && !strcmp((x)->text,(v)))
468 #define tok_isnt_(x,v) ((x) && ((x)->type!=TOK_OTHER || strcmp((x)->text,(v))))
471 * Handle TASM specific directives, which do not contain a % in
472 * front of them. We do it here because I could not find any other
473 * place to do it for the moment, and it is a hack (ideally it would
474 * be nice to be able to use the NASM pre-processor to do it).
476 static char *check_tasm_directive(char *line)
478 int32_t i, j, k, m, len;
479 char *p, *q, *oldline, oldchar;
481 p = nasm_skip_spaces(line);
483 /* Binary search for the directive name */
484 i = -1;
485 j = ARRAY_SIZE(tasm_directives);
486 q = nasm_skip_word(p);
487 len = q - p;
488 if (len) {
489 oldchar = p[len];
490 p[len] = 0;
491 while (j - i > 1) {
492 k = (j + i) / 2;
493 m = nasm_stricmp(p, tasm_directives[k]);
494 if (m == 0) {
495 /* We have found a directive, so jam a % in front of it
496 * so that NASM will then recognise it as one if it's own.
498 p[len] = oldchar;
499 len = strlen(p);
500 oldline = line;
501 line = nasm_malloc(len + 2);
502 line[0] = '%';
503 if (k == TM_IFDIFI) {
505 * NASM does not recognise IFDIFI, so we convert
506 * it to %if 0. This is not used in NASM
507 * compatible code, but does need to parse for the
508 * TASM macro package.
510 strcpy(line + 1, "if 0");
511 } else {
512 memcpy(line + 1, p, len + 1);
514 nasm_free(oldline);
515 return line;
516 } else if (m < 0) {
517 j = k;
518 } else
519 i = k;
521 p[len] = oldchar;
523 return line;
527 * The pre-preprocessing stage... This function translates line
528 * number indications as they emerge from GNU cpp (`# lineno "file"
529 * flags') into NASM preprocessor line number indications (`%line
530 * lineno file').
532 static char *prepreproc(char *line)
534 int lineno, fnlen;
535 char *fname, *oldline;
537 if (line[0] == '#' && line[1] == ' ') {
538 oldline = line;
539 fname = oldline + 2;
540 lineno = atoi(fname);
541 fname += strspn(fname, "0123456789 ");
542 if (*fname == '"')
543 fname++;
544 fnlen = strcspn(fname, "\"");
545 line = nasm_malloc(20 + fnlen);
546 snprintf(line, 20 + fnlen, "%%line %d %.*s", lineno, fnlen, fname);
547 nasm_free(oldline);
549 if (tasm_compatible_mode)
550 return check_tasm_directive(line);
551 return line;
555 * Free a linked list of tokens.
557 static void free_tlist(Token * list)
559 while (list)
560 list = delete_Token(list);
564 * Free a linked list of lines.
566 static void free_llist(Line * list)
568 Line *l, *tmp;
569 list_for_each_safe(l, tmp, list) {
570 free_tlist(l->first);
571 nasm_free(l);
576 * Free an MMacro
578 static void free_mmacro(MMacro * m)
580 nasm_free(m->name);
581 free_tlist(m->dlist);
582 nasm_free(m->defaults);
583 free_llist(m->expansion);
584 nasm_free(m);
588 * Free all currently defined macros, and free the hash tables
590 static void free_smacro_table(struct hash_table *smt)
592 SMacro *s, *tmp;
593 const char *key;
594 struct hash_tbl_node *it = NULL;
596 while ((s = hash_iterate(smt, &it, &key)) != NULL) {
597 nasm_free((void *)key);
598 list_for_each_safe(s, tmp, s) {
599 nasm_free(s->name);
600 free_tlist(s->expansion);
601 nasm_free(s);
604 hash_free(smt);
607 static void free_mmacro_table(struct hash_table *mmt)
609 MMacro *m, *tmp;
610 const char *key;
611 struct hash_tbl_node *it = NULL;
613 it = NULL;
614 while ((m = hash_iterate(mmt, &it, &key)) != NULL) {
615 nasm_free((void *)key);
616 list_for_each_safe(m ,tmp, m)
617 free_mmacro(m);
619 hash_free(mmt);
622 static void free_macros(void)
624 free_smacro_table(&smacros);
625 free_mmacro_table(&mmacros);
629 * Initialize the hash tables
631 static void init_macros(void)
633 hash_init(&smacros, HASH_LARGE);
634 hash_init(&mmacros, HASH_LARGE);
638 * Pop the context stack.
640 static void ctx_pop(void)
642 Context *c = cstk;
644 cstk = cstk->next;
645 free_smacro_table(&c->localmac);
646 nasm_free(c->name);
647 nasm_free(c);
651 * Search for a key in the hash index; adding it if necessary
652 * (in which case we initialize the data pointer to NULL.)
654 static void **
655 hash_findi_add(struct hash_table *hash, const char *str)
657 struct hash_insert hi;
658 void **r;
659 char *strx;
661 r = hash_findi(hash, str, &hi);
662 if (r)
663 return r;
665 strx = nasm_strdup(str); /* Use a more efficient allocator here? */
666 return hash_add(&hi, strx, NULL);
670 * Like hash_findi, but returns the data element rather than a pointer
671 * to it. Used only when not adding a new element, hence no third
672 * argument.
674 static void *
675 hash_findix(struct hash_table *hash, const char *str)
677 void **p;
679 p = hash_findi(hash, str, NULL);
680 return p ? *p : NULL;
684 * read line from standart macros set,
685 * if there no more left -- return NULL
687 static char *line_from_stdmac(void)
689 unsigned char c;
690 const unsigned char *p = stdmacpos;
691 char *line, *q;
692 size_t len = 0;
694 if (!stdmacpos)
695 return NULL;
697 while ((c = *p++)) {
698 if (c >= 0x80)
699 len += pp_directives_len[c - 0x80] + 1;
700 else
701 len++;
704 line = nasm_malloc(len + 1);
705 q = line;
706 while ((c = *stdmacpos++)) {
707 if (c >= 0x80) {
708 memcpy(q, pp_directives[c - 0x80], pp_directives_len[c - 0x80]);
709 q += pp_directives_len[c - 0x80];
710 *q++ = ' ';
711 } else {
712 *q++ = c;
715 stdmacpos = p;
716 *q = '\0';
718 if (!*stdmacpos) {
719 /* This was the last of the standard macro chain... */
720 stdmacpos = NULL;
721 if (any_extrastdmac) {
722 stdmacpos = extrastdmac;
723 any_extrastdmac = false;
724 } else if (do_predef) {
725 Line *pd, *l;
726 Token *head, **tail, *t;
729 * Nasty hack: here we push the contents of
730 * `predef' on to the top-level expansion stack,
731 * since this is the most convenient way to
732 * implement the pre-include and pre-define
733 * features.
735 list_for_each(pd, predef) {
736 head = NULL;
737 tail = &head;
738 list_for_each(t, pd->first) {
739 *tail = new_Token(NULL, t->type, t->text, 0);
740 tail = &(*tail)->next;
743 l = nasm_malloc(sizeof(Line));
744 l->next = istk->expansion;
745 l->first = head;
746 l->finishes = NULL;
748 istk->expansion = l;
750 do_predef = false;
754 return line;
757 #define BUF_DELTA 512
759 * Read a line from the top file in istk, handling multiple CR/LFs
760 * at the end of the line read, and handling spurious ^Zs. Will
761 * return lines from the standard macro set if this has not already
762 * been done.
764 static char *read_line(void)
766 char *buffer, *p, *q;
767 int bufsize, continued_count;
770 * standart macros set (predefined) goes first
772 p = line_from_stdmac();
773 if (p)
774 return p;
777 * regular read from a file
779 bufsize = BUF_DELTA;
780 buffer = nasm_malloc(BUF_DELTA);
781 p = buffer;
782 continued_count = 0;
783 while (1) {
784 q = fgets(p, bufsize - (p - buffer), istk->fp);
785 if (!q)
786 break;
787 p += strlen(p);
788 if (p > buffer && p[-1] == '\n') {
790 * Convert backslash-CRLF line continuation sequences into
791 * nothing at all (for DOS and Windows)
793 if (((p - 2) > buffer) && (p[-3] == '\\') && (p[-2] == '\r')) {
794 p -= 3;
795 *p = 0;
796 continued_count++;
799 * Also convert backslash-LF line continuation sequences into
800 * nothing at all (for Unix)
802 else if (((p - 1) > buffer) && (p[-2] == '\\')) {
803 p -= 2;
804 *p = 0;
805 continued_count++;
806 } else {
807 break;
810 if (p - buffer > bufsize - 10) {
811 int32_t offset = p - buffer;
812 bufsize += BUF_DELTA;
813 buffer = nasm_realloc(buffer, bufsize);
814 p = buffer + offset; /* prevent stale-pointer problems */
818 if (!q && p == buffer) {
819 nasm_free(buffer);
820 return NULL;
823 src_set_linnum(src_get_linnum() + istk->lineinc +
824 (continued_count * istk->lineinc));
827 * Play safe: remove CRs as well as LFs, if any of either are
828 * present at the end of the line.
830 while (--p >= buffer && (*p == '\n' || *p == '\r'))
831 *p = '\0';
834 * Handle spurious ^Z, which may be inserted into source files
835 * by some file transfer utilities.
837 buffer[strcspn(buffer, "\032")] = '\0';
839 list->line(LIST_READ, buffer);
841 return buffer;
845 * Tokenize a line of text. This is a very simple process since we
846 * don't need to parse the value out of e.g. numeric tokens: we
847 * simply split one string into many.
849 static Token *tokenize(char *line)
851 char c, *p = line;
852 enum pp_token_type type;
853 Token *list = NULL;
854 Token *t, **tail = &list;
856 while (*line) {
857 p = line;
858 if (*p == '%') {
859 p++;
860 if (*p == '+' && !nasm_isdigit(p[1])) {
861 p++;
862 type = TOK_PASTE;
863 } else if (nasm_isdigit(*p) ||
864 ((*p == '-' || *p == '+') && nasm_isdigit(p[1]))) {
865 do {
866 p++;
868 while (nasm_isdigit(*p));
869 type = TOK_PREPROC_ID;
870 } else if (*p == '{') {
871 p++;
872 while (*p && *p != '}') {
873 p[-1] = *p;
874 p++;
876 p[-1] = '\0';
877 if (*p)
878 p++;
879 type = TOK_PREPROC_ID;
880 } else if (*p == '[') {
881 int lvl = 1;
882 line += 2; /* Skip the leading %[ */
883 p++;
884 while (lvl && (c = *p++)) {
885 switch (c) {
886 case ']':
887 lvl--;
888 break;
889 case '%':
890 if (*p == '[')
891 lvl++;
892 break;
893 case '\'':
894 case '\"':
895 case '`':
896 p = nasm_skip_string(p - 1) + 1;
897 break;
898 default:
899 break;
902 p--;
903 if (*p)
904 *p++ = '\0';
905 if (lvl)
906 error(ERR_NONFATAL, "unterminated %[ construct");
907 type = TOK_INDIRECT;
908 } else if (*p == '?') {
909 type = TOK_PREPROC_Q; /* %? */
910 p++;
911 if (*p == '?') {
912 type = TOK_PREPROC_QQ; /* %?? */
913 p++;
915 } else if (isidchar(*p) ||
916 ((*p == '!' || *p == '%' || *p == '$') &&
917 isidchar(p[1]))) {
918 do {
919 p++;
921 while (isidchar(*p));
922 type = TOK_PREPROC_ID;
923 } else {
924 type = TOK_OTHER;
925 if (*p == '%')
926 p++;
928 } else if (isidstart(*p) || (*p == '$' && isidstart(p[1]))) {
929 type = TOK_ID;
930 p++;
931 while (*p && isidchar(*p))
932 p++;
933 } else if (*p == '\'' || *p == '"' || *p == '`') {
935 * A string token.
937 type = TOK_STRING;
938 p = nasm_skip_string(p);
940 if (*p) {
941 p++;
942 } else {
943 error(ERR_WARNING|ERR_PASS1, "unterminated string");
944 /* Handling unterminated strings by UNV */
945 /* type = -1; */
947 } else if (p[0] == '$' && p[1] == '$') {
948 type = TOK_OTHER; /* TOKEN_BASE */
949 p += 2;
950 } else if (isnumstart(*p)) {
951 bool is_hex = false;
952 bool is_float = false;
953 bool has_e = false;
954 char c, *r;
957 * A numeric token.
960 if (*p == '$') {
961 p++;
962 is_hex = true;
965 for (;;) {
966 c = *p++;
968 if (!is_hex && (c == 'e' || c == 'E')) {
969 has_e = true;
970 if (*p == '+' || *p == '-') {
972 * e can only be followed by +/- if it is either a
973 * prefixed hex number or a floating-point number
975 p++;
976 is_float = true;
978 } else if (c == 'H' || c == 'h' || c == 'X' || c == 'x') {
979 is_hex = true;
980 } else if (c == 'P' || c == 'p') {
981 is_float = true;
982 if (*p == '+' || *p == '-')
983 p++;
984 } else if (isnumchar(c) || c == '_')
985 ; /* just advance */
986 else if (c == '.') {
988 * we need to deal with consequences of the legacy
989 * parser, like "1.nolist" being two tokens
990 * (TOK_NUMBER, TOK_ID) here; at least give it
991 * a shot for now. In the future, we probably need
992 * a flex-based scanner with proper pattern matching
993 * to do it as well as it can be done. Nothing in
994 * the world is going to help the person who wants
995 * 0x123.p16 interpreted as two tokens, though.
997 r = p;
998 while (*r == '_')
999 r++;
1001 if (nasm_isdigit(*r) || (is_hex && nasm_isxdigit(*r)) ||
1002 (!is_hex && (*r == 'e' || *r == 'E')) ||
1003 (*r == 'p' || *r == 'P')) {
1004 p = r;
1005 is_float = true;
1006 } else
1007 break; /* Terminate the token */
1008 } else
1009 break;
1011 p--; /* Point to first character beyond number */
1013 if (p == line+1 && *line == '$') {
1014 type = TOK_OTHER; /* TOKEN_HERE */
1015 } else {
1016 if (has_e && !is_hex) {
1017 /* 1e13 is floating-point, but 1e13h is not */
1018 is_float = true;
1021 type = is_float ? TOK_FLOAT : TOK_NUMBER;
1023 } else if (nasm_isspace(*p)) {
1024 type = TOK_WHITESPACE;
1025 p = nasm_skip_spaces(p);
1027 * Whitespace just before end-of-line is discarded by
1028 * pretending it's a comment; whitespace just before a
1029 * comment gets lumped into the comment.
1031 if (!*p || *p == ';') {
1032 type = TOK_COMMENT;
1033 while (*p)
1034 p++;
1036 } else if (*p == ';') {
1037 type = TOK_COMMENT;
1038 while (*p)
1039 p++;
1040 } else {
1042 * Anything else is an operator of some kind. We check
1043 * for all the double-character operators (>>, <<, //,
1044 * %%, <=, >=, ==, !=, <>, &&, ||, ^^), but anything
1045 * else is a single-character operator.
1047 type = TOK_OTHER;
1048 if ((p[0] == '>' && p[1] == '>') ||
1049 (p[0] == '<' && p[1] == '<') ||
1050 (p[0] == '/' && p[1] == '/') ||
1051 (p[0] == '<' && p[1] == '=') ||
1052 (p[0] == '>' && p[1] == '=') ||
1053 (p[0] == '=' && p[1] == '=') ||
1054 (p[0] == '!' && p[1] == '=') ||
1055 (p[0] == '<' && p[1] == '>') ||
1056 (p[0] == '&' && p[1] == '&') ||
1057 (p[0] == '|' && p[1] == '|') ||
1058 (p[0] == '^' && p[1] == '^')) {
1059 p++;
1061 p++;
1064 /* Handling unterminated string by UNV */
1065 /*if (type == -1)
1067 *tail = t = new_Token(NULL, TOK_STRING, line, p-line+1);
1068 t->text[p-line] = *line;
1069 tail = &t->next;
1071 else */
1072 if (type != TOK_COMMENT) {
1073 *tail = t = new_Token(NULL, type, line, p - line);
1074 tail = &t->next;
1076 line = p;
1078 return list;
1082 * this function allocates a new managed block of memory and
1083 * returns a pointer to the block. The managed blocks are
1084 * deleted only all at once by the delete_Blocks function.
1086 static void *new_Block(size_t size)
1088 Blocks *b = &blocks;
1090 /* first, get to the end of the linked list */
1091 while (b->next)
1092 b = b->next;
1093 /* now allocate the requested chunk */
1094 b->chunk = nasm_malloc(size);
1096 /* now allocate a new block for the next request */
1097 b->next = nasm_malloc(sizeof(Blocks));
1098 /* and initialize the contents of the new block */
1099 b->next->next = NULL;
1100 b->next->chunk = NULL;
1101 return b->chunk;
1105 * this function deletes all managed blocks of memory
1107 static void delete_Blocks(void)
1109 Blocks *a, *b = &blocks;
1112 * keep in mind that the first block, pointed to by blocks
1113 * is a static and not dynamically allocated, so we don't
1114 * free it.
1116 while (b) {
1117 if (b->chunk)
1118 nasm_free(b->chunk);
1119 a = b;
1120 b = b->next;
1121 if (a != &blocks)
1122 nasm_free(a);
1127 * this function creates a new Token and passes a pointer to it
1128 * back to the caller. It sets the type and text elements, and
1129 * also the a.mac and next elements to NULL.
1131 static Token *new_Token(Token * next, enum pp_token_type type,
1132 const char *text, int txtlen)
1134 Token *t;
1135 int i;
1137 if (!freeTokens) {
1138 freeTokens = (Token *) new_Block(TOKEN_BLOCKSIZE * sizeof(Token));
1139 for (i = 0; i < TOKEN_BLOCKSIZE - 1; i++)
1140 freeTokens[i].next = &freeTokens[i + 1];
1141 freeTokens[i].next = NULL;
1143 t = freeTokens;
1144 freeTokens = t->next;
1145 t->next = next;
1146 t->a.mac = NULL;
1147 t->type = type;
1148 if (type == TOK_WHITESPACE || !text) {
1149 t->text = NULL;
1150 } else {
1151 if (txtlen == 0)
1152 txtlen = strlen(text);
1153 t->text = nasm_malloc(txtlen+1);
1154 memcpy(t->text, text, txtlen);
1155 t->text[txtlen] = '\0';
1157 return t;
1160 static Token *delete_Token(Token * t)
1162 Token *next = t->next;
1163 nasm_free(t->text);
1164 t->next = freeTokens;
1165 freeTokens = t;
1166 return next;
1170 * Convert a line of tokens back into text.
1171 * If expand_locals is not zero, identifiers of the form "%$*xxx"
1172 * will be transformed into ..@ctxnum.xxx
1174 static char *detoken(Token * tlist, bool expand_locals)
1176 Token *t;
1177 char *line, *p;
1178 const char *q;
1179 int len = 0;
1181 list_for_each(t, tlist) {
1182 if (t->type == TOK_PREPROC_ID && t->text[1] == '!') {
1183 char *p = getenv(t->text + 2);
1184 char *q = t->text;
1185 if (!p) {
1186 error(ERR_NONFATAL | ERR_PASS1,
1187 "nonexistent environment variable `%s'", q + 2);
1188 p = "";
1190 t->text = nasm_strdup(p);
1191 nasm_free(q);
1193 /* Expand local macros here and not during preprocessing */
1194 if (expand_locals &&
1195 t->type == TOK_PREPROC_ID && t->text &&
1196 t->text[0] == '%' && t->text[1] == '$') {
1197 const char *q;
1198 char *p;
1199 Context *ctx = get_ctx(t->text, &q, false);
1200 if (ctx) {
1201 char buffer[40];
1202 snprintf(buffer, sizeof(buffer), "..@%"PRIu32".", ctx->number);
1203 p = nasm_strcat(buffer, q);
1204 nasm_free(t->text);
1205 t->text = p;
1208 if (t->type == TOK_WHITESPACE)
1209 len++;
1210 else if (t->text)
1211 len += strlen(t->text);
1214 p = line = nasm_malloc(len + 1);
1216 list_for_each(t, tlist) {
1217 if (t->type == TOK_WHITESPACE) {
1218 *p++ = ' ';
1219 } else if (t->text) {
1220 q = t->text;
1221 while (*q)
1222 *p++ = *q++;
1225 *p = '\0';
1227 return line;
1231 * A scanner, suitable for use by the expression evaluator, which
1232 * operates on a line of Tokens. Expects a pointer to a pointer to
1233 * the first token in the line to be passed in as its private_data
1234 * field.
1236 * FIX: This really needs to be unified with stdscan.
1238 static int ppscan(void *private_data, struct tokenval *tokval)
1240 Token **tlineptr = private_data;
1241 Token *tline;
1242 char ourcopy[MAX_KEYWORD+1], *p, *r, *s;
1244 do {
1245 tline = *tlineptr;
1246 *tlineptr = tline ? tline->next : NULL;
1247 } while (tline && (tline->type == TOK_WHITESPACE ||
1248 tline->type == TOK_COMMENT));
1250 if (!tline)
1251 return tokval->t_type = TOKEN_EOS;
1253 tokval->t_charptr = tline->text;
1255 if (tline->text[0] == '$' && !tline->text[1])
1256 return tokval->t_type = TOKEN_HERE;
1257 if (tline->text[0] == '$' && tline->text[1] == '$' && !tline->text[2])
1258 return tokval->t_type = TOKEN_BASE;
1260 if (tline->type == TOK_ID) {
1261 p = tokval->t_charptr = tline->text;
1262 if (p[0] == '$') {
1263 tokval->t_charptr++;
1264 return tokval->t_type = TOKEN_ID;
1267 for (r = p, s = ourcopy; *r; r++) {
1268 if (r >= p+MAX_KEYWORD)
1269 return tokval->t_type = TOKEN_ID; /* Not a keyword */
1270 *s++ = nasm_tolower(*r);
1272 *s = '\0';
1273 /* right, so we have an identifier sitting in temp storage. now,
1274 * is it actually a register or instruction name, or what? */
1275 return nasm_token_hash(ourcopy, tokval);
1278 if (tline->type == TOK_NUMBER) {
1279 bool rn_error;
1280 tokval->t_integer = readnum(tline->text, &rn_error);
1281 tokval->t_charptr = tline->text;
1282 if (rn_error)
1283 return tokval->t_type = TOKEN_ERRNUM;
1284 else
1285 return tokval->t_type = TOKEN_NUM;
1288 if (tline->type == TOK_FLOAT) {
1289 return tokval->t_type = TOKEN_FLOAT;
1292 if (tline->type == TOK_STRING) {
1293 char bq, *ep;
1295 bq = tline->text[0];
1296 tokval->t_charptr = tline->text;
1297 tokval->t_inttwo = nasm_unquote(tline->text, &ep);
1299 if (ep[0] != bq || ep[1] != '\0')
1300 return tokval->t_type = TOKEN_ERRSTR;
1301 else
1302 return tokval->t_type = TOKEN_STR;
1305 if (tline->type == TOK_OTHER) {
1306 if (!strcmp(tline->text, "<<"))
1307 return tokval->t_type = TOKEN_SHL;
1308 if (!strcmp(tline->text, ">>"))
1309 return tokval->t_type = TOKEN_SHR;
1310 if (!strcmp(tline->text, "//"))
1311 return tokval->t_type = TOKEN_SDIV;
1312 if (!strcmp(tline->text, "%%"))
1313 return tokval->t_type = TOKEN_SMOD;
1314 if (!strcmp(tline->text, "=="))
1315 return tokval->t_type = TOKEN_EQ;
1316 if (!strcmp(tline->text, "<>"))
1317 return tokval->t_type = TOKEN_NE;
1318 if (!strcmp(tline->text, "!="))
1319 return tokval->t_type = TOKEN_NE;
1320 if (!strcmp(tline->text, "<="))
1321 return tokval->t_type = TOKEN_LE;
1322 if (!strcmp(tline->text, ">="))
1323 return tokval->t_type = TOKEN_GE;
1324 if (!strcmp(tline->text, "&&"))
1325 return tokval->t_type = TOKEN_DBL_AND;
1326 if (!strcmp(tline->text, "^^"))
1327 return tokval->t_type = TOKEN_DBL_XOR;
1328 if (!strcmp(tline->text, "||"))
1329 return tokval->t_type = TOKEN_DBL_OR;
1333 * We have no other options: just return the first character of
1334 * the token text.
1336 return tokval->t_type = tline->text[0];
1340 * Compare a string to the name of an existing macro; this is a
1341 * simple wrapper which calls either strcmp or nasm_stricmp
1342 * depending on the value of the `casesense' parameter.
1344 static int mstrcmp(const char *p, const char *q, bool casesense)
1346 return casesense ? strcmp(p, q) : nasm_stricmp(p, q);
1350 * Compare a string to the name of an existing macro; this is a
1351 * simple wrapper which calls either strcmp or nasm_stricmp
1352 * depending on the value of the `casesense' parameter.
1354 static int mmemcmp(const char *p, const char *q, size_t l, bool casesense)
1356 return casesense ? memcmp(p, q, l) : nasm_memicmp(p, q, l);
1360 * Return the Context structure associated with a %$ token. Return
1361 * NULL, having _already_ reported an error condition, if the
1362 * context stack isn't deep enough for the supplied number of $
1363 * signs.
1364 * If all_contexts == true, contexts that enclose current are
1365 * also scanned for such smacro, until it is found; if not -
1366 * only the context that directly results from the number of $'s
1367 * in variable's name.
1369 * If "namep" is non-NULL, set it to the pointer to the macro name
1370 * tail, i.e. the part beyond %$...
1372 static Context *get_ctx(const char *name, const char **namep,
1373 bool all_contexts)
1375 Context *ctx;
1376 SMacro *m;
1377 int i;
1379 if (namep)
1380 *namep = name;
1382 if (!name || name[0] != '%' || name[1] != '$')
1383 return NULL;
1385 if (!cstk) {
1386 error(ERR_NONFATAL, "`%s': context stack is empty", name);
1387 return NULL;
1390 name += 2;
1391 ctx = cstk;
1392 i = 0;
1393 while (ctx && *name == '$') {
1394 name++;
1395 i++;
1396 ctx = ctx->next;
1398 if (!ctx) {
1399 error(ERR_NONFATAL, "`%s': context stack is only"
1400 " %d level%s deep", name, i, (i == 1 ? "" : "s"));
1401 return NULL;
1404 if (namep)
1405 *namep = name;
1407 if (!all_contexts)
1408 return ctx;
1410 do {
1411 /* Search for this smacro in found context */
1412 m = hash_findix(&ctx->localmac, name);
1413 while (m) {
1414 if (!mstrcmp(m->name, name, m->casesense))
1415 return ctx;
1416 m = m->next;
1418 ctx = ctx->next;
1420 while (ctx);
1421 return NULL;
1425 * Check to see if a file is already in a string list
1427 static bool in_list(const StrList *list, const char *str)
1429 while (list) {
1430 if (!strcmp(list->str, str))
1431 return true;
1432 list = list->next;
1434 return false;
1438 * Open an include file. This routine must always return a valid
1439 * file pointer if it returns - it's responsible for throwing an
1440 * ERR_FATAL and bombing out completely if not. It should also try
1441 * the include path one by one until it finds the file or reaches
1442 * the end of the path.
1444 static FILE *inc_fopen(const char *file, StrList **dhead, StrList ***dtail,
1445 bool missing_ok)
1447 FILE *fp;
1448 char *prefix = "";
1449 IncPath *ip = ipath;
1450 int len = strlen(file);
1451 size_t prefix_len = 0;
1452 StrList *sl;
1454 while (1) {
1455 sl = nasm_malloc(prefix_len+len+1+sizeof sl->next);
1456 memcpy(sl->str, prefix, prefix_len);
1457 memcpy(sl->str+prefix_len, file, len+1);
1458 fp = fopen(sl->str, "r");
1459 if (fp && dhead && !in_list(*dhead, sl->str)) {
1460 sl->next = NULL;
1461 **dtail = sl;
1462 *dtail = &sl->next;
1463 } else {
1464 nasm_free(sl);
1466 if (fp)
1467 return fp;
1468 if (!ip) {
1469 if (!missing_ok)
1470 break;
1471 prefix = NULL;
1472 } else {
1473 prefix = ip->path;
1474 ip = ip->next;
1476 if (prefix) {
1477 prefix_len = strlen(prefix);
1478 } else {
1479 /* -MG given and file not found */
1480 if (dhead && !in_list(*dhead, file)) {
1481 sl = nasm_malloc(len+1+sizeof sl->next);
1482 sl->next = NULL;
1483 strcpy(sl->str, file);
1484 **dtail = sl;
1485 *dtail = &sl->next;
1487 return NULL;
1491 error(ERR_FATAL, "unable to open include file `%s'", file);
1492 return NULL;
1496 * Determine if we should warn on defining a single-line macro of
1497 * name `name', with `nparam' parameters. If nparam is 0 or -1, will
1498 * return true if _any_ single-line macro of that name is defined.
1499 * Otherwise, will return true if a single-line macro with either
1500 * `nparam' or no parameters is defined.
1502 * If a macro with precisely the right number of parameters is
1503 * defined, or nparam is -1, the address of the definition structure
1504 * will be returned in `defn'; otherwise NULL will be returned. If `defn'
1505 * is NULL, no action will be taken regarding its contents, and no
1506 * error will occur.
1508 * Note that this is also called with nparam zero to resolve
1509 * `ifdef'.
1511 * If you already know which context macro belongs to, you can pass
1512 * the context pointer as first parameter; if you won't but name begins
1513 * with %$ the context will be automatically computed. If all_contexts
1514 * is true, macro will be searched in outer contexts as well.
1516 static bool
1517 smacro_defined(Context * ctx, const char *name, int nparam, SMacro ** defn,
1518 bool nocase)
1520 struct hash_table *smtbl;
1521 SMacro *m;
1523 if (ctx) {
1524 smtbl = &ctx->localmac;
1525 } else if (name[0] == '%' && name[1] == '$') {
1526 if (cstk)
1527 ctx = get_ctx(name, &name, false);
1528 if (!ctx)
1529 return false; /* got to return _something_ */
1530 smtbl = &ctx->localmac;
1531 } else {
1532 smtbl = &smacros;
1534 m = (SMacro *) hash_findix(smtbl, name);
1536 while (m) {
1537 if (!mstrcmp(m->name, name, m->casesense && nocase) &&
1538 (nparam <= 0 || m->nparam == 0 || nparam == (int) m->nparam)) {
1539 if (defn) {
1540 if (nparam == (int) m->nparam || nparam == -1)
1541 *defn = m;
1542 else
1543 *defn = NULL;
1545 return true;
1547 m = m->next;
1550 return false;
1554 * Count and mark off the parameters in a multi-line macro call.
1555 * This is called both from within the multi-line macro expansion
1556 * code, and also to mark off the default parameters when provided
1557 * in a %macro definition line.
1559 static void count_mmac_params(Token * t, int *nparam, Token *** params)
1561 int paramsize, brace;
1563 *nparam = paramsize = 0;
1564 *params = NULL;
1565 while (t) {
1566 /* +1: we need space for the final NULL */
1567 if (*nparam+1 >= paramsize) {
1568 paramsize += PARAM_DELTA;
1569 *params = nasm_realloc(*params, sizeof(**params) * paramsize);
1571 skip_white_(t);
1572 brace = false;
1573 if (tok_is_(t, "{"))
1574 brace = true;
1575 (*params)[(*nparam)++] = t;
1576 while (tok_isnt_(t, brace ? "}" : ","))
1577 t = t->next;
1578 if (t) { /* got a comma/brace */
1579 t = t->next;
1580 if (brace) {
1582 * Now we've found the closing brace, look further
1583 * for the comma.
1585 skip_white_(t);
1586 if (tok_isnt_(t, ",")) {
1587 error(ERR_NONFATAL,
1588 "braces do not enclose all of macro parameter");
1589 while (tok_isnt_(t, ","))
1590 t = t->next;
1592 if (t)
1593 t = t->next; /* eat the comma */
1600 * Determine whether one of the various `if' conditions is true or
1601 * not.
1603 * We must free the tline we get passed.
1605 static bool if_condition(Token * tline, enum preproc_token ct)
1607 enum pp_conditional i = PP_COND(ct);
1608 bool j;
1609 Token *t, *tt, **tptr, *origline;
1610 struct tokenval tokval;
1611 expr *evalresult;
1612 enum pp_token_type needtype;
1613 const char *p;
1615 origline = tline;
1617 switch (i) {
1618 case PPC_IFCTX:
1619 j = false; /* have we matched yet? */
1620 while (true) {
1621 skip_white_(tline);
1622 if (!tline)
1623 break;
1624 if (tline->type != TOK_ID) {
1625 error(ERR_NONFATAL,
1626 "`%s' expects context identifiers", pp_directives[ct]);
1627 free_tlist(origline);
1628 return -1;
1630 if (cstk && cstk->name && !nasm_stricmp(tline->text, cstk->name))
1631 j = true;
1632 tline = tline->next;
1634 break;
1636 case PPC_IFDEF:
1637 j = false; /* have we matched yet? */
1638 while (tline) {
1639 skip_white_(tline);
1640 if (!tline || (tline->type != TOK_ID &&
1641 (tline->type != TOK_PREPROC_ID ||
1642 tline->text[1] != '$'))) {
1643 error(ERR_NONFATAL,
1644 "`%s' expects macro identifiers", pp_directives[ct]);
1645 goto fail;
1647 if (smacro_defined(NULL, tline->text, 0, NULL, true))
1648 j = true;
1649 tline = tline->next;
1651 break;
1653 case PPC_IFENV:
1654 tline = expand_smacro(tline);
1655 j = false; /* have we matched yet? */
1656 while (tline) {
1657 skip_white_(tline);
1658 if (!tline || (tline->type != TOK_ID &&
1659 (tline->type != TOK_PREPROC_ID ||
1660 tline->text[1] != '!'))) {
1661 error(ERR_NONFATAL,
1662 "`%s' expects environment variable names",
1663 pp_directives[ct]);
1664 goto fail;
1666 p = tline->type == TOK_ID ? tline->text : tline->text + 2;
1667 if (getenv(p))
1668 j = true;
1669 tline = tline->next;
1671 break;
1673 case PPC_IFIDN:
1674 case PPC_IFIDNI:
1675 tline = expand_smacro(tline);
1676 t = tt = tline;
1677 while (tok_isnt_(tt, ","))
1678 tt = tt->next;
1679 if (!tt) {
1680 error(ERR_NONFATAL,
1681 "`%s' expects two comma-separated arguments",
1682 pp_directives[ct]);
1683 goto fail;
1685 tt = tt->next;
1686 j = true; /* assume equality unless proved not */
1687 while ((t->type != TOK_OTHER || strcmp(t->text, ",")) && tt) {
1688 if (tt->type == TOK_OTHER && !strcmp(tt->text, ",")) {
1689 error(ERR_NONFATAL, "`%s': more than one comma on line",
1690 pp_directives[ct]);
1691 goto fail;
1693 if (t->type == TOK_WHITESPACE) {
1694 t = t->next;
1695 continue;
1697 if (tt->type == TOK_WHITESPACE) {
1698 tt = tt->next;
1699 continue;
1701 if (tt->type != t->type) {
1702 j = false; /* found mismatching tokens */
1703 break;
1705 /* When comparing strings, need to unquote them first */
1706 if (t->type == TOK_STRING) {
1707 size_t l1 = nasm_unquote(t->text, NULL);
1708 size_t l2 = nasm_unquote(tt->text, NULL);
1710 if (l1 != l2) {
1711 j = false;
1712 break;
1714 if (mmemcmp(t->text, tt->text, l1, i == PPC_IFIDN)) {
1715 j = false;
1716 break;
1718 } else if (mstrcmp(tt->text, t->text, i == PPC_IFIDN) != 0) {
1719 j = false; /* found mismatching tokens */
1720 break;
1723 t = t->next;
1724 tt = tt->next;
1726 if ((t->type != TOK_OTHER || strcmp(t->text, ",")) || tt)
1727 j = false; /* trailing gunk on one end or other */
1728 break;
1730 case PPC_IFMACRO:
1732 bool found = false;
1733 MMacro searching, *mmac;
1735 skip_white_(tline);
1736 tline = expand_id(tline);
1737 if (!tok_type_(tline, TOK_ID)) {
1738 error(ERR_NONFATAL,
1739 "`%s' expects a macro name", pp_directives[ct]);
1740 goto fail;
1742 searching.name = nasm_strdup(tline->text);
1743 searching.casesense = true;
1744 searching.plus = false;
1745 searching.nolist = false;
1746 searching.in_progress = 0;
1747 searching.max_depth = 0;
1748 searching.rep_nest = NULL;
1749 searching.nparam_min = 0;
1750 searching.nparam_max = INT_MAX;
1751 tline = expand_smacro(tline->next);
1752 skip_white_(tline);
1753 if (!tline) {
1754 } else if (!tok_type_(tline, TOK_NUMBER)) {
1755 error(ERR_NONFATAL,
1756 "`%s' expects a parameter count or nothing",
1757 pp_directives[ct]);
1758 } else {
1759 searching.nparam_min = searching.nparam_max =
1760 readnum(tline->text, &j);
1761 if (j)
1762 error(ERR_NONFATAL,
1763 "unable to parse parameter count `%s'",
1764 tline->text);
1766 if (tline && tok_is_(tline->next, "-")) {
1767 tline = tline->next->next;
1768 if (tok_is_(tline, "*"))
1769 searching.nparam_max = INT_MAX;
1770 else if (!tok_type_(tline, TOK_NUMBER))
1771 error(ERR_NONFATAL,
1772 "`%s' expects a parameter count after `-'",
1773 pp_directives[ct]);
1774 else {
1775 searching.nparam_max = readnum(tline->text, &j);
1776 if (j)
1777 error(ERR_NONFATAL,
1778 "unable to parse parameter count `%s'",
1779 tline->text);
1780 if (searching.nparam_min > searching.nparam_max)
1781 error(ERR_NONFATAL,
1782 "minimum parameter count exceeds maximum");
1785 if (tline && tok_is_(tline->next, "+")) {
1786 tline = tline->next;
1787 searching.plus = true;
1789 mmac = (MMacro *) hash_findix(&mmacros, searching.name);
1790 while (mmac) {
1791 if (!strcmp(mmac->name, searching.name) &&
1792 (mmac->nparam_min <= searching.nparam_max
1793 || searching.plus)
1794 && (searching.nparam_min <= mmac->nparam_max
1795 || mmac->plus)) {
1796 found = true;
1797 break;
1799 mmac = mmac->next;
1801 if (tline && tline->next)
1802 error(ERR_WARNING|ERR_PASS1,
1803 "trailing garbage after %%ifmacro ignored");
1804 nasm_free(searching.name);
1805 j = found;
1806 break;
1809 case PPC_IFID:
1810 needtype = TOK_ID;
1811 goto iftype;
1812 case PPC_IFNUM:
1813 needtype = TOK_NUMBER;
1814 goto iftype;
1815 case PPC_IFSTR:
1816 needtype = TOK_STRING;
1817 goto iftype;
1819 iftype:
1820 t = tline = expand_smacro(tline);
1822 while (tok_type_(t, TOK_WHITESPACE) ||
1823 (needtype == TOK_NUMBER &&
1824 tok_type_(t, TOK_OTHER) &&
1825 (t->text[0] == '-' || t->text[0] == '+') &&
1826 !t->text[1]))
1827 t = t->next;
1829 j = tok_type_(t, needtype);
1830 break;
1832 case PPC_IFTOKEN:
1833 t = tline = expand_smacro(tline);
1834 while (tok_type_(t, TOK_WHITESPACE))
1835 t = t->next;
1837 j = false;
1838 if (t) {
1839 t = t->next; /* Skip the actual token */
1840 while (tok_type_(t, TOK_WHITESPACE))
1841 t = t->next;
1842 j = !t; /* Should be nothing left */
1844 break;
1846 case PPC_IFEMPTY:
1847 t = tline = expand_smacro(tline);
1848 while (tok_type_(t, TOK_WHITESPACE))
1849 t = t->next;
1851 j = !t; /* Should be empty */
1852 break;
1854 case PPC_IF:
1855 t = tline = expand_smacro(tline);
1856 tptr = &t;
1857 tokval.t_type = TOKEN_INVALID;
1858 evalresult = evaluate(ppscan, tptr, &tokval,
1859 NULL, pass | CRITICAL, error, NULL);
1860 if (!evalresult)
1861 return -1;
1862 if (tokval.t_type)
1863 error(ERR_WARNING|ERR_PASS1,
1864 "trailing garbage after expression ignored");
1865 if (!is_simple(evalresult)) {
1866 error(ERR_NONFATAL,
1867 "non-constant value given to `%s'", pp_directives[ct]);
1868 goto fail;
1870 j = reloc_value(evalresult) != 0;
1871 break;
1873 default:
1874 error(ERR_FATAL,
1875 "preprocessor directive `%s' not yet implemented",
1876 pp_directives[ct]);
1877 goto fail;
1880 free_tlist(origline);
1881 return j ^ PP_NEGATIVE(ct);
1883 fail:
1884 free_tlist(origline);
1885 return -1;
1889 * Common code for defining an smacro
1891 static bool define_smacro(Context *ctx, const char *mname, bool casesense,
1892 int nparam, Token *expansion)
1894 SMacro *smac, **smhead;
1895 struct hash_table *smtbl;
1897 if (smacro_defined(ctx, mname, nparam, &smac, casesense)) {
1898 if (!smac) {
1899 error(ERR_WARNING|ERR_PASS1,
1900 "single-line macro `%s' defined both with and"
1901 " without parameters", mname);
1903 * Some instances of the old code considered this a failure,
1904 * some others didn't. What is the right thing to do here?
1906 free_tlist(expansion);
1907 return false; /* Failure */
1908 } else {
1910 * We're redefining, so we have to take over an
1911 * existing SMacro structure. This means freeing
1912 * what was already in it.
1914 nasm_free(smac->name);
1915 free_tlist(smac->expansion);
1917 } else {
1918 smtbl = ctx ? &ctx->localmac : &smacros;
1919 smhead = (SMacro **) hash_findi_add(smtbl, mname);
1920 smac = nasm_malloc(sizeof(SMacro));
1921 smac->next = *smhead;
1922 *smhead = smac;
1924 smac->name = nasm_strdup(mname);
1925 smac->casesense = casesense;
1926 smac->nparam = nparam;
1927 smac->expansion = expansion;
1928 smac->in_progress = false;
1929 return true; /* Success */
1933 * Undefine an smacro
1935 static void undef_smacro(Context *ctx, const char *mname)
1937 SMacro **smhead, *s, **sp;
1938 struct hash_table *smtbl;
1940 smtbl = ctx ? &ctx->localmac : &smacros;
1941 smhead = (SMacro **)hash_findi(smtbl, mname, NULL);
1943 if (smhead) {
1945 * We now have a macro name... go hunt for it.
1947 sp = smhead;
1948 while ((s = *sp) != NULL) {
1949 if (!mstrcmp(s->name, mname, s->casesense)) {
1950 *sp = s->next;
1951 nasm_free(s->name);
1952 free_tlist(s->expansion);
1953 nasm_free(s);
1954 } else {
1955 sp = &s->next;
1962 * Parse a mmacro specification.
1964 static bool parse_mmacro_spec(Token *tline, MMacro *def, const char *directive)
1966 bool err;
1968 tline = tline->next;
1969 skip_white_(tline);
1970 tline = expand_id(tline);
1971 if (!tok_type_(tline, TOK_ID)) {
1972 error(ERR_NONFATAL, "`%s' expects a macro name", directive);
1973 return false;
1976 def->prev = NULL;
1977 def->name = nasm_strdup(tline->text);
1978 def->plus = false;
1979 def->nolist = false;
1980 def->in_progress = 0;
1981 def->rep_nest = NULL;
1982 def->nparam_min = 0;
1983 def->nparam_max = 0;
1985 tline = expand_smacro(tline->next);
1986 skip_white_(tline);
1987 if (!tok_type_(tline, TOK_NUMBER)) {
1988 error(ERR_NONFATAL, "`%s' expects a parameter count", directive);
1989 } else {
1990 def->nparam_min = def->nparam_max =
1991 readnum(tline->text, &err);
1992 if (err)
1993 error(ERR_NONFATAL,
1994 "unable to parse parameter count `%s'", tline->text);
1996 if (tline && tok_is_(tline->next, "-")) {
1997 tline = tline->next->next;
1998 if (tok_is_(tline, "*")) {
1999 def->nparam_max = INT_MAX;
2000 } else if (!tok_type_(tline, TOK_NUMBER)) {
2001 error(ERR_NONFATAL,
2002 "`%s' expects a parameter count after `-'", directive);
2003 } else {
2004 def->nparam_max = readnum(tline->text, &err);
2005 if (err) {
2006 error(ERR_NONFATAL, "unable to parse parameter count `%s'",
2007 tline->text);
2009 if (def->nparam_min > def->nparam_max) {
2010 error(ERR_NONFATAL, "minimum parameter count exceeds maximum");
2014 if (tline && tok_is_(tline->next, "+")) {
2015 tline = tline->next;
2016 def->plus = true;
2018 if (tline && tok_type_(tline->next, TOK_ID) &&
2019 !nasm_stricmp(tline->next->text, ".nolist")) {
2020 tline = tline->next;
2021 def->nolist = true;
2025 * Handle default parameters.
2027 if (tline && tline->next) {
2028 def->dlist = tline->next;
2029 tline->next = NULL;
2030 count_mmac_params(def->dlist, &def->ndefs, &def->defaults);
2031 } else {
2032 def->dlist = NULL;
2033 def->defaults = NULL;
2035 def->expansion = NULL;
2037 if (def->defaults && def->ndefs > def->nparam_max - def->nparam_min &&
2038 !def->plus)
2039 error(ERR_WARNING|ERR_PASS1|ERR_WARN_MDP,
2040 "too many default macro parameters");
2042 return true;
2047 * Decode a size directive
2049 static int parse_size(const char *str) {
2050 static const char *size_names[] =
2051 { "byte", "dword", "oword", "qword", "tword", "word", "yword" };
2052 static const int sizes[] =
2053 { 0, 1, 4, 16, 8, 10, 2, 32 };
2055 return sizes[bsii(str, size_names, ARRAY_SIZE(size_names))+1];
2059 * nasm_unquote with error if the string contains NUL characters.
2060 * If the string contains NUL characters, issue an error and return
2061 * the C len, i.e. truncate at the NUL.
2063 static size_t nasm_unquote_cstr(char *qstr, enum preproc_token directive)
2065 size_t len = nasm_unquote(qstr, NULL);
2066 size_t clen = strlen(qstr);
2068 if (len != clen)
2069 error(ERR_NONFATAL, "NUL character in `%s' directive",
2070 pp_directives[directive]);
2072 return clen;
2076 * find and process preprocessor directive in passed line
2077 * Find out if a line contains a preprocessor directive, and deal
2078 * with it if so.
2080 * If a directive _is_ found, it is the responsibility of this routine
2081 * (and not the caller) to free_tlist() the line.
2083 * @param tline a pointer to the current tokeninzed line linked list
2084 * @return DIRECTIVE_FOUND or NO_DIRECTIVE_FOUND
2087 static int do_directive(Token * tline)
2089 enum preproc_token i;
2090 int j;
2091 bool err;
2092 int nparam;
2093 bool nolist;
2094 bool casesense;
2095 int k, m;
2096 int offset;
2097 char *p, *pp;
2098 const char *mname;
2099 Include *inc;
2100 Context *ctx;
2101 Cond *cond;
2102 MMacro *mmac, **mmhead;
2103 Token *t, *tt, *param_start, *macro_start, *last, **tptr, *origline;
2104 Line *l;
2105 struct tokenval tokval;
2106 expr *evalresult;
2107 MMacro *tmp_defining; /* Used when manipulating rep_nest */
2108 int64_t count;
2109 size_t len;
2110 int severity;
2112 origline = tline;
2114 skip_white_(tline);
2115 if (!tline || !tok_type_(tline, TOK_PREPROC_ID) ||
2116 (tline->text[1] == '%' || tline->text[1] == '$'
2117 || tline->text[1] == '!'))
2118 return NO_DIRECTIVE_FOUND;
2120 i = pp_token_hash(tline->text);
2123 * FIXME: We zap execution of PP_RMACRO, PP_IRMACRO, PP_EXITMACRO
2124 * since they are known to be buggy at moment, we need to fix them
2125 * in future release (2.09-2.10)
2127 if (i == PP_RMACRO || i == PP_RMACRO || i == PP_EXITMACRO) {
2128 error(ERR_NONFATAL, "unknown preprocessor directive `%s'",
2129 tline->text);
2130 return NO_DIRECTIVE_FOUND;
2134 * If we're in a non-emitting branch of a condition construct,
2135 * or walking to the end of an already terminated %rep block,
2136 * we should ignore all directives except for condition
2137 * directives.
2139 if (((istk->conds && !emitting(istk->conds->state)) ||
2140 (istk->mstk && !istk->mstk->in_progress)) && !is_condition(i)) {
2141 return NO_DIRECTIVE_FOUND;
2145 * If we're defining a macro or reading a %rep block, we should
2146 * ignore all directives except for %macro/%imacro (which nest),
2147 * %endm/%endmacro, and (only if we're in a %rep block) %endrep.
2148 * If we're in a %rep block, another %rep nests, so should be let through.
2150 if (defining && i != PP_MACRO && i != PP_IMACRO &&
2151 i != PP_RMACRO && i != PP_IRMACRO &&
2152 i != PP_ENDMACRO && i != PP_ENDM &&
2153 (defining->name || (i != PP_ENDREP && i != PP_REP))) {
2154 return NO_DIRECTIVE_FOUND;
2157 if (defining) {
2158 if (i == PP_MACRO || i == PP_IMACRO ||
2159 i == PP_RMACRO || i == PP_IRMACRO) {
2160 nested_mac_count++;
2161 return NO_DIRECTIVE_FOUND;
2162 } else if (nested_mac_count > 0) {
2163 if (i == PP_ENDMACRO) {
2164 nested_mac_count--;
2165 return NO_DIRECTIVE_FOUND;
2168 if (!defining->name) {
2169 if (i == PP_REP) {
2170 nested_rep_count++;
2171 return NO_DIRECTIVE_FOUND;
2172 } else if (nested_rep_count > 0) {
2173 if (i == PP_ENDREP) {
2174 nested_rep_count--;
2175 return NO_DIRECTIVE_FOUND;
2181 switch (i) {
2182 case PP_INVALID:
2183 error(ERR_NONFATAL, "unknown preprocessor directive `%s'",
2184 tline->text);
2185 return NO_DIRECTIVE_FOUND; /* didn't get it */
2187 case PP_STACKSIZE:
2188 /* Directive to tell NASM what the default stack size is. The
2189 * default is for a 16-bit stack, and this can be overriden with
2190 * %stacksize large.
2192 tline = tline->next;
2193 if (tline && tline->type == TOK_WHITESPACE)
2194 tline = tline->next;
2195 if (!tline || tline->type != TOK_ID) {
2196 error(ERR_NONFATAL, "`%%stacksize' missing size parameter");
2197 free_tlist(origline);
2198 return DIRECTIVE_FOUND;
2200 if (nasm_stricmp(tline->text, "flat") == 0) {
2201 /* All subsequent ARG directives are for a 32-bit stack */
2202 StackSize = 4;
2203 StackPointer = "ebp";
2204 ArgOffset = 8;
2205 LocalOffset = 0;
2206 } else if (nasm_stricmp(tline->text, "flat64") == 0) {
2207 /* All subsequent ARG directives are for a 64-bit stack */
2208 StackSize = 8;
2209 StackPointer = "rbp";
2210 ArgOffset = 16;
2211 LocalOffset = 0;
2212 } else if (nasm_stricmp(tline->text, "large") == 0) {
2213 /* All subsequent ARG directives are for a 16-bit stack,
2214 * far function call.
2216 StackSize = 2;
2217 StackPointer = "bp";
2218 ArgOffset = 4;
2219 LocalOffset = 0;
2220 } else if (nasm_stricmp(tline->text, "small") == 0) {
2221 /* All subsequent ARG directives are for a 16-bit stack,
2222 * far function call. We don't support near functions.
2224 StackSize = 2;
2225 StackPointer = "bp";
2226 ArgOffset = 6;
2227 LocalOffset = 0;
2228 } else {
2229 error(ERR_NONFATAL, "`%%stacksize' invalid size type");
2230 free_tlist(origline);
2231 return DIRECTIVE_FOUND;
2233 free_tlist(origline);
2234 return DIRECTIVE_FOUND;
2236 case PP_ARG:
2237 /* TASM like ARG directive to define arguments to functions, in
2238 * the following form:
2240 * ARG arg1:WORD, arg2:DWORD, arg4:QWORD
2242 offset = ArgOffset;
2243 do {
2244 char *arg, directive[256];
2245 int size = StackSize;
2247 /* Find the argument name */
2248 tline = tline->next;
2249 if (tline && tline->type == TOK_WHITESPACE)
2250 tline = tline->next;
2251 if (!tline || tline->type != TOK_ID) {
2252 error(ERR_NONFATAL, "`%%arg' missing argument parameter");
2253 free_tlist(origline);
2254 return DIRECTIVE_FOUND;
2256 arg = tline->text;
2258 /* Find the argument size type */
2259 tline = tline->next;
2260 if (!tline || tline->type != TOK_OTHER
2261 || tline->text[0] != ':') {
2262 error(ERR_NONFATAL,
2263 "Syntax error processing `%%arg' directive");
2264 free_tlist(origline);
2265 return DIRECTIVE_FOUND;
2267 tline = tline->next;
2268 if (!tline || tline->type != TOK_ID) {
2269 error(ERR_NONFATAL, "`%%arg' missing size type parameter");
2270 free_tlist(origline);
2271 return DIRECTIVE_FOUND;
2274 /* Allow macro expansion of type parameter */
2275 tt = tokenize(tline->text);
2276 tt = expand_smacro(tt);
2277 size = parse_size(tt->text);
2278 if (!size) {
2279 error(ERR_NONFATAL,
2280 "Invalid size type for `%%arg' missing directive");
2281 free_tlist(tt);
2282 free_tlist(origline);
2283 return DIRECTIVE_FOUND;
2285 free_tlist(tt);
2287 /* Round up to even stack slots */
2288 size = ALIGN(size, StackSize);
2290 /* Now define the macro for the argument */
2291 snprintf(directive, sizeof(directive), "%%define %s (%s+%d)",
2292 arg, StackPointer, offset);
2293 do_directive(tokenize(directive));
2294 offset += size;
2296 /* Move to the next argument in the list */
2297 tline = tline->next;
2298 if (tline && tline->type == TOK_WHITESPACE)
2299 tline = tline->next;
2300 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2301 ArgOffset = offset;
2302 free_tlist(origline);
2303 return DIRECTIVE_FOUND;
2305 case PP_LOCAL:
2306 /* TASM like LOCAL directive to define local variables for a
2307 * function, in the following form:
2309 * LOCAL local1:WORD, local2:DWORD, local4:QWORD = LocalSize
2311 * The '= LocalSize' at the end is ignored by NASM, but is
2312 * required by TASM to define the local parameter size (and used
2313 * by the TASM macro package).
2315 offset = LocalOffset;
2316 do {
2317 char *local, directive[256];
2318 int size = StackSize;
2320 /* Find the argument name */
2321 tline = tline->next;
2322 if (tline && tline->type == TOK_WHITESPACE)
2323 tline = tline->next;
2324 if (!tline || tline->type != TOK_ID) {
2325 error(ERR_NONFATAL,
2326 "`%%local' missing argument parameter");
2327 free_tlist(origline);
2328 return DIRECTIVE_FOUND;
2330 local = tline->text;
2332 /* Find the argument size type */
2333 tline = tline->next;
2334 if (!tline || tline->type != TOK_OTHER
2335 || tline->text[0] != ':') {
2336 error(ERR_NONFATAL,
2337 "Syntax error processing `%%local' directive");
2338 free_tlist(origline);
2339 return DIRECTIVE_FOUND;
2341 tline = tline->next;
2342 if (!tline || tline->type != TOK_ID) {
2343 error(ERR_NONFATAL,
2344 "`%%local' missing size type parameter");
2345 free_tlist(origline);
2346 return DIRECTIVE_FOUND;
2349 /* Allow macro expansion of type parameter */
2350 tt = tokenize(tline->text);
2351 tt = expand_smacro(tt);
2352 size = parse_size(tt->text);
2353 if (!size) {
2354 error(ERR_NONFATAL,
2355 "Invalid size type for `%%local' missing directive");
2356 free_tlist(tt);
2357 free_tlist(origline);
2358 return DIRECTIVE_FOUND;
2360 free_tlist(tt);
2362 /* Round up to even stack slots */
2363 size = ALIGN(size, StackSize);
2365 offset += size; /* Negative offset, increment before */
2367 /* Now define the macro for the argument */
2368 snprintf(directive, sizeof(directive), "%%define %s (%s-%d)",
2369 local, StackPointer, offset);
2370 do_directive(tokenize(directive));
2372 /* Now define the assign to setup the enter_c macro correctly */
2373 snprintf(directive, sizeof(directive),
2374 "%%assign %%$localsize %%$localsize+%d", size);
2375 do_directive(tokenize(directive));
2377 /* Move to the next argument in the list */
2378 tline = tline->next;
2379 if (tline && tline->type == TOK_WHITESPACE)
2380 tline = tline->next;
2381 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2382 LocalOffset = offset;
2383 free_tlist(origline);
2384 return DIRECTIVE_FOUND;
2386 case PP_CLEAR:
2387 if (tline->next)
2388 error(ERR_WARNING|ERR_PASS1,
2389 "trailing garbage after `%%clear' ignored");
2390 free_macros();
2391 init_macros();
2392 free_tlist(origline);
2393 return DIRECTIVE_FOUND;
2395 case PP_DEPEND:
2396 t = tline->next = expand_smacro(tline->next);
2397 skip_white_(t);
2398 if (!t || (t->type != TOK_STRING &&
2399 t->type != TOK_INTERNAL_STRING)) {
2400 error(ERR_NONFATAL, "`%%depend' expects a file name");
2401 free_tlist(origline);
2402 return DIRECTIVE_FOUND; /* but we did _something_ */
2404 if (t->next)
2405 error(ERR_WARNING|ERR_PASS1,
2406 "trailing garbage after `%%depend' ignored");
2407 p = t->text;
2408 if (t->type != TOK_INTERNAL_STRING)
2409 nasm_unquote_cstr(p, i);
2410 if (dephead && !in_list(*dephead, p)) {
2411 StrList *sl = nasm_malloc(strlen(p)+1+sizeof sl->next);
2412 sl->next = NULL;
2413 strcpy(sl->str, p);
2414 *deptail = sl;
2415 deptail = &sl->next;
2417 free_tlist(origline);
2418 return DIRECTIVE_FOUND;
2420 case PP_INCLUDE:
2421 t = tline->next = expand_smacro(tline->next);
2422 skip_white_(t);
2424 if (!t || (t->type != TOK_STRING &&
2425 t->type != TOK_INTERNAL_STRING)) {
2426 error(ERR_NONFATAL, "`%%include' expects a file name");
2427 free_tlist(origline);
2428 return DIRECTIVE_FOUND; /* but we did _something_ */
2430 if (t->next)
2431 error(ERR_WARNING|ERR_PASS1,
2432 "trailing garbage after `%%include' ignored");
2433 p = t->text;
2434 if (t->type != TOK_INTERNAL_STRING)
2435 nasm_unquote_cstr(p, i);
2436 inc = nasm_malloc(sizeof(Include));
2437 inc->next = istk;
2438 inc->conds = NULL;
2439 inc->fp = inc_fopen(p, dephead, &deptail, pass == 0);
2440 if (!inc->fp) {
2441 /* -MG given but file not found */
2442 nasm_free(inc);
2443 } else {
2444 inc->fname = src_set_fname(nasm_strdup(p));
2445 inc->lineno = src_set_linnum(0);
2446 inc->lineinc = 1;
2447 inc->expansion = NULL;
2448 inc->mstk = NULL;
2449 istk = inc;
2450 list->uplevel(LIST_INCLUDE);
2452 free_tlist(origline);
2453 return DIRECTIVE_FOUND;
2455 case PP_USE:
2457 static macros_t *use_pkg;
2458 const char *pkg_macro = NULL;
2460 tline = tline->next;
2461 skip_white_(tline);
2462 tline = expand_id(tline);
2464 if (!tline || (tline->type != TOK_STRING &&
2465 tline->type != TOK_INTERNAL_STRING &&
2466 tline->type != TOK_ID)) {
2467 error(ERR_NONFATAL, "`%%use' expects a package name");
2468 free_tlist(origline);
2469 return DIRECTIVE_FOUND; /* but we did _something_ */
2471 if (tline->next)
2472 error(ERR_WARNING|ERR_PASS1,
2473 "trailing garbage after `%%use' ignored");
2474 if (tline->type == TOK_STRING)
2475 nasm_unquote_cstr(tline->text, i);
2476 use_pkg = nasm_stdmac_find_package(tline->text);
2477 if (!use_pkg)
2478 error(ERR_NONFATAL, "unknown `%%use' package: %s", tline->text);
2479 else
2480 pkg_macro = (char *)use_pkg + 1; /* The first string will be <%define>__USE_*__ */
2481 if (use_pkg && ! smacro_defined(NULL, pkg_macro, 0, NULL, true)) {
2482 /* Not already included, go ahead and include it */
2483 stdmacpos = use_pkg;
2485 free_tlist(origline);
2486 return DIRECTIVE_FOUND;
2488 case PP_PUSH:
2489 case PP_REPL:
2490 case PP_POP:
2491 tline = tline->next;
2492 skip_white_(tline);
2493 tline = expand_id(tline);
2494 if (tline) {
2495 if (!tok_type_(tline, TOK_ID)) {
2496 error(ERR_NONFATAL, "`%s' expects a context identifier",
2497 pp_directives[i]);
2498 free_tlist(origline);
2499 return DIRECTIVE_FOUND; /* but we did _something_ */
2501 if (tline->next)
2502 error(ERR_WARNING|ERR_PASS1,
2503 "trailing garbage after `%s' ignored",
2504 pp_directives[i]);
2505 p = nasm_strdup(tline->text);
2506 } else {
2507 p = NULL; /* Anonymous */
2510 if (i == PP_PUSH) {
2511 ctx = nasm_malloc(sizeof(Context));
2512 ctx->next = cstk;
2513 hash_init(&ctx->localmac, HASH_SMALL);
2514 ctx->name = p;
2515 ctx->number = unique++;
2516 cstk = ctx;
2517 } else {
2518 /* %pop or %repl */
2519 if (!cstk) {
2520 error(ERR_NONFATAL, "`%s': context stack is empty",
2521 pp_directives[i]);
2522 } else if (i == PP_POP) {
2523 if (p && (!cstk->name || nasm_stricmp(p, cstk->name)))
2524 error(ERR_NONFATAL, "`%%pop' in wrong context: %s, "
2525 "expected %s",
2526 cstk->name ? cstk->name : "anonymous", p);
2527 else
2528 ctx_pop();
2529 } else {
2530 /* i == PP_REPL */
2531 nasm_free(cstk->name);
2532 cstk->name = p;
2533 p = NULL;
2535 nasm_free(p);
2537 free_tlist(origline);
2538 return DIRECTIVE_FOUND;
2539 case PP_FATAL:
2540 severity = ERR_FATAL;
2541 goto issue_error;
2542 case PP_ERROR:
2543 severity = ERR_NONFATAL;
2544 goto issue_error;
2545 case PP_WARNING:
2546 severity = ERR_WARNING|ERR_WARN_USER;
2547 goto issue_error;
2549 issue_error:
2551 /* Only error out if this is the final pass */
2552 if (pass != 2 && i != PP_FATAL)
2553 return DIRECTIVE_FOUND;
2555 tline->next = expand_smacro(tline->next);
2556 tline = tline->next;
2557 skip_white_(tline);
2558 t = tline ? tline->next : NULL;
2559 skip_white_(t);
2560 if (tok_type_(tline, TOK_STRING) && !t) {
2561 /* The line contains only a quoted string */
2562 p = tline->text;
2563 nasm_unquote(p, NULL); /* Ignore NUL character truncation */
2564 error(severity, "%s", p);
2565 } else {
2566 /* Not a quoted string, or more than a quoted string */
2567 p = detoken(tline, false);
2568 error(severity, "%s", p);
2569 nasm_free(p);
2571 free_tlist(origline);
2572 return DIRECTIVE_FOUND;
2575 CASE_PP_IF:
2576 if (istk->conds && !emitting(istk->conds->state))
2577 j = COND_NEVER;
2578 else {
2579 j = if_condition(tline->next, i);
2580 tline->next = NULL; /* it got freed */
2581 j = j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2583 cond = nasm_malloc(sizeof(Cond));
2584 cond->next = istk->conds;
2585 cond->state = j;
2586 istk->conds = cond;
2587 if(istk->mstk)
2588 istk->mstk->condcnt ++;
2589 free_tlist(origline);
2590 return DIRECTIVE_FOUND;
2592 CASE_PP_ELIF:
2593 if (!istk->conds)
2594 error(ERR_FATAL, "`%s': no matching `%%if'", pp_directives[i]);
2595 switch(istk->conds->state) {
2596 case COND_IF_TRUE:
2597 istk->conds->state = COND_DONE;
2598 break;
2600 case COND_DONE:
2601 case COND_NEVER:
2602 break;
2604 case COND_ELSE_TRUE:
2605 case COND_ELSE_FALSE:
2606 error_precond(ERR_WARNING|ERR_PASS1,
2607 "`%%elif' after `%%else' ignored");
2608 istk->conds->state = COND_NEVER;
2609 break;
2611 case COND_IF_FALSE:
2613 * IMPORTANT: In the case of %if, we will already have
2614 * called expand_mmac_params(); however, if we're
2615 * processing an %elif we must have been in a
2616 * non-emitting mode, which would have inhibited
2617 * the normal invocation of expand_mmac_params().
2618 * Therefore, we have to do it explicitly here.
2620 j = if_condition(expand_mmac_params(tline->next), i);
2621 tline->next = NULL; /* it got freed */
2622 istk->conds->state =
2623 j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2624 break;
2626 free_tlist(origline);
2627 return DIRECTIVE_FOUND;
2629 case PP_ELSE:
2630 if (tline->next)
2631 error_precond(ERR_WARNING|ERR_PASS1,
2632 "trailing garbage after `%%else' ignored");
2633 if (!istk->conds)
2634 error(ERR_FATAL, "`%%else': no matching `%%if'");
2635 switch(istk->conds->state) {
2636 case COND_IF_TRUE:
2637 case COND_DONE:
2638 istk->conds->state = COND_ELSE_FALSE;
2639 break;
2641 case COND_NEVER:
2642 break;
2644 case COND_IF_FALSE:
2645 istk->conds->state = COND_ELSE_TRUE;
2646 break;
2648 case COND_ELSE_TRUE:
2649 case COND_ELSE_FALSE:
2650 error_precond(ERR_WARNING|ERR_PASS1,
2651 "`%%else' after `%%else' ignored.");
2652 istk->conds->state = COND_NEVER;
2653 break;
2655 free_tlist(origline);
2656 return DIRECTIVE_FOUND;
2658 case PP_ENDIF:
2659 if (tline->next)
2660 error_precond(ERR_WARNING|ERR_PASS1,
2661 "trailing garbage after `%%endif' ignored");
2662 if (!istk->conds)
2663 error(ERR_FATAL, "`%%endif': no matching `%%if'");
2664 cond = istk->conds;
2665 istk->conds = cond->next;
2666 nasm_free(cond);
2667 if(istk->mstk)
2668 istk->mstk->condcnt --;
2669 free_tlist(origline);
2670 return DIRECTIVE_FOUND;
2672 case PP_RMACRO:
2673 case PP_IRMACRO:
2674 case PP_MACRO:
2675 case PP_IMACRO:
2676 if (defining) {
2677 error(ERR_FATAL, "`%s': already defining a macro",
2678 pp_directives[i]);
2679 return DIRECTIVE_FOUND;
2681 defining = nasm_malloc(sizeof(MMacro));
2682 defining->max_depth =
2683 (i == PP_RMACRO) || (i == PP_IRMACRO) ? DEADMAN_LIMIT : 0;
2684 defining->casesense = (i == PP_MACRO) || (i == PP_RMACRO);
2685 if (!parse_mmacro_spec(tline, defining, pp_directives[i])) {
2686 nasm_free(defining);
2687 defining = NULL;
2688 return DIRECTIVE_FOUND;
2691 mmac = (MMacro *) hash_findix(&mmacros, defining->name);
2692 while (mmac) {
2693 if (!strcmp(mmac->name, defining->name) &&
2694 (mmac->nparam_min <= defining->nparam_max
2695 || defining->plus)
2696 && (defining->nparam_min <= mmac->nparam_max
2697 || mmac->plus)) {
2698 error(ERR_WARNING|ERR_PASS1,
2699 "redefining multi-line macro `%s'", defining->name);
2700 return DIRECTIVE_FOUND;
2702 mmac = mmac->next;
2704 free_tlist(origline);
2705 return DIRECTIVE_FOUND;
2707 case PP_ENDM:
2708 case PP_ENDMACRO:
2709 if (! (defining && defining->name)) {
2710 error(ERR_NONFATAL, "`%s': not defining a macro", tline->text);
2711 return DIRECTIVE_FOUND;
2713 mmhead = (MMacro **) hash_findi_add(&mmacros, defining->name);
2714 defining->next = *mmhead;
2715 *mmhead = defining;
2716 defining = NULL;
2717 free_tlist(origline);
2718 return DIRECTIVE_FOUND;
2720 case PP_EXITMACRO:
2722 * We must search along istk->expansion until we hit a
2723 * macro-end marker for a macro with a name. Then we
2724 * bypass all lines between exitmacro and endmacro.
2726 list_for_each(l, istk->expansion)
2727 if (l->finishes && l->finishes->name)
2728 break;
2730 if (l) {
2732 * Remove all conditional entries relative to this
2733 * macro invocation. (safe to do in this context)
2735 for ( ; l->finishes->condcnt > 0; l->finishes->condcnt --) {
2736 cond = istk->conds;
2737 istk->conds = cond->next;
2738 nasm_free(cond);
2740 istk->expansion = l;
2741 } else {
2742 error(ERR_NONFATAL, "`%%exitmacro' not within `%%macro' block");
2744 free_tlist(origline);
2745 return DIRECTIVE_FOUND;
2747 case PP_UNMACRO:
2748 case PP_UNIMACRO:
2750 MMacro **mmac_p;
2751 MMacro spec;
2753 spec.casesense = (i == PP_UNMACRO);
2754 if (!parse_mmacro_spec(tline, &spec, pp_directives[i])) {
2755 return DIRECTIVE_FOUND;
2757 mmac_p = (MMacro **) hash_findi(&mmacros, spec.name, NULL);
2758 while (mmac_p && *mmac_p) {
2759 mmac = *mmac_p;
2760 if (mmac->casesense == spec.casesense &&
2761 !mstrcmp(mmac->name, spec.name, spec.casesense) &&
2762 mmac->nparam_min == spec.nparam_min &&
2763 mmac->nparam_max == spec.nparam_max &&
2764 mmac->plus == spec.plus) {
2765 *mmac_p = mmac->next;
2766 free_mmacro(mmac);
2767 } else {
2768 mmac_p = &mmac->next;
2771 free_tlist(origline);
2772 free_tlist(spec.dlist);
2773 return DIRECTIVE_FOUND;
2776 case PP_ROTATE:
2777 if (tline->next && tline->next->type == TOK_WHITESPACE)
2778 tline = tline->next;
2779 if (!tline->next) {
2780 free_tlist(origline);
2781 error(ERR_NONFATAL, "`%%rotate' missing rotate count");
2782 return DIRECTIVE_FOUND;
2784 t = expand_smacro(tline->next);
2785 tline->next = NULL;
2786 free_tlist(origline);
2787 tline = t;
2788 tptr = &t;
2789 tokval.t_type = TOKEN_INVALID;
2790 evalresult =
2791 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
2792 free_tlist(tline);
2793 if (!evalresult)
2794 return DIRECTIVE_FOUND;
2795 if (tokval.t_type)
2796 error(ERR_WARNING|ERR_PASS1,
2797 "trailing garbage after expression ignored");
2798 if (!is_simple(evalresult)) {
2799 error(ERR_NONFATAL, "non-constant value given to `%%rotate'");
2800 return DIRECTIVE_FOUND;
2802 mmac = istk->mstk;
2803 while (mmac && !mmac->name) /* avoid mistaking %reps for macros */
2804 mmac = mmac->next_active;
2805 if (!mmac) {
2806 error(ERR_NONFATAL, "`%%rotate' invoked outside a macro call");
2807 } else if (mmac->nparam == 0) {
2808 error(ERR_NONFATAL,
2809 "`%%rotate' invoked within macro without parameters");
2810 } else {
2811 int rotate = mmac->rotate + reloc_value(evalresult);
2813 rotate %= (int)mmac->nparam;
2814 if (rotate < 0)
2815 rotate += mmac->nparam;
2817 mmac->rotate = rotate;
2819 return DIRECTIVE_FOUND;
2821 case PP_REP:
2822 nolist = false;
2823 do {
2824 tline = tline->next;
2825 } while (tok_type_(tline, TOK_WHITESPACE));
2827 if (tok_type_(tline, TOK_ID) &&
2828 nasm_stricmp(tline->text, ".nolist") == 0) {
2829 nolist = true;
2830 do {
2831 tline = tline->next;
2832 } while (tok_type_(tline, TOK_WHITESPACE));
2835 if (tline) {
2836 t = expand_smacro(tline);
2837 tptr = &t;
2838 tokval.t_type = TOKEN_INVALID;
2839 evalresult =
2840 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
2841 if (!evalresult) {
2842 free_tlist(origline);
2843 return DIRECTIVE_FOUND;
2845 if (tokval.t_type)
2846 error(ERR_WARNING|ERR_PASS1,
2847 "trailing garbage after expression ignored");
2848 if (!is_simple(evalresult)) {
2849 error(ERR_NONFATAL, "non-constant value given to `%%rep'");
2850 return DIRECTIVE_FOUND;
2852 count = reloc_value(evalresult) + 1;
2853 } else {
2854 error(ERR_NONFATAL, "`%%rep' expects a repeat count");
2855 count = 0;
2857 free_tlist(origline);
2859 tmp_defining = defining;
2860 defining = nasm_malloc(sizeof(MMacro));
2861 defining->prev = NULL;
2862 defining->name = NULL; /* flags this macro as a %rep block */
2863 defining->casesense = false;
2864 defining->plus = false;
2865 defining->nolist = nolist;
2866 defining->in_progress = count;
2867 defining->max_depth = 0;
2868 defining->nparam_min = defining->nparam_max = 0;
2869 defining->defaults = NULL;
2870 defining->dlist = NULL;
2871 defining->expansion = NULL;
2872 defining->next_active = istk->mstk;
2873 defining->rep_nest = tmp_defining;
2874 return DIRECTIVE_FOUND;
2876 case PP_ENDREP:
2877 if (!defining || defining->name) {
2878 error(ERR_NONFATAL, "`%%endrep': no matching `%%rep'");
2879 return DIRECTIVE_FOUND;
2883 * Now we have a "macro" defined - although it has no name
2884 * and we won't be entering it in the hash tables - we must
2885 * push a macro-end marker for it on to istk->expansion.
2886 * After that, it will take care of propagating itself (a
2887 * macro-end marker line for a macro which is really a %rep
2888 * block will cause the macro to be re-expanded, complete
2889 * with another macro-end marker to ensure the process
2890 * continues) until the whole expansion is forcibly removed
2891 * from istk->expansion by a %exitrep.
2893 l = nasm_malloc(sizeof(Line));
2894 l->next = istk->expansion;
2895 l->finishes = defining;
2896 l->first = NULL;
2897 istk->expansion = l;
2899 istk->mstk = defining;
2901 list->uplevel(defining->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
2902 tmp_defining = defining;
2903 defining = defining->rep_nest;
2904 free_tlist(origline);
2905 return DIRECTIVE_FOUND;
2907 case PP_EXITREP:
2909 * We must search along istk->expansion until we hit a
2910 * macro-end marker for a macro with no name. Then we set
2911 * its `in_progress' flag to 0.
2913 list_for_each(l, istk->expansion)
2914 if (l->finishes && !l->finishes->name)
2915 break;
2917 if (l)
2918 l->finishes->in_progress = 1;
2919 else
2920 error(ERR_NONFATAL, "`%%exitrep' not within `%%rep' block");
2921 free_tlist(origline);
2922 return DIRECTIVE_FOUND;
2924 case PP_XDEFINE:
2925 case PP_IXDEFINE:
2926 case PP_DEFINE:
2927 case PP_IDEFINE:
2928 casesense = (i == PP_DEFINE || i == PP_XDEFINE);
2930 tline = tline->next;
2931 skip_white_(tline);
2932 tline = expand_id(tline);
2933 if (!tline || (tline->type != TOK_ID &&
2934 (tline->type != TOK_PREPROC_ID ||
2935 tline->text[1] != '$'))) {
2936 error(ERR_NONFATAL, "`%s' expects a macro identifier",
2937 pp_directives[i]);
2938 free_tlist(origline);
2939 return DIRECTIVE_FOUND;
2942 ctx = get_ctx(tline->text, &mname, false);
2943 last = tline;
2944 param_start = tline = tline->next;
2945 nparam = 0;
2947 /* Expand the macro definition now for %xdefine and %ixdefine */
2948 if ((i == PP_XDEFINE) || (i == PP_IXDEFINE))
2949 tline = expand_smacro(tline);
2951 if (tok_is_(tline, "(")) {
2953 * This macro has parameters.
2956 tline = tline->next;
2957 while (1) {
2958 skip_white_(tline);
2959 if (!tline) {
2960 error(ERR_NONFATAL, "parameter identifier expected");
2961 free_tlist(origline);
2962 return DIRECTIVE_FOUND;
2964 if (tline->type != TOK_ID) {
2965 error(ERR_NONFATAL,
2966 "`%s': parameter identifier expected",
2967 tline->text);
2968 free_tlist(origline);
2969 return DIRECTIVE_FOUND;
2971 tline->type = TOK_SMAC_PARAM + nparam++;
2972 tline = tline->next;
2973 skip_white_(tline);
2974 if (tok_is_(tline, ",")) {
2975 tline = tline->next;
2976 } else {
2977 if (!tok_is_(tline, ")")) {
2978 error(ERR_NONFATAL,
2979 "`)' expected to terminate macro template");
2980 free_tlist(origline);
2981 return DIRECTIVE_FOUND;
2983 break;
2986 last = tline;
2987 tline = tline->next;
2989 if (tok_type_(tline, TOK_WHITESPACE))
2990 last = tline, tline = tline->next;
2991 macro_start = NULL;
2992 last->next = NULL;
2993 t = tline;
2994 while (t) {
2995 if (t->type == TOK_ID) {
2996 list_for_each(tt, param_start)
2997 if (tt->type >= TOK_SMAC_PARAM &&
2998 !strcmp(tt->text, t->text))
2999 t->type = tt->type;
3001 tt = t->next;
3002 t->next = macro_start;
3003 macro_start = t;
3004 t = tt;
3007 * Good. We now have a macro name, a parameter count, and a
3008 * token list (in reverse order) for an expansion. We ought
3009 * to be OK just to create an SMacro, store it, and let
3010 * free_tlist have the rest of the line (which we have
3011 * carefully re-terminated after chopping off the expansion
3012 * from the end).
3014 define_smacro(ctx, mname, casesense, nparam, macro_start);
3015 free_tlist(origline);
3016 return DIRECTIVE_FOUND;
3018 case PP_UNDEF:
3019 tline = tline->next;
3020 skip_white_(tline);
3021 tline = expand_id(tline);
3022 if (!tline || (tline->type != TOK_ID &&
3023 (tline->type != TOK_PREPROC_ID ||
3024 tline->text[1] != '$'))) {
3025 error(ERR_NONFATAL, "`%%undef' expects a macro identifier");
3026 free_tlist(origline);
3027 return DIRECTIVE_FOUND;
3029 if (tline->next) {
3030 error(ERR_WARNING|ERR_PASS1,
3031 "trailing garbage after macro name ignored");
3034 /* Find the context that symbol belongs to */
3035 ctx = get_ctx(tline->text, &mname, false);
3036 undef_smacro(ctx, mname);
3037 free_tlist(origline);
3038 return DIRECTIVE_FOUND;
3040 case PP_DEFSTR:
3041 case PP_IDEFSTR:
3042 casesense = (i == PP_DEFSTR);
3044 tline = tline->next;
3045 skip_white_(tline);
3046 tline = expand_id(tline);
3047 if (!tline || (tline->type != TOK_ID &&
3048 (tline->type != TOK_PREPROC_ID ||
3049 tline->text[1] != '$'))) {
3050 error(ERR_NONFATAL, "`%s' expects a macro identifier",
3051 pp_directives[i]);
3052 free_tlist(origline);
3053 return DIRECTIVE_FOUND;
3056 ctx = get_ctx(tline->text, &mname, false);
3057 last = tline;
3058 tline = expand_smacro(tline->next);
3059 last->next = NULL;
3061 while (tok_type_(tline, TOK_WHITESPACE))
3062 tline = delete_Token(tline);
3064 p = detoken(tline, false);
3065 macro_start = nasm_malloc(sizeof(*macro_start));
3066 macro_start->next = NULL;
3067 macro_start->text = nasm_quote(p, strlen(p));
3068 macro_start->type = TOK_STRING;
3069 macro_start->a.mac = NULL;
3070 nasm_free(p);
3073 * We now have a macro name, an implicit parameter count of
3074 * zero, and a string token to use as an expansion. Create
3075 * and store an SMacro.
3077 define_smacro(ctx, mname, casesense, 0, macro_start);
3078 free_tlist(origline);
3079 return DIRECTIVE_FOUND;
3081 case PP_DEFTOK:
3082 case PP_IDEFTOK:
3083 casesense = (i == PP_DEFTOK);
3085 tline = tline->next;
3086 skip_white_(tline);
3087 tline = expand_id(tline);
3088 if (!tline || (tline->type != TOK_ID &&
3089 (tline->type != TOK_PREPROC_ID ||
3090 tline->text[1] != '$'))) {
3091 error(ERR_NONFATAL,
3092 "`%s' expects a macro identifier as first parameter",
3093 pp_directives[i]);
3094 free_tlist(origline);
3095 return DIRECTIVE_FOUND;
3097 ctx = get_ctx(tline->text, &mname, false);
3098 last = tline;
3099 tline = expand_smacro(tline->next);
3100 last->next = NULL;
3102 t = tline;
3103 while (tok_type_(t, TOK_WHITESPACE))
3104 t = t->next;
3105 /* t should now point to the string */
3106 if (t->type != TOK_STRING) {
3107 error(ERR_NONFATAL,
3108 "`%s` requires string as second parameter",
3109 pp_directives[i]);
3110 free_tlist(tline);
3111 free_tlist(origline);
3112 return DIRECTIVE_FOUND;
3115 nasm_unquote_cstr(t->text, i);
3116 macro_start = tokenize(t->text);
3119 * We now have a macro name, an implicit parameter count of
3120 * zero, and a numeric token to use as an expansion. Create
3121 * and store an SMacro.
3123 define_smacro(ctx, mname, casesense, 0, macro_start);
3124 free_tlist(tline);
3125 free_tlist(origline);
3126 return DIRECTIVE_FOUND;
3128 case PP_PATHSEARCH:
3130 FILE *fp;
3131 StrList *xsl = NULL;
3132 StrList **xst = &xsl;
3134 casesense = true;
3136 tline = tline->next;
3137 skip_white_(tline);
3138 tline = expand_id(tline);
3139 if (!tline || (tline->type != TOK_ID &&
3140 (tline->type != TOK_PREPROC_ID ||
3141 tline->text[1] != '$'))) {
3142 error(ERR_NONFATAL,
3143 "`%%pathsearch' expects a macro identifier as first parameter");
3144 free_tlist(origline);
3145 return DIRECTIVE_FOUND;
3147 ctx = get_ctx(tline->text, &mname, false);
3148 last = tline;
3149 tline = expand_smacro(tline->next);
3150 last->next = NULL;
3152 t = tline;
3153 while (tok_type_(t, TOK_WHITESPACE))
3154 t = t->next;
3156 if (!t || (t->type != TOK_STRING &&
3157 t->type != TOK_INTERNAL_STRING)) {
3158 error(ERR_NONFATAL, "`%%pathsearch' expects a file name");
3159 free_tlist(tline);
3160 free_tlist(origline);
3161 return DIRECTIVE_FOUND; /* but we did _something_ */
3163 if (t->next)
3164 error(ERR_WARNING|ERR_PASS1,
3165 "trailing garbage after `%%pathsearch' ignored");
3166 p = t->text;
3167 if (t->type != TOK_INTERNAL_STRING)
3168 nasm_unquote(p, NULL);
3170 fp = inc_fopen(p, &xsl, &xst, true);
3171 if (fp) {
3172 p = xsl->str;
3173 fclose(fp); /* Don't actually care about the file */
3175 macro_start = nasm_malloc(sizeof(*macro_start));
3176 macro_start->next = NULL;
3177 macro_start->text = nasm_quote(p, strlen(p));
3178 macro_start->type = TOK_STRING;
3179 macro_start->a.mac = NULL;
3180 if (xsl)
3181 nasm_free(xsl);
3184 * We now have a macro name, an implicit parameter count of
3185 * zero, and a string token to use as an expansion. Create
3186 * and store an SMacro.
3188 define_smacro(ctx, mname, casesense, 0, macro_start);
3189 free_tlist(tline);
3190 free_tlist(origline);
3191 return DIRECTIVE_FOUND;
3194 case PP_STRLEN:
3195 casesense = true;
3197 tline = tline->next;
3198 skip_white_(tline);
3199 tline = expand_id(tline);
3200 if (!tline || (tline->type != TOK_ID &&
3201 (tline->type != TOK_PREPROC_ID ||
3202 tline->text[1] != '$'))) {
3203 error(ERR_NONFATAL,
3204 "`%%strlen' expects a macro identifier as first parameter");
3205 free_tlist(origline);
3206 return DIRECTIVE_FOUND;
3208 ctx = get_ctx(tline->text, &mname, false);
3209 last = tline;
3210 tline = expand_smacro(tline->next);
3211 last->next = NULL;
3213 t = tline;
3214 while (tok_type_(t, TOK_WHITESPACE))
3215 t = t->next;
3216 /* t should now point to the string */
3217 if (t->type != TOK_STRING) {
3218 error(ERR_NONFATAL,
3219 "`%%strlen` requires string as second parameter");
3220 free_tlist(tline);
3221 free_tlist(origline);
3222 return DIRECTIVE_FOUND;
3225 macro_start = nasm_malloc(sizeof(*macro_start));
3226 macro_start->next = NULL;
3227 make_tok_num(macro_start, nasm_unquote(t->text, NULL));
3228 macro_start->a.mac = NULL;
3231 * We now have a macro name, an implicit parameter count of
3232 * zero, and a numeric token to use as an expansion. Create
3233 * and store an SMacro.
3235 define_smacro(ctx, mname, casesense, 0, macro_start);
3236 free_tlist(tline);
3237 free_tlist(origline);
3238 return DIRECTIVE_FOUND;
3240 case PP_STRCAT:
3241 casesense = true;
3243 tline = tline->next;
3244 skip_white_(tline);
3245 tline = expand_id(tline);
3246 if (!tline || (tline->type != TOK_ID &&
3247 (tline->type != TOK_PREPROC_ID ||
3248 tline->text[1] != '$'))) {
3249 error(ERR_NONFATAL,
3250 "`%%strcat' expects a macro identifier as first parameter");
3251 free_tlist(origline);
3252 return DIRECTIVE_FOUND;
3254 ctx = get_ctx(tline->text, &mname, false);
3255 last = tline;
3256 tline = expand_smacro(tline->next);
3257 last->next = NULL;
3259 len = 0;
3260 list_for_each(t, tline) {
3261 switch (t->type) {
3262 case TOK_WHITESPACE:
3263 break;
3264 case TOK_STRING:
3265 len += t->a.len = nasm_unquote(t->text, NULL);
3266 break;
3267 case TOK_OTHER:
3268 if (!strcmp(t->text, ",")) /* permit comma separators */
3269 break;
3270 /* else fall through */
3271 default:
3272 error(ERR_NONFATAL,
3273 "non-string passed to `%%strcat' (%d)", t->type);
3274 free_tlist(tline);
3275 free_tlist(origline);
3276 return DIRECTIVE_FOUND;
3280 p = pp = nasm_malloc(len);
3281 list_for_each(t, tline) {
3282 if (t->type == TOK_STRING) {
3283 memcpy(p, t->text, t->a.len);
3284 p += t->a.len;
3289 * We now have a macro name, an implicit parameter count of
3290 * zero, and a numeric token to use as an expansion. Create
3291 * and store an SMacro.
3293 macro_start = new_Token(NULL, TOK_STRING, NULL, 0);
3294 macro_start->text = nasm_quote(pp, len);
3295 nasm_free(pp);
3296 define_smacro(ctx, mname, casesense, 0, macro_start);
3297 free_tlist(tline);
3298 free_tlist(origline);
3299 return DIRECTIVE_FOUND;
3301 case PP_SUBSTR:
3303 int64_t a1, a2;
3304 size_t len;
3306 casesense = true;
3308 tline = tline->next;
3309 skip_white_(tline);
3310 tline = expand_id(tline);
3311 if (!tline || (tline->type != TOK_ID &&
3312 (tline->type != TOK_PREPROC_ID ||
3313 tline->text[1] != '$'))) {
3314 error(ERR_NONFATAL,
3315 "`%%substr' expects a macro identifier as first parameter");
3316 free_tlist(origline);
3317 return DIRECTIVE_FOUND;
3319 ctx = get_ctx(tline->text, &mname, false);
3320 last = tline;
3321 tline = expand_smacro(tline->next);
3322 last->next = NULL;
3324 t = tline->next;
3325 while (tok_type_(t, TOK_WHITESPACE))
3326 t = t->next;
3328 /* t should now point to the string */
3329 if (t->type != TOK_STRING) {
3330 error(ERR_NONFATAL,
3331 "`%%substr` requires string as second parameter");
3332 free_tlist(tline);
3333 free_tlist(origline);
3334 return DIRECTIVE_FOUND;
3337 tt = t->next;
3338 tptr = &tt;
3339 tokval.t_type = TOKEN_INVALID;
3340 evalresult = evaluate(ppscan, tptr, &tokval, NULL,
3341 pass, error, NULL);
3342 if (!evalresult) {
3343 free_tlist(tline);
3344 free_tlist(origline);
3345 return DIRECTIVE_FOUND;
3346 } else if (!is_simple(evalresult)) {
3347 error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3348 free_tlist(tline);
3349 free_tlist(origline);
3350 return DIRECTIVE_FOUND;
3352 a1 = evalresult->value-1;
3354 while (tok_type_(tt, TOK_WHITESPACE))
3355 tt = tt->next;
3356 if (!tt) {
3357 a2 = 1; /* Backwards compatibility: one character */
3358 } else {
3359 tokval.t_type = TOKEN_INVALID;
3360 evalresult = evaluate(ppscan, tptr, &tokval, NULL,
3361 pass, error, NULL);
3362 if (!evalresult) {
3363 free_tlist(tline);
3364 free_tlist(origline);
3365 return DIRECTIVE_FOUND;
3366 } else if (!is_simple(evalresult)) {
3367 error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3368 free_tlist(tline);
3369 free_tlist(origline);
3370 return DIRECTIVE_FOUND;
3372 a2 = evalresult->value;
3375 len = nasm_unquote(t->text, NULL);
3376 if (a2 < 0)
3377 a2 = a2+1+len-a1;
3378 if (a1+a2 > (int64_t)len)
3379 a2 = len-a1;
3381 macro_start = nasm_malloc(sizeof(*macro_start));
3382 macro_start->next = NULL;
3383 macro_start->text = nasm_quote((a1 < 0) ? "" : t->text+a1, a2);
3384 macro_start->type = TOK_STRING;
3385 macro_start->a.mac = NULL;
3388 * We now have a macro name, an implicit parameter count of
3389 * zero, and a numeric token to use as an expansion. Create
3390 * and store an SMacro.
3392 define_smacro(ctx, mname, casesense, 0, macro_start);
3393 free_tlist(tline);
3394 free_tlist(origline);
3395 return DIRECTIVE_FOUND;
3398 case PP_ASSIGN:
3399 case PP_IASSIGN:
3400 casesense = (i == PP_ASSIGN);
3402 tline = tline->next;
3403 skip_white_(tline);
3404 tline = expand_id(tline);
3405 if (!tline || (tline->type != TOK_ID &&
3406 (tline->type != TOK_PREPROC_ID ||
3407 tline->text[1] != '$'))) {
3408 error(ERR_NONFATAL,
3409 "`%%%sassign' expects a macro identifier",
3410 (i == PP_IASSIGN ? "i" : ""));
3411 free_tlist(origline);
3412 return DIRECTIVE_FOUND;
3414 ctx = get_ctx(tline->text, &mname, false);
3415 last = tline;
3416 tline = expand_smacro(tline->next);
3417 last->next = NULL;
3419 t = tline;
3420 tptr = &t;
3421 tokval.t_type = TOKEN_INVALID;
3422 evalresult =
3423 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
3424 free_tlist(tline);
3425 if (!evalresult) {
3426 free_tlist(origline);
3427 return DIRECTIVE_FOUND;
3430 if (tokval.t_type)
3431 error(ERR_WARNING|ERR_PASS1,
3432 "trailing garbage after expression ignored");
3434 if (!is_simple(evalresult)) {
3435 error(ERR_NONFATAL,
3436 "non-constant value given to `%%%sassign'",
3437 (i == PP_IASSIGN ? "i" : ""));
3438 free_tlist(origline);
3439 return DIRECTIVE_FOUND;
3442 macro_start = nasm_malloc(sizeof(*macro_start));
3443 macro_start->next = NULL;
3444 make_tok_num(macro_start, reloc_value(evalresult));
3445 macro_start->a.mac = NULL;
3448 * We now have a macro name, an implicit parameter count of
3449 * zero, and a numeric token to use as an expansion. Create
3450 * and store an SMacro.
3452 define_smacro(ctx, mname, casesense, 0, macro_start);
3453 free_tlist(origline);
3454 return DIRECTIVE_FOUND;
3456 case PP_LINE:
3458 * Syntax is `%line nnn[+mmm] [filename]'
3460 tline = tline->next;
3461 skip_white_(tline);
3462 if (!tok_type_(tline, TOK_NUMBER)) {
3463 error(ERR_NONFATAL, "`%%line' expects line number");
3464 free_tlist(origline);
3465 return DIRECTIVE_FOUND;
3467 k = readnum(tline->text, &err);
3468 m = 1;
3469 tline = tline->next;
3470 if (tok_is_(tline, "+")) {
3471 tline = tline->next;
3472 if (!tok_type_(tline, TOK_NUMBER)) {
3473 error(ERR_NONFATAL, "`%%line' expects line increment");
3474 free_tlist(origline);
3475 return DIRECTIVE_FOUND;
3477 m = readnum(tline->text, &err);
3478 tline = tline->next;
3480 skip_white_(tline);
3481 src_set_linnum(k);
3482 istk->lineinc = m;
3483 if (tline) {
3484 nasm_free(src_set_fname(detoken(tline, false)));
3486 free_tlist(origline);
3487 return DIRECTIVE_FOUND;
3489 default:
3490 error(ERR_FATAL,
3491 "preprocessor directive `%s' not yet implemented",
3492 pp_directives[i]);
3493 return DIRECTIVE_FOUND;
3498 * Ensure that a macro parameter contains a condition code and
3499 * nothing else. Return the condition code index if so, or -1
3500 * otherwise.
3502 static int find_cc(Token * t)
3504 Token *tt;
3505 int i, j, k, m;
3507 if (!t)
3508 return -1; /* Probably a %+ without a space */
3510 skip_white_(t);
3511 if (t->type != TOK_ID)
3512 return -1;
3513 tt = t->next;
3514 skip_white_(tt);
3515 if (tt && (tt->type != TOK_OTHER || strcmp(tt->text, ",")))
3516 return -1;
3518 i = -1;
3519 j = ARRAY_SIZE(conditions);
3520 while (j - i > 1) {
3521 k = (j + i) / 2;
3522 m = nasm_stricmp(t->text, conditions[k]);
3523 if (m == 0) {
3524 i = k;
3525 j = -2;
3526 break;
3527 } else if (m < 0) {
3528 j = k;
3529 } else
3530 i = k;
3532 if (j != -2)
3533 return -1;
3534 return i;
3537 static bool paste_tokens(Token **head, bool handle_paste_tokens)
3539 Token **tail, *t, *tt;
3540 Token **paste_head;
3541 bool did_paste = false;
3542 char *tmp;
3544 /* Now handle token pasting... */
3545 paste_head = NULL;
3546 tail = head;
3547 while ((t = *tail) && (tt = t->next)) {
3548 switch (t->type) {
3549 case TOK_WHITESPACE:
3550 if (tt->type == TOK_WHITESPACE) {
3551 /* Zap adjacent whitespace tokens */
3552 t->next = delete_Token(tt);
3553 } else {
3554 /* Do not advance paste_head here */
3555 tail = &t->next;
3557 break;
3558 case TOK_ID:
3559 case TOK_NUMBER:
3560 case TOK_FLOAT:
3562 size_t len = 0;
3563 char *tmp, *p;
3565 while (tt && (tt->type == TOK_ID || tt->type == TOK_PREPROC_ID ||
3566 tt->type == TOK_NUMBER || tt->type == TOK_FLOAT ||
3567 tt->type == TOK_OTHER)) {
3568 len += strlen(tt->text);
3569 tt = tt->next;
3573 * Now tt points to the first token after
3574 * the potential paste area...
3576 if (tt != t->next) {
3577 /* We have at least two tokens... */
3578 len += strlen(t->text);
3579 p = tmp = nasm_malloc(len+1);
3581 while (t != tt) {
3582 strcpy(p, t->text);
3583 p = strchr(p, '\0');
3584 t = delete_Token(t);
3587 t = *tail = tokenize(tmp);
3588 nasm_free(tmp);
3590 while (t->next) {
3591 tail = &t->next;
3592 t = t->next;
3594 t->next = tt; /* Attach the remaining token chain */
3596 did_paste = true;
3598 paste_head = tail;
3599 tail = &t->next;
3600 break;
3602 case TOK_PASTE: /* %+ */
3603 if (handle_paste_tokens) {
3604 /* Zap %+ and whitespace tokens to the right */
3605 while (t && (t->type == TOK_WHITESPACE ||
3606 t->type == TOK_PASTE))
3607 t = *tail = delete_Token(t);
3608 if (!paste_head || !t)
3609 break; /* Nothing to paste with */
3610 tail = paste_head;
3611 t = *tail;
3612 tt = t->next;
3613 while (tok_type_(tt, TOK_WHITESPACE))
3614 tt = t->next = delete_Token(tt);
3616 if (tt) {
3617 tmp = nasm_strcat(t->text, tt->text);
3618 delete_Token(t);
3619 tt = delete_Token(tt);
3620 t = *tail = tokenize(tmp);
3621 nasm_free(tmp);
3622 while (t->next) {
3623 tail = &t->next;
3624 t = t->next;
3626 t->next = tt; /* Attach the remaining token chain */
3627 did_paste = true;
3629 paste_head = tail;
3630 tail = &t->next;
3631 break;
3633 /* else fall through */
3634 default:
3635 tail = &t->next;
3636 if (!tok_type_(t->next, TOK_WHITESPACE))
3637 paste_head = tail;
3638 break;
3641 return did_paste;
3645 * expands to a list of tokens from %{x:y}
3647 static Token *expand_mmac_params_range(MMacro *mac, Token *tline, Token ***last)
3649 Token *t = tline, **tt, *tm, *head;
3650 char *pos;
3651 int fst, lst, j, i;
3653 pos = strchr(tline->text, ':');
3654 nasm_assert(pos);
3656 lst = atoi(pos + 1);
3657 fst = atoi(tline->text + 1);
3660 * only macros params are accounted so
3661 * if someone passes %0 -- we reject such
3662 * value(s)
3664 if (lst == 0 || fst == 0)
3665 goto err;
3667 /* the values should be sane */
3668 if ((fst > (int)mac->nparam || fst < (-(int)mac->nparam)) ||
3669 (lst > (int)mac->nparam || lst < (-(int)mac->nparam)))
3670 goto err;
3672 fst = fst < 0 ? fst + (int)mac->nparam + 1: fst;
3673 lst = lst < 0 ? lst + (int)mac->nparam + 1: lst;
3675 /* counted from zero */
3676 fst--, lst--;
3679 * it will be at least one token
3681 tm = mac->params[(fst + mac->rotate) % mac->nparam];
3682 t = new_Token(NULL, tm->type, tm->text, 0);
3683 head = t, tt = &t->next;
3684 if (fst < lst) {
3685 for (i = fst + 1; i <= lst; i++) {
3686 t = new_Token(NULL, TOK_OTHER, ",", 0);
3687 *tt = t, tt = &t->next;
3688 j = (i + mac->rotate) % mac->nparam;
3689 tm = mac->params[j];
3690 t = new_Token(NULL, tm->type, tm->text, 0);
3691 *tt = t, tt = &t->next;
3693 } else {
3694 for (i = fst - 1; i >= lst; i--) {
3695 t = new_Token(NULL, TOK_OTHER, ",", 0);
3696 *tt = t, tt = &t->next;
3697 j = (i + mac->rotate) % mac->nparam;
3698 tm = mac->params[j];
3699 t = new_Token(NULL, tm->type, tm->text, 0);
3700 *tt = t, tt = &t->next;
3704 *last = tt;
3705 return head;
3707 err:
3708 error(ERR_NONFATAL, "`%%{%s}': macro parameters out of range",
3709 &tline->text[1]);
3710 return tline;
3714 * Expand MMacro-local things: parameter references (%0, %n, %+n,
3715 * %-n) and MMacro-local identifiers (%%foo) as well as
3716 * macro indirection (%[...]) and range (%{..:..}).
3718 static Token *expand_mmac_params(Token * tline)
3720 Token *t, *tt, **tail, *thead;
3721 bool changed = false;
3722 char *pos;
3724 tail = &thead;
3725 thead = NULL;
3727 while (tline) {
3728 if (tline->type == TOK_PREPROC_ID &&
3729 (((tline->text[1] == '+' || tline->text[1] == '-') && tline->text[2]) ||
3730 (tline->text[1] >= '0' && tline->text[1] <= '9') ||
3731 tline->text[1] == '%')) {
3732 char *text = NULL;
3733 int type = 0, cc; /* type = 0 to placate optimisers */
3734 char tmpbuf[30];
3735 unsigned int n;
3736 int i;
3737 MMacro *mac;
3739 t = tline;
3740 tline = tline->next;
3742 mac = istk->mstk;
3743 while (mac && !mac->name) /* avoid mistaking %reps for macros */
3744 mac = mac->next_active;
3745 if (!mac) {
3746 error(ERR_NONFATAL, "`%s': not in a macro call", t->text);
3747 } else {
3748 pos = strchr(t->text, ':');
3749 if (!pos) {
3750 switch (t->text[1]) {
3752 * We have to make a substitution of one of the
3753 * forms %1, %-1, %+1, %%foo, %0.
3755 case '0':
3756 type = TOK_NUMBER;
3757 snprintf(tmpbuf, sizeof(tmpbuf), "%d", mac->nparam);
3758 text = nasm_strdup(tmpbuf);
3759 break;
3760 case '%':
3761 type = TOK_ID;
3762 snprintf(tmpbuf, sizeof(tmpbuf), "..@%"PRIu64".",
3763 mac->unique);
3764 text = nasm_strcat(tmpbuf, t->text + 2);
3765 break;
3766 case '-':
3767 n = atoi(t->text + 2) - 1;
3768 if (n >= mac->nparam)
3769 tt = NULL;
3770 else {
3771 if (mac->nparam > 1)
3772 n = (n + mac->rotate) % mac->nparam;
3773 tt = mac->params[n];
3775 cc = find_cc(tt);
3776 if (cc == -1) {
3777 error(ERR_NONFATAL,
3778 "macro parameter %d is not a condition code",
3779 n + 1);
3780 text = NULL;
3781 } else {
3782 type = TOK_ID;
3783 if (inverse_ccs[cc] == -1) {
3784 error(ERR_NONFATAL,
3785 "condition code `%s' is not invertible",
3786 conditions[cc]);
3787 text = NULL;
3788 } else
3789 text = nasm_strdup(conditions[inverse_ccs[cc]]);
3791 break;
3792 case '+':
3793 n = atoi(t->text + 2) - 1;
3794 if (n >= mac->nparam)
3795 tt = NULL;
3796 else {
3797 if (mac->nparam > 1)
3798 n = (n + mac->rotate) % mac->nparam;
3799 tt = mac->params[n];
3801 cc = find_cc(tt);
3802 if (cc == -1) {
3803 error(ERR_NONFATAL,
3804 "macro parameter %d is not a condition code",
3805 n + 1);
3806 text = NULL;
3807 } else {
3808 type = TOK_ID;
3809 text = nasm_strdup(conditions[cc]);
3811 break;
3812 default:
3813 n = atoi(t->text + 1) - 1;
3814 if (n >= mac->nparam)
3815 tt = NULL;
3816 else {
3817 if (mac->nparam > 1)
3818 n = (n + mac->rotate) % mac->nparam;
3819 tt = mac->params[n];
3821 if (tt) {
3822 for (i = 0; i < mac->paramlen[n]; i++) {
3823 *tail = new_Token(NULL, tt->type, tt->text, 0);
3824 tail = &(*tail)->next;
3825 tt = tt->next;
3828 text = NULL; /* we've done it here */
3829 break;
3831 } else {
3833 * seems we have a parameters range here
3835 Token *head, **last;
3836 head = expand_mmac_params_range(mac, t, &last);
3837 if (head != t) {
3838 *tail = head;
3839 *last = tline;
3840 tline = head;
3841 text = NULL;
3845 if (!text) {
3846 delete_Token(t);
3847 } else {
3848 *tail = t;
3849 tail = &t->next;
3850 t->type = type;
3851 nasm_free(t->text);
3852 t->text = text;
3853 t->a.mac = NULL;
3855 changed = true;
3856 continue;
3857 } else if (tline->type == TOK_INDIRECT) {
3858 t = tline;
3859 tline = tline->next;
3860 tt = tokenize(t->text);
3861 tt = expand_mmac_params(tt);
3862 tt = expand_smacro(tt);
3863 *tail = tt;
3864 while (tt) {
3865 tt->a.mac = NULL; /* Necessary? */
3866 tail = &tt->next;
3867 tt = tt->next;
3869 delete_Token(t);
3870 changed = true;
3871 } else {
3872 t = *tail = tline;
3873 tline = tline->next;
3874 t->a.mac = NULL;
3875 tail = &t->next;
3878 *tail = NULL;
3880 if (changed)
3881 paste_tokens(&thead, false);
3883 return thead;
3887 * Expand all single-line macro calls made in the given line.
3888 * Return the expanded version of the line. The original is deemed
3889 * to be destroyed in the process. (In reality we'll just move
3890 * Tokens from input to output a lot of the time, rather than
3891 * actually bothering to destroy and replicate.)
3894 static Token *expand_smacro(Token * tline)
3896 Token *t, *tt, *mstart, **tail, *thead;
3897 SMacro *head = NULL, *m;
3898 Token **params;
3899 int *paramsize;
3900 unsigned int nparam, sparam;
3901 int brackets;
3902 Token *org_tline = tline;
3903 Context *ctx;
3904 const char *mname;
3905 int deadman = DEADMAN_LIMIT;
3906 bool expanded;
3909 * Trick: we should avoid changing the start token pointer since it can
3910 * be contained in "next" field of other token. Because of this
3911 * we allocate a copy of first token and work with it; at the end of
3912 * routine we copy it back
3914 if (org_tline) {
3915 tline = new_Token(org_tline->next, org_tline->type,
3916 org_tline->text, 0);
3917 tline->a.mac = org_tline->a.mac;
3918 nasm_free(org_tline->text);
3919 org_tline->text = NULL;
3922 expanded = true; /* Always expand %+ at least once */
3924 again:
3925 thead = NULL;
3926 tail = &thead;
3928 while (tline) { /* main token loop */
3929 if (!--deadman) {
3930 error(ERR_NONFATAL, "interminable macro recursion");
3931 goto err;
3934 if ((mname = tline->text)) {
3935 /* if this token is a local macro, look in local context */
3936 if (tline->type == TOK_ID) {
3937 head = (SMacro *)hash_findix(&smacros, mname);
3938 } else if (tline->type == TOK_PREPROC_ID) {
3939 ctx = get_ctx(mname, &mname, true);
3940 head = ctx ? (SMacro *)hash_findix(&ctx->localmac, mname) : NULL;
3941 } else
3942 head = NULL;
3945 * We've hit an identifier. As in is_mmacro below, we first
3946 * check whether the identifier is a single-line macro at
3947 * all, then think about checking for parameters if
3948 * necessary.
3950 list_for_each(m, head)
3951 if (!mstrcmp(m->name, mname, m->casesense))
3952 break;
3953 if (m) {
3954 mstart = tline;
3955 params = NULL;
3956 paramsize = NULL;
3957 if (m->nparam == 0) {
3959 * Simple case: the macro is parameterless. Discard the
3960 * one token that the macro call took, and push the
3961 * expansion back on the to-do stack.
3963 if (!m->expansion) {
3964 if (!strcmp("__FILE__", m->name)) {
3965 int32_t num = 0;
3966 char *file = NULL;
3967 src_get(&num, &file);
3968 tline->text = nasm_quote(file, strlen(file));
3969 tline->type = TOK_STRING;
3970 nasm_free(file);
3971 continue;
3973 if (!strcmp("__LINE__", m->name)) {
3974 nasm_free(tline->text);
3975 make_tok_num(tline, src_get_linnum());
3976 continue;
3978 if (!strcmp("__BITS__", m->name)) {
3979 nasm_free(tline->text);
3980 make_tok_num(tline, globalbits);
3981 continue;
3983 tline = delete_Token(tline);
3984 continue;
3986 } else {
3988 * Complicated case: at least one macro with this name
3989 * exists and takes parameters. We must find the
3990 * parameters in the call, count them, find the SMacro
3991 * that corresponds to that form of the macro call, and
3992 * substitute for the parameters when we expand. What a
3993 * pain.
3995 /*tline = tline->next;
3996 skip_white_(tline); */
3997 do {
3998 t = tline->next;
3999 while (tok_type_(t, TOK_SMAC_END)) {
4000 t->a.mac->in_progress = false;
4001 t->text = NULL;
4002 t = tline->next = delete_Token(t);
4004 tline = t;
4005 } while (tok_type_(tline, TOK_WHITESPACE));
4006 if (!tok_is_(tline, "(")) {
4008 * This macro wasn't called with parameters: ignore
4009 * the call. (Behaviour borrowed from gnu cpp.)
4011 tline = mstart;
4012 m = NULL;
4013 } else {
4014 int paren = 0;
4015 int white = 0;
4016 brackets = 0;
4017 nparam = 0;
4018 sparam = PARAM_DELTA;
4019 params = nasm_malloc(sparam * sizeof(Token *));
4020 params[0] = tline->next;
4021 paramsize = nasm_malloc(sparam * sizeof(int));
4022 paramsize[0] = 0;
4023 while (true) { /* parameter loop */
4025 * For some unusual expansions
4026 * which concatenates function call
4028 t = tline->next;
4029 while (tok_type_(t, TOK_SMAC_END)) {
4030 t->a.mac->in_progress = false;
4031 t->text = NULL;
4032 t = tline->next = delete_Token(t);
4034 tline = t;
4036 if (!tline) {
4037 error(ERR_NONFATAL,
4038 "macro call expects terminating `)'");
4039 break;
4041 if (tline->type == TOK_WHITESPACE
4042 && brackets <= 0) {
4043 if (paramsize[nparam])
4044 white++;
4045 else
4046 params[nparam] = tline->next;
4047 continue; /* parameter loop */
4049 if (tline->type == TOK_OTHER
4050 && tline->text[1] == 0) {
4051 char ch = tline->text[0];
4052 if (ch == ',' && !paren && brackets <= 0) {
4053 if (++nparam >= sparam) {
4054 sparam += PARAM_DELTA;
4055 params = nasm_realloc(params,
4056 sparam * sizeof(Token *));
4057 paramsize = nasm_realloc(paramsize,
4058 sparam * sizeof(int));
4060 params[nparam] = tline->next;
4061 paramsize[nparam] = 0;
4062 white = 0;
4063 continue; /* parameter loop */
4065 if (ch == '{' &&
4066 (brackets > 0 || (brackets == 0 &&
4067 !paramsize[nparam])))
4069 if (!(brackets++)) {
4070 params[nparam] = tline->next;
4071 continue; /* parameter loop */
4074 if (ch == '}' && brackets > 0)
4075 if (--brackets == 0) {
4076 brackets = -1;
4077 continue; /* parameter loop */
4079 if (ch == '(' && !brackets)
4080 paren++;
4081 if (ch == ')' && brackets <= 0)
4082 if (--paren < 0)
4083 break;
4085 if (brackets < 0) {
4086 brackets = 0;
4087 error(ERR_NONFATAL, "braces do not "
4088 "enclose all of macro parameter");
4090 paramsize[nparam] += white + 1;
4091 white = 0;
4092 } /* parameter loop */
4093 nparam++;
4094 while (m && (m->nparam != nparam ||
4095 mstrcmp(m->name, mname,
4096 m->casesense)))
4097 m = m->next;
4098 if (!m)
4099 error(ERR_WARNING|ERR_PASS1|ERR_WARN_MNP,
4100 "macro `%s' exists, "
4101 "but not taking %d parameters",
4102 mstart->text, nparam);
4105 if (m && m->in_progress)
4106 m = NULL;
4107 if (!m) { /* in progess or didn't find '(' or wrong nparam */
4109 * Design question: should we handle !tline, which
4110 * indicates missing ')' here, or expand those
4111 * macros anyway, which requires the (t) test a few
4112 * lines down?
4114 nasm_free(params);
4115 nasm_free(paramsize);
4116 tline = mstart;
4117 } else {
4119 * Expand the macro: we are placed on the last token of the
4120 * call, so that we can easily split the call from the
4121 * following tokens. We also start by pushing an SMAC_END
4122 * token for the cycle removal.
4124 t = tline;
4125 if (t) {
4126 tline = t->next;
4127 t->next = NULL;
4129 tt = new_Token(tline, TOK_SMAC_END, NULL, 0);
4130 tt->a.mac = m;
4131 m->in_progress = true;
4132 tline = tt;
4133 list_for_each(t, m->expansion) {
4134 if (t->type >= TOK_SMAC_PARAM) {
4135 Token *pcopy = tline, **ptail = &pcopy;
4136 Token *ttt, *pt;
4137 int i;
4139 ttt = params[t->type - TOK_SMAC_PARAM];
4140 i = paramsize[t->type - TOK_SMAC_PARAM];
4141 while (--i >= 0) {
4142 pt = *ptail = new_Token(tline, ttt->type,
4143 ttt->text, 0);
4144 ptail = &pt->next;
4145 ttt = ttt->next;
4147 tline = pcopy;
4148 } else if (t->type == TOK_PREPROC_Q) {
4149 tt = new_Token(tline, TOK_ID, mname, 0);
4150 tline = tt;
4151 } else if (t->type == TOK_PREPROC_QQ) {
4152 tt = new_Token(tline, TOK_ID, m->name, 0);
4153 tline = tt;
4154 } else {
4155 tt = new_Token(tline, t->type, t->text, 0);
4156 tline = tt;
4161 * Having done that, get rid of the macro call, and clean
4162 * up the parameters.
4164 nasm_free(params);
4165 nasm_free(paramsize);
4166 free_tlist(mstart);
4167 expanded = true;
4168 continue; /* main token loop */
4173 if (tline->type == TOK_SMAC_END) {
4174 tline->a.mac->in_progress = false;
4175 tline = delete_Token(tline);
4176 } else {
4177 t = *tail = tline;
4178 tline = tline->next;
4179 t->a.mac = NULL;
4180 t->next = NULL;
4181 tail = &t->next;
4186 * Now scan the entire line and look for successive TOK_IDs that resulted
4187 * after expansion (they can't be produced by tokenize()). The successive
4188 * TOK_IDs should be concatenated.
4189 * Also we look for %+ tokens and concatenate the tokens before and after
4190 * them (without white spaces in between).
4192 if (expanded && paste_tokens(&thead, true)) {
4194 * If we concatenated something, *and* we had previously expanded
4195 * an actual macro, scan the lines again for macros...
4197 tline = thead;
4198 expanded = false;
4199 goto again;
4202 err:
4203 if (org_tline) {
4204 if (thead) {
4205 *org_tline = *thead;
4206 /* since we just gave text to org_line, don't free it */
4207 thead->text = NULL;
4208 delete_Token(thead);
4209 } else {
4210 /* the expression expanded to empty line;
4211 we can't return NULL for some reasons
4212 we just set the line to a single WHITESPACE token. */
4213 memset(org_tline, 0, sizeof(*org_tline));
4214 org_tline->text = NULL;
4215 org_tline->type = TOK_WHITESPACE;
4217 thead = org_tline;
4220 return thead;
4224 * Similar to expand_smacro but used exclusively with macro identifiers
4225 * right before they are fetched in. The reason is that there can be
4226 * identifiers consisting of several subparts. We consider that if there
4227 * are more than one element forming the name, user wants a expansion,
4228 * otherwise it will be left as-is. Example:
4230 * %define %$abc cde
4232 * the identifier %$abc will be left as-is so that the handler for %define
4233 * will suck it and define the corresponding value. Other case:
4235 * %define _%$abc cde
4237 * In this case user wants name to be expanded *before* %define starts
4238 * working, so we'll expand %$abc into something (if it has a value;
4239 * otherwise it will be left as-is) then concatenate all successive
4240 * PP_IDs into one.
4242 static Token *expand_id(Token * tline)
4244 Token *cur, *oldnext = NULL;
4246 if (!tline || !tline->next)
4247 return tline;
4249 cur = tline;
4250 while (cur->next &&
4251 (cur->next->type == TOK_ID ||
4252 cur->next->type == TOK_PREPROC_ID
4253 || cur->next->type == TOK_NUMBER))
4254 cur = cur->next;
4256 /* If identifier consists of just one token, don't expand */
4257 if (cur == tline)
4258 return tline;
4260 if (cur) {
4261 oldnext = cur->next; /* Detach the tail past identifier */
4262 cur->next = NULL; /* so that expand_smacro stops here */
4265 tline = expand_smacro(tline);
4267 if (cur) {
4268 /* expand_smacro possibly changhed tline; re-scan for EOL */
4269 cur = tline;
4270 while (cur && cur->next)
4271 cur = cur->next;
4272 if (cur)
4273 cur->next = oldnext;
4276 return tline;
4280 * Determine whether the given line constitutes a multi-line macro
4281 * call, and return the MMacro structure called if so. Doesn't have
4282 * to check for an initial label - that's taken care of in
4283 * expand_mmacro - but must check numbers of parameters. Guaranteed
4284 * to be called with tline->type == TOK_ID, so the putative macro
4285 * name is easy to find.
4287 static MMacro *is_mmacro(Token * tline, Token *** params_array)
4289 MMacro *head, *m;
4290 Token **params;
4291 int nparam;
4293 head = (MMacro *) hash_findix(&mmacros, tline->text);
4296 * Efficiency: first we see if any macro exists with the given
4297 * name. If not, we can return NULL immediately. _Then_ we
4298 * count the parameters, and then we look further along the
4299 * list if necessary to find the proper MMacro.
4301 list_for_each(m, head)
4302 if (!mstrcmp(m->name, tline->text, m->casesense))
4303 break;
4304 if (!m)
4305 return NULL;
4308 * OK, we have a potential macro. Count and demarcate the
4309 * parameters.
4311 count_mmac_params(tline->next, &nparam, &params);
4314 * So we know how many parameters we've got. Find the MMacro
4315 * structure that handles this number.
4317 while (m) {
4318 if (m->nparam_min <= nparam
4319 && (m->plus || nparam <= m->nparam_max)) {
4321 * This one is right. Just check if cycle removal
4322 * prohibits us using it before we actually celebrate...
4324 if (m->in_progress > m->max_depth) {
4325 if (m->max_depth > 0) {
4326 error(ERR_WARNING,
4327 "reached maximum recursion depth of %i",
4328 m->max_depth);
4330 nasm_free(params);
4331 return NULL;
4334 * It's right, and we can use it. Add its default
4335 * parameters to the end of our list if necessary.
4337 if (m->defaults && nparam < m->nparam_min + m->ndefs) {
4338 params =
4339 nasm_realloc(params,
4340 ((m->nparam_min + m->ndefs +
4341 1) * sizeof(*params)));
4342 while (nparam < m->nparam_min + m->ndefs) {
4343 params[nparam] = m->defaults[nparam - m->nparam_min];
4344 nparam++;
4348 * If we've gone over the maximum parameter count (and
4349 * we're in Plus mode), ignore parameters beyond
4350 * nparam_max.
4352 if (m->plus && nparam > m->nparam_max)
4353 nparam = m->nparam_max;
4355 * Then terminate the parameter list, and leave.
4357 if (!params) { /* need this special case */
4358 params = nasm_malloc(sizeof(*params));
4359 nparam = 0;
4361 params[nparam] = NULL;
4362 *params_array = params;
4363 return m;
4366 * This one wasn't right: look for the next one with the
4367 * same name.
4369 list_for_each(m, m->next)
4370 if (!mstrcmp(m->name, tline->text, m->casesense))
4371 break;
4375 * After all that, we didn't find one with the right number of
4376 * parameters. Issue a warning, and fail to expand the macro.
4378 error(ERR_WARNING|ERR_PASS1|ERR_WARN_MNP,
4379 "macro `%s' exists, but not taking %d parameters",
4380 tline->text, nparam);
4381 nasm_free(params);
4382 return NULL;
4387 * Save MMacro invocation specific fields in
4388 * preparation for a recursive macro expansion
4390 static void push_mmacro(MMacro *m)
4392 MMacroInvocation *i;
4394 i = nasm_malloc(sizeof(MMacroInvocation));
4395 i->prev = m->prev;
4396 i->params = m->params;
4397 i->iline = m->iline;
4398 i->nparam = m->nparam;
4399 i->rotate = m->rotate;
4400 i->paramlen = m->paramlen;
4401 i->unique = m->unique;
4402 i->condcnt = m->condcnt;
4403 m->prev = i;
4408 * Restore MMacro invocation specific fields that were
4409 * saved during a previous recursive macro expansion
4411 static void pop_mmacro(MMacro *m)
4413 MMacroInvocation *i;
4415 if (m->prev) {
4416 i = m->prev;
4417 m->prev = i->prev;
4418 m->params = i->params;
4419 m->iline = i->iline;
4420 m->nparam = i->nparam;
4421 m->rotate = i->rotate;
4422 m->paramlen = i->paramlen;
4423 m->unique = i->unique;
4424 m->condcnt = i->condcnt;
4425 nasm_free(i);
4431 * Expand the multi-line macro call made by the given line, if
4432 * there is one to be expanded. If there is, push the expansion on
4433 * istk->expansion and return 1. Otherwise return 0.
4435 static int expand_mmacro(Token * tline)
4437 Token *startline = tline;
4438 Token *label = NULL;
4439 int dont_prepend = 0;
4440 Token **params, *t, *mtok, *tt;
4441 MMacro *m;
4442 Line *l, *ll;
4443 int i, nparam, *paramlen;
4444 const char *mname;
4446 t = tline;
4447 skip_white_(t);
4448 /* if (!tok_type_(t, TOK_ID)) Lino 02/25/02 */
4449 if (!tok_type_(t, TOK_ID) && !tok_type_(t, TOK_PREPROC_ID))
4450 return 0;
4451 mtok = t;
4452 m = is_mmacro(t, &params);
4453 if (m) {
4454 mname = t->text;
4455 } else {
4456 Token *last;
4458 * We have an id which isn't a macro call. We'll assume
4459 * it might be a label; we'll also check to see if a
4460 * colon follows it. Then, if there's another id after
4461 * that lot, we'll check it again for macro-hood.
4463 label = last = t;
4464 t = t->next;
4465 if (tok_type_(t, TOK_WHITESPACE))
4466 last = t, t = t->next;
4467 if (tok_is_(t, ":")) {
4468 dont_prepend = 1;
4469 last = t, t = t->next;
4470 if (tok_type_(t, TOK_WHITESPACE))
4471 last = t, t = t->next;
4473 if (!tok_type_(t, TOK_ID) || !(m = is_mmacro(t, &params)))
4474 return 0;
4475 last->next = NULL;
4476 mname = t->text;
4477 tline = t;
4481 * Fix up the parameters: this involves stripping leading and
4482 * trailing whitespace, then stripping braces if they are
4483 * present.
4485 for (nparam = 0; params[nparam]; nparam++) ;
4486 paramlen = nparam ? nasm_malloc(nparam * sizeof(*paramlen)) : NULL;
4488 for (i = 0; params[i]; i++) {
4489 int brace = false;
4490 int comma = (!m->plus || i < nparam - 1);
4492 t = params[i];
4493 skip_white_(t);
4494 if (tok_is_(t, "{"))
4495 t = t->next, brace = true, comma = false;
4496 params[i] = t;
4497 paramlen[i] = 0;
4498 while (t) {
4499 if (comma && t->type == TOK_OTHER && !strcmp(t->text, ","))
4500 break; /* ... because we have hit a comma */
4501 if (comma && t->type == TOK_WHITESPACE
4502 && tok_is_(t->next, ","))
4503 break; /* ... or a space then a comma */
4504 if (brace && t->type == TOK_OTHER && !strcmp(t->text, "}"))
4505 break; /* ... or a brace */
4506 t = t->next;
4507 paramlen[i]++;
4512 * OK, we have a MMacro structure together with a set of
4513 * parameters. We must now go through the expansion and push
4514 * copies of each Line on to istk->expansion. Substitution of
4515 * parameter tokens and macro-local tokens doesn't get done
4516 * until the single-line macro substitution process; this is
4517 * because delaying them allows us to change the semantics
4518 * later through %rotate.
4520 * First, push an end marker on to istk->expansion, mark this
4521 * macro as in progress, and set up its invocation-specific
4522 * variables.
4524 ll = nasm_malloc(sizeof(Line));
4525 ll->next = istk->expansion;
4526 ll->finishes = m;
4527 ll->first = NULL;
4528 istk->expansion = ll;
4531 * Save the previous MMacro expansion in the case of
4532 * macro recursion
4534 if (m->max_depth && m->in_progress)
4535 push_mmacro(m);
4537 m->in_progress ++;
4538 m->params = params;
4539 m->iline = tline;
4540 m->nparam = nparam;
4541 m->rotate = 0;
4542 m->paramlen = paramlen;
4543 m->unique = unique++;
4544 m->lineno = 0;
4545 m->condcnt = 0;
4547 m->next_active = istk->mstk;
4548 istk->mstk = m;
4550 list_for_each(l, m->expansion) {
4551 Token **tail;
4553 ll = nasm_malloc(sizeof(Line));
4554 ll->finishes = NULL;
4555 ll->next = istk->expansion;
4556 istk->expansion = ll;
4557 tail = &ll->first;
4559 list_for_each(t, l->first) {
4560 Token *x = t;
4561 switch (t->type) {
4562 case TOK_PREPROC_Q:
4563 tt = *tail = new_Token(NULL, TOK_ID, mname, 0);
4564 break;
4565 case TOK_PREPROC_QQ:
4566 tt = *tail = new_Token(NULL, TOK_ID, m->name, 0);
4567 break;
4568 case TOK_PREPROC_ID:
4569 if (t->text[1] == '0' && t->text[2] == '0') {
4570 dont_prepend = -1;
4571 x = label;
4572 if (!x)
4573 continue;
4575 /* fall through */
4576 default:
4577 tt = *tail = new_Token(NULL, x->type, x->text, 0);
4578 break;
4580 tail = &tt->next;
4582 *tail = NULL;
4586 * If we had a label, push it on as the first line of
4587 * the macro expansion.
4589 if (label) {
4590 if (dont_prepend < 0)
4591 free_tlist(startline);
4592 else {
4593 ll = nasm_malloc(sizeof(Line));
4594 ll->finishes = NULL;
4595 ll->next = istk->expansion;
4596 istk->expansion = ll;
4597 ll->first = startline;
4598 if (!dont_prepend) {
4599 while (label->next)
4600 label = label->next;
4601 label->next = tt = new_Token(NULL, TOK_OTHER, ":", 0);
4606 list->uplevel(m->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
4608 return 1;
4611 /* The function that actually does the error reporting */
4612 static void verror(int severity, const char *fmt, va_list arg)
4614 char buff[1024];
4616 vsnprintf(buff, sizeof(buff), fmt, arg);
4618 if (istk && istk->mstk && istk->mstk->name)
4619 nasm_error(severity, "(%s:%d) %s", istk->mstk->name,
4620 istk->mstk->lineno, buff);
4621 else
4622 nasm_error(severity, "%s", buff);
4626 * Since preprocessor always operate only on the line that didn't
4627 * arrived yet, we should always use ERR_OFFBY1.
4629 static void error(int severity, const char *fmt, ...)
4631 va_list arg;
4633 /* If we're in a dead branch of IF or something like it, ignore the error */
4634 if (istk && istk->conds && !emitting(istk->conds->state))
4635 return;
4637 va_start(arg, fmt);
4638 verror(severity, fmt, arg);
4639 va_end(arg);
4643 * Because %else etc are evaluated in the state context
4644 * of the previous branch, errors might get lost with error():
4645 * %if 0 ... %else trailing garbage ... %endif
4646 * So %else etc should report errors with this function.
4648 static void error_precond(int severity, const char *fmt, ...)
4650 va_list arg;
4652 /* Only ignore the error if it's really in a dead branch */
4653 if (istk && istk->conds && istk->conds->state == COND_NEVER)
4654 return;
4656 va_start(arg, fmt);
4657 verror(severity, fmt, arg);
4658 va_end(arg);
4661 static void
4662 pp_reset(char *file, int apass, ListGen * listgen, StrList **deplist)
4664 Token *t;
4666 cstk = NULL;
4667 istk = nasm_malloc(sizeof(Include));
4668 istk->next = NULL;
4669 istk->conds = NULL;
4670 istk->expansion = NULL;
4671 istk->mstk = NULL;
4672 istk->fp = fopen(file, "r");
4673 istk->fname = NULL;
4674 src_set_fname(nasm_strdup(file));
4675 src_set_linnum(0);
4676 istk->lineinc = 1;
4677 if (!istk->fp)
4678 error(ERR_FATAL|ERR_NOFILE, "unable to open input file `%s'",
4679 file);
4680 defining = NULL;
4681 nested_mac_count = 0;
4682 nested_rep_count = 0;
4683 init_macros();
4684 unique = 0;
4685 if (tasm_compatible_mode) {
4686 stdmacpos = nasm_stdmac;
4687 } else {
4688 stdmacpos = nasm_stdmac_after_tasm;
4690 any_extrastdmac = extrastdmac && *extrastdmac;
4691 do_predef = true;
4692 list = listgen;
4695 * 0 for dependencies, 1 for preparatory passes, 2 for final pass.
4696 * The caller, however, will also pass in 3 for preprocess-only so
4697 * we can set __PASS__ accordingly.
4699 pass = apass > 2 ? 2 : apass;
4701 dephead = deptail = deplist;
4702 if (deplist) {
4703 StrList *sl = nasm_malloc(strlen(file)+1+sizeof sl->next);
4704 sl->next = NULL;
4705 strcpy(sl->str, file);
4706 *deptail = sl;
4707 deptail = &sl->next;
4711 * Define the __PASS__ macro. This is defined here unlike
4712 * all the other builtins, because it is special -- it varies between
4713 * passes.
4715 t = nasm_malloc(sizeof(*t));
4716 t->next = NULL;
4717 make_tok_num(t, apass);
4718 t->a.mac = NULL;
4719 define_smacro(NULL, "__PASS__", true, 0, t);
4722 static char *pp_getline(void)
4724 char *line;
4725 Token *tline;
4727 while (1) {
4729 * Fetch a tokenized line, either from the macro-expansion
4730 * buffer or from the input file.
4732 tline = NULL;
4733 while (istk->expansion && istk->expansion->finishes) {
4734 Line *l = istk->expansion;
4735 if (!l->finishes->name && l->finishes->in_progress > 1) {
4736 Line *ll;
4739 * This is a macro-end marker for a macro with no
4740 * name, which means it's not really a macro at all
4741 * but a %rep block, and the `in_progress' field is
4742 * more than 1, meaning that we still need to
4743 * repeat. (1 means the natural last repetition; 0
4744 * means termination by %exitrep.) We have
4745 * therefore expanded up to the %endrep, and must
4746 * push the whole block on to the expansion buffer
4747 * again. We don't bother to remove the macro-end
4748 * marker: we'd only have to generate another one
4749 * if we did.
4751 l->finishes->in_progress--;
4752 list_for_each(l, l->finishes->expansion) {
4753 Token *t, *tt, **tail;
4755 ll = nasm_malloc(sizeof(Line));
4756 ll->next = istk->expansion;
4757 ll->finishes = NULL;
4758 ll->first = NULL;
4759 tail = &ll->first;
4761 list_for_each(t, l->first) {
4762 if (t->text || t->type == TOK_WHITESPACE) {
4763 tt = *tail = new_Token(NULL, t->type, t->text, 0);
4764 tail = &tt->next;
4768 istk->expansion = ll;
4770 } else {
4772 * Check whether a `%rep' was started and not ended
4773 * within this macro expansion. This can happen and
4774 * should be detected. It's a fatal error because
4775 * I'm too confused to work out how to recover
4776 * sensibly from it.
4778 if (defining) {
4779 if (defining->name)
4780 error(ERR_PANIC,
4781 "defining with name in expansion");
4782 else if (istk->mstk->name)
4783 error(ERR_FATAL,
4784 "`%%rep' without `%%endrep' within"
4785 " expansion of macro `%s'",
4786 istk->mstk->name);
4790 * FIXME: investigate the relationship at this point between
4791 * istk->mstk and l->finishes
4794 MMacro *m = istk->mstk;
4795 istk->mstk = m->next_active;
4796 if (m->name) {
4798 * This was a real macro call, not a %rep, and
4799 * therefore the parameter information needs to
4800 * be freed.
4802 if (m->prev) {
4803 pop_mmacro(m);
4804 l->finishes->in_progress --;
4805 } else {
4806 nasm_free(m->params);
4807 free_tlist(m->iline);
4808 nasm_free(m->paramlen);
4809 l->finishes->in_progress = 0;
4811 } else
4812 free_mmacro(m);
4814 istk->expansion = l->next;
4815 nasm_free(l);
4816 list->downlevel(LIST_MACRO);
4819 while (1) { /* until we get a line we can use */
4821 if (istk->expansion) { /* from a macro expansion */
4822 char *p;
4823 Line *l = istk->expansion;
4824 if (istk->mstk)
4825 istk->mstk->lineno++;
4826 tline = l->first;
4827 istk->expansion = l->next;
4828 nasm_free(l);
4829 p = detoken(tline, false);
4830 list->line(LIST_MACRO, p);
4831 nasm_free(p);
4832 break;
4834 line = read_line();
4835 if (line) { /* from the current input file */
4836 line = prepreproc(line);
4837 tline = tokenize(line);
4838 nasm_free(line);
4839 break;
4842 * The current file has ended; work down the istk
4845 Include *i = istk;
4846 fclose(i->fp);
4847 if (i->conds)
4848 error(ERR_FATAL,
4849 "expected `%%endif' before end of file");
4850 /* only set line and file name if there's a next node */
4851 if (i->next) {
4852 src_set_linnum(i->lineno);
4853 nasm_free(src_set_fname(i->fname));
4855 istk = i->next;
4856 list->downlevel(LIST_INCLUDE);
4857 nasm_free(i);
4858 if (!istk)
4859 return NULL;
4860 if (istk->expansion && istk->expansion->finishes)
4861 break;
4866 * We must expand MMacro parameters and MMacro-local labels
4867 * _before_ we plunge into directive processing, to cope
4868 * with things like `%define something %1' such as STRUC
4869 * uses. Unless we're _defining_ a MMacro, in which case
4870 * those tokens should be left alone to go into the
4871 * definition; and unless we're in a non-emitting
4872 * condition, in which case we don't want to meddle with
4873 * anything.
4875 if (!defining && !(istk->conds && !emitting(istk->conds->state))
4876 && !(istk->mstk && !istk->mstk->in_progress)) {
4877 tline = expand_mmac_params(tline);
4881 * Check the line to see if it's a preprocessor directive.
4883 if (do_directive(tline) == DIRECTIVE_FOUND) {
4884 continue;
4885 } else if (defining) {
4887 * We're defining a multi-line macro. We emit nothing
4888 * at all, and just
4889 * shove the tokenized line on to the macro definition.
4891 Line *l = nasm_malloc(sizeof(Line));
4892 l->next = defining->expansion;
4893 l->first = tline;
4894 l->finishes = NULL;
4895 defining->expansion = l;
4896 continue;
4897 } else if (istk->conds && !emitting(istk->conds->state)) {
4899 * We're in a non-emitting branch of a condition block.
4900 * Emit nothing at all, not even a blank line: when we
4901 * emerge from the condition we'll give a line-number
4902 * directive so we keep our place correctly.
4904 free_tlist(tline);
4905 continue;
4906 } else if (istk->mstk && !istk->mstk->in_progress) {
4908 * We're in a %rep block which has been terminated, so
4909 * we're walking through to the %endrep without
4910 * emitting anything. Emit nothing at all, not even a
4911 * blank line: when we emerge from the %rep block we'll
4912 * give a line-number directive so we keep our place
4913 * correctly.
4915 free_tlist(tline);
4916 continue;
4917 } else {
4918 tline = expand_smacro(tline);
4919 if (!expand_mmacro(tline)) {
4921 * De-tokenize the line again, and emit it.
4923 line = detoken(tline, true);
4924 free_tlist(tline);
4925 break;
4926 } else {
4927 continue; /* expand_mmacro calls free_tlist */
4932 return line;
4935 static void pp_cleanup(int pass)
4937 if (defining) {
4938 if (defining->name) {
4939 error(ERR_NONFATAL,
4940 "end of file while still defining macro `%s'",
4941 defining->name);
4942 } else {
4943 error(ERR_NONFATAL, "end of file while still in %%rep");
4946 free_mmacro(defining);
4947 defining = NULL;
4949 while (cstk)
4950 ctx_pop();
4951 free_macros();
4952 while (istk) {
4953 Include *i = istk;
4954 istk = istk->next;
4955 fclose(i->fp);
4956 nasm_free(i->fname);
4957 nasm_free(i);
4959 while (cstk)
4960 ctx_pop();
4961 nasm_free(src_set_fname(NULL));
4962 if (pass == 0) {
4963 IncPath *i;
4964 free_llist(predef);
4965 delete_Blocks();
4966 while ((i = ipath)) {
4967 ipath = i->next;
4968 if (i->path)
4969 nasm_free(i->path);
4970 nasm_free(i);
4975 void pp_include_path(char *path)
4977 IncPath *i;
4979 i = nasm_malloc(sizeof(IncPath));
4980 i->path = path ? nasm_strdup(path) : NULL;
4981 i->next = NULL;
4983 if (ipath) {
4984 IncPath *j = ipath;
4985 while (j->next)
4986 j = j->next;
4987 j->next = i;
4988 } else {
4989 ipath = i;
4993 void pp_pre_include(char *fname)
4995 Token *inc, *space, *name;
4996 Line *l;
4998 name = new_Token(NULL, TOK_INTERNAL_STRING, fname, 0);
4999 space = new_Token(name, TOK_WHITESPACE, NULL, 0);
5000 inc = new_Token(space, TOK_PREPROC_ID, "%include", 0);
5002 l = nasm_malloc(sizeof(Line));
5003 l->next = predef;
5004 l->first = inc;
5005 l->finishes = NULL;
5006 predef = l;
5009 void pp_pre_define(char *definition)
5011 Token *def, *space;
5012 Line *l;
5013 char *equals;
5015 equals = strchr(definition, '=');
5016 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
5017 def = new_Token(space, TOK_PREPROC_ID, "%define", 0);
5018 if (equals)
5019 *equals = ' ';
5020 space->next = tokenize(definition);
5021 if (equals)
5022 *equals = '=';
5024 l = nasm_malloc(sizeof(Line));
5025 l->next = predef;
5026 l->first = def;
5027 l->finishes = NULL;
5028 predef = l;
5031 void pp_pre_undefine(char *definition)
5033 Token *def, *space;
5034 Line *l;
5036 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
5037 def = new_Token(space, TOK_PREPROC_ID, "%undef", 0);
5038 space->next = tokenize(definition);
5040 l = nasm_malloc(sizeof(Line));
5041 l->next = predef;
5042 l->first = def;
5043 l->finishes = NULL;
5044 predef = l;
5048 * Added by Keith Kanios:
5050 * This function is used to assist with "runtime" preprocessor
5051 * directives. (e.g. pp_runtime("%define __BITS__ 64");)
5053 * ERRORS ARE IGNORED HERE, SO MAKE COMPLETELY SURE THAT YOU
5054 * PASS A VALID STRING TO THIS FUNCTION!!!!!
5057 void pp_runtime(char *definition)
5059 Token *def;
5061 def = tokenize(definition);
5062 if (do_directive(def) == NO_DIRECTIVE_FOUND)
5063 free_tlist(def);
5067 void pp_extra_stdmac(macros_t *macros)
5069 extrastdmac = macros;
5072 static void make_tok_num(Token * tok, int64_t val)
5074 char numbuf[20];
5075 snprintf(numbuf, sizeof(numbuf), "%"PRId64"", val);
5076 tok->text = nasm_strdup(numbuf);
5077 tok->type = TOK_NUMBER;
5080 Preproc nasmpp = {
5081 pp_reset,
5082 pp_getline,
5083 pp_cleanup