BR 2672117: handle long lines for db/dw/etc.
[nasm/perl-rewrite.git] / preproc.c
blob0579b0f42bfc9730d61c60c5405161efe8f32f53
1 /* preproc.c macro preprocessor for the Netwide Assembler
3 * The Netwide Assembler is copyright (C) 1996 Simon Tatham and
4 * Julian Hall. All rights reserved. The software is
5 * redistributable under the license given in the file "LICENSE"
6 * distributed in the NASM archive.
8 * initial version 18/iii/97 by Simon Tatham
9 */
11 /* Typical flow of text through preproc
13 * pp_getline gets tokenized lines, either
15 * from a macro expansion
17 * or
18 * {
19 * read_line gets raw text from stdmacpos, or predef, or current input file
20 * tokenize converts to tokens
21 * }
23 * expand_mmac_params is used to expand %1 etc., unless a macro is being
24 * defined or a false conditional is being processed
25 * (%0, %1, %+1, %-1, %%foo
27 * do_directive checks for directives
29 * expand_smacro is used to expand single line macros
31 * expand_mmacro is used to expand multi-line macros
33 * detoken is used to convert the line back to text
36 #include "compiler.h"
38 #include <stdio.h>
39 #include <stdarg.h>
40 #include <stdlib.h>
41 #include <stddef.h>
42 #include <string.h>
43 #include <ctype.h>
44 #include <limits.h>
45 #include <inttypes.h>
47 #include "nasm.h"
48 #include "nasmlib.h"
49 #include "preproc.h"
50 #include "hashtbl.h"
51 #include "quote.h"
52 #include "stdscan.h"
53 #include "tokens.h"
54 #include "tables.h"
56 typedef struct SMacro SMacro;
57 typedef struct MMacro MMacro;
58 typedef struct Context Context;
59 typedef struct Token Token;
60 typedef struct Blocks Blocks;
61 typedef struct Line Line;
62 typedef struct Include Include;
63 typedef struct Cond Cond;
64 typedef struct IncPath IncPath;
67 * Note on the storage of both SMacro and MMacros: the hash table
68 * indexes them case-insensitively, and we then have to go through a
69 * linked list of potential case aliases (and, for MMacros, parameter
70 * ranges); this is to preserve the matching semantics of the earlier
71 * code. If the number of case aliases for a specific macro is a
72 * performance issue, you may want to reconsider your coding style.
76 * Store the definition of a single-line macro.
78 struct SMacro {
79 SMacro *next;
80 char *name;
81 bool casesense;
82 bool in_progress;
83 unsigned int nparam;
84 Token *expansion;
88 * Store the definition of a multi-line macro. This is also used to
89 * store the interiors of `%rep...%endrep' blocks, which are
90 * effectively self-re-invoking multi-line macros which simply
91 * don't have a name or bother to appear in the hash tables. %rep
92 * blocks are signified by having a NULL `name' field.
94 * In a MMacro describing a `%rep' block, the `in_progress' field
95 * isn't merely boolean, but gives the number of repeats left to
96 * run.
98 * The `next' field is used for storing MMacros in hash tables; the
99 * `next_active' field is for stacking them on istk entries.
101 * When a MMacro is being expanded, `params', `iline', `nparam',
102 * `paramlen', `rotate' and `unique' are local to the invocation.
104 struct MMacro {
105 MMacro *next;
106 char *name;
107 int nparam_min, nparam_max;
108 bool casesense;
109 bool plus; /* is the last parameter greedy? */
110 bool nolist; /* is this macro listing-inhibited? */
111 int64_t in_progress;
112 Token *dlist; /* All defaults as one list */
113 Token **defaults; /* Parameter default pointers */
114 int ndefs; /* number of default parameters */
115 Line *expansion;
117 MMacro *next_active;
118 MMacro *rep_nest; /* used for nesting %rep */
119 Token **params; /* actual parameters */
120 Token *iline; /* invocation line */
121 unsigned int nparam, rotate;
122 int *paramlen;
123 uint64_t unique;
124 int lineno; /* Current line number on expansion */
128 * The context stack is composed of a linked list of these.
130 struct Context {
131 Context *next;
132 char *name;
133 struct hash_table localmac;
134 uint32_t number;
138 * This is the internal form which we break input lines up into.
139 * Typically stored in linked lists.
141 * Note that `type' serves a double meaning: TOK_SMAC_PARAM is not
142 * necessarily used as-is, but is intended to denote the number of
143 * the substituted parameter. So in the definition
145 * %define a(x,y) ( (x) & ~(y) )
147 * the token representing `x' will have its type changed to
148 * TOK_SMAC_PARAM, but the one representing `y' will be
149 * TOK_SMAC_PARAM+1.
151 * TOK_INTERNAL_STRING is a dirty hack: it's a single string token
152 * which doesn't need quotes around it. Used in the pre-include
153 * mechanism as an alternative to trying to find a sensible type of
154 * quote to use on the filename we were passed.
156 enum pp_token_type {
157 TOK_NONE = 0, TOK_WHITESPACE, TOK_COMMENT, TOK_ID,
158 TOK_PREPROC_ID, TOK_STRING,
159 TOK_NUMBER, TOK_FLOAT, TOK_SMAC_END, TOK_OTHER,
160 TOK_INTERNAL_STRING,
161 TOK_PREPROC_Q, TOK_PREPROC_QQ,
162 TOK_INDIRECT, /* %[...] */
163 TOK_SMAC_PARAM, /* MUST BE LAST IN THE LIST!!! */
164 TOK_MAX = INT_MAX /* Keep compiler from reducing the range */
167 struct Token {
168 Token *next;
169 char *text;
170 union {
171 SMacro *mac; /* associated macro for TOK_SMAC_END */
172 size_t len; /* scratch length field */
173 } a; /* Auxiliary data */
174 enum pp_token_type type;
178 * Multi-line macro definitions are stored as a linked list of
179 * these, which is essentially a container to allow several linked
180 * lists of Tokens.
182 * Note that in this module, linked lists are treated as stacks
183 * wherever possible. For this reason, Lines are _pushed_ on to the
184 * `expansion' field in MMacro structures, so that the linked list,
185 * if walked, would give the macro lines in reverse order; this
186 * means that we can walk the list when expanding a macro, and thus
187 * push the lines on to the `expansion' field in _istk_ in reverse
188 * order (so that when popped back off they are in the right
189 * order). It may seem cockeyed, and it relies on my design having
190 * an even number of steps in, but it works...
192 * Some of these structures, rather than being actual lines, are
193 * markers delimiting the end of the expansion of a given macro.
194 * This is for use in the cycle-tracking and %rep-handling code.
195 * Such structures have `finishes' non-NULL, and `first' NULL. All
196 * others have `finishes' NULL, but `first' may still be NULL if
197 * the line is blank.
199 struct Line {
200 Line *next;
201 MMacro *finishes;
202 Token *first;
206 * To handle an arbitrary level of file inclusion, we maintain a
207 * stack (ie linked list) of these things.
209 struct Include {
210 Include *next;
211 FILE *fp;
212 Cond *conds;
213 Line *expansion;
214 char *fname;
215 int lineno, lineinc;
216 MMacro *mstk; /* stack of active macros/reps */
220 * Include search path. This is simply a list of strings which get
221 * prepended, in turn, to the name of an include file, in an
222 * attempt to find the file if it's not in the current directory.
224 struct IncPath {
225 IncPath *next;
226 char *path;
230 * Conditional assembly: we maintain a separate stack of these for
231 * each level of file inclusion. (The only reason we keep the
232 * stacks separate is to ensure that a stray `%endif' in a file
233 * included from within the true branch of a `%if' won't terminate
234 * it and cause confusion: instead, rightly, it'll cause an error.)
236 struct Cond {
237 Cond *next;
238 int state;
240 enum {
242 * These states are for use just after %if or %elif: IF_TRUE
243 * means the condition has evaluated to truth so we are
244 * currently emitting, whereas IF_FALSE means we are not
245 * currently emitting but will start doing so if a %else comes
246 * up. In these states, all directives are admissible: %elif,
247 * %else and %endif. (And of course %if.)
249 COND_IF_TRUE, COND_IF_FALSE,
251 * These states come up after a %else: ELSE_TRUE means we're
252 * emitting, and ELSE_FALSE means we're not. In ELSE_* states,
253 * any %elif or %else will cause an error.
255 COND_ELSE_TRUE, COND_ELSE_FALSE,
257 * These states mean that we're not emitting now, and also that
258 * nothing until %endif will be emitted at all. COND_DONE is
259 * used when we've had our moment of emission
260 * and have now started seeing %elifs. COND_NEVER is used when
261 * the condition construct in question is contained within a
262 * non-emitting branch of a larger condition construct,
263 * or if there is an error.
265 COND_DONE, COND_NEVER
267 #define emitting(x) ( (x) == COND_IF_TRUE || (x) == COND_ELSE_TRUE )
270 * These defines are used as the possible return values for do_directive
272 #define NO_DIRECTIVE_FOUND 0
273 #define DIRECTIVE_FOUND 1
276 * Condition codes. Note that we use c_ prefix not C_ because C_ is
277 * used in nasm.h for the "real" condition codes. At _this_ level,
278 * we treat CXZ and ECXZ as condition codes, albeit non-invertible
279 * ones, so we need a different enum...
281 static const char * const conditions[] = {
282 "a", "ae", "b", "be", "c", "cxz", "e", "ecxz", "g", "ge", "l", "le",
283 "na", "nae", "nb", "nbe", "nc", "ne", "ng", "nge", "nl", "nle", "no",
284 "np", "ns", "nz", "o", "p", "pe", "po", "rcxz", "s", "z"
286 enum pp_conds {
287 c_A, c_AE, c_B, c_BE, c_C, c_CXZ, c_E, c_ECXZ, c_G, c_GE, c_L, c_LE,
288 c_NA, c_NAE, c_NB, c_NBE, c_NC, c_NE, c_NG, c_NGE, c_NL, c_NLE, c_NO,
289 c_NP, c_NS, c_NZ, c_O, c_P, c_PE, c_PO, c_RCXZ, c_S, c_Z,
290 c_none = -1
292 static const enum pp_conds inverse_ccs[] = {
293 c_NA, c_NAE, c_NB, c_NBE, c_NC, -1, c_NE, -1, c_NG, c_NGE, c_NL, c_NLE,
294 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,
295 c_Z, c_NO, c_NP, c_PO, c_PE, -1, c_NS, c_NZ
299 * Directive names.
301 /* If this is a an IF, ELIF, ELSE or ENDIF keyword */
302 static int is_condition(enum preproc_token arg)
304 return PP_IS_COND(arg) || (arg == PP_ELSE) || (arg == PP_ENDIF);
307 /* For TASM compatibility we need to be able to recognise TASM compatible
308 * conditional compilation directives. Using the NASM pre-processor does
309 * not work, so we look for them specifically from the following list and
310 * then jam in the equivalent NASM directive into the input stream.
313 enum {
314 TM_ARG, TM_ELIF, TM_ELSE, TM_ENDIF, TM_IF, TM_IFDEF, TM_IFDIFI,
315 TM_IFNDEF, TM_INCLUDE, TM_LOCAL
318 static const char * const tasm_directives[] = {
319 "arg", "elif", "else", "endif", "if", "ifdef", "ifdifi",
320 "ifndef", "include", "local"
323 static int StackSize = 4;
324 static char *StackPointer = "ebp";
325 static int ArgOffset = 8;
326 static int LocalOffset = 0;
328 static Context *cstk;
329 static Include *istk;
330 static IncPath *ipath = NULL;
332 static efunc _error; /* Pointer to client-provided error reporting function */
333 static evalfunc evaluate;
335 static int pass; /* HACK: pass 0 = generate dependencies only */
336 static StrList **dephead, **deptail; /* Dependency list */
338 static uint64_t unique; /* unique identifier numbers */
340 static Line *predef = NULL;
341 static bool do_predef;
343 static ListGen *list;
346 * The current set of multi-line macros we have defined.
348 static struct hash_table mmacros;
351 * The current set of single-line macros we have defined.
353 static struct hash_table smacros;
356 * The multi-line macro we are currently defining, or the %rep
357 * block we are currently reading, if any.
359 static MMacro *defining;
361 static uint64_t nested_mac_count;
362 static uint64_t nested_rep_count;
365 * The number of macro parameters to allocate space for at a time.
367 #define PARAM_DELTA 16
370 * The standard macro set: defined in macros.c in the array nasm_stdmac.
371 * This gives our position in the macro set, when we're processing it.
373 static macros_t *stdmacpos;
376 * The extra standard macros that come from the object format, if
377 * any.
379 static macros_t *extrastdmac = NULL;
380 static bool any_extrastdmac;
383 * Tokens are allocated in blocks to improve speed
385 #define TOKEN_BLOCKSIZE 4096
386 static Token *freeTokens = NULL;
387 struct Blocks {
388 Blocks *next;
389 void *chunk;
392 static Blocks blocks = { NULL, NULL };
395 * Forward declarations.
397 static Token *expand_mmac_params(Token * tline);
398 static Token *expand_smacro(Token * tline);
399 static Token *expand_id(Token * tline);
400 static Context *get_ctx(const char *name, const char **namep,
401 bool all_contexts);
402 static void make_tok_num(Token * tok, int64_t val);
403 static void error(int severity, const char *fmt, ...);
404 static void error_precond(int severity, const char *fmt, ...);
405 static void *new_Block(size_t size);
406 static void delete_Blocks(void);
407 static Token *new_Token(Token * next, enum pp_token_type type,
408 const char *text, int txtlen);
409 static Token *delete_Token(Token * t);
412 * Macros for safe checking of token pointers, avoid *(NULL)
414 #define tok_type_(x,t) ((x) && (x)->type == (t))
415 #define skip_white_(x) if (tok_type_((x), TOK_WHITESPACE)) (x)=(x)->next
416 #define tok_is_(x,v) (tok_type_((x), TOK_OTHER) && !strcmp((x)->text,(v)))
417 #define tok_isnt_(x,v) ((x) && ((x)->type!=TOK_OTHER || strcmp((x)->text,(v))))
419 /* Handle TASM specific directives, which do not contain a % in
420 * front of them. We do it here because I could not find any other
421 * place to do it for the moment, and it is a hack (ideally it would
422 * be nice to be able to use the NASM pre-processor to do it).
424 static char *check_tasm_directive(char *line)
426 int32_t i, j, k, m, len;
427 char *p = line, *oldline, oldchar;
429 /* Skip whitespace */
430 while (nasm_isspace(*p) && *p != 0)
431 p++;
433 /* Binary search for the directive name */
434 i = -1;
435 j = elements(tasm_directives);
436 len = 0;
437 while (!nasm_isspace(p[len]) && p[len] != 0)
438 len++;
439 if (len) {
440 oldchar = p[len];
441 p[len] = 0;
442 while (j - i > 1) {
443 k = (j + i) / 2;
444 m = nasm_stricmp(p, tasm_directives[k]);
445 if (m == 0) {
446 /* We have found a directive, so jam a % in front of it
447 * so that NASM will then recognise it as one if it's own.
449 p[len] = oldchar;
450 len = strlen(p);
451 oldline = line;
452 line = nasm_malloc(len + 2);
453 line[0] = '%';
454 if (k == TM_IFDIFI) {
455 /* NASM does not recognise IFDIFI, so we convert it to
456 * %ifdef BOGUS. This is not used in NASM comaptible
457 * code, but does need to parse for the TASM macro
458 * package.
460 strcpy(line + 1, "ifdef BOGUS");
461 } else {
462 memcpy(line + 1, p, len + 1);
464 nasm_free(oldline);
465 return line;
466 } else if (m < 0) {
467 j = k;
468 } else
469 i = k;
471 p[len] = oldchar;
473 return line;
477 * The pre-preprocessing stage... This function translates line
478 * number indications as they emerge from GNU cpp (`# lineno "file"
479 * flags') into NASM preprocessor line number indications (`%line
480 * lineno file').
482 static char *prepreproc(char *line)
484 int lineno, fnlen;
485 char *fname, *oldline;
487 if (line[0] == '#' && line[1] == ' ') {
488 oldline = line;
489 fname = oldline + 2;
490 lineno = atoi(fname);
491 fname += strspn(fname, "0123456789 ");
492 if (*fname == '"')
493 fname++;
494 fnlen = strcspn(fname, "\"");
495 line = nasm_malloc(20 + fnlen);
496 snprintf(line, 20 + fnlen, "%%line %d %.*s", lineno, fnlen, fname);
497 nasm_free(oldline);
499 if (tasm_compatible_mode)
500 return check_tasm_directive(line);
501 return line;
505 * Free a linked list of tokens.
507 static void free_tlist(Token * list)
509 while (list) {
510 list = delete_Token(list);
515 * Free a linked list of lines.
517 static void free_llist(Line * list)
519 Line *l;
520 while (list) {
521 l = list;
522 list = list->next;
523 free_tlist(l->first);
524 nasm_free(l);
529 * Free an MMacro
531 static void free_mmacro(MMacro * m)
533 nasm_free(m->name);
534 free_tlist(m->dlist);
535 nasm_free(m->defaults);
536 free_llist(m->expansion);
537 nasm_free(m);
541 * Free all currently defined macros, and free the hash tables
543 static void free_smacro_table(struct hash_table *smt)
545 SMacro *s;
546 const char *key;
547 struct hash_tbl_node *it = NULL;
549 while ((s = hash_iterate(smt, &it, &key)) != NULL) {
550 nasm_free((void *)key);
551 while (s) {
552 SMacro *ns = s->next;
553 nasm_free(s->name);
554 free_tlist(s->expansion);
555 nasm_free(s);
556 s = ns;
559 hash_free(smt);
562 static void free_mmacro_table(struct hash_table *mmt)
564 MMacro *m;
565 const char *key;
566 struct hash_tbl_node *it = NULL;
568 it = NULL;
569 while ((m = hash_iterate(mmt, &it, &key)) != NULL) {
570 nasm_free((void *)key);
571 while (m) {
572 MMacro *nm = m->next;
573 free_mmacro(m);
574 m = nm;
577 hash_free(mmt);
580 static void free_macros(void)
582 free_smacro_table(&smacros);
583 free_mmacro_table(&mmacros);
587 * Initialize the hash tables
589 static void init_macros(void)
591 hash_init(&smacros, HASH_LARGE);
592 hash_init(&mmacros, HASH_LARGE);
596 * Pop the context stack.
598 static void ctx_pop(void)
600 Context *c = cstk;
602 cstk = cstk->next;
603 free_smacro_table(&c->localmac);
604 nasm_free(c->name);
605 nasm_free(c);
609 * Search for a key in the hash index; adding it if necessary
610 * (in which case we initialize the data pointer to NULL.)
612 static void **
613 hash_findi_add(struct hash_table *hash, const char *str)
615 struct hash_insert hi;
616 void **r;
617 char *strx;
619 r = hash_findi(hash, str, &hi);
620 if (r)
621 return r;
623 strx = nasm_strdup(str); /* Use a more efficient allocator here? */
624 return hash_add(&hi, strx, NULL);
628 * Like hash_findi, but returns the data element rather than a pointer
629 * to it. Used only when not adding a new element, hence no third
630 * argument.
632 static void *
633 hash_findix(struct hash_table *hash, const char *str)
635 void **p;
637 p = hash_findi(hash, str, NULL);
638 return p ? *p : NULL;
641 #define BUF_DELTA 512
643 * Read a line from the top file in istk, handling multiple CR/LFs
644 * at the end of the line read, and handling spurious ^Zs. Will
645 * return lines from the standard macro set if this has not already
646 * been done.
648 static char *read_line(void)
650 char *buffer, *p, *q;
651 int bufsize, continued_count;
653 if (stdmacpos) {
654 unsigned char c;
655 const unsigned char *p = stdmacpos;
656 char *ret, *q;
657 size_t len = 0;
658 while ((c = *p++)) {
659 if (c >= 0x80)
660 len += pp_directives_len[c-0x80]+1;
661 else
662 len++;
664 ret = nasm_malloc(len+1);
665 q = ret;
666 while ((c = *stdmacpos++)) {
667 if (c >= 0x80) {
668 memcpy(q, pp_directives[c-0x80], pp_directives_len[c-0x80]);
669 q += pp_directives_len[c-0x80];
670 *q++ = ' ';
671 } else {
672 *q++ = c;
675 stdmacpos = p;
676 *q = '\0';
678 if (!*stdmacpos) {
679 /* This was the last of the standard macro chain... */
680 stdmacpos = NULL;
681 if (any_extrastdmac) {
682 stdmacpos = extrastdmac;
683 any_extrastdmac = false;
684 } else if (do_predef) {
685 Line *pd, *l;
686 Token *head, **tail, *t;
689 * Nasty hack: here we push the contents of
690 * `predef' on to the top-level expansion stack,
691 * since this is the most convenient way to
692 * implement the pre-include and pre-define
693 * features.
695 for (pd = predef; pd; pd = pd->next) {
696 head = NULL;
697 tail = &head;
698 for (t = pd->first; t; t = t->next) {
699 *tail = new_Token(NULL, t->type, t->text, 0);
700 tail = &(*tail)->next;
702 l = nasm_malloc(sizeof(Line));
703 l->next = istk->expansion;
704 l->first = head;
705 l->finishes = NULL;
706 istk->expansion = l;
708 do_predef = false;
711 return ret;
714 bufsize = BUF_DELTA;
715 buffer = nasm_malloc(BUF_DELTA);
716 p = buffer;
717 continued_count = 0;
718 while (1) {
719 q = fgets(p, bufsize - (p - buffer), istk->fp);
720 if (!q)
721 break;
722 p += strlen(p);
723 if (p > buffer && p[-1] == '\n') {
724 /* Convert backslash-CRLF line continuation sequences into
725 nothing at all (for DOS and Windows) */
726 if (((p - 2) > buffer) && (p[-3] == '\\') && (p[-2] == '\r')) {
727 p -= 3;
728 *p = 0;
729 continued_count++;
731 /* Also convert backslash-LF line continuation sequences into
732 nothing at all (for Unix) */
733 else if (((p - 1) > buffer) && (p[-2] == '\\')) {
734 p -= 2;
735 *p = 0;
736 continued_count++;
737 } else {
738 break;
741 if (p - buffer > bufsize - 10) {
742 int32_t offset = p - buffer;
743 bufsize += BUF_DELTA;
744 buffer = nasm_realloc(buffer, bufsize);
745 p = buffer + offset; /* prevent stale-pointer problems */
749 if (!q && p == buffer) {
750 nasm_free(buffer);
751 return NULL;
754 src_set_linnum(src_get_linnum() + istk->lineinc +
755 (continued_count * istk->lineinc));
758 * Play safe: remove CRs as well as LFs, if any of either are
759 * present at the end of the line.
761 while (--p >= buffer && (*p == '\n' || *p == '\r'))
762 *p = '\0';
765 * Handle spurious ^Z, which may be inserted into source files
766 * by some file transfer utilities.
768 buffer[strcspn(buffer, "\032")] = '\0';
770 list->line(LIST_READ, buffer);
772 return buffer;
776 * Tokenize a line of text. This is a very simple process since we
777 * don't need to parse the value out of e.g. numeric tokens: we
778 * simply split one string into many.
780 static Token *tokenize(char *line)
782 char c, *p = line;
783 enum pp_token_type type;
784 Token *list = NULL;
785 Token *t, **tail = &list;
787 while (*line) {
788 p = line;
789 if (*p == '%') {
790 p++;
791 if (nasm_isdigit(*p) ||
792 ((*p == '-' || *p == '+') && nasm_isdigit(p[1])) ||
793 ((*p == '+') && (nasm_isspace(p[1]) || !p[1]))) {
794 do {
795 p++;
797 while (nasm_isdigit(*p));
798 type = TOK_PREPROC_ID;
799 } else if (*p == '{') {
800 p++;
801 while (*p && *p != '}') {
802 p[-1] = *p;
803 p++;
805 p[-1] = '\0';
806 if (*p)
807 p++;
808 type = TOK_PREPROC_ID;
809 } else if (*p == '[') {
810 int lvl = 1;
811 line += 2; /* Skip the leading %[ */
812 p++;
813 while (lvl && (c = *p++)) {
814 switch (c) {
815 case ']':
816 lvl--;
817 break;
818 case '%':
819 if (*p == '[')
820 lvl++;
821 break;
822 case '\'':
823 case '\"':
824 case '`':
825 p = nasm_skip_string(p)+1;
826 break;
827 default:
828 break;
831 p--;
832 if (*p)
833 *p++ = '\0';
834 if (lvl)
835 error(ERR_NONFATAL, "unterminated %[ construct");
836 type = TOK_INDIRECT;
837 } else if (*p == '?') {
838 type = TOK_PREPROC_Q; /* %? */
839 p++;
840 if (*p == '?') {
841 type = TOK_PREPROC_QQ; /* %?? */
842 p++;
844 } else if (isidchar(*p) ||
845 ((*p == '!' || *p == '%' || *p == '$') &&
846 isidchar(p[1]))) {
847 do {
848 p++;
850 while (isidchar(*p));
851 type = TOK_PREPROC_ID;
852 } else {
853 type = TOK_OTHER;
854 if (*p == '%')
855 p++;
857 } else if (isidstart(*p) || (*p == '$' && isidstart(p[1]))) {
858 type = TOK_ID;
859 p++;
860 while (*p && isidchar(*p))
861 p++;
862 } else if (*p == '\'' || *p == '"' || *p == '`') {
864 * A string token.
866 type = TOK_STRING;
867 p = nasm_skip_string(p);
869 if (*p) {
870 p++;
871 } else {
872 error(ERR_WARNING|ERR_PASS1, "unterminated string");
873 /* Handling unterminated strings by UNV */
874 /* type = -1; */
876 } else if (isnumstart(*p)) {
877 bool is_hex = false;
878 bool is_float = false;
879 bool has_e = false;
880 char c, *r;
883 * A numeric token.
886 if (*p == '$') {
887 p++;
888 is_hex = true;
891 for (;;) {
892 c = *p++;
894 if (!is_hex && (c == 'e' || c == 'E')) {
895 has_e = true;
896 if (*p == '+' || *p == '-') {
897 /* e can only be followed by +/- if it is either a
898 prefixed hex number or a floating-point number */
899 p++;
900 is_float = true;
902 } else if (c == 'H' || c == 'h' || c == 'X' || c == 'x') {
903 is_hex = true;
904 } else if (c == 'P' || c == 'p') {
905 is_float = true;
906 if (*p == '+' || *p == '-')
907 p++;
908 } else if (isnumchar(c) || c == '_')
909 ; /* just advance */
910 else if (c == '.') {
911 /* we need to deal with consequences of the legacy
912 parser, like "1.nolist" being two tokens
913 (TOK_NUMBER, TOK_ID) here; at least give it
914 a shot for now. In the future, we probably need
915 a flex-based scanner with proper pattern matching
916 to do it as well as it can be done. Nothing in
917 the world is going to help the person who wants
918 0x123.p16 interpreted as two tokens, though. */
919 r = p;
920 while (*r == '_')
921 r++;
923 if (nasm_isdigit(*r) || (is_hex && nasm_isxdigit(*r)) ||
924 (!is_hex && (*r == 'e' || *r == 'E')) ||
925 (*r == 'p' || *r == 'P')) {
926 p = r;
927 is_float = true;
928 } else
929 break; /* Terminate the token */
930 } else
931 break;
933 p--; /* Point to first character beyond number */
935 if (has_e && !is_hex) {
936 /* 1e13 is floating-point, but 1e13h is not */
937 is_float = true;
940 type = is_float ? TOK_FLOAT : TOK_NUMBER;
941 } else if (nasm_isspace(*p)) {
942 type = TOK_WHITESPACE;
943 p++;
944 while (*p && nasm_isspace(*p))
945 p++;
947 * Whitespace just before end-of-line is discarded by
948 * pretending it's a comment; whitespace just before a
949 * comment gets lumped into the comment.
951 if (!*p || *p == ';') {
952 type = TOK_COMMENT;
953 while (*p)
954 p++;
956 } else if (*p == ';') {
957 type = TOK_COMMENT;
958 while (*p)
959 p++;
960 } else {
962 * Anything else is an operator of some kind. We check
963 * for all the double-character operators (>>, <<, //,
964 * %%, <=, >=, ==, !=, <>, &&, ||, ^^), but anything
965 * else is a single-character operator.
967 type = TOK_OTHER;
968 if ((p[0] == '>' && p[1] == '>') ||
969 (p[0] == '<' && p[1] == '<') ||
970 (p[0] == '/' && p[1] == '/') ||
971 (p[0] == '<' && p[1] == '=') ||
972 (p[0] == '>' && p[1] == '=') ||
973 (p[0] == '=' && p[1] == '=') ||
974 (p[0] == '!' && p[1] == '=') ||
975 (p[0] == '<' && p[1] == '>') ||
976 (p[0] == '&' && p[1] == '&') ||
977 (p[0] == '|' && p[1] == '|') ||
978 (p[0] == '^' && p[1] == '^')) {
979 p++;
981 p++;
984 /* Handling unterminated string by UNV */
985 /*if (type == -1)
987 *tail = t = new_Token(NULL, TOK_STRING, line, p-line+1);
988 t->text[p-line] = *line;
989 tail = &t->next;
991 else */
992 if (type != TOK_COMMENT) {
993 *tail = t = new_Token(NULL, type, line, p - line);
994 tail = &t->next;
996 line = p;
998 return list;
1002 * this function allocates a new managed block of memory and
1003 * returns a pointer to the block. The managed blocks are
1004 * deleted only all at once by the delete_Blocks function.
1006 static void *new_Block(size_t size)
1008 Blocks *b = &blocks;
1010 /* first, get to the end of the linked list */
1011 while (b->next)
1012 b = b->next;
1013 /* now allocate the requested chunk */
1014 b->chunk = nasm_malloc(size);
1016 /* now allocate a new block for the next request */
1017 b->next = nasm_malloc(sizeof(Blocks));
1018 /* and initialize the contents of the new block */
1019 b->next->next = NULL;
1020 b->next->chunk = NULL;
1021 return b->chunk;
1025 * this function deletes all managed blocks of memory
1027 static void delete_Blocks(void)
1029 Blocks *a, *b = &blocks;
1032 * keep in mind that the first block, pointed to by blocks
1033 * is a static and not dynamically allocated, so we don't
1034 * free it.
1036 while (b) {
1037 if (b->chunk)
1038 nasm_free(b->chunk);
1039 a = b;
1040 b = b->next;
1041 if (a != &blocks)
1042 nasm_free(a);
1047 * this function creates a new Token and passes a pointer to it
1048 * back to the caller. It sets the type and text elements, and
1049 * also the a.mac and next elements to NULL.
1051 static Token *new_Token(Token * next, enum pp_token_type type,
1052 const char *text, int txtlen)
1054 Token *t;
1055 int i;
1057 if (freeTokens == NULL) {
1058 freeTokens = (Token *) new_Block(TOKEN_BLOCKSIZE * sizeof(Token));
1059 for (i = 0; i < TOKEN_BLOCKSIZE - 1; i++)
1060 freeTokens[i].next = &freeTokens[i + 1];
1061 freeTokens[i].next = NULL;
1063 t = freeTokens;
1064 freeTokens = t->next;
1065 t->next = next;
1066 t->a.mac = NULL;
1067 t->type = type;
1068 if (type == TOK_WHITESPACE || text == NULL) {
1069 t->text = NULL;
1070 } else {
1071 if (txtlen == 0)
1072 txtlen = strlen(text);
1073 t->text = nasm_malloc(txtlen+1);
1074 memcpy(t->text, text, txtlen);
1075 t->text[txtlen] = '\0';
1077 return t;
1080 static Token *delete_Token(Token * t)
1082 Token *next = t->next;
1083 nasm_free(t->text);
1084 t->next = freeTokens;
1085 freeTokens = t;
1086 return next;
1090 * Convert a line of tokens back into text.
1091 * If expand_locals is not zero, identifiers of the form "%$*xxx"
1092 * will be transformed into ..@ctxnum.xxx
1094 static char *detoken(Token * tlist, bool expand_locals)
1096 Token *t;
1097 int len;
1098 char *line, *p;
1099 const char *q;
1101 len = 0;
1102 for (t = tlist; t; t = t->next) {
1103 if (t->type == TOK_PREPROC_ID && t->text[1] == '!') {
1104 char *p = getenv(t->text + 2);
1105 nasm_free(t->text);
1106 if (p)
1107 t->text = nasm_strdup(p);
1108 else
1109 t->text = NULL;
1111 /* Expand local macros here and not during preprocessing */
1112 if (expand_locals &&
1113 t->type == TOK_PREPROC_ID && t->text &&
1114 t->text[0] == '%' && t->text[1] == '$') {
1115 const char *q;
1116 char *p;
1117 Context *ctx = get_ctx(t->text, &q, false);
1118 if (ctx) {
1119 char buffer[40];
1120 snprintf(buffer, sizeof(buffer), "..@%"PRIu32".", ctx->number);
1121 p = nasm_strcat(buffer, q);
1122 nasm_free(t->text);
1123 t->text = p;
1126 if (t->type == TOK_WHITESPACE) {
1127 len++;
1128 } else if (t->text) {
1129 len += strlen(t->text);
1132 p = line = nasm_malloc(len + 1);
1133 for (t = tlist; t; t = t->next) {
1134 if (t->type == TOK_WHITESPACE) {
1135 *p++ = ' ';
1136 } else if (t->text) {
1137 q = t->text;
1138 while (*q)
1139 *p++ = *q++;
1142 *p = '\0';
1143 return line;
1147 * A scanner, suitable for use by the expression evaluator, which
1148 * operates on a line of Tokens. Expects a pointer to a pointer to
1149 * the first token in the line to be passed in as its private_data
1150 * field.
1152 * FIX: This really needs to be unified with stdscan.
1154 static int ppscan(void *private_data, struct tokenval *tokval)
1156 Token **tlineptr = private_data;
1157 Token *tline;
1158 char ourcopy[MAX_KEYWORD+1], *p, *r, *s;
1160 do {
1161 tline = *tlineptr;
1162 *tlineptr = tline ? tline->next : NULL;
1164 while (tline && (tline->type == TOK_WHITESPACE ||
1165 tline->type == TOK_COMMENT));
1167 if (!tline)
1168 return tokval->t_type = TOKEN_EOS;
1170 tokval->t_charptr = tline->text;
1172 if (tline->text[0] == '$' && !tline->text[1])
1173 return tokval->t_type = TOKEN_HERE;
1174 if (tline->text[0] == '$' && tline->text[1] == '$' && !tline->text[2])
1175 return tokval->t_type = TOKEN_BASE;
1177 if (tline->type == TOK_ID) {
1178 p = tokval->t_charptr = tline->text;
1179 if (p[0] == '$') {
1180 tokval->t_charptr++;
1181 return tokval->t_type = TOKEN_ID;
1184 for (r = p, s = ourcopy; *r; r++) {
1185 if (r >= p+MAX_KEYWORD)
1186 return tokval->t_type = TOKEN_ID; /* Not a keyword */
1187 *s++ = nasm_tolower(*r);
1189 *s = '\0';
1190 /* right, so we have an identifier sitting in temp storage. now,
1191 * is it actually a register or instruction name, or what? */
1192 return nasm_token_hash(ourcopy, tokval);
1195 if (tline->type == TOK_NUMBER) {
1196 bool rn_error;
1197 tokval->t_integer = readnum(tline->text, &rn_error);
1198 tokval->t_charptr = tline->text;
1199 if (rn_error)
1200 return tokval->t_type = TOKEN_ERRNUM;
1201 else
1202 return tokval->t_type = TOKEN_NUM;
1205 if (tline->type == TOK_FLOAT) {
1206 return tokval->t_type = TOKEN_FLOAT;
1209 if (tline->type == TOK_STRING) {
1210 char bq, *ep;
1212 bq = tline->text[0];
1213 tokval->t_charptr = tline->text;
1214 tokval->t_inttwo = nasm_unquote(tline->text, &ep);
1216 if (ep[0] != bq || ep[1] != '\0')
1217 return tokval->t_type = TOKEN_ERRSTR;
1218 else
1219 return tokval->t_type = TOKEN_STR;
1222 if (tline->type == TOK_OTHER) {
1223 if (!strcmp(tline->text, "<<"))
1224 return tokval->t_type = TOKEN_SHL;
1225 if (!strcmp(tline->text, ">>"))
1226 return tokval->t_type = TOKEN_SHR;
1227 if (!strcmp(tline->text, "//"))
1228 return tokval->t_type = TOKEN_SDIV;
1229 if (!strcmp(tline->text, "%%"))
1230 return tokval->t_type = TOKEN_SMOD;
1231 if (!strcmp(tline->text, "=="))
1232 return tokval->t_type = TOKEN_EQ;
1233 if (!strcmp(tline->text, "<>"))
1234 return tokval->t_type = TOKEN_NE;
1235 if (!strcmp(tline->text, "!="))
1236 return tokval->t_type = TOKEN_NE;
1237 if (!strcmp(tline->text, "<="))
1238 return tokval->t_type = TOKEN_LE;
1239 if (!strcmp(tline->text, ">="))
1240 return tokval->t_type = TOKEN_GE;
1241 if (!strcmp(tline->text, "&&"))
1242 return tokval->t_type = TOKEN_DBL_AND;
1243 if (!strcmp(tline->text, "^^"))
1244 return tokval->t_type = TOKEN_DBL_XOR;
1245 if (!strcmp(tline->text, "||"))
1246 return tokval->t_type = TOKEN_DBL_OR;
1250 * We have no other options: just return the first character of
1251 * the token text.
1253 return tokval->t_type = tline->text[0];
1257 * Compare a string to the name of an existing macro; this is a
1258 * simple wrapper which calls either strcmp or nasm_stricmp
1259 * depending on the value of the `casesense' parameter.
1261 static int mstrcmp(const char *p, const char *q, bool casesense)
1263 return casesense ? strcmp(p, q) : nasm_stricmp(p, q);
1267 * Compare a string to the name of an existing macro; this is a
1268 * simple wrapper which calls either strcmp or nasm_stricmp
1269 * depending on the value of the `casesense' parameter.
1271 static int mmemcmp(const char *p, const char *q, size_t l, bool casesense)
1273 return casesense ? memcmp(p, q, l) : nasm_memicmp(p, q, l);
1277 * Return the Context structure associated with a %$ token. Return
1278 * NULL, having _already_ reported an error condition, if the
1279 * context stack isn't deep enough for the supplied number of $
1280 * signs.
1281 * If all_contexts == true, contexts that enclose current are
1282 * also scanned for such smacro, until it is found; if not -
1283 * only the context that directly results from the number of $'s
1284 * in variable's name.
1286 * If "namep" is non-NULL, set it to the pointer to the macro name
1287 * tail, i.e. the part beyond %$...
1289 static Context *get_ctx(const char *name, const char **namep,
1290 bool all_contexts)
1292 Context *ctx;
1293 SMacro *m;
1294 int i;
1296 if (namep)
1297 *namep = name;
1299 if (!name || name[0] != '%' || name[1] != '$')
1300 return NULL;
1302 if (!cstk) {
1303 error(ERR_NONFATAL, "`%s': context stack is empty", name);
1304 return NULL;
1307 name += 2;
1308 ctx = cstk;
1309 i = 0;
1310 while (ctx && *name == '$') {
1311 name++;
1312 i++;
1313 ctx = ctx->next;
1315 if (!ctx) {
1316 error(ERR_NONFATAL, "`%s': context stack is only"
1317 " %d level%s deep", name, i, (i == 1 ? "" : "s"));
1318 return NULL;
1321 if (namep)
1322 *namep = name;
1324 if (!all_contexts)
1325 return ctx;
1327 do {
1328 /* Search for this smacro in found context */
1329 m = hash_findix(&ctx->localmac, name);
1330 while (m) {
1331 if (!mstrcmp(m->name, name, m->casesense))
1332 return ctx;
1333 m = m->next;
1335 ctx = ctx->next;
1337 while (ctx);
1338 return NULL;
1342 * Check to see if a file is already in a string list
1344 static bool in_list(const StrList *list, const char *str)
1346 while (list) {
1347 if (!strcmp(list->str, str))
1348 return true;
1349 list = list->next;
1351 return false;
1355 * Open an include file. This routine must always return a valid
1356 * file pointer if it returns - it's responsible for throwing an
1357 * ERR_FATAL and bombing out completely if not. It should also try
1358 * the include path one by one until it finds the file or reaches
1359 * the end of the path.
1361 static FILE *inc_fopen(const char *file, StrList **dhead, StrList ***dtail,
1362 bool missing_ok)
1364 FILE *fp;
1365 char *prefix = "";
1366 IncPath *ip = ipath;
1367 int len = strlen(file);
1368 size_t prefix_len = 0;
1369 StrList *sl;
1371 while (1) {
1372 sl = nasm_malloc(prefix_len+len+1+sizeof sl->next);
1373 memcpy(sl->str, prefix, prefix_len);
1374 memcpy(sl->str+prefix_len, file, len+1);
1375 fp = fopen(sl->str, "r");
1376 if (fp && dhead && !in_list(*dhead, sl->str)) {
1377 sl->next = NULL;
1378 **dtail = sl;
1379 *dtail = &sl->next;
1380 } else {
1381 nasm_free(sl);
1383 if (fp)
1384 return fp;
1385 if (!ip) {
1386 if (!missing_ok)
1387 break;
1388 prefix = NULL;
1389 } else {
1390 prefix = ip->path;
1391 ip = ip->next;
1393 if (prefix) {
1394 prefix_len = strlen(prefix);
1395 } else {
1396 /* -MG given and file not found */
1397 if (dhead && !in_list(*dhead, file)) {
1398 sl = nasm_malloc(len+1+sizeof sl->next);
1399 sl->next = NULL;
1400 strcpy(sl->str, file);
1401 **dtail = sl;
1402 *dtail = &sl->next;
1404 return NULL;
1408 error(ERR_FATAL, "unable to open include file `%s'", file);
1409 return NULL; /* never reached - placate compilers */
1413 * Determine if we should warn on defining a single-line macro of
1414 * name `name', with `nparam' parameters. If nparam is 0 or -1, will
1415 * return true if _any_ single-line macro of that name is defined.
1416 * Otherwise, will return true if a single-line macro with either
1417 * `nparam' or no parameters is defined.
1419 * If a macro with precisely the right number of parameters is
1420 * defined, or nparam is -1, the address of the definition structure
1421 * will be returned in `defn'; otherwise NULL will be returned. If `defn'
1422 * is NULL, no action will be taken regarding its contents, and no
1423 * error will occur.
1425 * Note that this is also called with nparam zero to resolve
1426 * `ifdef'.
1428 * If you already know which context macro belongs to, you can pass
1429 * the context pointer as first parameter; if you won't but name begins
1430 * with %$ the context will be automatically computed. If all_contexts
1431 * is true, macro will be searched in outer contexts as well.
1433 static bool
1434 smacro_defined(Context * ctx, const char *name, int nparam, SMacro ** defn,
1435 bool nocase)
1437 struct hash_table *smtbl;
1438 SMacro *m;
1440 if (ctx) {
1441 smtbl = &ctx->localmac;
1442 } else if (name[0] == '%' && name[1] == '$') {
1443 if (cstk)
1444 ctx = get_ctx(name, &name, false);
1445 if (!ctx)
1446 return false; /* got to return _something_ */
1447 smtbl = &ctx->localmac;
1448 } else {
1449 smtbl = &smacros;
1451 m = (SMacro *) hash_findix(smtbl, name);
1453 while (m) {
1454 if (!mstrcmp(m->name, name, m->casesense && nocase) &&
1455 (nparam <= 0 || m->nparam == 0 || nparam == (int) m->nparam)) {
1456 if (defn) {
1457 if (nparam == (int) m->nparam || nparam == -1)
1458 *defn = m;
1459 else
1460 *defn = NULL;
1462 return true;
1464 m = m->next;
1467 return false;
1471 * Count and mark off the parameters in a multi-line macro call.
1472 * This is called both from within the multi-line macro expansion
1473 * code, and also to mark off the default parameters when provided
1474 * in a %macro definition line.
1476 static void count_mmac_params(Token * t, int *nparam, Token *** params)
1478 int paramsize, brace;
1480 *nparam = paramsize = 0;
1481 *params = NULL;
1482 while (t) {
1483 /* +1: we need space for the final NULL */
1484 if (*nparam+1 >= paramsize) {
1485 paramsize += PARAM_DELTA;
1486 *params = nasm_realloc(*params, sizeof(**params) * paramsize);
1488 skip_white_(t);
1489 brace = false;
1490 if (tok_is_(t, "{"))
1491 brace = true;
1492 (*params)[(*nparam)++] = t;
1493 while (tok_isnt_(t, brace ? "}" : ","))
1494 t = t->next;
1495 if (t) { /* got a comma/brace */
1496 t = t->next;
1497 if (brace) {
1499 * Now we've found the closing brace, look further
1500 * for the comma.
1502 skip_white_(t);
1503 if (tok_isnt_(t, ",")) {
1504 error(ERR_NONFATAL,
1505 "braces do not enclose all of macro parameter");
1506 while (tok_isnt_(t, ","))
1507 t = t->next;
1509 if (t)
1510 t = t->next; /* eat the comma */
1517 * Determine whether one of the various `if' conditions is true or
1518 * not.
1520 * We must free the tline we get passed.
1522 static bool if_condition(Token * tline, enum preproc_token ct)
1524 enum pp_conditional i = PP_COND(ct);
1525 bool j;
1526 Token *t, *tt, **tptr, *origline;
1527 struct tokenval tokval;
1528 expr *evalresult;
1529 enum pp_token_type needtype;
1531 origline = tline;
1533 switch (i) {
1534 case PPC_IFCTX:
1535 j = false; /* have we matched yet? */
1536 while (true) {
1537 skip_white_(tline);
1538 if (!tline)
1539 break;
1540 if (tline->type != TOK_ID) {
1541 error(ERR_NONFATAL,
1542 "`%s' expects context identifiers", pp_directives[ct]);
1543 free_tlist(origline);
1544 return -1;
1546 if (cstk && cstk->name && !nasm_stricmp(tline->text, cstk->name))
1547 j = true;
1548 tline = tline->next;
1550 break;
1552 case PPC_IFDEF:
1553 j = false; /* have we matched yet? */
1554 while (tline) {
1555 skip_white_(tline);
1556 if (!tline || (tline->type != TOK_ID &&
1557 (tline->type != TOK_PREPROC_ID ||
1558 tline->text[1] != '$'))) {
1559 error(ERR_NONFATAL,
1560 "`%s' expects macro identifiers", pp_directives[ct]);
1561 goto fail;
1563 if (smacro_defined(NULL, tline->text, 0, NULL, true))
1564 j = true;
1565 tline = tline->next;
1567 break;
1569 case PPC_IFIDN:
1570 case PPC_IFIDNI:
1571 tline = expand_smacro(tline);
1572 t = tt = tline;
1573 while (tok_isnt_(tt, ","))
1574 tt = tt->next;
1575 if (!tt) {
1576 error(ERR_NONFATAL,
1577 "`%s' expects two comma-separated arguments",
1578 pp_directives[ct]);
1579 goto fail;
1581 tt = tt->next;
1582 j = true; /* assume equality unless proved not */
1583 while ((t->type != TOK_OTHER || strcmp(t->text, ",")) && tt) {
1584 if (tt->type == TOK_OTHER && !strcmp(tt->text, ",")) {
1585 error(ERR_NONFATAL, "`%s': more than one comma on line",
1586 pp_directives[ct]);
1587 goto fail;
1589 if (t->type == TOK_WHITESPACE) {
1590 t = t->next;
1591 continue;
1593 if (tt->type == TOK_WHITESPACE) {
1594 tt = tt->next;
1595 continue;
1597 if (tt->type != t->type) {
1598 j = false; /* found mismatching tokens */
1599 break;
1601 /* When comparing strings, need to unquote them first */
1602 if (t->type == TOK_STRING) {
1603 size_t l1 = nasm_unquote(t->text, NULL);
1604 size_t l2 = nasm_unquote(tt->text, NULL);
1606 if (l1 != l2) {
1607 j = false;
1608 break;
1610 if (mmemcmp(t->text, tt->text, l1, i == PPC_IFIDN)) {
1611 j = false;
1612 break;
1614 } else if (mstrcmp(tt->text, t->text, i == PPC_IFIDN) != 0) {
1615 j = false; /* found mismatching tokens */
1616 break;
1619 t = t->next;
1620 tt = tt->next;
1622 if ((t->type != TOK_OTHER || strcmp(t->text, ",")) || tt)
1623 j = false; /* trailing gunk on one end or other */
1624 break;
1626 case PPC_IFMACRO:
1628 bool found = false;
1629 MMacro searching, *mmac;
1631 skip_white_(tline);
1632 tline = expand_id(tline);
1633 if (!tok_type_(tline, TOK_ID)) {
1634 error(ERR_NONFATAL,
1635 "`%s' expects a macro name", pp_directives[ct]);
1636 goto fail;
1638 searching.name = nasm_strdup(tline->text);
1639 searching.casesense = true;
1640 searching.plus = false;
1641 searching.nolist = false;
1642 searching.in_progress = 0;
1643 searching.rep_nest = NULL;
1644 searching.nparam_min = 0;
1645 searching.nparam_max = INT_MAX;
1646 tline = expand_smacro(tline->next);
1647 skip_white_(tline);
1648 if (!tline) {
1649 } else if (!tok_type_(tline, TOK_NUMBER)) {
1650 error(ERR_NONFATAL,
1651 "`%s' expects a parameter count or nothing",
1652 pp_directives[ct]);
1653 } else {
1654 searching.nparam_min = searching.nparam_max =
1655 readnum(tline->text, &j);
1656 if (j)
1657 error(ERR_NONFATAL,
1658 "unable to parse parameter count `%s'",
1659 tline->text);
1661 if (tline && tok_is_(tline->next, "-")) {
1662 tline = tline->next->next;
1663 if (tok_is_(tline, "*"))
1664 searching.nparam_max = INT_MAX;
1665 else if (!tok_type_(tline, TOK_NUMBER))
1666 error(ERR_NONFATAL,
1667 "`%s' expects a parameter count after `-'",
1668 pp_directives[ct]);
1669 else {
1670 searching.nparam_max = readnum(tline->text, &j);
1671 if (j)
1672 error(ERR_NONFATAL,
1673 "unable to parse parameter count `%s'",
1674 tline->text);
1675 if (searching.nparam_min > searching.nparam_max)
1676 error(ERR_NONFATAL,
1677 "minimum parameter count exceeds maximum");
1680 if (tline && tok_is_(tline->next, "+")) {
1681 tline = tline->next;
1682 searching.plus = true;
1684 mmac = (MMacro *) hash_findix(&mmacros, searching.name);
1685 while (mmac) {
1686 if (!strcmp(mmac->name, searching.name) &&
1687 (mmac->nparam_min <= searching.nparam_max
1688 || searching.plus)
1689 && (searching.nparam_min <= mmac->nparam_max
1690 || mmac->plus)) {
1691 found = true;
1692 break;
1694 mmac = mmac->next;
1696 if(tline && tline->next)
1697 error(ERR_WARNING|ERR_PASS1,
1698 "trailing garbage after %%ifmacro ignored");
1699 nasm_free(searching.name);
1700 j = found;
1701 break;
1704 case PPC_IFID:
1705 needtype = TOK_ID;
1706 goto iftype;
1707 case PPC_IFNUM:
1708 needtype = TOK_NUMBER;
1709 goto iftype;
1710 case PPC_IFSTR:
1711 needtype = TOK_STRING;
1712 goto iftype;
1714 iftype:
1715 t = tline = expand_smacro(tline);
1717 while (tok_type_(t, TOK_WHITESPACE) ||
1718 (needtype == TOK_NUMBER &&
1719 tok_type_(t, TOK_OTHER) &&
1720 (t->text[0] == '-' || t->text[0] == '+') &&
1721 !t->text[1]))
1722 t = t->next;
1724 j = tok_type_(t, needtype);
1725 break;
1727 case PPC_IFTOKEN:
1728 t = tline = expand_smacro(tline);
1729 while (tok_type_(t, TOK_WHITESPACE))
1730 t = t->next;
1732 j = false;
1733 if (t) {
1734 t = t->next; /* Skip the actual token */
1735 while (tok_type_(t, TOK_WHITESPACE))
1736 t = t->next;
1737 j = !t; /* Should be nothing left */
1739 break;
1741 case PPC_IFEMPTY:
1742 t = tline = expand_smacro(tline);
1743 while (tok_type_(t, TOK_WHITESPACE))
1744 t = t->next;
1746 j = !t; /* Should be empty */
1747 break;
1749 case PPC_IF:
1750 t = tline = expand_smacro(tline);
1751 tptr = &t;
1752 tokval.t_type = TOKEN_INVALID;
1753 evalresult = evaluate(ppscan, tptr, &tokval,
1754 NULL, pass | CRITICAL, error, NULL);
1755 if (!evalresult)
1756 return -1;
1757 if (tokval.t_type)
1758 error(ERR_WARNING|ERR_PASS1,
1759 "trailing garbage after expression ignored");
1760 if (!is_simple(evalresult)) {
1761 error(ERR_NONFATAL,
1762 "non-constant value given to `%s'", pp_directives[ct]);
1763 goto fail;
1765 j = reloc_value(evalresult) != 0;
1766 break;
1768 default:
1769 error(ERR_FATAL,
1770 "preprocessor directive `%s' not yet implemented",
1771 pp_directives[ct]);
1772 goto fail;
1775 free_tlist(origline);
1776 return j ^ PP_NEGATIVE(ct);
1778 fail:
1779 free_tlist(origline);
1780 return -1;
1784 * Common code for defining an smacro
1786 static bool define_smacro(Context *ctx, const char *mname, bool casesense,
1787 int nparam, Token *expansion)
1789 SMacro *smac, **smhead;
1790 struct hash_table *smtbl;
1792 if (smacro_defined(ctx, mname, nparam, &smac, casesense)) {
1793 if (!smac) {
1794 error(ERR_WARNING|ERR_PASS1,
1795 "single-line macro `%s' defined both with and"
1796 " without parameters", mname);
1798 /* Some instances of the old code considered this a failure,
1799 some others didn't. What is the right thing to do here? */
1800 free_tlist(expansion);
1801 return false; /* Failure */
1802 } else {
1804 * We're redefining, so we have to take over an
1805 * existing SMacro structure. This means freeing
1806 * what was already in it.
1808 nasm_free(smac->name);
1809 free_tlist(smac->expansion);
1811 } else {
1812 smtbl = ctx ? &ctx->localmac : &smacros;
1813 smhead = (SMacro **) hash_findi_add(smtbl, mname);
1814 smac = nasm_malloc(sizeof(SMacro));
1815 smac->next = *smhead;
1816 *smhead = smac;
1818 smac->name = nasm_strdup(mname);
1819 smac->casesense = casesense;
1820 smac->nparam = nparam;
1821 smac->expansion = expansion;
1822 smac->in_progress = false;
1823 return true; /* Success */
1827 * Undefine an smacro
1829 static void undef_smacro(Context *ctx, const char *mname)
1831 SMacro **smhead, *s, **sp;
1832 struct hash_table *smtbl;
1834 smtbl = ctx ? &ctx->localmac : &smacros;
1835 smhead = (SMacro **)hash_findi(smtbl, mname, NULL);
1837 if (smhead) {
1839 * We now have a macro name... go hunt for it.
1841 sp = smhead;
1842 while ((s = *sp) != NULL) {
1843 if (!mstrcmp(s->name, mname, s->casesense)) {
1844 *sp = s->next;
1845 nasm_free(s->name);
1846 free_tlist(s->expansion);
1847 nasm_free(s);
1848 } else {
1849 sp = &s->next;
1856 * Parse a mmacro specification.
1858 static bool parse_mmacro_spec(Token *tline, MMacro *def, const char *directive)
1860 bool err;
1862 tline = tline->next;
1863 skip_white_(tline);
1864 tline = expand_id(tline);
1865 if (!tok_type_(tline, TOK_ID)) {
1866 error(ERR_NONFATAL, "`%s' expects a macro name", directive);
1867 return false;
1870 def->name = nasm_strdup(tline->text);
1871 def->plus = false;
1872 def->nolist = false;
1873 def->in_progress = 0;
1874 def->rep_nest = NULL;
1875 def->nparam_min = 0;
1876 def->nparam_max = 0;
1878 tline = expand_smacro(tline->next);
1879 skip_white_(tline);
1880 if (!tok_type_(tline, TOK_NUMBER)) {
1881 error(ERR_NONFATAL, "`%s' expects a parameter count", directive);
1882 } else {
1883 def->nparam_min = def->nparam_max =
1884 readnum(tline->text, &err);
1885 if (err)
1886 error(ERR_NONFATAL,
1887 "unable to parse parameter count `%s'", tline->text);
1889 if (tline && tok_is_(tline->next, "-")) {
1890 tline = tline->next->next;
1891 if (tok_is_(tline, "*")) {
1892 def->nparam_max = INT_MAX;
1893 } else if (!tok_type_(tline, TOK_NUMBER)) {
1894 error(ERR_NONFATAL,
1895 "`%s' expects a parameter count after `-'", directive);
1896 } else {
1897 def->nparam_max = readnum(tline->text, &err);
1898 if (err) {
1899 error(ERR_NONFATAL, "unable to parse parameter count `%s'",
1900 tline->text);
1902 if (def->nparam_min > def->nparam_max) {
1903 error(ERR_NONFATAL, "minimum parameter count exceeds maximum");
1907 if (tline && tok_is_(tline->next, "+")) {
1908 tline = tline->next;
1909 def->plus = true;
1911 if (tline && tok_type_(tline->next, TOK_ID) &&
1912 !nasm_stricmp(tline->next->text, ".nolist")) {
1913 tline = tline->next;
1914 def->nolist = true;
1918 * Handle default parameters.
1920 if (tline && tline->next) {
1921 def->dlist = tline->next;
1922 tline->next = NULL;
1923 count_mmac_params(def->dlist, &def->ndefs, &def->defaults);
1924 } else {
1925 def->dlist = NULL;
1926 def->defaults = NULL;
1928 def->expansion = NULL;
1930 if(def->defaults &&
1931 def->ndefs > def->nparam_max - def->nparam_min &&
1932 !def->plus)
1933 error(ERR_WARNING|ERR_PASS1|ERR_WARN_MDP,
1934 "too many default macro parameters");
1936 return true;
1941 * Decode a size directive
1943 static int parse_size(const char *str) {
1944 static const char *size_names[] =
1945 { "byte", "dword", "oword", "qword", "tword", "word", "yword" };
1946 static const int sizes[] =
1947 { 0, 1, 4, 16, 8, 10, 2, 32 };
1949 return sizes[bsii(str, size_names, elements(size_names))+1];
1953 * find and process preprocessor directive in passed line
1954 * Find out if a line contains a preprocessor directive, and deal
1955 * with it if so.
1957 * If a directive _is_ found, it is the responsibility of this routine
1958 * (and not the caller) to free_tlist() the line.
1960 * @param tline a pointer to the current tokeninzed line linked list
1961 * @return DIRECTIVE_FOUND or NO_DIRECTIVE_FOUND
1964 static int do_directive(Token * tline)
1966 enum preproc_token i;
1967 int j;
1968 bool err;
1969 int nparam;
1970 bool nolist;
1971 bool casesense;
1972 int k, m;
1973 int offset;
1974 char *p, *pp;
1975 const char *mname;
1976 Include *inc;
1977 Context *ctx;
1978 Cond *cond;
1979 MMacro *mmac, **mmhead;
1980 Token *t, *tt, *param_start, *macro_start, *last, **tptr, *origline;
1981 Line *l;
1982 struct tokenval tokval;
1983 expr *evalresult;
1984 MMacro *tmp_defining; /* Used when manipulating rep_nest */
1985 int64_t count;
1986 size_t len;
1987 int severity;
1989 origline = tline;
1991 skip_white_(tline);
1992 if (!tline || !tok_type_(tline, TOK_PREPROC_ID) ||
1993 (tline->text[1] == '%' || tline->text[1] == '$'
1994 || tline->text[1] == '!'))
1995 return NO_DIRECTIVE_FOUND;
1997 i = pp_token_hash(tline->text);
2000 * If we're in a non-emitting branch of a condition construct,
2001 * or walking to the end of an already terminated %rep block,
2002 * we should ignore all directives except for condition
2003 * directives.
2005 if (((istk->conds && !emitting(istk->conds->state)) ||
2006 (istk->mstk && !istk->mstk->in_progress)) && !is_condition(i)) {
2007 return NO_DIRECTIVE_FOUND;
2011 * If we're defining a macro or reading a %rep block, we should
2012 * ignore all directives except for %macro/%imacro (which nest),
2013 * %endm/%endmacro, and (only if we're in a %rep block) %endrep.
2014 * If we're in a %rep block, another %rep nests, so should be let through.
2016 if (defining && i != PP_MACRO && i != PP_IMACRO &&
2017 i != PP_ENDMACRO && i != PP_ENDM &&
2018 (defining->name || (i != PP_ENDREP && i != PP_REP))) {
2019 return NO_DIRECTIVE_FOUND;
2022 if (defining) {
2023 if (i == PP_MACRO || i == PP_IMACRO) {
2024 nested_mac_count++;
2025 return NO_DIRECTIVE_FOUND;
2026 } else if (nested_mac_count > 0) {
2027 if (i == PP_ENDMACRO) {
2028 nested_mac_count--;
2029 return NO_DIRECTIVE_FOUND;
2032 if (!defining->name) {
2033 if (i == PP_REP) {
2034 nested_rep_count++;
2035 return NO_DIRECTIVE_FOUND;
2036 } else if (nested_rep_count > 0) {
2037 if (i == PP_ENDREP) {
2038 nested_rep_count--;
2039 return NO_DIRECTIVE_FOUND;
2045 switch (i) {
2046 case PP_INVALID:
2047 error(ERR_NONFATAL, "unknown preprocessor directive `%s'",
2048 tline->text);
2049 return NO_DIRECTIVE_FOUND; /* didn't get it */
2051 case PP_STACKSIZE:
2052 /* Directive to tell NASM what the default stack size is. The
2053 * default is for a 16-bit stack, and this can be overriden with
2054 * %stacksize large.
2055 * the following form:
2057 * ARG arg1:WORD, arg2:DWORD, arg4:QWORD
2059 tline = tline->next;
2060 if (tline && tline->type == TOK_WHITESPACE)
2061 tline = tline->next;
2062 if (!tline || tline->type != TOK_ID) {
2063 error(ERR_NONFATAL, "`%%stacksize' missing size parameter");
2064 free_tlist(origline);
2065 return DIRECTIVE_FOUND;
2067 if (nasm_stricmp(tline->text, "flat") == 0) {
2068 /* All subsequent ARG directives are for a 32-bit stack */
2069 StackSize = 4;
2070 StackPointer = "ebp";
2071 ArgOffset = 8;
2072 LocalOffset = 0;
2073 } else if (nasm_stricmp(tline->text, "flat64") == 0) {
2074 /* All subsequent ARG directives are for a 64-bit stack */
2075 StackSize = 8;
2076 StackPointer = "rbp";
2077 ArgOffset = 8;
2078 LocalOffset = 0;
2079 } else if (nasm_stricmp(tline->text, "large") == 0) {
2080 /* All subsequent ARG directives are for a 16-bit stack,
2081 * far function call.
2083 StackSize = 2;
2084 StackPointer = "bp";
2085 ArgOffset = 4;
2086 LocalOffset = 0;
2087 } else if (nasm_stricmp(tline->text, "small") == 0) {
2088 /* All subsequent ARG directives are for a 16-bit stack,
2089 * far function call. We don't support near functions.
2091 StackSize = 2;
2092 StackPointer = "bp";
2093 ArgOffset = 6;
2094 LocalOffset = 0;
2095 } else {
2096 error(ERR_NONFATAL, "`%%stacksize' invalid size type");
2097 free_tlist(origline);
2098 return DIRECTIVE_FOUND;
2100 free_tlist(origline);
2101 return DIRECTIVE_FOUND;
2103 case PP_ARG:
2104 /* TASM like ARG directive to define arguments to functions, in
2105 * the following form:
2107 * ARG arg1:WORD, arg2:DWORD, arg4:QWORD
2109 offset = ArgOffset;
2110 do {
2111 char *arg, directive[256];
2112 int size = StackSize;
2114 /* Find the argument name */
2115 tline = tline->next;
2116 if (tline && tline->type == TOK_WHITESPACE)
2117 tline = tline->next;
2118 if (!tline || tline->type != TOK_ID) {
2119 error(ERR_NONFATAL, "`%%arg' missing argument parameter");
2120 free_tlist(origline);
2121 return DIRECTIVE_FOUND;
2123 arg = tline->text;
2125 /* Find the argument size type */
2126 tline = tline->next;
2127 if (!tline || tline->type != TOK_OTHER
2128 || tline->text[0] != ':') {
2129 error(ERR_NONFATAL,
2130 "Syntax error processing `%%arg' directive");
2131 free_tlist(origline);
2132 return DIRECTIVE_FOUND;
2134 tline = tline->next;
2135 if (!tline || tline->type != TOK_ID) {
2136 error(ERR_NONFATAL, "`%%arg' missing size type parameter");
2137 free_tlist(origline);
2138 return DIRECTIVE_FOUND;
2141 /* Allow macro expansion of type parameter */
2142 tt = tokenize(tline->text);
2143 tt = expand_smacro(tt);
2144 size = parse_size(tt->text);
2145 if (!size) {
2146 error(ERR_NONFATAL,
2147 "Invalid size type for `%%arg' missing directive");
2148 free_tlist(tt);
2149 free_tlist(origline);
2150 return DIRECTIVE_FOUND;
2152 free_tlist(tt);
2154 /* Round up to even stack slots */
2155 size = (size+StackSize-1) & ~(StackSize-1);
2157 /* Now define the macro for the argument */
2158 snprintf(directive, sizeof(directive), "%%define %s (%s+%d)",
2159 arg, StackPointer, offset);
2160 do_directive(tokenize(directive));
2161 offset += size;
2163 /* Move to the next argument in the list */
2164 tline = tline->next;
2165 if (tline && tline->type == TOK_WHITESPACE)
2166 tline = tline->next;
2167 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2168 ArgOffset = offset;
2169 free_tlist(origline);
2170 return DIRECTIVE_FOUND;
2172 case PP_LOCAL:
2173 /* TASM like LOCAL directive to define local variables for a
2174 * function, in the following form:
2176 * LOCAL local1:WORD, local2:DWORD, local4:QWORD = LocalSize
2178 * The '= LocalSize' at the end is ignored by NASM, but is
2179 * required by TASM to define the local parameter size (and used
2180 * by the TASM macro package).
2182 offset = LocalOffset;
2183 do {
2184 char *local, directive[256];
2185 int size = StackSize;
2187 /* Find the argument name */
2188 tline = tline->next;
2189 if (tline && tline->type == TOK_WHITESPACE)
2190 tline = tline->next;
2191 if (!tline || tline->type != TOK_ID) {
2192 error(ERR_NONFATAL,
2193 "`%%local' missing argument parameter");
2194 free_tlist(origline);
2195 return DIRECTIVE_FOUND;
2197 local = tline->text;
2199 /* Find the argument size type */
2200 tline = tline->next;
2201 if (!tline || tline->type != TOK_OTHER
2202 || tline->text[0] != ':') {
2203 error(ERR_NONFATAL,
2204 "Syntax error processing `%%local' directive");
2205 free_tlist(origline);
2206 return DIRECTIVE_FOUND;
2208 tline = tline->next;
2209 if (!tline || tline->type != TOK_ID) {
2210 error(ERR_NONFATAL,
2211 "`%%local' missing size type parameter");
2212 free_tlist(origline);
2213 return DIRECTIVE_FOUND;
2216 /* Allow macro expansion of type parameter */
2217 tt = tokenize(tline->text);
2218 tt = expand_smacro(tt);
2219 size = parse_size(tt->text);
2220 if (!size) {
2221 error(ERR_NONFATAL,
2222 "Invalid size type for `%%local' missing directive");
2223 free_tlist(tt);
2224 free_tlist(origline);
2225 return DIRECTIVE_FOUND;
2227 free_tlist(tt);
2229 /* Round up to even stack slots */
2230 size = (size+StackSize-1) & ~(StackSize-1);
2232 offset += size; /* Negative offset, increment before */
2234 /* Now define the macro for the argument */
2235 snprintf(directive, sizeof(directive), "%%define %s (%s-%d)",
2236 local, StackPointer, offset);
2237 do_directive(tokenize(directive));
2239 /* Now define the assign to setup the enter_c macro correctly */
2240 snprintf(directive, sizeof(directive),
2241 "%%assign %%$localsize %%$localsize+%d", size);
2242 do_directive(tokenize(directive));
2244 /* Move to the next argument in the list */
2245 tline = tline->next;
2246 if (tline && tline->type == TOK_WHITESPACE)
2247 tline = tline->next;
2248 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2249 LocalOffset = offset;
2250 free_tlist(origline);
2251 return DIRECTIVE_FOUND;
2253 case PP_CLEAR:
2254 if (tline->next)
2255 error(ERR_WARNING|ERR_PASS1,
2256 "trailing garbage after `%%clear' ignored");
2257 free_macros();
2258 init_macros();
2259 free_tlist(origline);
2260 return DIRECTIVE_FOUND;
2262 case PP_DEPEND:
2263 t = tline->next = expand_smacro(tline->next);
2264 skip_white_(t);
2265 if (!t || (t->type != TOK_STRING &&
2266 t->type != TOK_INTERNAL_STRING)) {
2267 error(ERR_NONFATAL, "`%%depend' expects a file name");
2268 free_tlist(origline);
2269 return DIRECTIVE_FOUND; /* but we did _something_ */
2271 if (t->next)
2272 error(ERR_WARNING|ERR_PASS1,
2273 "trailing garbage after `%%depend' ignored");
2274 p = t->text;
2275 if (t->type != TOK_INTERNAL_STRING)
2276 nasm_unquote(p, NULL);
2277 if (dephead && !in_list(*dephead, p)) {
2278 StrList *sl = nasm_malloc(strlen(p)+1+sizeof sl->next);
2279 sl->next = NULL;
2280 strcpy(sl->str, p);
2281 *deptail = sl;
2282 deptail = &sl->next;
2284 free_tlist(origline);
2285 return DIRECTIVE_FOUND;
2287 case PP_INCLUDE:
2288 t = tline->next = expand_smacro(tline->next);
2289 skip_white_(t);
2291 if (!t || (t->type != TOK_STRING &&
2292 t->type != TOK_INTERNAL_STRING)) {
2293 error(ERR_NONFATAL, "`%%include' expects a file name");
2294 free_tlist(origline);
2295 return DIRECTIVE_FOUND; /* but we did _something_ */
2297 if (t->next)
2298 error(ERR_WARNING|ERR_PASS1,
2299 "trailing garbage after `%%include' ignored");
2300 p = t->text;
2301 if (t->type != TOK_INTERNAL_STRING)
2302 nasm_unquote(p, NULL);
2303 inc = nasm_malloc(sizeof(Include));
2304 inc->next = istk;
2305 inc->conds = NULL;
2306 inc->fp = inc_fopen(p, dephead, &deptail, pass == 0);
2307 if (!inc->fp) {
2308 /* -MG given but file not found */
2309 nasm_free(inc);
2310 } else {
2311 inc->fname = src_set_fname(nasm_strdup(p));
2312 inc->lineno = src_set_linnum(0);
2313 inc->lineinc = 1;
2314 inc->expansion = NULL;
2315 inc->mstk = NULL;
2316 istk = inc;
2317 list->uplevel(LIST_INCLUDE);
2319 free_tlist(origline);
2320 return DIRECTIVE_FOUND;
2322 case PP_USE:
2324 static macros_t *use_pkg;
2325 const char *pkg_macro;
2327 tline = tline->next;
2328 skip_white_(tline);
2329 tline = expand_id(tline);
2331 if (!tline || (tline->type != TOK_STRING &&
2332 tline->type != TOK_INTERNAL_STRING &&
2333 tline->type != TOK_ID)) {
2334 error(ERR_NONFATAL, "`%%use' expects a package name");
2335 free_tlist(origline);
2336 return DIRECTIVE_FOUND; /* but we did _something_ */
2338 if (tline->next)
2339 error(ERR_WARNING|ERR_PASS1,
2340 "trailing garbage after `%%use' ignored");
2341 if (tline->type == TOK_STRING)
2342 nasm_unquote(tline->text, NULL);
2343 use_pkg = nasm_stdmac_find_package(tline->text);
2344 if (!use_pkg)
2345 error(ERR_NONFATAL, "unknown `%%use' package: %s", tline->text);
2346 /* The first string will be <%define>__USE_*__ */
2347 pkg_macro = (char *)use_pkg + 1;
2348 if (!smacro_defined(NULL, pkg_macro, 0, NULL, true)) {
2349 /* Not already included, go ahead and include it */
2350 stdmacpos = use_pkg;
2352 free_tlist(origline);
2353 return DIRECTIVE_FOUND;
2355 case PP_PUSH:
2356 case PP_REPL:
2357 case PP_POP:
2358 tline = tline->next;
2359 skip_white_(tline);
2360 tline = expand_id(tline);
2361 if (tline) {
2362 if (!tok_type_(tline, TOK_ID)) {
2363 error(ERR_NONFATAL, "`%s' expects a context identifier",
2364 pp_directives[i]);
2365 free_tlist(origline);
2366 return DIRECTIVE_FOUND; /* but we did _something_ */
2368 if (tline->next)
2369 error(ERR_WARNING|ERR_PASS1,
2370 "trailing garbage after `%s' ignored",
2371 pp_directives[i]);
2372 p = nasm_strdup(tline->text);
2373 } else {
2374 p = NULL; /* Anonymous */
2377 if (i == PP_PUSH) {
2378 ctx = nasm_malloc(sizeof(Context));
2379 ctx->next = cstk;
2380 hash_init(&ctx->localmac, HASH_SMALL);
2381 ctx->name = p;
2382 ctx->number = unique++;
2383 cstk = ctx;
2384 } else {
2385 /* %pop or %repl */
2386 if (!cstk) {
2387 error(ERR_NONFATAL, "`%s': context stack is empty",
2388 pp_directives[i]);
2389 } else if (i == PP_POP) {
2390 if (p && (!cstk->name || nasm_stricmp(p, cstk->name)))
2391 error(ERR_NONFATAL, "`%%pop' in wrong context: %s, "
2392 "expected %s",
2393 cstk->name ? cstk->name : "anonymous", p);
2394 else
2395 ctx_pop();
2396 } else {
2397 /* i == PP_REPL */
2398 nasm_free(cstk->name);
2399 cstk->name = p;
2400 p = NULL;
2402 nasm_free(p);
2404 free_tlist(origline);
2405 return DIRECTIVE_FOUND;
2406 case PP_FATAL:
2407 severity = ERR_FATAL|ERR_NO_SEVERITY;
2408 goto issue_error;
2409 case PP_ERROR:
2410 severity = ERR_NONFATAL|ERR_NO_SEVERITY;
2411 goto issue_error;
2412 case PP_WARNING:
2413 severity = ERR_WARNING|ERR_NO_SEVERITY|ERR_WARN_USER;
2414 goto issue_error;
2416 issue_error:
2418 /* Only error out if this is the final pass */
2419 if (pass != 2 && i != PP_FATAL)
2420 return DIRECTIVE_FOUND;
2422 tline->next = expand_smacro(tline->next);
2423 tline = tline->next;
2424 skip_white_(tline);
2425 t = tline ? tline->next : NULL;
2426 skip_white_(t);
2427 if (tok_type_(tline, TOK_STRING) && !t) {
2428 /* The line contains only a quoted string */
2429 p = tline->text;
2430 nasm_unquote(p, NULL);
2431 error(severity, "%s: %s", pp_directives[i], p);
2432 } else {
2433 /* Not a quoted string, or more than a quoted string */
2434 p = detoken(tline, false);
2435 error(severity, "%s: %s", pp_directives[i], p);
2436 nasm_free(p);
2438 free_tlist(origline);
2439 return DIRECTIVE_FOUND;
2442 CASE_PP_IF:
2443 if (istk->conds && !emitting(istk->conds->state))
2444 j = COND_NEVER;
2445 else {
2446 j = if_condition(tline->next, i);
2447 tline->next = NULL; /* it got freed */
2448 j = j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2450 cond = nasm_malloc(sizeof(Cond));
2451 cond->next = istk->conds;
2452 cond->state = j;
2453 istk->conds = cond;
2454 free_tlist(origline);
2455 return DIRECTIVE_FOUND;
2457 CASE_PP_ELIF:
2458 if (!istk->conds)
2459 error(ERR_FATAL, "`%s': no matching `%%if'", pp_directives[i]);
2460 switch(istk->conds->state) {
2461 case COND_IF_TRUE:
2462 istk->conds->state = COND_DONE;
2463 break;
2465 case COND_DONE:
2466 case COND_NEVER:
2467 break;
2469 case COND_ELSE_TRUE:
2470 case COND_ELSE_FALSE:
2471 error_precond(ERR_WARNING|ERR_PASS1,
2472 "`%%elif' after `%%else' ignored");
2473 istk->conds->state = COND_NEVER;
2474 break;
2476 case COND_IF_FALSE:
2478 * IMPORTANT: In the case of %if, we will already have
2479 * called expand_mmac_params(); however, if we're
2480 * processing an %elif we must have been in a
2481 * non-emitting mode, which would have inhibited
2482 * the normal invocation of expand_mmac_params().
2483 * Therefore, we have to do it explicitly here.
2485 j = if_condition(expand_mmac_params(tline->next), i);
2486 tline->next = NULL; /* it got freed */
2487 istk->conds->state =
2488 j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2489 break;
2491 free_tlist(origline);
2492 return DIRECTIVE_FOUND;
2494 case PP_ELSE:
2495 if (tline->next)
2496 error_precond(ERR_WARNING|ERR_PASS1,
2497 "trailing garbage after `%%else' ignored");
2498 if (!istk->conds)
2499 error(ERR_FATAL, "`%%else': no matching `%%if'");
2500 switch(istk->conds->state) {
2501 case COND_IF_TRUE:
2502 case COND_DONE:
2503 istk->conds->state = COND_ELSE_FALSE;
2504 break;
2506 case COND_NEVER:
2507 break;
2509 case COND_IF_FALSE:
2510 istk->conds->state = COND_ELSE_TRUE;
2511 break;
2513 case COND_ELSE_TRUE:
2514 case COND_ELSE_FALSE:
2515 error_precond(ERR_WARNING|ERR_PASS1,
2516 "`%%else' after `%%else' ignored.");
2517 istk->conds->state = COND_NEVER;
2518 break;
2520 free_tlist(origline);
2521 return DIRECTIVE_FOUND;
2523 case PP_ENDIF:
2524 if (tline->next)
2525 error_precond(ERR_WARNING|ERR_PASS1,
2526 "trailing garbage after `%%endif' ignored");
2527 if (!istk->conds)
2528 error(ERR_FATAL, "`%%endif': no matching `%%if'");
2529 cond = istk->conds;
2530 istk->conds = cond->next;
2531 nasm_free(cond);
2532 free_tlist(origline);
2533 return DIRECTIVE_FOUND;
2535 case PP_MACRO:
2536 case PP_IMACRO:
2537 if (defining) {
2538 error(ERR_FATAL,
2539 "`%%%smacro': already defining a macro",
2540 (i == PP_IMACRO ? "i" : ""));
2541 return DIRECTIVE_FOUND;
2543 defining = nasm_malloc(sizeof(MMacro));
2544 defining->casesense = (i == PP_MACRO);
2545 if (!parse_mmacro_spec(tline, defining, pp_directives[i])) {
2546 nasm_free(defining);
2547 defining = NULL;
2548 return DIRECTIVE_FOUND;
2551 mmac = (MMacro *) hash_findix(&mmacros, defining->name);
2552 while (mmac) {
2553 if (!strcmp(mmac->name, defining->name) &&
2554 (mmac->nparam_min <= defining->nparam_max
2555 || defining->plus)
2556 && (defining->nparam_min <= mmac->nparam_max
2557 || mmac->plus)) {
2558 error(ERR_WARNING|ERR_PASS1,
2559 "redefining multi-line macro `%s'", defining->name);
2560 return DIRECTIVE_FOUND;
2562 mmac = mmac->next;
2564 free_tlist(origline);
2565 return DIRECTIVE_FOUND;
2567 case PP_ENDM:
2568 case PP_ENDMACRO:
2569 if (! (defining && defining->name)) {
2570 error(ERR_NONFATAL, "`%s': not defining a macro", tline->text);
2571 return DIRECTIVE_FOUND;
2573 mmhead = (MMacro **) hash_findi_add(&mmacros, defining->name);
2574 defining->next = *mmhead;
2575 *mmhead = defining;
2576 defining = NULL;
2577 free_tlist(origline);
2578 return DIRECTIVE_FOUND;
2580 case PP_UNMACRO:
2581 case PP_UNIMACRO:
2583 MMacro **mmac_p;
2584 MMacro spec;
2586 spec.casesense = (i == PP_UNMACRO);
2587 if (!parse_mmacro_spec(tline, &spec, pp_directives[i])) {
2588 return DIRECTIVE_FOUND;
2590 mmac_p = (MMacro **) hash_findi(&mmacros, spec.name, NULL);
2591 while (mmac_p && *mmac_p) {
2592 mmac = *mmac_p;
2593 if (mmac->casesense == spec.casesense &&
2594 !mstrcmp(mmac->name, spec.name, spec.casesense) &&
2595 mmac->nparam_min == spec.nparam_min &&
2596 mmac->nparam_max == spec.nparam_max &&
2597 mmac->plus == spec.plus) {
2598 *mmac_p = mmac->next;
2599 free_mmacro(mmac);
2600 } else {
2601 mmac_p = &mmac->next;
2604 free_tlist(origline);
2605 free_tlist(spec.dlist);
2606 return DIRECTIVE_FOUND;
2609 case PP_ROTATE:
2610 if (tline->next && tline->next->type == TOK_WHITESPACE)
2611 tline = tline->next;
2612 if (tline->next == NULL) {
2613 free_tlist(origline);
2614 error(ERR_NONFATAL, "`%%rotate' missing rotate count");
2615 return DIRECTIVE_FOUND;
2617 t = expand_smacro(tline->next);
2618 tline->next = NULL;
2619 free_tlist(origline);
2620 tline = t;
2621 tptr = &t;
2622 tokval.t_type = TOKEN_INVALID;
2623 evalresult =
2624 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
2625 free_tlist(tline);
2626 if (!evalresult)
2627 return DIRECTIVE_FOUND;
2628 if (tokval.t_type)
2629 error(ERR_WARNING|ERR_PASS1,
2630 "trailing garbage after expression ignored");
2631 if (!is_simple(evalresult)) {
2632 error(ERR_NONFATAL, "non-constant value given to `%%rotate'");
2633 return DIRECTIVE_FOUND;
2635 mmac = istk->mstk;
2636 while (mmac && !mmac->name) /* avoid mistaking %reps for macros */
2637 mmac = mmac->next_active;
2638 if (!mmac) {
2639 error(ERR_NONFATAL, "`%%rotate' invoked outside a macro call");
2640 } else if (mmac->nparam == 0) {
2641 error(ERR_NONFATAL,
2642 "`%%rotate' invoked within macro without parameters");
2643 } else {
2644 int rotate = mmac->rotate + reloc_value(evalresult);
2646 rotate %= (int)mmac->nparam;
2647 if (rotate < 0)
2648 rotate += mmac->nparam;
2650 mmac->rotate = rotate;
2652 return DIRECTIVE_FOUND;
2654 case PP_REP:
2655 nolist = false;
2656 do {
2657 tline = tline->next;
2658 } while (tok_type_(tline, TOK_WHITESPACE));
2660 if (tok_type_(tline, TOK_ID) &&
2661 nasm_stricmp(tline->text, ".nolist") == 0) {
2662 nolist = true;
2663 do {
2664 tline = tline->next;
2665 } while (tok_type_(tline, TOK_WHITESPACE));
2668 if (tline) {
2669 t = expand_smacro(tline);
2670 tptr = &t;
2671 tokval.t_type = TOKEN_INVALID;
2672 evalresult =
2673 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
2674 if (!evalresult) {
2675 free_tlist(origline);
2676 return DIRECTIVE_FOUND;
2678 if (tokval.t_type)
2679 error(ERR_WARNING|ERR_PASS1,
2680 "trailing garbage after expression ignored");
2681 if (!is_simple(evalresult)) {
2682 error(ERR_NONFATAL, "non-constant value given to `%%rep'");
2683 return DIRECTIVE_FOUND;
2685 count = reloc_value(evalresult) + 1;
2686 } else {
2687 error(ERR_NONFATAL, "`%%rep' expects a repeat count");
2688 count = 0;
2690 free_tlist(origline);
2692 tmp_defining = defining;
2693 defining = nasm_malloc(sizeof(MMacro));
2694 defining->name = NULL; /* flags this macro as a %rep block */
2695 defining->casesense = false;
2696 defining->plus = false;
2697 defining->nolist = nolist;
2698 defining->in_progress = count;
2699 defining->nparam_min = defining->nparam_max = 0;
2700 defining->defaults = NULL;
2701 defining->dlist = NULL;
2702 defining->expansion = NULL;
2703 defining->next_active = istk->mstk;
2704 defining->rep_nest = tmp_defining;
2705 return DIRECTIVE_FOUND;
2707 case PP_ENDREP:
2708 if (!defining || defining->name) {
2709 error(ERR_NONFATAL, "`%%endrep': no matching `%%rep'");
2710 return DIRECTIVE_FOUND;
2714 * Now we have a "macro" defined - although it has no name
2715 * and we won't be entering it in the hash tables - we must
2716 * push a macro-end marker for it on to istk->expansion.
2717 * After that, it will take care of propagating itself (a
2718 * macro-end marker line for a macro which is really a %rep
2719 * block will cause the macro to be re-expanded, complete
2720 * with another macro-end marker to ensure the process
2721 * continues) until the whole expansion is forcibly removed
2722 * from istk->expansion by a %exitrep.
2724 l = nasm_malloc(sizeof(Line));
2725 l->next = istk->expansion;
2726 l->finishes = defining;
2727 l->first = NULL;
2728 istk->expansion = l;
2730 istk->mstk = defining;
2732 list->uplevel(defining->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
2733 tmp_defining = defining;
2734 defining = defining->rep_nest;
2735 free_tlist(origline);
2736 return DIRECTIVE_FOUND;
2738 case PP_EXITREP:
2740 * We must search along istk->expansion until we hit a
2741 * macro-end marker for a macro with no name. Then we set
2742 * its `in_progress' flag to 0.
2744 for (l = istk->expansion; l; l = l->next)
2745 if (l->finishes && !l->finishes->name)
2746 break;
2748 if (l)
2749 l->finishes->in_progress = 1;
2750 else
2751 error(ERR_NONFATAL, "`%%exitrep' not within `%%rep' block");
2752 free_tlist(origline);
2753 return DIRECTIVE_FOUND;
2755 case PP_XDEFINE:
2756 case PP_IXDEFINE:
2757 case PP_DEFINE:
2758 case PP_IDEFINE:
2759 casesense = (i == PP_DEFINE || i == PP_XDEFINE);
2761 tline = tline->next;
2762 skip_white_(tline);
2763 tline = expand_id(tline);
2764 if (!tline || (tline->type != TOK_ID &&
2765 (tline->type != TOK_PREPROC_ID ||
2766 tline->text[1] != '$'))) {
2767 error(ERR_NONFATAL, "`%s' expects a macro identifier",
2768 pp_directives[i]);
2769 free_tlist(origline);
2770 return DIRECTIVE_FOUND;
2773 ctx = get_ctx(tline->text, &mname, false);
2774 last = tline;
2775 param_start = tline = tline->next;
2776 nparam = 0;
2778 /* Expand the macro definition now for %xdefine and %ixdefine */
2779 if ((i == PP_XDEFINE) || (i == PP_IXDEFINE))
2780 tline = expand_smacro(tline);
2782 if (tok_is_(tline, "(")) {
2784 * This macro has parameters.
2787 tline = tline->next;
2788 while (1) {
2789 skip_white_(tline);
2790 if (!tline) {
2791 error(ERR_NONFATAL, "parameter identifier expected");
2792 free_tlist(origline);
2793 return DIRECTIVE_FOUND;
2795 if (tline->type != TOK_ID) {
2796 error(ERR_NONFATAL,
2797 "`%s': parameter identifier expected",
2798 tline->text);
2799 free_tlist(origline);
2800 return DIRECTIVE_FOUND;
2802 tline->type = TOK_SMAC_PARAM + nparam++;
2803 tline = tline->next;
2804 skip_white_(tline);
2805 if (tok_is_(tline, ",")) {
2806 tline = tline->next;
2807 } else {
2808 if (!tok_is_(tline, ")")) {
2809 error(ERR_NONFATAL,
2810 "`)' expected to terminate macro template");
2811 free_tlist(origline);
2812 return DIRECTIVE_FOUND;
2814 break;
2817 last = tline;
2818 tline = tline->next;
2820 if (tok_type_(tline, TOK_WHITESPACE))
2821 last = tline, tline = tline->next;
2822 macro_start = NULL;
2823 last->next = NULL;
2824 t = tline;
2825 while (t) {
2826 if (t->type == TOK_ID) {
2827 for (tt = param_start; tt; tt = tt->next)
2828 if (tt->type >= TOK_SMAC_PARAM &&
2829 !strcmp(tt->text, t->text))
2830 t->type = tt->type;
2832 tt = t->next;
2833 t->next = macro_start;
2834 macro_start = t;
2835 t = tt;
2838 * Good. We now have a macro name, a parameter count, and a
2839 * token list (in reverse order) for an expansion. We ought
2840 * to be OK just to create an SMacro, store it, and let
2841 * free_tlist have the rest of the line (which we have
2842 * carefully re-terminated after chopping off the expansion
2843 * from the end).
2845 define_smacro(ctx, mname, casesense, nparam, macro_start);
2846 free_tlist(origline);
2847 return DIRECTIVE_FOUND;
2849 case PP_UNDEF:
2850 tline = tline->next;
2851 skip_white_(tline);
2852 tline = expand_id(tline);
2853 if (!tline || (tline->type != TOK_ID &&
2854 (tline->type != TOK_PREPROC_ID ||
2855 tline->text[1] != '$'))) {
2856 error(ERR_NONFATAL, "`%%undef' expects a macro identifier");
2857 free_tlist(origline);
2858 return DIRECTIVE_FOUND;
2860 if (tline->next) {
2861 error(ERR_WARNING|ERR_PASS1,
2862 "trailing garbage after macro name ignored");
2865 /* Find the context that symbol belongs to */
2866 ctx = get_ctx(tline->text, &mname, false);
2867 undef_smacro(ctx, mname);
2868 free_tlist(origline);
2869 return DIRECTIVE_FOUND;
2871 case PP_DEFSTR:
2872 case PP_IDEFSTR:
2873 casesense = (i == PP_DEFSTR);
2875 tline = tline->next;
2876 skip_white_(tline);
2877 tline = expand_id(tline);
2878 if (!tline || (tline->type != TOK_ID &&
2879 (tline->type != TOK_PREPROC_ID ||
2880 tline->text[1] != '$'))) {
2881 error(ERR_NONFATAL, "`%s' expects a macro identifier",
2882 pp_directives[i]);
2883 free_tlist(origline);
2884 return DIRECTIVE_FOUND;
2887 ctx = get_ctx(tline->text, &mname, false);
2888 last = tline;
2889 tline = expand_smacro(tline->next);
2890 last->next = NULL;
2892 while (tok_type_(tline, TOK_WHITESPACE))
2893 tline = delete_Token(tline);
2895 p = detoken(tline, false);
2896 macro_start = nasm_malloc(sizeof(*macro_start));
2897 macro_start->next = NULL;
2898 macro_start->text = nasm_quote(p, strlen(p));
2899 macro_start->type = TOK_STRING;
2900 macro_start->a.mac = NULL;
2901 nasm_free(p);
2904 * We now have a macro name, an implicit parameter count of
2905 * zero, and a string token to use as an expansion. Create
2906 * and store an SMacro.
2908 define_smacro(ctx, mname, casesense, 0, macro_start);
2909 free_tlist(origline);
2910 return DIRECTIVE_FOUND;
2912 case PP_PATHSEARCH:
2914 FILE *fp;
2915 StrList *xsl = NULL;
2916 StrList **xst = &xsl;
2918 casesense = true;
2920 tline = tline->next;
2921 skip_white_(tline);
2922 tline = expand_id(tline);
2923 if (!tline || (tline->type != TOK_ID &&
2924 (tline->type != TOK_PREPROC_ID ||
2925 tline->text[1] != '$'))) {
2926 error(ERR_NONFATAL,
2927 "`%%pathsearch' expects a macro identifier as first parameter");
2928 free_tlist(origline);
2929 return DIRECTIVE_FOUND;
2931 ctx = get_ctx(tline->text, &mname, false);
2932 last = tline;
2933 tline = expand_smacro(tline->next);
2934 last->next = NULL;
2936 t = tline;
2937 while (tok_type_(t, TOK_WHITESPACE))
2938 t = t->next;
2940 if (!t || (t->type != TOK_STRING &&
2941 t->type != TOK_INTERNAL_STRING)) {
2942 error(ERR_NONFATAL, "`%%pathsearch' expects a file name");
2943 free_tlist(tline);
2944 free_tlist(origline);
2945 return DIRECTIVE_FOUND; /* but we did _something_ */
2947 if (t->next)
2948 error(ERR_WARNING|ERR_PASS1,
2949 "trailing garbage after `%%pathsearch' ignored");
2950 p = t->text;
2951 if (t->type != TOK_INTERNAL_STRING)
2952 nasm_unquote(p, NULL);
2954 fp = inc_fopen(p, &xsl, &xst, true);
2955 if (fp) {
2956 p = xsl->str;
2957 fclose(fp); /* Don't actually care about the file */
2959 macro_start = nasm_malloc(sizeof(*macro_start));
2960 macro_start->next = NULL;
2961 macro_start->text = nasm_quote(p, strlen(p));
2962 macro_start->type = TOK_STRING;
2963 macro_start->a.mac = NULL;
2964 if (xsl)
2965 nasm_free(xsl);
2968 * We now have a macro name, an implicit parameter count of
2969 * zero, and a string token to use as an expansion. Create
2970 * and store an SMacro.
2972 define_smacro(ctx, mname, casesense, 0, macro_start);
2973 free_tlist(tline);
2974 free_tlist(origline);
2975 return DIRECTIVE_FOUND;
2978 case PP_STRLEN:
2979 casesense = true;
2981 tline = tline->next;
2982 skip_white_(tline);
2983 tline = expand_id(tline);
2984 if (!tline || (tline->type != TOK_ID &&
2985 (tline->type != TOK_PREPROC_ID ||
2986 tline->text[1] != '$'))) {
2987 error(ERR_NONFATAL,
2988 "`%%strlen' expects a macro identifier as first parameter");
2989 free_tlist(origline);
2990 return DIRECTIVE_FOUND;
2992 ctx = get_ctx(tline->text, &mname, false);
2993 last = tline;
2994 tline = expand_smacro(tline->next);
2995 last->next = NULL;
2997 t = tline;
2998 while (tok_type_(t, TOK_WHITESPACE))
2999 t = t->next;
3000 /* t should now point to the string */
3001 if (t->type != TOK_STRING) {
3002 error(ERR_NONFATAL,
3003 "`%%strlen` requires string as second parameter");
3004 free_tlist(tline);
3005 free_tlist(origline);
3006 return DIRECTIVE_FOUND;
3009 macro_start = nasm_malloc(sizeof(*macro_start));
3010 macro_start->next = NULL;
3011 make_tok_num(macro_start, nasm_unquote(t->text, NULL));
3012 macro_start->a.mac = NULL;
3015 * We now have a macro name, an implicit parameter count of
3016 * zero, and a numeric token to use as an expansion. Create
3017 * and store an SMacro.
3019 define_smacro(ctx, mname, casesense, 0, macro_start);
3020 free_tlist(tline);
3021 free_tlist(origline);
3022 return DIRECTIVE_FOUND;
3024 case PP_STRCAT:
3025 casesense = true;
3027 tline = tline->next;
3028 skip_white_(tline);
3029 tline = expand_id(tline);
3030 if (!tline || (tline->type != TOK_ID &&
3031 (tline->type != TOK_PREPROC_ID ||
3032 tline->text[1] != '$'))) {
3033 error(ERR_NONFATAL,
3034 "`%%strcat' expects a macro identifier as first parameter");
3035 free_tlist(origline);
3036 return DIRECTIVE_FOUND;
3038 ctx = get_ctx(tline->text, &mname, false);
3039 last = tline;
3040 tline = expand_smacro(tline->next);
3041 last->next = NULL;
3043 len = 0;
3044 for (t = tline; t; t = t->next) {
3045 switch (t->type) {
3046 case TOK_WHITESPACE:
3047 break;
3048 case TOK_STRING:
3049 len += t->a.len = nasm_unquote(t->text, NULL);
3050 break;
3051 case TOK_OTHER:
3052 if (!strcmp(t->text, ",")) /* permit comma separators */
3053 break;
3054 /* else fall through */
3055 default:
3056 error(ERR_NONFATAL,
3057 "non-string passed to `%%strcat' (%d)", t->type);
3058 free_tlist(tline);
3059 free_tlist(origline);
3060 return DIRECTIVE_FOUND;
3064 p = pp = nasm_malloc(len);
3065 t = tline;
3066 for (t = tline; t; t = t->next) {
3067 if (t->type == TOK_STRING) {
3068 memcpy(p, t->text, t->a.len);
3069 p += t->a.len;
3074 * We now have a macro name, an implicit parameter count of
3075 * zero, and a numeric token to use as an expansion. Create
3076 * and store an SMacro.
3078 macro_start = new_Token(NULL, TOK_STRING, NULL, 0);
3079 macro_start->text = nasm_quote(pp, len);
3080 nasm_free(pp);
3081 define_smacro(ctx, mname, casesense, 0, macro_start);
3082 free_tlist(tline);
3083 free_tlist(origline);
3084 return DIRECTIVE_FOUND;
3086 case PP_SUBSTR:
3088 int64_t a1, a2;
3089 size_t len;
3091 casesense = true;
3093 tline = tline->next;
3094 skip_white_(tline);
3095 tline = expand_id(tline);
3096 if (!tline || (tline->type != TOK_ID &&
3097 (tline->type != TOK_PREPROC_ID ||
3098 tline->text[1] != '$'))) {
3099 error(ERR_NONFATAL,
3100 "`%%substr' expects a macro identifier as first parameter");
3101 free_tlist(origline);
3102 return DIRECTIVE_FOUND;
3104 ctx = get_ctx(tline->text, &mname, false);
3105 last = tline;
3106 tline = expand_smacro(tline->next);
3107 last->next = NULL;
3109 t = tline->next;
3110 while (tok_type_(t, TOK_WHITESPACE))
3111 t = t->next;
3113 /* t should now point to the string */
3114 if (t->type != TOK_STRING) {
3115 error(ERR_NONFATAL,
3116 "`%%substr` requires string as second parameter");
3117 free_tlist(tline);
3118 free_tlist(origline);
3119 return DIRECTIVE_FOUND;
3122 tt = t->next;
3123 tptr = &tt;
3124 tokval.t_type = TOKEN_INVALID;
3125 evalresult = evaluate(ppscan, tptr, &tokval, NULL,
3126 pass, error, NULL);
3127 if (!evalresult) {
3128 free_tlist(tline);
3129 free_tlist(origline);
3130 return DIRECTIVE_FOUND;
3131 } else if (!is_simple(evalresult)) {
3132 error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3133 free_tlist(tline);
3134 free_tlist(origline);
3135 return DIRECTIVE_FOUND;
3137 a1 = evalresult->value-1;
3139 while (tok_type_(tt, TOK_WHITESPACE))
3140 tt = tt->next;
3141 if (!tt) {
3142 a2 = 1; /* Backwards compatibility: one character */
3143 } else {
3144 tokval.t_type = TOKEN_INVALID;
3145 evalresult = evaluate(ppscan, tptr, &tokval, NULL,
3146 pass, error, NULL);
3147 if (!evalresult) {
3148 free_tlist(tline);
3149 free_tlist(origline);
3150 return DIRECTIVE_FOUND;
3151 } else if (!is_simple(evalresult)) {
3152 error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3153 free_tlist(tline);
3154 free_tlist(origline);
3155 return DIRECTIVE_FOUND;
3157 a2 = evalresult->value;
3160 len = nasm_unquote(t->text, NULL);
3161 if (a2 < 0)
3162 a2 = a2+1+len-a1;
3163 if (a1+a2 > (int64_t)len)
3164 a2 = len-a1;
3166 macro_start = nasm_malloc(sizeof(*macro_start));
3167 macro_start->next = NULL;
3168 macro_start->text = nasm_quote((a1 < 0) ? "" : t->text+a1, a2);
3169 macro_start->type = TOK_STRING;
3170 macro_start->a.mac = NULL;
3173 * We now have a macro name, an implicit parameter count of
3174 * zero, and a numeric token to use as an expansion. Create
3175 * and store an SMacro.
3177 define_smacro(ctx, mname, casesense, 0, macro_start);
3178 free_tlist(tline);
3179 free_tlist(origline);
3180 return DIRECTIVE_FOUND;
3183 case PP_ASSIGN:
3184 case PP_IASSIGN:
3185 casesense = (i == PP_ASSIGN);
3187 tline = tline->next;
3188 skip_white_(tline);
3189 tline = expand_id(tline);
3190 if (!tline || (tline->type != TOK_ID &&
3191 (tline->type != TOK_PREPROC_ID ||
3192 tline->text[1] != '$'))) {
3193 error(ERR_NONFATAL,
3194 "`%%%sassign' expects a macro identifier",
3195 (i == PP_IASSIGN ? "i" : ""));
3196 free_tlist(origline);
3197 return DIRECTIVE_FOUND;
3199 ctx = get_ctx(tline->text, &mname, false);
3200 last = tline;
3201 tline = expand_smacro(tline->next);
3202 last->next = NULL;
3204 t = tline;
3205 tptr = &t;
3206 tokval.t_type = TOKEN_INVALID;
3207 evalresult =
3208 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
3209 free_tlist(tline);
3210 if (!evalresult) {
3211 free_tlist(origline);
3212 return DIRECTIVE_FOUND;
3215 if (tokval.t_type)
3216 error(ERR_WARNING|ERR_PASS1,
3217 "trailing garbage after expression ignored");
3219 if (!is_simple(evalresult)) {
3220 error(ERR_NONFATAL,
3221 "non-constant value given to `%%%sassign'",
3222 (i == PP_IASSIGN ? "i" : ""));
3223 free_tlist(origline);
3224 return DIRECTIVE_FOUND;
3227 macro_start = nasm_malloc(sizeof(*macro_start));
3228 macro_start->next = NULL;
3229 make_tok_num(macro_start, reloc_value(evalresult));
3230 macro_start->a.mac = NULL;
3233 * We now have a macro name, an implicit parameter count of
3234 * zero, and a numeric token to use as an expansion. Create
3235 * and store an SMacro.
3237 define_smacro(ctx, mname, casesense, 0, macro_start);
3238 free_tlist(origline);
3239 return DIRECTIVE_FOUND;
3241 case PP_LINE:
3243 * Syntax is `%line nnn[+mmm] [filename]'
3245 tline = tline->next;
3246 skip_white_(tline);
3247 if (!tok_type_(tline, TOK_NUMBER)) {
3248 error(ERR_NONFATAL, "`%%line' expects line number");
3249 free_tlist(origline);
3250 return DIRECTIVE_FOUND;
3252 k = readnum(tline->text, &err);
3253 m = 1;
3254 tline = tline->next;
3255 if (tok_is_(tline, "+")) {
3256 tline = tline->next;
3257 if (!tok_type_(tline, TOK_NUMBER)) {
3258 error(ERR_NONFATAL, "`%%line' expects line increment");
3259 free_tlist(origline);
3260 return DIRECTIVE_FOUND;
3262 m = readnum(tline->text, &err);
3263 tline = tline->next;
3265 skip_white_(tline);
3266 src_set_linnum(k);
3267 istk->lineinc = m;
3268 if (tline) {
3269 nasm_free(src_set_fname(detoken(tline, false)));
3271 free_tlist(origline);
3272 return DIRECTIVE_FOUND;
3274 default:
3275 error(ERR_FATAL,
3276 "preprocessor directive `%s' not yet implemented",
3277 pp_directives[i]);
3278 return DIRECTIVE_FOUND;
3283 * Ensure that a macro parameter contains a condition code and
3284 * nothing else. Return the condition code index if so, or -1
3285 * otherwise.
3287 static int find_cc(Token * t)
3289 Token *tt;
3290 int i, j, k, m;
3292 if (!t)
3293 return -1; /* Probably a %+ without a space */
3295 skip_white_(t);
3296 if (t->type != TOK_ID)
3297 return -1;
3298 tt = t->next;
3299 skip_white_(tt);
3300 if (tt && (tt->type != TOK_OTHER || strcmp(tt->text, ",")))
3301 return -1;
3303 i = -1;
3304 j = elements(conditions);
3305 while (j - i > 1) {
3306 k = (j + i) / 2;
3307 m = nasm_stricmp(t->text, conditions[k]);
3308 if (m == 0) {
3309 i = k;
3310 j = -2;
3311 break;
3312 } else if (m < 0) {
3313 j = k;
3314 } else
3315 i = k;
3317 if (j != -2)
3318 return -1;
3319 return i;
3323 * Expand MMacro-local things: parameter references (%0, %n, %+n,
3324 * %-n) and MMacro-local identifiers (%%foo) as well as
3325 * macro indirection (%[...]).
3327 static Token *expand_mmac_params(Token * tline)
3329 Token *t, *tt, **tail, *thead;
3331 tail = &thead;
3332 thead = NULL;
3334 while (tline) {
3335 if (tline->type == TOK_PREPROC_ID &&
3336 (((tline->text[1] == '+' || tline->text[1] == '-')
3337 && tline->text[2]) || tline->text[1] == '%'
3338 || (tline->text[1] >= '0' && tline->text[1] <= '9'))) {
3339 char *text = NULL;
3340 int type = 0, cc; /* type = 0 to placate optimisers */
3341 char tmpbuf[30];
3342 unsigned int n;
3343 int i;
3344 MMacro *mac;
3346 t = tline;
3347 tline = tline->next;
3349 mac = istk->mstk;
3350 while (mac && !mac->name) /* avoid mistaking %reps for macros */
3351 mac = mac->next_active;
3352 if (!mac)
3353 error(ERR_NONFATAL, "`%s': not in a macro call", t->text);
3354 else
3355 switch (t->text[1]) {
3357 * We have to make a substitution of one of the
3358 * forms %1, %-1, %+1, %%foo, %0.
3360 case '0':
3361 type = TOK_NUMBER;
3362 snprintf(tmpbuf, sizeof(tmpbuf), "%d", mac->nparam);
3363 text = nasm_strdup(tmpbuf);
3364 break;
3365 case '%':
3366 type = TOK_ID;
3367 snprintf(tmpbuf, sizeof(tmpbuf), "..@%"PRIu64".",
3368 mac->unique);
3369 text = nasm_strcat(tmpbuf, t->text + 2);
3370 break;
3371 case '-':
3372 n = atoi(t->text + 2) - 1;
3373 if (n >= mac->nparam)
3374 tt = NULL;
3375 else {
3376 if (mac->nparam > 1)
3377 n = (n + mac->rotate) % mac->nparam;
3378 tt = mac->params[n];
3380 cc = find_cc(tt);
3381 if (cc == -1) {
3382 error(ERR_NONFATAL,
3383 "macro parameter %d is not a condition code",
3384 n + 1);
3385 text = NULL;
3386 } else {
3387 type = TOK_ID;
3388 if (inverse_ccs[cc] == -1) {
3389 error(ERR_NONFATAL,
3390 "condition code `%s' is not invertible",
3391 conditions[cc]);
3392 text = NULL;
3393 } else
3394 text = nasm_strdup(conditions[inverse_ccs[cc]]);
3396 break;
3397 case '+':
3398 n = atoi(t->text + 2) - 1;
3399 if (n >= mac->nparam)
3400 tt = NULL;
3401 else {
3402 if (mac->nparam > 1)
3403 n = (n + mac->rotate) % mac->nparam;
3404 tt = mac->params[n];
3406 cc = find_cc(tt);
3407 if (cc == -1) {
3408 error(ERR_NONFATAL,
3409 "macro parameter %d is not a condition code",
3410 n + 1);
3411 text = NULL;
3412 } else {
3413 type = TOK_ID;
3414 text = nasm_strdup(conditions[cc]);
3416 break;
3417 default:
3418 n = atoi(t->text + 1) - 1;
3419 if (n >= mac->nparam)
3420 tt = NULL;
3421 else {
3422 if (mac->nparam > 1)
3423 n = (n + mac->rotate) % mac->nparam;
3424 tt = mac->params[n];
3426 if (tt) {
3427 for (i = 0; i < mac->paramlen[n]; i++) {
3428 *tail = new_Token(NULL, tt->type, tt->text, 0);
3429 tail = &(*tail)->next;
3430 tt = tt->next;
3433 text = NULL; /* we've done it here */
3434 break;
3436 if (!text) {
3437 delete_Token(t);
3438 } else {
3439 *tail = t;
3440 tail = &t->next;
3441 t->type = type;
3442 nasm_free(t->text);
3443 t->text = text;
3444 t->a.mac = NULL;
3446 continue;
3447 } else if (tline->type == TOK_INDIRECT) {
3448 t = tline;
3449 tline = tline->next;
3450 tt = tokenize(t->text);
3451 tt = expand_mmac_params(tt);
3452 tt = expand_smacro(tt);
3453 *tail = tt;
3454 while (tt) {
3455 tt->a.mac = NULL; /* Necessary? */
3456 tail = &tt->next;
3457 tt = tt->next;
3459 delete_Token(t);
3460 } else {
3461 t = *tail = tline;
3462 tline = tline->next;
3463 t->a.mac = NULL;
3464 tail = &t->next;
3467 *tail = NULL;
3469 /* Now handle token pasting... */
3470 t = thead;
3471 while (t && (tt = t->next)) {
3472 switch (t->type) {
3473 case TOK_WHITESPACE:
3474 if (tt->type == TOK_WHITESPACE) {
3475 t->next = delete_Token(tt);
3476 } else {
3477 t = tt;
3479 break;
3480 case TOK_ID:
3481 case TOK_NUMBER:
3482 if (tt->type == t->type || tt->type == TOK_NUMBER) {
3483 char *tmp = nasm_strcat(t->text, tt->text);
3484 nasm_free(t->text);
3485 t->text = tmp;
3486 t->next = delete_Token(tt);
3487 } else {
3488 t = tt;
3490 break;
3491 default:
3492 t = tt;
3493 break;
3496 return thead;
3500 * Expand all single-line macro calls made in the given line.
3501 * Return the expanded version of the line. The original is deemed
3502 * to be destroyed in the process. (In reality we'll just move
3503 * Tokens from input to output a lot of the time, rather than
3504 * actually bothering to destroy and replicate.)
3506 #define DEADMAN_LIMIT (1 << 20)
3508 static Token *expand_smacro(Token * tline)
3510 Token *t, *tt, *mstart, **tail, *thead;
3511 struct hash_table *smtbl;
3512 SMacro *head = NULL, *m;
3513 Token **params;
3514 int *paramsize;
3515 unsigned int nparam, sparam;
3516 int brackets, rescan;
3517 Token *org_tline = tline;
3518 Context *ctx;
3519 const char *mname;
3520 int deadman = DEADMAN_LIMIT;
3523 * Trick: we should avoid changing the start token pointer since it can
3524 * be contained in "next" field of other token. Because of this
3525 * we allocate a copy of first token and work with it; at the end of
3526 * routine we copy it back
3528 if (org_tline) {
3529 tline =
3530 new_Token(org_tline->next, org_tline->type, org_tline->text,
3532 tline->a.mac = org_tline->a.mac;
3533 nasm_free(org_tline->text);
3534 org_tline->text = NULL;
3537 again:
3538 tail = &thead;
3539 thead = NULL;
3541 while (tline) { /* main token loop */
3542 if (!--deadman) {
3543 error(ERR_NONFATAL, "interminable macro recursion");
3544 break;
3547 if ((mname = tline->text)) {
3548 /* if this token is a local macro, look in local context */
3549 if (tline->type == TOK_ID || tline->type == TOK_PREPROC_ID)
3550 ctx = get_ctx(mname, &mname, true);
3551 else
3552 ctx = NULL;
3553 smtbl = ctx ? &ctx->localmac : &smacros;
3554 head = (SMacro *) hash_findix(smtbl, mname);
3557 * We've hit an identifier. As in is_mmacro below, we first
3558 * check whether the identifier is a single-line macro at
3559 * all, then think about checking for parameters if
3560 * necessary.
3562 for (m = head; m; m = m->next)
3563 if (!mstrcmp(m->name, mname, m->casesense))
3564 break;
3565 if (m) {
3566 mstart = tline;
3567 params = NULL;
3568 paramsize = NULL;
3569 if (m->nparam == 0) {
3571 * Simple case: the macro is parameterless. Discard the
3572 * one token that the macro call took, and push the
3573 * expansion back on the to-do stack.
3575 if (!m->expansion) {
3576 if (!strcmp("__FILE__", m->name)) {
3577 int32_t num = 0;
3578 char *file = NULL;
3579 src_get(&num, &file);
3580 tline->text = nasm_quote(file, strlen(file));
3581 tline->type = TOK_STRING;
3582 nasm_free(file);
3583 continue;
3585 if (!strcmp("__LINE__", m->name)) {
3586 nasm_free(tline->text);
3587 make_tok_num(tline, src_get_linnum());
3588 continue;
3590 if (!strcmp("__BITS__", m->name)) {
3591 nasm_free(tline->text);
3592 make_tok_num(tline, globalbits);
3593 continue;
3595 tline = delete_Token(tline);
3596 continue;
3598 } else {
3600 * Complicated case: at least one macro with this name
3601 * exists and takes parameters. We must find the
3602 * parameters in the call, count them, find the SMacro
3603 * that corresponds to that form of the macro call, and
3604 * substitute for the parameters when we expand. What a
3605 * pain.
3607 /*tline = tline->next;
3608 skip_white_(tline); */
3609 do {
3610 t = tline->next;
3611 while (tok_type_(t, TOK_SMAC_END)) {
3612 t->a.mac->in_progress = false;
3613 t->text = NULL;
3614 t = tline->next = delete_Token(t);
3616 tline = t;
3617 } while (tok_type_(tline, TOK_WHITESPACE));
3618 if (!tok_is_(tline, "(")) {
3620 * This macro wasn't called with parameters: ignore
3621 * the call. (Behaviour borrowed from gnu cpp.)
3623 tline = mstart;
3624 m = NULL;
3625 } else {
3626 int paren = 0;
3627 int white = 0;
3628 brackets = 0;
3629 nparam = 0;
3630 sparam = PARAM_DELTA;
3631 params = nasm_malloc(sparam * sizeof(Token *));
3632 params[0] = tline->next;
3633 paramsize = nasm_malloc(sparam * sizeof(int));
3634 paramsize[0] = 0;
3635 while (true) { /* parameter loop */
3637 * For some unusual expansions
3638 * which concatenates function call
3640 t = tline->next;
3641 while (tok_type_(t, TOK_SMAC_END)) {
3642 t->a.mac->in_progress = false;
3643 t->text = NULL;
3644 t = tline->next = delete_Token(t);
3646 tline = t;
3648 if (!tline) {
3649 error(ERR_NONFATAL,
3650 "macro call expects terminating `)'");
3651 break;
3653 if (tline->type == TOK_WHITESPACE
3654 && brackets <= 0) {
3655 if (paramsize[nparam])
3656 white++;
3657 else
3658 params[nparam] = tline->next;
3659 continue; /* parameter loop */
3661 if (tline->type == TOK_OTHER
3662 && tline->text[1] == 0) {
3663 char ch = tline->text[0];
3664 if (ch == ',' && !paren && brackets <= 0) {
3665 if (++nparam >= sparam) {
3666 sparam += PARAM_DELTA;
3667 params = nasm_realloc(params,
3668 sparam *
3669 sizeof(Token
3670 *));
3671 paramsize =
3672 nasm_realloc(paramsize,
3673 sparam *
3674 sizeof(int));
3676 params[nparam] = tline->next;
3677 paramsize[nparam] = 0;
3678 white = 0;
3679 continue; /* parameter loop */
3681 if (ch == '{' &&
3682 (brackets > 0 || (brackets == 0 &&
3683 !paramsize[nparam])))
3685 if (!(brackets++)) {
3686 params[nparam] = tline->next;
3687 continue; /* parameter loop */
3690 if (ch == '}' && brackets > 0)
3691 if (--brackets == 0) {
3692 brackets = -1;
3693 continue; /* parameter loop */
3695 if (ch == '(' && !brackets)
3696 paren++;
3697 if (ch == ')' && brackets <= 0)
3698 if (--paren < 0)
3699 break;
3701 if (brackets < 0) {
3702 brackets = 0;
3703 error(ERR_NONFATAL, "braces do not "
3704 "enclose all of macro parameter");
3706 paramsize[nparam] += white + 1;
3707 white = 0;
3708 } /* parameter loop */
3709 nparam++;
3710 while (m && (m->nparam != nparam ||
3711 mstrcmp(m->name, mname,
3712 m->casesense)))
3713 m = m->next;
3714 if (!m)
3715 error(ERR_WARNING|ERR_PASS1|ERR_WARN_MNP,
3716 "macro `%s' exists, "
3717 "but not taking %d parameters",
3718 mstart->text, nparam);
3721 if (m && m->in_progress)
3722 m = NULL;
3723 if (!m) { /* in progess or didn't find '(' or wrong nparam */
3725 * Design question: should we handle !tline, which
3726 * indicates missing ')' here, or expand those
3727 * macros anyway, which requires the (t) test a few
3728 * lines down?
3730 nasm_free(params);
3731 nasm_free(paramsize);
3732 tline = mstart;
3733 } else {
3735 * Expand the macro: we are placed on the last token of the
3736 * call, so that we can easily split the call from the
3737 * following tokens. We also start by pushing an SMAC_END
3738 * token for the cycle removal.
3740 t = tline;
3741 if (t) {
3742 tline = t->next;
3743 t->next = NULL;
3745 tt = new_Token(tline, TOK_SMAC_END, NULL, 0);
3746 tt->a.mac = m;
3747 m->in_progress = true;
3748 tline = tt;
3749 for (t = m->expansion; t; t = t->next) {
3750 if (t->type >= TOK_SMAC_PARAM) {
3751 Token *pcopy = tline, **ptail = &pcopy;
3752 Token *ttt, *pt;
3753 int i;
3755 ttt = params[t->type - TOK_SMAC_PARAM];
3756 for (i = paramsize[t->type - TOK_SMAC_PARAM];
3757 --i >= 0;) {
3758 pt = *ptail =
3759 new_Token(tline, ttt->type, ttt->text,
3761 ptail = &pt->next;
3762 ttt = ttt->next;
3764 tline = pcopy;
3765 } else if (t->type == TOK_PREPROC_Q) {
3766 tt = new_Token(tline, TOK_ID, mname, 0);
3767 tline = tt;
3768 } else if (t->type == TOK_PREPROC_QQ) {
3769 tt = new_Token(tline, TOK_ID, m->name, 0);
3770 tline = tt;
3771 } else {
3772 tt = new_Token(tline, t->type, t->text, 0);
3773 tline = tt;
3778 * Having done that, get rid of the macro call, and clean
3779 * up the parameters.
3781 nasm_free(params);
3782 nasm_free(paramsize);
3783 free_tlist(mstart);
3784 continue; /* main token loop */
3789 if (tline->type == TOK_SMAC_END) {
3790 tline->a.mac->in_progress = false;
3791 tline = delete_Token(tline);
3792 } else {
3793 t = *tail = tline;
3794 tline = tline->next;
3795 t->a.mac = NULL;
3796 t->next = NULL;
3797 tail = &t->next;
3802 * Now scan the entire line and look for successive TOK_IDs that resulted
3803 * after expansion (they can't be produced by tokenize()). The successive
3804 * TOK_IDs should be concatenated.
3805 * Also we look for %+ tokens and concatenate the tokens before and after
3806 * them (without white spaces in between).
3808 t = thead;
3809 rescan = 0;
3810 while (t) {
3811 while (t && t->type != TOK_ID && t->type != TOK_PREPROC_ID)
3812 t = t->next;
3813 if (!t || !t->next)
3814 break;
3815 if (t->next->type == TOK_ID ||
3816 t->next->type == TOK_PREPROC_ID ||
3817 t->next->type == TOK_NUMBER) {
3818 char *p = nasm_strcat(t->text, t->next->text);
3819 nasm_free(t->text);
3820 t->next = delete_Token(t->next);
3821 t->text = p;
3822 rescan = 1;
3823 } else if (t->next->type == TOK_WHITESPACE && t->next->next &&
3824 t->next->next->type == TOK_PREPROC_ID &&
3825 strcmp(t->next->next->text, "%+") == 0) {
3826 /* free the next whitespace, the %+ token and next whitespace */
3827 int i;
3828 for (i = 1; i <= 3; i++) {
3829 if (!t->next
3830 || (i != 2 && t->next->type != TOK_WHITESPACE))
3831 break;
3832 t->next = delete_Token(t->next);
3833 } /* endfor */
3834 } else
3835 t = t->next;
3837 /* If we concatenaded something, re-scan the line for macros */
3838 if (rescan) {
3839 tline = thead;
3840 goto again;
3843 if (org_tline) {
3844 if (thead) {
3845 *org_tline = *thead;
3846 /* since we just gave text to org_line, don't free it */
3847 thead->text = NULL;
3848 delete_Token(thead);
3849 } else {
3850 /* the expression expanded to empty line;
3851 we can't return NULL for some reasons
3852 we just set the line to a single WHITESPACE token. */
3853 memset(org_tline, 0, sizeof(*org_tline));
3854 org_tline->text = NULL;
3855 org_tline->type = TOK_WHITESPACE;
3857 thead = org_tline;
3860 return thead;
3864 * Similar to expand_smacro but used exclusively with macro identifiers
3865 * right before they are fetched in. The reason is that there can be
3866 * identifiers consisting of several subparts. We consider that if there
3867 * are more than one element forming the name, user wants a expansion,
3868 * otherwise it will be left as-is. Example:
3870 * %define %$abc cde
3872 * the identifier %$abc will be left as-is so that the handler for %define
3873 * will suck it and define the corresponding value. Other case:
3875 * %define _%$abc cde
3877 * In this case user wants name to be expanded *before* %define starts
3878 * working, so we'll expand %$abc into something (if it has a value;
3879 * otherwise it will be left as-is) then concatenate all successive
3880 * PP_IDs into one.
3882 static Token *expand_id(Token * tline)
3884 Token *cur, *oldnext = NULL;
3886 if (!tline || !tline->next)
3887 return tline;
3889 cur = tline;
3890 while (cur->next &&
3891 (cur->next->type == TOK_ID ||
3892 cur->next->type == TOK_PREPROC_ID
3893 || cur->next->type == TOK_NUMBER))
3894 cur = cur->next;
3896 /* If identifier consists of just one token, don't expand */
3897 if (cur == tline)
3898 return tline;
3900 if (cur) {
3901 oldnext = cur->next; /* Detach the tail past identifier */
3902 cur->next = NULL; /* so that expand_smacro stops here */
3905 tline = expand_smacro(tline);
3907 if (cur) {
3908 /* expand_smacro possibly changhed tline; re-scan for EOL */
3909 cur = tline;
3910 while (cur && cur->next)
3911 cur = cur->next;
3912 if (cur)
3913 cur->next = oldnext;
3916 return tline;
3920 * Determine whether the given line constitutes a multi-line macro
3921 * call, and return the MMacro structure called if so. Doesn't have
3922 * to check for an initial label - that's taken care of in
3923 * expand_mmacro - but must check numbers of parameters. Guaranteed
3924 * to be called with tline->type == TOK_ID, so the putative macro
3925 * name is easy to find.
3927 static MMacro *is_mmacro(Token * tline, Token *** params_array)
3929 MMacro *head, *m;
3930 Token **params;
3931 int nparam;
3933 head = (MMacro *) hash_findix(&mmacros, tline->text);
3936 * Efficiency: first we see if any macro exists with the given
3937 * name. If not, we can return NULL immediately. _Then_ we
3938 * count the parameters, and then we look further along the
3939 * list if necessary to find the proper MMacro.
3941 for (m = head; m; m = m->next)
3942 if (!mstrcmp(m->name, tline->text, m->casesense))
3943 break;
3944 if (!m)
3945 return NULL;
3948 * OK, we have a potential macro. Count and demarcate the
3949 * parameters.
3951 count_mmac_params(tline->next, &nparam, &params);
3954 * So we know how many parameters we've got. Find the MMacro
3955 * structure that handles this number.
3957 while (m) {
3958 if (m->nparam_min <= nparam
3959 && (m->plus || nparam <= m->nparam_max)) {
3961 * This one is right. Just check if cycle removal
3962 * prohibits us using it before we actually celebrate...
3964 if (m->in_progress) {
3965 #if 0
3966 error(ERR_NONFATAL,
3967 "self-reference in multi-line macro `%s'", m->name);
3968 #endif
3969 nasm_free(params);
3970 return NULL;
3973 * It's right, and we can use it. Add its default
3974 * parameters to the end of our list if necessary.
3976 if (m->defaults && nparam < m->nparam_min + m->ndefs) {
3977 params =
3978 nasm_realloc(params,
3979 ((m->nparam_min + m->ndefs +
3980 1) * sizeof(*params)));
3981 while (nparam < m->nparam_min + m->ndefs) {
3982 params[nparam] = m->defaults[nparam - m->nparam_min];
3983 nparam++;
3987 * If we've gone over the maximum parameter count (and
3988 * we're in Plus mode), ignore parameters beyond
3989 * nparam_max.
3991 if (m->plus && nparam > m->nparam_max)
3992 nparam = m->nparam_max;
3994 * Then terminate the parameter list, and leave.
3996 if (!params) { /* need this special case */
3997 params = nasm_malloc(sizeof(*params));
3998 nparam = 0;
4000 params[nparam] = NULL;
4001 *params_array = params;
4002 return m;
4005 * This one wasn't right: look for the next one with the
4006 * same name.
4008 for (m = m->next; m; m = m->next)
4009 if (!mstrcmp(m->name, tline->text, m->casesense))
4010 break;
4014 * After all that, we didn't find one with the right number of
4015 * parameters. Issue a warning, and fail to expand the macro.
4017 error(ERR_WARNING|ERR_PASS1|ERR_WARN_MNP,
4018 "macro `%s' exists, but not taking %d parameters",
4019 tline->text, nparam);
4020 nasm_free(params);
4021 return NULL;
4025 * Expand the multi-line macro call made by the given line, if
4026 * there is one to be expanded. If there is, push the expansion on
4027 * istk->expansion and return 1. Otherwise return 0.
4029 static int expand_mmacro(Token * tline)
4031 Token *startline = tline;
4032 Token *label = NULL;
4033 int dont_prepend = 0;
4034 Token **params, *t, *mtok, *tt;
4035 MMacro *m;
4036 Line *l, *ll;
4037 int i, nparam, *paramlen;
4038 const char *mname;
4040 t = tline;
4041 skip_white_(t);
4042 /* if (!tok_type_(t, TOK_ID)) Lino 02/25/02 */
4043 if (!tok_type_(t, TOK_ID) && !tok_type_(t, TOK_PREPROC_ID))
4044 return 0;
4045 mtok = t;
4046 m = is_mmacro(t, &params);
4047 if (m) {
4048 mname = t->text;
4049 } else {
4050 Token *last;
4052 * We have an id which isn't a macro call. We'll assume
4053 * it might be a label; we'll also check to see if a
4054 * colon follows it. Then, if there's another id after
4055 * that lot, we'll check it again for macro-hood.
4057 label = last = t;
4058 t = t->next;
4059 if (tok_type_(t, TOK_WHITESPACE))
4060 last = t, t = t->next;
4061 if (tok_is_(t, ":")) {
4062 dont_prepend = 1;
4063 last = t, t = t->next;
4064 if (tok_type_(t, TOK_WHITESPACE))
4065 last = t, t = t->next;
4067 if (!tok_type_(t, TOK_ID) || (m = is_mmacro(t, &params)) == NULL)
4068 return 0;
4069 last->next = NULL;
4070 mname = t->text;
4071 tline = t;
4075 * Fix up the parameters: this involves stripping leading and
4076 * trailing whitespace, then stripping braces if they are
4077 * present.
4079 for (nparam = 0; params[nparam]; nparam++) ;
4080 paramlen = nparam ? nasm_malloc(nparam * sizeof(*paramlen)) : NULL;
4082 for (i = 0; params[i]; i++) {
4083 int brace = false;
4084 int comma = (!m->plus || i < nparam - 1);
4086 t = params[i];
4087 skip_white_(t);
4088 if (tok_is_(t, "{"))
4089 t = t->next, brace = true, comma = false;
4090 params[i] = t;
4091 paramlen[i] = 0;
4092 while (t) {
4093 if (comma && t->type == TOK_OTHER && !strcmp(t->text, ","))
4094 break; /* ... because we have hit a comma */
4095 if (comma && t->type == TOK_WHITESPACE
4096 && tok_is_(t->next, ","))
4097 break; /* ... or a space then a comma */
4098 if (brace && t->type == TOK_OTHER && !strcmp(t->text, "}"))
4099 break; /* ... or a brace */
4100 t = t->next;
4101 paramlen[i]++;
4106 * OK, we have a MMacro structure together with a set of
4107 * parameters. We must now go through the expansion and push
4108 * copies of each Line on to istk->expansion. Substitution of
4109 * parameter tokens and macro-local tokens doesn't get done
4110 * until the single-line macro substitution process; this is
4111 * because delaying them allows us to change the semantics
4112 * later through %rotate.
4114 * First, push an end marker on to istk->expansion, mark this
4115 * macro as in progress, and set up its invocation-specific
4116 * variables.
4118 ll = nasm_malloc(sizeof(Line));
4119 ll->next = istk->expansion;
4120 ll->finishes = m;
4121 ll->first = NULL;
4122 istk->expansion = ll;
4124 m->in_progress = true;
4125 m->params = params;
4126 m->iline = tline;
4127 m->nparam = nparam;
4128 m->rotate = 0;
4129 m->paramlen = paramlen;
4130 m->unique = unique++;
4131 m->lineno = 0;
4133 m->next_active = istk->mstk;
4134 istk->mstk = m;
4136 for (l = m->expansion; l; l = l->next) {
4137 Token **tail;
4139 ll = nasm_malloc(sizeof(Line));
4140 ll->finishes = NULL;
4141 ll->next = istk->expansion;
4142 istk->expansion = ll;
4143 tail = &ll->first;
4145 for (t = l->first; t; t = t->next) {
4146 Token *x = t;
4147 switch (t->type) {
4148 case TOK_PREPROC_Q:
4149 tt = *tail = new_Token(NULL, TOK_ID, mname, 0);
4150 break;
4151 case TOK_PREPROC_QQ:
4152 tt = *tail = new_Token(NULL, TOK_ID, m->name, 0);
4153 break;
4154 case TOK_PREPROC_ID:
4155 if (t->text[1] == '0' && t->text[2] == '0') {
4156 dont_prepend = -1;
4157 x = label;
4158 if (!x)
4159 continue;
4161 /* fall through */
4162 default:
4163 tt = *tail = new_Token(NULL, x->type, x->text, 0);
4164 break;
4166 tail = &tt->next;
4168 *tail = NULL;
4172 * If we had a label, push it on as the first line of
4173 * the macro expansion.
4175 if (label) {
4176 if (dont_prepend < 0)
4177 free_tlist(startline);
4178 else {
4179 ll = nasm_malloc(sizeof(Line));
4180 ll->finishes = NULL;
4181 ll->next = istk->expansion;
4182 istk->expansion = ll;
4183 ll->first = startline;
4184 if (!dont_prepend) {
4185 while (label->next)
4186 label = label->next;
4187 label->next = tt = new_Token(NULL, TOK_OTHER, ":", 0);
4192 list->uplevel(m->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
4194 return 1;
4197 /* The function that actually does the error reporting */
4198 static void verror(int severity, const char *fmt, va_list arg)
4200 char buff[1024];
4202 vsnprintf(buff, sizeof(buff), fmt, arg);
4204 if (istk && istk->mstk && istk->mstk->name)
4205 _error(severity, "(%s:%d) %s", istk->mstk->name,
4206 istk->mstk->lineno, buff);
4207 else
4208 _error(severity, "%s", buff);
4212 * Since preprocessor always operate only on the line that didn't
4213 * arrived yet, we should always use ERR_OFFBY1.
4215 static void error(int severity, const char *fmt, ...)
4217 va_list arg;
4219 /* If we're in a dead branch of IF or something like it, ignore the error */
4220 if (istk && istk->conds && !emitting(istk->conds->state))
4221 return;
4223 va_start(arg, fmt);
4224 verror(severity, fmt, arg);
4225 va_end(arg);
4229 * Because %else etc are evaluated in the state context
4230 * of the previous branch, errors might get lost with error():
4231 * %if 0 ... %else trailing garbage ... %endif
4232 * So %else etc should report errors with this function.
4234 static void error_precond(int severity, const char *fmt, ...)
4236 va_list arg;
4238 /* Only ignore the error if it's really in a dead branch */
4239 if (istk && istk->conds && istk->conds->state == COND_NEVER)
4240 return;
4242 va_start(arg, fmt);
4243 verror(severity, fmt, arg);
4244 va_end(arg);
4247 static void
4248 pp_reset(char *file, int apass, efunc errfunc, evalfunc eval,
4249 ListGen * listgen, StrList **deplist)
4251 Token *t;
4253 _error = errfunc;
4254 cstk = NULL;
4255 istk = nasm_malloc(sizeof(Include));
4256 istk->next = NULL;
4257 istk->conds = NULL;
4258 istk->expansion = NULL;
4259 istk->mstk = NULL;
4260 istk->fp = fopen(file, "r");
4261 istk->fname = NULL;
4262 src_set_fname(nasm_strdup(file));
4263 src_set_linnum(0);
4264 istk->lineinc = 1;
4265 if (!istk->fp)
4266 error(ERR_FATAL|ERR_NOFILE, "unable to open input file `%s'",
4267 file);
4268 defining = NULL;
4269 nested_mac_count = 0;
4270 nested_rep_count = 0;
4271 init_macros();
4272 unique = 0;
4273 if (tasm_compatible_mode) {
4274 stdmacpos = nasm_stdmac;
4275 } else {
4276 stdmacpos = nasm_stdmac_after_tasm;
4278 any_extrastdmac = extrastdmac && *extrastdmac;
4279 do_predef = true;
4280 list = listgen;
4281 evaluate = eval;
4284 * 0 for dependencies, 1 for preparatory passes, 2 for final pass.
4285 * The caller, however, will also pass in 3 for preprocess-only so
4286 * we can set __PASS__ accordingly.
4288 pass = apass > 2 ? 2 : apass;
4290 dephead = deptail = deplist;
4291 if (deplist) {
4292 StrList *sl = nasm_malloc(strlen(file)+1+sizeof sl->next);
4293 sl->next = NULL;
4294 strcpy(sl->str, file);
4295 *deptail = sl;
4296 deptail = &sl->next;
4300 * Define the __PASS__ macro. This is defined here unlike
4301 * all the other builtins, because it is special -- it varies between
4302 * passes.
4304 t = nasm_malloc(sizeof(*t));
4305 t->next = NULL;
4306 make_tok_num(t, apass);
4307 t->a.mac = NULL;
4308 define_smacro(NULL, "__PASS__", true, 0, t);
4311 static char *pp_getline(void)
4313 char *line;
4314 Token *tline;
4316 while (1) {
4318 * Fetch a tokenized line, either from the macro-expansion
4319 * buffer or from the input file.
4321 tline = NULL;
4322 while (istk->expansion && istk->expansion->finishes) {
4323 Line *l = istk->expansion;
4324 if (!l->finishes->name && l->finishes->in_progress > 1) {
4325 Line *ll;
4328 * This is a macro-end marker for a macro with no
4329 * name, which means it's not really a macro at all
4330 * but a %rep block, and the `in_progress' field is
4331 * more than 1, meaning that we still need to
4332 * repeat. (1 means the natural last repetition; 0
4333 * means termination by %exitrep.) We have
4334 * therefore expanded up to the %endrep, and must
4335 * push the whole block on to the expansion buffer
4336 * again. We don't bother to remove the macro-end
4337 * marker: we'd only have to generate another one
4338 * if we did.
4340 l->finishes->in_progress--;
4341 for (l = l->finishes->expansion; l; l = l->next) {
4342 Token *t, *tt, **tail;
4344 ll = nasm_malloc(sizeof(Line));
4345 ll->next = istk->expansion;
4346 ll->finishes = NULL;
4347 ll->first = NULL;
4348 tail = &ll->first;
4350 for (t = l->first; t; t = t->next) {
4351 if (t->text || t->type == TOK_WHITESPACE) {
4352 tt = *tail =
4353 new_Token(NULL, t->type, t->text, 0);
4354 tail = &tt->next;
4358 istk->expansion = ll;
4360 } else {
4362 * Check whether a `%rep' was started and not ended
4363 * within this macro expansion. This can happen and
4364 * should be detected. It's a fatal error because
4365 * I'm too confused to work out how to recover
4366 * sensibly from it.
4368 if (defining) {
4369 if (defining->name)
4370 error(ERR_PANIC,
4371 "defining with name in expansion");
4372 else if (istk->mstk->name)
4373 error(ERR_FATAL,
4374 "`%%rep' without `%%endrep' within"
4375 " expansion of macro `%s'",
4376 istk->mstk->name);
4380 * FIXME: investigate the relationship at this point between
4381 * istk->mstk and l->finishes
4384 MMacro *m = istk->mstk;
4385 istk->mstk = m->next_active;
4386 if (m->name) {
4388 * This was a real macro call, not a %rep, and
4389 * therefore the parameter information needs to
4390 * be freed.
4392 nasm_free(m->params);
4393 free_tlist(m->iline);
4394 nasm_free(m->paramlen);
4395 l->finishes->in_progress = false;
4396 } else
4397 free_mmacro(m);
4399 istk->expansion = l->next;
4400 nasm_free(l);
4401 list->downlevel(LIST_MACRO);
4404 while (1) { /* until we get a line we can use */
4406 if (istk->expansion) { /* from a macro expansion */
4407 char *p;
4408 Line *l = istk->expansion;
4409 if (istk->mstk)
4410 istk->mstk->lineno++;
4411 tline = l->first;
4412 istk->expansion = l->next;
4413 nasm_free(l);
4414 p = detoken(tline, false);
4415 list->line(LIST_MACRO, p);
4416 nasm_free(p);
4417 break;
4419 line = read_line();
4420 if (line) { /* from the current input file */
4421 line = prepreproc(line);
4422 tline = tokenize(line);
4423 nasm_free(line);
4424 break;
4427 * The current file has ended; work down the istk
4430 Include *i = istk;
4431 fclose(i->fp);
4432 if (i->conds)
4433 error(ERR_FATAL,
4434 "expected `%%endif' before end of file");
4435 /* only set line and file name if there's a next node */
4436 if (i->next) {
4437 src_set_linnum(i->lineno);
4438 nasm_free(src_set_fname(i->fname));
4440 istk = i->next;
4441 list->downlevel(LIST_INCLUDE);
4442 nasm_free(i);
4443 if (!istk)
4444 return NULL;
4445 if (istk->expansion && istk->expansion->finishes)
4446 break;
4451 * We must expand MMacro parameters and MMacro-local labels
4452 * _before_ we plunge into directive processing, to cope
4453 * with things like `%define something %1' such as STRUC
4454 * uses. Unless we're _defining_ a MMacro, in which case
4455 * those tokens should be left alone to go into the
4456 * definition; and unless we're in a non-emitting
4457 * condition, in which case we don't want to meddle with
4458 * anything.
4460 if (!defining && !(istk->conds && !emitting(istk->conds->state))
4461 && !(istk->mstk && !istk->mstk->in_progress)) {
4462 tline = expand_mmac_params(tline);
4466 * Check the line to see if it's a preprocessor directive.
4468 if (do_directive(tline) == DIRECTIVE_FOUND) {
4469 continue;
4470 } else if (defining) {
4472 * We're defining a multi-line macro. We emit nothing
4473 * at all, and just
4474 * shove the tokenized line on to the macro definition.
4476 Line *l = nasm_malloc(sizeof(Line));
4477 l->next = defining->expansion;
4478 l->first = tline;
4479 l->finishes = NULL;
4480 defining->expansion = l;
4481 continue;
4482 } else if (istk->conds && !emitting(istk->conds->state)) {
4484 * We're in a non-emitting branch of a condition block.
4485 * Emit nothing at all, not even a blank line: when we
4486 * emerge from the condition we'll give a line-number
4487 * directive so we keep our place correctly.
4489 free_tlist(tline);
4490 continue;
4491 } else if (istk->mstk && !istk->mstk->in_progress) {
4493 * We're in a %rep block which has been terminated, so
4494 * we're walking through to the %endrep without
4495 * emitting anything. Emit nothing at all, not even a
4496 * blank line: when we emerge from the %rep block we'll
4497 * give a line-number directive so we keep our place
4498 * correctly.
4500 free_tlist(tline);
4501 continue;
4502 } else {
4503 tline = expand_smacro(tline);
4504 if (!expand_mmacro(tline)) {
4506 * De-tokenize the line again, and emit it.
4508 line = detoken(tline, true);
4509 free_tlist(tline);
4510 break;
4511 } else {
4512 continue; /* expand_mmacro calls free_tlist */
4517 return line;
4520 static void pp_cleanup(int pass)
4522 if (defining) {
4523 if(defining->name) {
4524 error(ERR_NONFATAL,
4525 "end of file while still defining macro `%s'",
4526 defining->name);
4527 } else {
4528 error(ERR_NONFATAL, "end of file while still in %%rep");
4531 free_mmacro(defining);
4533 while (cstk)
4534 ctx_pop();
4535 free_macros();
4536 while (istk) {
4537 Include *i = istk;
4538 istk = istk->next;
4539 fclose(i->fp);
4540 nasm_free(i->fname);
4541 nasm_free(i);
4543 while (cstk)
4544 ctx_pop();
4545 nasm_free(src_set_fname(NULL));
4546 if (pass == 0) {
4547 IncPath *i;
4548 free_llist(predef);
4549 delete_Blocks();
4550 while ((i = ipath)) {
4551 ipath = i->next;
4552 if (i->path)
4553 nasm_free(i->path);
4554 nasm_free(i);
4559 void pp_include_path(char *path)
4561 IncPath *i;
4563 i = nasm_malloc(sizeof(IncPath));
4564 i->path = path ? nasm_strdup(path) : NULL;
4565 i->next = NULL;
4567 if (ipath != NULL) {
4568 IncPath *j = ipath;
4569 while (j->next != NULL)
4570 j = j->next;
4571 j->next = i;
4572 } else {
4573 ipath = i;
4577 void pp_pre_include(char *fname)
4579 Token *inc, *space, *name;
4580 Line *l;
4582 name = new_Token(NULL, TOK_INTERNAL_STRING, fname, 0);
4583 space = new_Token(name, TOK_WHITESPACE, NULL, 0);
4584 inc = new_Token(space, TOK_PREPROC_ID, "%include", 0);
4586 l = nasm_malloc(sizeof(Line));
4587 l->next = predef;
4588 l->first = inc;
4589 l->finishes = NULL;
4590 predef = l;
4593 void pp_pre_define(char *definition)
4595 Token *def, *space;
4596 Line *l;
4597 char *equals;
4599 equals = strchr(definition, '=');
4600 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
4601 def = new_Token(space, TOK_PREPROC_ID, "%define", 0);
4602 if (equals)
4603 *equals = ' ';
4604 space->next = tokenize(definition);
4605 if (equals)
4606 *equals = '=';
4608 l = nasm_malloc(sizeof(Line));
4609 l->next = predef;
4610 l->first = def;
4611 l->finishes = NULL;
4612 predef = l;
4615 void pp_pre_undefine(char *definition)
4617 Token *def, *space;
4618 Line *l;
4620 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
4621 def = new_Token(space, TOK_PREPROC_ID, "%undef", 0);
4622 space->next = tokenize(definition);
4624 l = nasm_malloc(sizeof(Line));
4625 l->next = predef;
4626 l->first = def;
4627 l->finishes = NULL;
4628 predef = l;
4632 * Added by Keith Kanios:
4634 * This function is used to assist with "runtime" preprocessor
4635 * directives. (e.g. pp_runtime("%define __BITS__ 64");)
4637 * ERRORS ARE IGNORED HERE, SO MAKE COMPLETELY SURE THAT YOU
4638 * PASS A VALID STRING TO THIS FUNCTION!!!!!
4641 void pp_runtime(char *definition)
4643 Token *def;
4645 def = tokenize(definition);
4646 if(do_directive(def) == NO_DIRECTIVE_FOUND)
4647 free_tlist(def);
4651 void pp_extra_stdmac(macros_t *macros)
4653 extrastdmac = macros;
4656 static void make_tok_num(Token * tok, int64_t val)
4658 char numbuf[20];
4659 snprintf(numbuf, sizeof(numbuf), "%"PRId64"", val);
4660 tok->text = nasm_strdup(numbuf);
4661 tok->type = TOK_NUMBER;
4664 Preproc nasmpp = {
4665 pp_reset,
4666 pp_getline,
4667 pp_cleanup