Merge branch 'master' of git://repo.or.cz/nasm
[nasm/perl-rewrite.git] / preproc.c
blobc6e0fb60f95925b380d44e20c93eae464fa24efc
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 (p[0] == '$' && p[1] == '$') {
877 type = TOK_OTHER; /* TOKEN_BASE */
878 p += 2;
879 } else if (isnumstart(*p)) {
880 bool is_hex = false;
881 bool is_float = false;
882 bool has_e = false;
883 char c, *r;
886 * A numeric token.
889 if (*p == '$') {
890 p++;
891 is_hex = true;
894 for (;;) {
895 c = *p++;
897 if (!is_hex && (c == 'e' || c == 'E')) {
898 has_e = true;
899 if (*p == '+' || *p == '-') {
900 /* e can only be followed by +/- if it is either a
901 prefixed hex number or a floating-point number */
902 p++;
903 is_float = true;
905 } else if (c == 'H' || c == 'h' || c == 'X' || c == 'x') {
906 is_hex = true;
907 } else if (c == 'P' || c == 'p') {
908 is_float = true;
909 if (*p == '+' || *p == '-')
910 p++;
911 } else if (isnumchar(c) || c == '_')
912 ; /* just advance */
913 else if (c == '.') {
914 /* we need to deal with consequences of the legacy
915 parser, like "1.nolist" being two tokens
916 (TOK_NUMBER, TOK_ID) here; at least give it
917 a shot for now. In the future, we probably need
918 a flex-based scanner with proper pattern matching
919 to do it as well as it can be done. Nothing in
920 the world is going to help the person who wants
921 0x123.p16 interpreted as two tokens, though. */
922 r = p;
923 while (*r == '_')
924 r++;
926 if (nasm_isdigit(*r) || (is_hex && nasm_isxdigit(*r)) ||
927 (!is_hex && (*r == 'e' || *r == 'E')) ||
928 (*r == 'p' || *r == 'P')) {
929 p = r;
930 is_float = true;
931 } else
932 break; /* Terminate the token */
933 } else
934 break;
936 p--; /* Point to first character beyond number */
938 if (p == line+1 && *line == '$') {
939 type = TOK_OTHER; /* TOKEN_HERE */
940 } else {
941 if (has_e && !is_hex) {
942 /* 1e13 is floating-point, but 1e13h is not */
943 is_float = true;
946 type = is_float ? TOK_FLOAT : TOK_NUMBER;
948 } else if (nasm_isspace(*p)) {
949 type = TOK_WHITESPACE;
950 p++;
951 while (*p && nasm_isspace(*p))
952 p++;
954 * Whitespace just before end-of-line is discarded by
955 * pretending it's a comment; whitespace just before a
956 * comment gets lumped into the comment.
958 if (!*p || *p == ';') {
959 type = TOK_COMMENT;
960 while (*p)
961 p++;
963 } else if (*p == ';') {
964 type = TOK_COMMENT;
965 while (*p)
966 p++;
967 } else {
969 * Anything else is an operator of some kind. We check
970 * for all the double-character operators (>>, <<, //,
971 * %%, <=, >=, ==, !=, <>, &&, ||, ^^), but anything
972 * else is a single-character operator.
974 type = TOK_OTHER;
975 if ((p[0] == '>' && p[1] == '>') ||
976 (p[0] == '<' && p[1] == '<') ||
977 (p[0] == '/' && p[1] == '/') ||
978 (p[0] == '<' && p[1] == '=') ||
979 (p[0] == '>' && p[1] == '=') ||
980 (p[0] == '=' && p[1] == '=') ||
981 (p[0] == '!' && p[1] == '=') ||
982 (p[0] == '<' && p[1] == '>') ||
983 (p[0] == '&' && p[1] == '&') ||
984 (p[0] == '|' && p[1] == '|') ||
985 (p[0] == '^' && p[1] == '^')) {
986 p++;
988 p++;
991 /* Handling unterminated string by UNV */
992 /*if (type == -1)
994 *tail = t = new_Token(NULL, TOK_STRING, line, p-line+1);
995 t->text[p-line] = *line;
996 tail = &t->next;
998 else */
999 if (type != TOK_COMMENT) {
1000 *tail = t = new_Token(NULL, type, line, p - line);
1001 tail = &t->next;
1003 line = p;
1005 return list;
1009 * this function allocates a new managed block of memory and
1010 * returns a pointer to the block. The managed blocks are
1011 * deleted only all at once by the delete_Blocks function.
1013 static void *new_Block(size_t size)
1015 Blocks *b = &blocks;
1017 /* first, get to the end of the linked list */
1018 while (b->next)
1019 b = b->next;
1020 /* now allocate the requested chunk */
1021 b->chunk = nasm_malloc(size);
1023 /* now allocate a new block for the next request */
1024 b->next = nasm_malloc(sizeof(Blocks));
1025 /* and initialize the contents of the new block */
1026 b->next->next = NULL;
1027 b->next->chunk = NULL;
1028 return b->chunk;
1032 * this function deletes all managed blocks of memory
1034 static void delete_Blocks(void)
1036 Blocks *a, *b = &blocks;
1039 * keep in mind that the first block, pointed to by blocks
1040 * is a static and not dynamically allocated, so we don't
1041 * free it.
1043 while (b) {
1044 if (b->chunk)
1045 nasm_free(b->chunk);
1046 a = b;
1047 b = b->next;
1048 if (a != &blocks)
1049 nasm_free(a);
1054 * this function creates a new Token and passes a pointer to it
1055 * back to the caller. It sets the type and text elements, and
1056 * also the a.mac and next elements to NULL.
1058 static Token *new_Token(Token * next, enum pp_token_type type,
1059 const char *text, int txtlen)
1061 Token *t;
1062 int i;
1064 if (freeTokens == NULL) {
1065 freeTokens = (Token *) new_Block(TOKEN_BLOCKSIZE * sizeof(Token));
1066 for (i = 0; i < TOKEN_BLOCKSIZE - 1; i++)
1067 freeTokens[i].next = &freeTokens[i + 1];
1068 freeTokens[i].next = NULL;
1070 t = freeTokens;
1071 freeTokens = t->next;
1072 t->next = next;
1073 t->a.mac = NULL;
1074 t->type = type;
1075 if (type == TOK_WHITESPACE || text == NULL) {
1076 t->text = NULL;
1077 } else {
1078 if (txtlen == 0)
1079 txtlen = strlen(text);
1080 t->text = nasm_malloc(txtlen+1);
1081 memcpy(t->text, text, txtlen);
1082 t->text[txtlen] = '\0';
1084 return t;
1087 static Token *delete_Token(Token * t)
1089 Token *next = t->next;
1090 nasm_free(t->text);
1091 t->next = freeTokens;
1092 freeTokens = t;
1093 return next;
1097 * Convert a line of tokens back into text.
1098 * If expand_locals is not zero, identifiers of the form "%$*xxx"
1099 * will be transformed into ..@ctxnum.xxx
1101 static char *detoken(Token * tlist, bool expand_locals)
1103 Token *t;
1104 int len;
1105 char *line, *p;
1106 const char *q;
1108 len = 0;
1109 for (t = tlist; t; t = t->next) {
1110 if (t->type == TOK_PREPROC_ID && t->text[1] == '!') {
1111 char *p = getenv(t->text + 2);
1112 nasm_free(t->text);
1113 if (p)
1114 t->text = nasm_strdup(p);
1115 else
1116 t->text = NULL;
1118 /* Expand local macros here and not during preprocessing */
1119 if (expand_locals &&
1120 t->type == TOK_PREPROC_ID && t->text &&
1121 t->text[0] == '%' && t->text[1] == '$') {
1122 const char *q;
1123 char *p;
1124 Context *ctx = get_ctx(t->text, &q, false);
1125 if (ctx) {
1126 char buffer[40];
1127 snprintf(buffer, sizeof(buffer), "..@%"PRIu32".", ctx->number);
1128 p = nasm_strcat(buffer, q);
1129 nasm_free(t->text);
1130 t->text = p;
1133 if (t->type == TOK_WHITESPACE) {
1134 len++;
1135 } else if (t->text) {
1136 len += strlen(t->text);
1139 p = line = nasm_malloc(len + 1);
1140 for (t = tlist; t; t = t->next) {
1141 if (t->type == TOK_WHITESPACE) {
1142 *p++ = ' ';
1143 } else if (t->text) {
1144 q = t->text;
1145 while (*q)
1146 *p++ = *q++;
1149 *p = '\0';
1150 return line;
1154 * A scanner, suitable for use by the expression evaluator, which
1155 * operates on a line of Tokens. Expects a pointer to a pointer to
1156 * the first token in the line to be passed in as its private_data
1157 * field.
1159 * FIX: This really needs to be unified with stdscan.
1161 static int ppscan(void *private_data, struct tokenval *tokval)
1163 Token **tlineptr = private_data;
1164 Token *tline;
1165 char ourcopy[MAX_KEYWORD+1], *p, *r, *s;
1167 do {
1168 tline = *tlineptr;
1169 *tlineptr = tline ? tline->next : NULL;
1171 while (tline && (tline->type == TOK_WHITESPACE ||
1172 tline->type == TOK_COMMENT));
1174 if (!tline)
1175 return tokval->t_type = TOKEN_EOS;
1177 tokval->t_charptr = tline->text;
1179 if (tline->text[0] == '$' && !tline->text[1])
1180 return tokval->t_type = TOKEN_HERE;
1181 if (tline->text[0] == '$' && tline->text[1] == '$' && !tline->text[2])
1182 return tokval->t_type = TOKEN_BASE;
1184 if (tline->type == TOK_ID) {
1185 p = tokval->t_charptr = tline->text;
1186 if (p[0] == '$') {
1187 tokval->t_charptr++;
1188 return tokval->t_type = TOKEN_ID;
1191 for (r = p, s = ourcopy; *r; r++) {
1192 if (r >= p+MAX_KEYWORD)
1193 return tokval->t_type = TOKEN_ID; /* Not a keyword */
1194 *s++ = nasm_tolower(*r);
1196 *s = '\0';
1197 /* right, so we have an identifier sitting in temp storage. now,
1198 * is it actually a register or instruction name, or what? */
1199 return nasm_token_hash(ourcopy, tokval);
1202 if (tline->type == TOK_NUMBER) {
1203 bool rn_error;
1204 tokval->t_integer = readnum(tline->text, &rn_error);
1205 tokval->t_charptr = tline->text;
1206 if (rn_error)
1207 return tokval->t_type = TOKEN_ERRNUM;
1208 else
1209 return tokval->t_type = TOKEN_NUM;
1212 if (tline->type == TOK_FLOAT) {
1213 return tokval->t_type = TOKEN_FLOAT;
1216 if (tline->type == TOK_STRING) {
1217 char bq, *ep;
1219 bq = tline->text[0];
1220 tokval->t_charptr = tline->text;
1221 tokval->t_inttwo = nasm_unquote(tline->text, &ep);
1223 if (ep[0] != bq || ep[1] != '\0')
1224 return tokval->t_type = TOKEN_ERRSTR;
1225 else
1226 return tokval->t_type = TOKEN_STR;
1229 if (tline->type == TOK_OTHER) {
1230 if (!strcmp(tline->text, "<<"))
1231 return tokval->t_type = TOKEN_SHL;
1232 if (!strcmp(tline->text, ">>"))
1233 return tokval->t_type = TOKEN_SHR;
1234 if (!strcmp(tline->text, "//"))
1235 return tokval->t_type = TOKEN_SDIV;
1236 if (!strcmp(tline->text, "%%"))
1237 return tokval->t_type = TOKEN_SMOD;
1238 if (!strcmp(tline->text, "=="))
1239 return tokval->t_type = TOKEN_EQ;
1240 if (!strcmp(tline->text, "<>"))
1241 return tokval->t_type = TOKEN_NE;
1242 if (!strcmp(tline->text, "!="))
1243 return tokval->t_type = TOKEN_NE;
1244 if (!strcmp(tline->text, "<="))
1245 return tokval->t_type = TOKEN_LE;
1246 if (!strcmp(tline->text, ">="))
1247 return tokval->t_type = TOKEN_GE;
1248 if (!strcmp(tline->text, "&&"))
1249 return tokval->t_type = TOKEN_DBL_AND;
1250 if (!strcmp(tline->text, "^^"))
1251 return tokval->t_type = TOKEN_DBL_XOR;
1252 if (!strcmp(tline->text, "||"))
1253 return tokval->t_type = TOKEN_DBL_OR;
1257 * We have no other options: just return the first character of
1258 * the token text.
1260 return tokval->t_type = tline->text[0];
1264 * Compare a string to the name of an existing macro; this is a
1265 * simple wrapper which calls either strcmp or nasm_stricmp
1266 * depending on the value of the `casesense' parameter.
1268 static int mstrcmp(const char *p, const char *q, bool casesense)
1270 return casesense ? strcmp(p, q) : nasm_stricmp(p, q);
1274 * Compare a string to the name of an existing macro; this is a
1275 * simple wrapper which calls either strcmp or nasm_stricmp
1276 * depending on the value of the `casesense' parameter.
1278 static int mmemcmp(const char *p, const char *q, size_t l, bool casesense)
1280 return casesense ? memcmp(p, q, l) : nasm_memicmp(p, q, l);
1284 * Return the Context structure associated with a %$ token. Return
1285 * NULL, having _already_ reported an error condition, if the
1286 * context stack isn't deep enough for the supplied number of $
1287 * signs.
1288 * If all_contexts == true, contexts that enclose current are
1289 * also scanned for such smacro, until it is found; if not -
1290 * only the context that directly results from the number of $'s
1291 * in variable's name.
1293 * If "namep" is non-NULL, set it to the pointer to the macro name
1294 * tail, i.e. the part beyond %$...
1296 static Context *get_ctx(const char *name, const char **namep,
1297 bool all_contexts)
1299 Context *ctx;
1300 SMacro *m;
1301 int i;
1303 if (namep)
1304 *namep = name;
1306 if (!name || name[0] != '%' || name[1] != '$')
1307 return NULL;
1309 if (!cstk) {
1310 error(ERR_NONFATAL, "`%s': context stack is empty", name);
1311 return NULL;
1314 name += 2;
1315 ctx = cstk;
1316 i = 0;
1317 while (ctx && *name == '$') {
1318 name++;
1319 i++;
1320 ctx = ctx->next;
1322 if (!ctx) {
1323 error(ERR_NONFATAL, "`%s': context stack is only"
1324 " %d level%s deep", name, i, (i == 1 ? "" : "s"));
1325 return NULL;
1328 if (namep)
1329 *namep = name;
1331 if (!all_contexts)
1332 return ctx;
1334 do {
1335 /* Search for this smacro in found context */
1336 m = hash_findix(&ctx->localmac, name);
1337 while (m) {
1338 if (!mstrcmp(m->name, name, m->casesense))
1339 return ctx;
1340 m = m->next;
1342 ctx = ctx->next;
1344 while (ctx);
1345 return NULL;
1349 * Check to see if a file is already in a string list
1351 static bool in_list(const StrList *list, const char *str)
1353 while (list) {
1354 if (!strcmp(list->str, str))
1355 return true;
1356 list = list->next;
1358 return false;
1362 * Open an include file. This routine must always return a valid
1363 * file pointer if it returns - it's responsible for throwing an
1364 * ERR_FATAL and bombing out completely if not. It should also try
1365 * the include path one by one until it finds the file or reaches
1366 * the end of the path.
1368 static FILE *inc_fopen(const char *file, StrList **dhead, StrList ***dtail,
1369 bool missing_ok)
1371 FILE *fp;
1372 char *prefix = "";
1373 IncPath *ip = ipath;
1374 int len = strlen(file);
1375 size_t prefix_len = 0;
1376 StrList *sl;
1378 while (1) {
1379 sl = nasm_malloc(prefix_len+len+1+sizeof sl->next);
1380 memcpy(sl->str, prefix, prefix_len);
1381 memcpy(sl->str+prefix_len, file, len+1);
1382 fp = fopen(sl->str, "r");
1383 if (fp && dhead && !in_list(*dhead, sl->str)) {
1384 sl->next = NULL;
1385 **dtail = sl;
1386 *dtail = &sl->next;
1387 } else {
1388 nasm_free(sl);
1390 if (fp)
1391 return fp;
1392 if (!ip) {
1393 if (!missing_ok)
1394 break;
1395 prefix = NULL;
1396 } else {
1397 prefix = ip->path;
1398 ip = ip->next;
1400 if (prefix) {
1401 prefix_len = strlen(prefix);
1402 } else {
1403 /* -MG given and file not found */
1404 if (dhead && !in_list(*dhead, file)) {
1405 sl = nasm_malloc(len+1+sizeof sl->next);
1406 sl->next = NULL;
1407 strcpy(sl->str, file);
1408 **dtail = sl;
1409 *dtail = &sl->next;
1411 return NULL;
1415 error(ERR_FATAL, "unable to open include file `%s'", file);
1416 return NULL; /* never reached - placate compilers */
1420 * Determine if we should warn on defining a single-line macro of
1421 * name `name', with `nparam' parameters. If nparam is 0 or -1, will
1422 * return true if _any_ single-line macro of that name is defined.
1423 * Otherwise, will return true if a single-line macro with either
1424 * `nparam' or no parameters is defined.
1426 * If a macro with precisely the right number of parameters is
1427 * defined, or nparam is -1, the address of the definition structure
1428 * will be returned in `defn'; otherwise NULL will be returned. If `defn'
1429 * is NULL, no action will be taken regarding its contents, and no
1430 * error will occur.
1432 * Note that this is also called with nparam zero to resolve
1433 * `ifdef'.
1435 * If you already know which context macro belongs to, you can pass
1436 * the context pointer as first parameter; if you won't but name begins
1437 * with %$ the context will be automatically computed. If all_contexts
1438 * is true, macro will be searched in outer contexts as well.
1440 static bool
1441 smacro_defined(Context * ctx, const char *name, int nparam, SMacro ** defn,
1442 bool nocase)
1444 struct hash_table *smtbl;
1445 SMacro *m;
1447 if (ctx) {
1448 smtbl = &ctx->localmac;
1449 } else if (name[0] == '%' && name[1] == '$') {
1450 if (cstk)
1451 ctx = get_ctx(name, &name, false);
1452 if (!ctx)
1453 return false; /* got to return _something_ */
1454 smtbl = &ctx->localmac;
1455 } else {
1456 smtbl = &smacros;
1458 m = (SMacro *) hash_findix(smtbl, name);
1460 while (m) {
1461 if (!mstrcmp(m->name, name, m->casesense && nocase) &&
1462 (nparam <= 0 || m->nparam == 0 || nparam == (int) m->nparam)) {
1463 if (defn) {
1464 if (nparam == (int) m->nparam || nparam == -1)
1465 *defn = m;
1466 else
1467 *defn = NULL;
1469 return true;
1471 m = m->next;
1474 return false;
1478 * Count and mark off the parameters in a multi-line macro call.
1479 * This is called both from within the multi-line macro expansion
1480 * code, and also to mark off the default parameters when provided
1481 * in a %macro definition line.
1483 static void count_mmac_params(Token * t, int *nparam, Token *** params)
1485 int paramsize, brace;
1487 *nparam = paramsize = 0;
1488 *params = NULL;
1489 while (t) {
1490 /* +1: we need space for the final NULL */
1491 if (*nparam+1 >= paramsize) {
1492 paramsize += PARAM_DELTA;
1493 *params = nasm_realloc(*params, sizeof(**params) * paramsize);
1495 skip_white_(t);
1496 brace = false;
1497 if (tok_is_(t, "{"))
1498 brace = true;
1499 (*params)[(*nparam)++] = t;
1500 while (tok_isnt_(t, brace ? "}" : ","))
1501 t = t->next;
1502 if (t) { /* got a comma/brace */
1503 t = t->next;
1504 if (brace) {
1506 * Now we've found the closing brace, look further
1507 * for the comma.
1509 skip_white_(t);
1510 if (tok_isnt_(t, ",")) {
1511 error(ERR_NONFATAL,
1512 "braces do not enclose all of macro parameter");
1513 while (tok_isnt_(t, ","))
1514 t = t->next;
1516 if (t)
1517 t = t->next; /* eat the comma */
1524 * Determine whether one of the various `if' conditions is true or
1525 * not.
1527 * We must free the tline we get passed.
1529 static bool if_condition(Token * tline, enum preproc_token ct)
1531 enum pp_conditional i = PP_COND(ct);
1532 bool j;
1533 Token *t, *tt, **tptr, *origline;
1534 struct tokenval tokval;
1535 expr *evalresult;
1536 enum pp_token_type needtype;
1538 origline = tline;
1540 switch (i) {
1541 case PPC_IFCTX:
1542 j = false; /* have we matched yet? */
1543 while (true) {
1544 skip_white_(tline);
1545 if (!tline)
1546 break;
1547 if (tline->type != TOK_ID) {
1548 error(ERR_NONFATAL,
1549 "`%s' expects context identifiers", pp_directives[ct]);
1550 free_tlist(origline);
1551 return -1;
1553 if (cstk && cstk->name && !nasm_stricmp(tline->text, cstk->name))
1554 j = true;
1555 tline = tline->next;
1557 break;
1559 case PPC_IFDEF:
1560 j = false; /* have we matched yet? */
1561 while (tline) {
1562 skip_white_(tline);
1563 if (!tline || (tline->type != TOK_ID &&
1564 (tline->type != TOK_PREPROC_ID ||
1565 tline->text[1] != '$'))) {
1566 error(ERR_NONFATAL,
1567 "`%s' expects macro identifiers", pp_directives[ct]);
1568 goto fail;
1570 if (smacro_defined(NULL, tline->text, 0, NULL, true))
1571 j = true;
1572 tline = tline->next;
1574 break;
1576 case PPC_IFIDN:
1577 case PPC_IFIDNI:
1578 tline = expand_smacro(tline);
1579 t = tt = tline;
1580 while (tok_isnt_(tt, ","))
1581 tt = tt->next;
1582 if (!tt) {
1583 error(ERR_NONFATAL,
1584 "`%s' expects two comma-separated arguments",
1585 pp_directives[ct]);
1586 goto fail;
1588 tt = tt->next;
1589 j = true; /* assume equality unless proved not */
1590 while ((t->type != TOK_OTHER || strcmp(t->text, ",")) && tt) {
1591 if (tt->type == TOK_OTHER && !strcmp(tt->text, ",")) {
1592 error(ERR_NONFATAL, "`%s': more than one comma on line",
1593 pp_directives[ct]);
1594 goto fail;
1596 if (t->type == TOK_WHITESPACE) {
1597 t = t->next;
1598 continue;
1600 if (tt->type == TOK_WHITESPACE) {
1601 tt = tt->next;
1602 continue;
1604 if (tt->type != t->type) {
1605 j = false; /* found mismatching tokens */
1606 break;
1608 /* When comparing strings, need to unquote them first */
1609 if (t->type == TOK_STRING) {
1610 size_t l1 = nasm_unquote(t->text, NULL);
1611 size_t l2 = nasm_unquote(tt->text, NULL);
1613 if (l1 != l2) {
1614 j = false;
1615 break;
1617 if (mmemcmp(t->text, tt->text, l1, i == PPC_IFIDN)) {
1618 j = false;
1619 break;
1621 } else if (mstrcmp(tt->text, t->text, i == PPC_IFIDN) != 0) {
1622 j = false; /* found mismatching tokens */
1623 break;
1626 t = t->next;
1627 tt = tt->next;
1629 if ((t->type != TOK_OTHER || strcmp(t->text, ",")) || tt)
1630 j = false; /* trailing gunk on one end or other */
1631 break;
1633 case PPC_IFMACRO:
1635 bool found = false;
1636 MMacro searching, *mmac;
1638 skip_white_(tline);
1639 tline = expand_id(tline);
1640 if (!tok_type_(tline, TOK_ID)) {
1641 error(ERR_NONFATAL,
1642 "`%s' expects a macro name", pp_directives[ct]);
1643 goto fail;
1645 searching.name = nasm_strdup(tline->text);
1646 searching.casesense = true;
1647 searching.plus = false;
1648 searching.nolist = false;
1649 searching.in_progress = 0;
1650 searching.rep_nest = NULL;
1651 searching.nparam_min = 0;
1652 searching.nparam_max = INT_MAX;
1653 tline = expand_smacro(tline->next);
1654 skip_white_(tline);
1655 if (!tline) {
1656 } else if (!tok_type_(tline, TOK_NUMBER)) {
1657 error(ERR_NONFATAL,
1658 "`%s' expects a parameter count or nothing",
1659 pp_directives[ct]);
1660 } else {
1661 searching.nparam_min = searching.nparam_max =
1662 readnum(tline->text, &j);
1663 if (j)
1664 error(ERR_NONFATAL,
1665 "unable to parse parameter count `%s'",
1666 tline->text);
1668 if (tline && tok_is_(tline->next, "-")) {
1669 tline = tline->next->next;
1670 if (tok_is_(tline, "*"))
1671 searching.nparam_max = INT_MAX;
1672 else if (!tok_type_(tline, TOK_NUMBER))
1673 error(ERR_NONFATAL,
1674 "`%s' expects a parameter count after `-'",
1675 pp_directives[ct]);
1676 else {
1677 searching.nparam_max = readnum(tline->text, &j);
1678 if (j)
1679 error(ERR_NONFATAL,
1680 "unable to parse parameter count `%s'",
1681 tline->text);
1682 if (searching.nparam_min > searching.nparam_max)
1683 error(ERR_NONFATAL,
1684 "minimum parameter count exceeds maximum");
1687 if (tline && tok_is_(tline->next, "+")) {
1688 tline = tline->next;
1689 searching.plus = true;
1691 mmac = (MMacro *) hash_findix(&mmacros, searching.name);
1692 while (mmac) {
1693 if (!strcmp(mmac->name, searching.name) &&
1694 (mmac->nparam_min <= searching.nparam_max
1695 || searching.plus)
1696 && (searching.nparam_min <= mmac->nparam_max
1697 || mmac->plus)) {
1698 found = true;
1699 break;
1701 mmac = mmac->next;
1703 if(tline && tline->next)
1704 error(ERR_WARNING|ERR_PASS1,
1705 "trailing garbage after %%ifmacro ignored");
1706 nasm_free(searching.name);
1707 j = found;
1708 break;
1711 case PPC_IFID:
1712 needtype = TOK_ID;
1713 goto iftype;
1714 case PPC_IFNUM:
1715 needtype = TOK_NUMBER;
1716 goto iftype;
1717 case PPC_IFSTR:
1718 needtype = TOK_STRING;
1719 goto iftype;
1721 iftype:
1722 t = tline = expand_smacro(tline);
1724 while (tok_type_(t, TOK_WHITESPACE) ||
1725 (needtype == TOK_NUMBER &&
1726 tok_type_(t, TOK_OTHER) &&
1727 (t->text[0] == '-' || t->text[0] == '+') &&
1728 !t->text[1]))
1729 t = t->next;
1731 j = tok_type_(t, needtype);
1732 break;
1734 case PPC_IFTOKEN:
1735 t = tline = expand_smacro(tline);
1736 while (tok_type_(t, TOK_WHITESPACE))
1737 t = t->next;
1739 j = false;
1740 if (t) {
1741 t = t->next; /* Skip the actual token */
1742 while (tok_type_(t, TOK_WHITESPACE))
1743 t = t->next;
1744 j = !t; /* Should be nothing left */
1746 break;
1748 case PPC_IFEMPTY:
1749 t = tline = expand_smacro(tline);
1750 while (tok_type_(t, TOK_WHITESPACE))
1751 t = t->next;
1753 j = !t; /* Should be empty */
1754 break;
1756 case PPC_IF:
1757 t = tline = expand_smacro(tline);
1758 tptr = &t;
1759 tokval.t_type = TOKEN_INVALID;
1760 evalresult = evaluate(ppscan, tptr, &tokval,
1761 NULL, pass | CRITICAL, error, NULL);
1762 if (!evalresult)
1763 return -1;
1764 if (tokval.t_type)
1765 error(ERR_WARNING|ERR_PASS1,
1766 "trailing garbage after expression ignored");
1767 if (!is_simple(evalresult)) {
1768 error(ERR_NONFATAL,
1769 "non-constant value given to `%s'", pp_directives[ct]);
1770 goto fail;
1772 j = reloc_value(evalresult) != 0;
1773 break;
1775 default:
1776 error(ERR_FATAL,
1777 "preprocessor directive `%s' not yet implemented",
1778 pp_directives[ct]);
1779 goto fail;
1782 free_tlist(origline);
1783 return j ^ PP_NEGATIVE(ct);
1785 fail:
1786 free_tlist(origline);
1787 return -1;
1791 * Common code for defining an smacro
1793 static bool define_smacro(Context *ctx, const char *mname, bool casesense,
1794 int nparam, Token *expansion)
1796 SMacro *smac, **smhead;
1797 struct hash_table *smtbl;
1799 if (smacro_defined(ctx, mname, nparam, &smac, casesense)) {
1800 if (!smac) {
1801 error(ERR_WARNING|ERR_PASS1,
1802 "single-line macro `%s' defined both with and"
1803 " without parameters", mname);
1805 /* Some instances of the old code considered this a failure,
1806 some others didn't. What is the right thing to do here? */
1807 free_tlist(expansion);
1808 return false; /* Failure */
1809 } else {
1811 * We're redefining, so we have to take over an
1812 * existing SMacro structure. This means freeing
1813 * what was already in it.
1815 nasm_free(smac->name);
1816 free_tlist(smac->expansion);
1818 } else {
1819 smtbl = ctx ? &ctx->localmac : &smacros;
1820 smhead = (SMacro **) hash_findi_add(smtbl, mname);
1821 smac = nasm_malloc(sizeof(SMacro));
1822 smac->next = *smhead;
1823 *smhead = smac;
1825 smac->name = nasm_strdup(mname);
1826 smac->casesense = casesense;
1827 smac->nparam = nparam;
1828 smac->expansion = expansion;
1829 smac->in_progress = false;
1830 return true; /* Success */
1834 * Undefine an smacro
1836 static void undef_smacro(Context *ctx, const char *mname)
1838 SMacro **smhead, *s, **sp;
1839 struct hash_table *smtbl;
1841 smtbl = ctx ? &ctx->localmac : &smacros;
1842 smhead = (SMacro **)hash_findi(smtbl, mname, NULL);
1844 if (smhead) {
1846 * We now have a macro name... go hunt for it.
1848 sp = smhead;
1849 while ((s = *sp) != NULL) {
1850 if (!mstrcmp(s->name, mname, s->casesense)) {
1851 *sp = s->next;
1852 nasm_free(s->name);
1853 free_tlist(s->expansion);
1854 nasm_free(s);
1855 } else {
1856 sp = &s->next;
1863 * Parse a mmacro specification.
1865 static bool parse_mmacro_spec(Token *tline, MMacro *def, const char *directive)
1867 bool err;
1869 tline = tline->next;
1870 skip_white_(tline);
1871 tline = expand_id(tline);
1872 if (!tok_type_(tline, TOK_ID)) {
1873 error(ERR_NONFATAL, "`%s' expects a macro name", directive);
1874 return false;
1877 def->name = nasm_strdup(tline->text);
1878 def->plus = false;
1879 def->nolist = false;
1880 def->in_progress = 0;
1881 def->rep_nest = NULL;
1882 def->nparam_min = 0;
1883 def->nparam_max = 0;
1885 tline = expand_smacro(tline->next);
1886 skip_white_(tline);
1887 if (!tok_type_(tline, TOK_NUMBER)) {
1888 error(ERR_NONFATAL, "`%s' expects a parameter count", directive);
1889 } else {
1890 def->nparam_min = def->nparam_max =
1891 readnum(tline->text, &err);
1892 if (err)
1893 error(ERR_NONFATAL,
1894 "unable to parse parameter count `%s'", tline->text);
1896 if (tline && tok_is_(tline->next, "-")) {
1897 tline = tline->next->next;
1898 if (tok_is_(tline, "*")) {
1899 def->nparam_max = INT_MAX;
1900 } else if (!tok_type_(tline, TOK_NUMBER)) {
1901 error(ERR_NONFATAL,
1902 "`%s' expects a parameter count after `-'", directive);
1903 } else {
1904 def->nparam_max = readnum(tline->text, &err);
1905 if (err) {
1906 error(ERR_NONFATAL, "unable to parse parameter count `%s'",
1907 tline->text);
1909 if (def->nparam_min > def->nparam_max) {
1910 error(ERR_NONFATAL, "minimum parameter count exceeds maximum");
1914 if (tline && tok_is_(tline->next, "+")) {
1915 tline = tline->next;
1916 def->plus = true;
1918 if (tline && tok_type_(tline->next, TOK_ID) &&
1919 !nasm_stricmp(tline->next->text, ".nolist")) {
1920 tline = tline->next;
1921 def->nolist = true;
1925 * Handle default parameters.
1927 if (tline && tline->next) {
1928 def->dlist = tline->next;
1929 tline->next = NULL;
1930 count_mmac_params(def->dlist, &def->ndefs, &def->defaults);
1931 } else {
1932 def->dlist = NULL;
1933 def->defaults = NULL;
1935 def->expansion = NULL;
1937 if(def->defaults &&
1938 def->ndefs > def->nparam_max - def->nparam_min &&
1939 !def->plus)
1940 error(ERR_WARNING|ERR_PASS1|ERR_WARN_MDP,
1941 "too many default macro parameters");
1943 return true;
1948 * Decode a size directive
1950 static int parse_size(const char *str) {
1951 static const char *size_names[] =
1952 { "byte", "dword", "oword", "qword", "tword", "word", "yword" };
1953 static const int sizes[] =
1954 { 0, 1, 4, 16, 8, 10, 2, 32 };
1956 return sizes[bsii(str, size_names, elements(size_names))+1];
1960 * find and process preprocessor directive in passed line
1961 * Find out if a line contains a preprocessor directive, and deal
1962 * with it if so.
1964 * If a directive _is_ found, it is the responsibility of this routine
1965 * (and not the caller) to free_tlist() the line.
1967 * @param tline a pointer to the current tokeninzed line linked list
1968 * @return DIRECTIVE_FOUND or NO_DIRECTIVE_FOUND
1971 static int do_directive(Token * tline)
1973 enum preproc_token i;
1974 int j;
1975 bool err;
1976 int nparam;
1977 bool nolist;
1978 bool casesense;
1979 int k, m;
1980 int offset;
1981 char *p, *pp;
1982 const char *mname;
1983 Include *inc;
1984 Context *ctx;
1985 Cond *cond;
1986 MMacro *mmac, **mmhead;
1987 Token *t, *tt, *param_start, *macro_start, *last, **tptr, *origline;
1988 Line *l;
1989 struct tokenval tokval;
1990 expr *evalresult;
1991 MMacro *tmp_defining; /* Used when manipulating rep_nest */
1992 int64_t count;
1993 size_t len;
1994 int severity;
1996 origline = tline;
1998 skip_white_(tline);
1999 if (!tline || !tok_type_(tline, TOK_PREPROC_ID) ||
2000 (tline->text[1] == '%' || tline->text[1] == '$'
2001 || tline->text[1] == '!'))
2002 return NO_DIRECTIVE_FOUND;
2004 i = pp_token_hash(tline->text);
2007 * If we're in a non-emitting branch of a condition construct,
2008 * or walking to the end of an already terminated %rep block,
2009 * we should ignore all directives except for condition
2010 * directives.
2012 if (((istk->conds && !emitting(istk->conds->state)) ||
2013 (istk->mstk && !istk->mstk->in_progress)) && !is_condition(i)) {
2014 return NO_DIRECTIVE_FOUND;
2018 * If we're defining a macro or reading a %rep block, we should
2019 * ignore all directives except for %macro/%imacro (which nest),
2020 * %endm/%endmacro, and (only if we're in a %rep block) %endrep.
2021 * If we're in a %rep block, another %rep nests, so should be let through.
2023 if (defining && i != PP_MACRO && i != PP_IMACRO &&
2024 i != PP_ENDMACRO && i != PP_ENDM &&
2025 (defining->name || (i != PP_ENDREP && i != PP_REP))) {
2026 return NO_DIRECTIVE_FOUND;
2029 if (defining) {
2030 if (i == PP_MACRO || i == PP_IMACRO) {
2031 nested_mac_count++;
2032 return NO_DIRECTIVE_FOUND;
2033 } else if (nested_mac_count > 0) {
2034 if (i == PP_ENDMACRO) {
2035 nested_mac_count--;
2036 return NO_DIRECTIVE_FOUND;
2039 if (!defining->name) {
2040 if (i == PP_REP) {
2041 nested_rep_count++;
2042 return NO_DIRECTIVE_FOUND;
2043 } else if (nested_rep_count > 0) {
2044 if (i == PP_ENDREP) {
2045 nested_rep_count--;
2046 return NO_DIRECTIVE_FOUND;
2052 switch (i) {
2053 case PP_INVALID:
2054 error(ERR_NONFATAL, "unknown preprocessor directive `%s'",
2055 tline->text);
2056 return NO_DIRECTIVE_FOUND; /* didn't get it */
2058 case PP_STACKSIZE:
2059 /* Directive to tell NASM what the default stack size is. The
2060 * default is for a 16-bit stack, and this can be overriden with
2061 * %stacksize large.
2062 * the following form:
2064 * ARG arg1:WORD, arg2:DWORD, arg4:QWORD
2066 tline = tline->next;
2067 if (tline && tline->type == TOK_WHITESPACE)
2068 tline = tline->next;
2069 if (!tline || tline->type != TOK_ID) {
2070 error(ERR_NONFATAL, "`%%stacksize' missing size parameter");
2071 free_tlist(origline);
2072 return DIRECTIVE_FOUND;
2074 if (nasm_stricmp(tline->text, "flat") == 0) {
2075 /* All subsequent ARG directives are for a 32-bit stack */
2076 StackSize = 4;
2077 StackPointer = "ebp";
2078 ArgOffset = 8;
2079 LocalOffset = 0;
2080 } else if (nasm_stricmp(tline->text, "flat64") == 0) {
2081 /* All subsequent ARG directives are for a 64-bit stack */
2082 StackSize = 8;
2083 StackPointer = "rbp";
2084 ArgOffset = 8;
2085 LocalOffset = 0;
2086 } else if (nasm_stricmp(tline->text, "large") == 0) {
2087 /* All subsequent ARG directives are for a 16-bit stack,
2088 * far function call.
2090 StackSize = 2;
2091 StackPointer = "bp";
2092 ArgOffset = 4;
2093 LocalOffset = 0;
2094 } else if (nasm_stricmp(tline->text, "small") == 0) {
2095 /* All subsequent ARG directives are for a 16-bit stack,
2096 * far function call. We don't support near functions.
2098 StackSize = 2;
2099 StackPointer = "bp";
2100 ArgOffset = 6;
2101 LocalOffset = 0;
2102 } else {
2103 error(ERR_NONFATAL, "`%%stacksize' invalid size type");
2104 free_tlist(origline);
2105 return DIRECTIVE_FOUND;
2107 free_tlist(origline);
2108 return DIRECTIVE_FOUND;
2110 case PP_ARG:
2111 /* TASM like ARG directive to define arguments to functions, in
2112 * the following form:
2114 * ARG arg1:WORD, arg2:DWORD, arg4:QWORD
2116 offset = ArgOffset;
2117 do {
2118 char *arg, directive[256];
2119 int size = StackSize;
2121 /* Find the argument name */
2122 tline = tline->next;
2123 if (tline && tline->type == TOK_WHITESPACE)
2124 tline = tline->next;
2125 if (!tline || tline->type != TOK_ID) {
2126 error(ERR_NONFATAL, "`%%arg' missing argument parameter");
2127 free_tlist(origline);
2128 return DIRECTIVE_FOUND;
2130 arg = tline->text;
2132 /* Find the argument size type */
2133 tline = tline->next;
2134 if (!tline || tline->type != TOK_OTHER
2135 || tline->text[0] != ':') {
2136 error(ERR_NONFATAL,
2137 "Syntax error processing `%%arg' directive");
2138 free_tlist(origline);
2139 return DIRECTIVE_FOUND;
2141 tline = tline->next;
2142 if (!tline || tline->type != TOK_ID) {
2143 error(ERR_NONFATAL, "`%%arg' missing size type parameter");
2144 free_tlist(origline);
2145 return DIRECTIVE_FOUND;
2148 /* Allow macro expansion of type parameter */
2149 tt = tokenize(tline->text);
2150 tt = expand_smacro(tt);
2151 size = parse_size(tt->text);
2152 if (!size) {
2153 error(ERR_NONFATAL,
2154 "Invalid size type for `%%arg' missing directive");
2155 free_tlist(tt);
2156 free_tlist(origline);
2157 return DIRECTIVE_FOUND;
2159 free_tlist(tt);
2161 /* Round up to even stack slots */
2162 size = (size+StackSize-1) & ~(StackSize-1);
2164 /* Now define the macro for the argument */
2165 snprintf(directive, sizeof(directive), "%%define %s (%s+%d)",
2166 arg, StackPointer, offset);
2167 do_directive(tokenize(directive));
2168 offset += size;
2170 /* Move to the next argument in the list */
2171 tline = tline->next;
2172 if (tline && tline->type == TOK_WHITESPACE)
2173 tline = tline->next;
2174 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2175 ArgOffset = offset;
2176 free_tlist(origline);
2177 return DIRECTIVE_FOUND;
2179 case PP_LOCAL:
2180 /* TASM like LOCAL directive to define local variables for a
2181 * function, in the following form:
2183 * LOCAL local1:WORD, local2:DWORD, local4:QWORD = LocalSize
2185 * The '= LocalSize' at the end is ignored by NASM, but is
2186 * required by TASM to define the local parameter size (and used
2187 * by the TASM macro package).
2189 offset = LocalOffset;
2190 do {
2191 char *local, directive[256];
2192 int size = StackSize;
2194 /* Find the argument name */
2195 tline = tline->next;
2196 if (tline && tline->type == TOK_WHITESPACE)
2197 tline = tline->next;
2198 if (!tline || tline->type != TOK_ID) {
2199 error(ERR_NONFATAL,
2200 "`%%local' missing argument parameter");
2201 free_tlist(origline);
2202 return DIRECTIVE_FOUND;
2204 local = tline->text;
2206 /* Find the argument size type */
2207 tline = tline->next;
2208 if (!tline || tline->type != TOK_OTHER
2209 || tline->text[0] != ':') {
2210 error(ERR_NONFATAL,
2211 "Syntax error processing `%%local' directive");
2212 free_tlist(origline);
2213 return DIRECTIVE_FOUND;
2215 tline = tline->next;
2216 if (!tline || tline->type != TOK_ID) {
2217 error(ERR_NONFATAL,
2218 "`%%local' missing size type parameter");
2219 free_tlist(origline);
2220 return DIRECTIVE_FOUND;
2223 /* Allow macro expansion of type parameter */
2224 tt = tokenize(tline->text);
2225 tt = expand_smacro(tt);
2226 size = parse_size(tt->text);
2227 if (!size) {
2228 error(ERR_NONFATAL,
2229 "Invalid size type for `%%local' missing directive");
2230 free_tlist(tt);
2231 free_tlist(origline);
2232 return DIRECTIVE_FOUND;
2234 free_tlist(tt);
2236 /* Round up to even stack slots */
2237 size = (size+StackSize-1) & ~(StackSize-1);
2239 offset += size; /* Negative offset, increment before */
2241 /* Now define the macro for the argument */
2242 snprintf(directive, sizeof(directive), "%%define %s (%s-%d)",
2243 local, StackPointer, offset);
2244 do_directive(tokenize(directive));
2246 /* Now define the assign to setup the enter_c macro correctly */
2247 snprintf(directive, sizeof(directive),
2248 "%%assign %%$localsize %%$localsize+%d", size);
2249 do_directive(tokenize(directive));
2251 /* Move to the next argument in the list */
2252 tline = tline->next;
2253 if (tline && tline->type == TOK_WHITESPACE)
2254 tline = tline->next;
2255 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2256 LocalOffset = offset;
2257 free_tlist(origline);
2258 return DIRECTIVE_FOUND;
2260 case PP_CLEAR:
2261 if (tline->next)
2262 error(ERR_WARNING|ERR_PASS1,
2263 "trailing garbage after `%%clear' ignored");
2264 free_macros();
2265 init_macros();
2266 free_tlist(origline);
2267 return DIRECTIVE_FOUND;
2269 case PP_DEPEND:
2270 t = tline->next = expand_smacro(tline->next);
2271 skip_white_(t);
2272 if (!t || (t->type != TOK_STRING &&
2273 t->type != TOK_INTERNAL_STRING)) {
2274 error(ERR_NONFATAL, "`%%depend' expects a file name");
2275 free_tlist(origline);
2276 return DIRECTIVE_FOUND; /* but we did _something_ */
2278 if (t->next)
2279 error(ERR_WARNING|ERR_PASS1,
2280 "trailing garbage after `%%depend' ignored");
2281 p = t->text;
2282 if (t->type != TOK_INTERNAL_STRING)
2283 nasm_unquote(p, NULL);
2284 if (dephead && !in_list(*dephead, p)) {
2285 StrList *sl = nasm_malloc(strlen(p)+1+sizeof sl->next);
2286 sl->next = NULL;
2287 strcpy(sl->str, p);
2288 *deptail = sl;
2289 deptail = &sl->next;
2291 free_tlist(origline);
2292 return DIRECTIVE_FOUND;
2294 case PP_INCLUDE:
2295 t = tline->next = expand_smacro(tline->next);
2296 skip_white_(t);
2298 if (!t || (t->type != TOK_STRING &&
2299 t->type != TOK_INTERNAL_STRING)) {
2300 error(ERR_NONFATAL, "`%%include' expects a file name");
2301 free_tlist(origline);
2302 return DIRECTIVE_FOUND; /* but we did _something_ */
2304 if (t->next)
2305 error(ERR_WARNING|ERR_PASS1,
2306 "trailing garbage after `%%include' ignored");
2307 p = t->text;
2308 if (t->type != TOK_INTERNAL_STRING)
2309 nasm_unquote(p, NULL);
2310 inc = nasm_malloc(sizeof(Include));
2311 inc->next = istk;
2312 inc->conds = NULL;
2313 inc->fp = inc_fopen(p, dephead, &deptail, pass == 0);
2314 if (!inc->fp) {
2315 /* -MG given but file not found */
2316 nasm_free(inc);
2317 } else {
2318 inc->fname = src_set_fname(nasm_strdup(p));
2319 inc->lineno = src_set_linnum(0);
2320 inc->lineinc = 1;
2321 inc->expansion = NULL;
2322 inc->mstk = NULL;
2323 istk = inc;
2324 list->uplevel(LIST_INCLUDE);
2326 free_tlist(origline);
2327 return DIRECTIVE_FOUND;
2329 case PP_USE:
2331 static macros_t *use_pkg;
2332 const char *pkg_macro;
2334 tline = tline->next;
2335 skip_white_(tline);
2336 tline = expand_id(tline);
2338 if (!tline || (tline->type != TOK_STRING &&
2339 tline->type != TOK_INTERNAL_STRING &&
2340 tline->type != TOK_ID)) {
2341 error(ERR_NONFATAL, "`%%use' expects a package name");
2342 free_tlist(origline);
2343 return DIRECTIVE_FOUND; /* but we did _something_ */
2345 if (tline->next)
2346 error(ERR_WARNING|ERR_PASS1,
2347 "trailing garbage after `%%use' ignored");
2348 if (tline->type == TOK_STRING)
2349 nasm_unquote(tline->text, NULL);
2350 use_pkg = nasm_stdmac_find_package(tline->text);
2351 if (!use_pkg)
2352 error(ERR_NONFATAL, "unknown `%%use' package: %s", tline->text);
2353 /* The first string will be <%define>__USE_*__ */
2354 pkg_macro = (char *)use_pkg + 1;
2355 if (!smacro_defined(NULL, pkg_macro, 0, NULL, true)) {
2356 /* Not already included, go ahead and include it */
2357 stdmacpos = use_pkg;
2359 free_tlist(origline);
2360 return DIRECTIVE_FOUND;
2362 case PP_PUSH:
2363 case PP_REPL:
2364 case PP_POP:
2365 tline = tline->next;
2366 skip_white_(tline);
2367 tline = expand_id(tline);
2368 if (tline) {
2369 if (!tok_type_(tline, TOK_ID)) {
2370 error(ERR_NONFATAL, "`%s' expects a context identifier",
2371 pp_directives[i]);
2372 free_tlist(origline);
2373 return DIRECTIVE_FOUND; /* but we did _something_ */
2375 if (tline->next)
2376 error(ERR_WARNING|ERR_PASS1,
2377 "trailing garbage after `%s' ignored",
2378 pp_directives[i]);
2379 p = nasm_strdup(tline->text);
2380 } else {
2381 p = NULL; /* Anonymous */
2384 if (i == PP_PUSH) {
2385 ctx = nasm_malloc(sizeof(Context));
2386 ctx->next = cstk;
2387 hash_init(&ctx->localmac, HASH_SMALL);
2388 ctx->name = p;
2389 ctx->number = unique++;
2390 cstk = ctx;
2391 } else {
2392 /* %pop or %repl */
2393 if (!cstk) {
2394 error(ERR_NONFATAL, "`%s': context stack is empty",
2395 pp_directives[i]);
2396 } else if (i == PP_POP) {
2397 if (p && (!cstk->name || nasm_stricmp(p, cstk->name)))
2398 error(ERR_NONFATAL, "`%%pop' in wrong context: %s, "
2399 "expected %s",
2400 cstk->name ? cstk->name : "anonymous", p);
2401 else
2402 ctx_pop();
2403 } else {
2404 /* i == PP_REPL */
2405 nasm_free(cstk->name);
2406 cstk->name = p;
2407 p = NULL;
2409 nasm_free(p);
2411 free_tlist(origline);
2412 return DIRECTIVE_FOUND;
2413 case PP_FATAL:
2414 severity = ERR_FATAL|ERR_NO_SEVERITY;
2415 goto issue_error;
2416 case PP_ERROR:
2417 severity = ERR_NONFATAL|ERR_NO_SEVERITY;
2418 goto issue_error;
2419 case PP_WARNING:
2420 severity = ERR_WARNING|ERR_NO_SEVERITY|ERR_WARN_USER;
2421 goto issue_error;
2423 issue_error:
2425 /* Only error out if this is the final pass */
2426 if (pass != 2 && i != PP_FATAL)
2427 return DIRECTIVE_FOUND;
2429 tline->next = expand_smacro(tline->next);
2430 tline = tline->next;
2431 skip_white_(tline);
2432 t = tline ? tline->next : NULL;
2433 skip_white_(t);
2434 if (tok_type_(tline, TOK_STRING) && !t) {
2435 /* The line contains only a quoted string */
2436 p = tline->text;
2437 nasm_unquote(p, NULL);
2438 error(severity, "%s: %s", pp_directives[i], p);
2439 } else {
2440 /* Not a quoted string, or more than a quoted string */
2441 p = detoken(tline, false);
2442 error(severity, "%s: %s", pp_directives[i], p);
2443 nasm_free(p);
2445 free_tlist(origline);
2446 return DIRECTIVE_FOUND;
2449 CASE_PP_IF:
2450 if (istk->conds && !emitting(istk->conds->state))
2451 j = COND_NEVER;
2452 else {
2453 j = if_condition(tline->next, i);
2454 tline->next = NULL; /* it got freed */
2455 j = j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2457 cond = nasm_malloc(sizeof(Cond));
2458 cond->next = istk->conds;
2459 cond->state = j;
2460 istk->conds = cond;
2461 free_tlist(origline);
2462 return DIRECTIVE_FOUND;
2464 CASE_PP_ELIF:
2465 if (!istk->conds)
2466 error(ERR_FATAL, "`%s': no matching `%%if'", pp_directives[i]);
2467 switch(istk->conds->state) {
2468 case COND_IF_TRUE:
2469 istk->conds->state = COND_DONE;
2470 break;
2472 case COND_DONE:
2473 case COND_NEVER:
2474 break;
2476 case COND_ELSE_TRUE:
2477 case COND_ELSE_FALSE:
2478 error_precond(ERR_WARNING|ERR_PASS1,
2479 "`%%elif' after `%%else' ignored");
2480 istk->conds->state = COND_NEVER;
2481 break;
2483 case COND_IF_FALSE:
2485 * IMPORTANT: In the case of %if, we will already have
2486 * called expand_mmac_params(); however, if we're
2487 * processing an %elif we must have been in a
2488 * non-emitting mode, which would have inhibited
2489 * the normal invocation of expand_mmac_params().
2490 * Therefore, we have to do it explicitly here.
2492 j = if_condition(expand_mmac_params(tline->next), i);
2493 tline->next = NULL; /* it got freed */
2494 istk->conds->state =
2495 j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2496 break;
2498 free_tlist(origline);
2499 return DIRECTIVE_FOUND;
2501 case PP_ELSE:
2502 if (tline->next)
2503 error_precond(ERR_WARNING|ERR_PASS1,
2504 "trailing garbage after `%%else' ignored");
2505 if (!istk->conds)
2506 error(ERR_FATAL, "`%%else': no matching `%%if'");
2507 switch(istk->conds->state) {
2508 case COND_IF_TRUE:
2509 case COND_DONE:
2510 istk->conds->state = COND_ELSE_FALSE;
2511 break;
2513 case COND_NEVER:
2514 break;
2516 case COND_IF_FALSE:
2517 istk->conds->state = COND_ELSE_TRUE;
2518 break;
2520 case COND_ELSE_TRUE:
2521 case COND_ELSE_FALSE:
2522 error_precond(ERR_WARNING|ERR_PASS1,
2523 "`%%else' after `%%else' ignored.");
2524 istk->conds->state = COND_NEVER;
2525 break;
2527 free_tlist(origline);
2528 return DIRECTIVE_FOUND;
2530 case PP_ENDIF:
2531 if (tline->next)
2532 error_precond(ERR_WARNING|ERR_PASS1,
2533 "trailing garbage after `%%endif' ignored");
2534 if (!istk->conds)
2535 error(ERR_FATAL, "`%%endif': no matching `%%if'");
2536 cond = istk->conds;
2537 istk->conds = cond->next;
2538 nasm_free(cond);
2539 free_tlist(origline);
2540 return DIRECTIVE_FOUND;
2542 case PP_MACRO:
2543 case PP_IMACRO:
2544 if (defining) {
2545 error(ERR_FATAL,
2546 "`%%%smacro': already defining a macro",
2547 (i == PP_IMACRO ? "i" : ""));
2548 return DIRECTIVE_FOUND;
2550 defining = nasm_malloc(sizeof(MMacro));
2551 defining->casesense = (i == PP_MACRO);
2552 if (!parse_mmacro_spec(tline, defining, pp_directives[i])) {
2553 nasm_free(defining);
2554 defining = NULL;
2555 return DIRECTIVE_FOUND;
2558 mmac = (MMacro *) hash_findix(&mmacros, defining->name);
2559 while (mmac) {
2560 if (!strcmp(mmac->name, defining->name) &&
2561 (mmac->nparam_min <= defining->nparam_max
2562 || defining->plus)
2563 && (defining->nparam_min <= mmac->nparam_max
2564 || mmac->plus)) {
2565 error(ERR_WARNING|ERR_PASS1,
2566 "redefining multi-line macro `%s'", defining->name);
2567 return DIRECTIVE_FOUND;
2569 mmac = mmac->next;
2571 free_tlist(origline);
2572 return DIRECTIVE_FOUND;
2574 case PP_ENDM:
2575 case PP_ENDMACRO:
2576 if (! (defining && defining->name)) {
2577 error(ERR_NONFATAL, "`%s': not defining a macro", tline->text);
2578 return DIRECTIVE_FOUND;
2580 mmhead = (MMacro **) hash_findi_add(&mmacros, defining->name);
2581 defining->next = *mmhead;
2582 *mmhead = defining;
2583 defining = NULL;
2584 free_tlist(origline);
2585 return DIRECTIVE_FOUND;
2587 case PP_UNMACRO:
2588 case PP_UNIMACRO:
2590 MMacro **mmac_p;
2591 MMacro spec;
2593 spec.casesense = (i == PP_UNMACRO);
2594 if (!parse_mmacro_spec(tline, &spec, pp_directives[i])) {
2595 return DIRECTIVE_FOUND;
2597 mmac_p = (MMacro **) hash_findi(&mmacros, spec.name, NULL);
2598 while (mmac_p && *mmac_p) {
2599 mmac = *mmac_p;
2600 if (mmac->casesense == spec.casesense &&
2601 !mstrcmp(mmac->name, spec.name, spec.casesense) &&
2602 mmac->nparam_min == spec.nparam_min &&
2603 mmac->nparam_max == spec.nparam_max &&
2604 mmac->plus == spec.plus) {
2605 *mmac_p = mmac->next;
2606 free_mmacro(mmac);
2607 } else {
2608 mmac_p = &mmac->next;
2611 free_tlist(origline);
2612 free_tlist(spec.dlist);
2613 return DIRECTIVE_FOUND;
2616 case PP_ROTATE:
2617 if (tline->next && tline->next->type == TOK_WHITESPACE)
2618 tline = tline->next;
2619 if (tline->next == NULL) {
2620 free_tlist(origline);
2621 error(ERR_NONFATAL, "`%%rotate' missing rotate count");
2622 return DIRECTIVE_FOUND;
2624 t = expand_smacro(tline->next);
2625 tline->next = NULL;
2626 free_tlist(origline);
2627 tline = t;
2628 tptr = &t;
2629 tokval.t_type = TOKEN_INVALID;
2630 evalresult =
2631 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
2632 free_tlist(tline);
2633 if (!evalresult)
2634 return DIRECTIVE_FOUND;
2635 if (tokval.t_type)
2636 error(ERR_WARNING|ERR_PASS1,
2637 "trailing garbage after expression ignored");
2638 if (!is_simple(evalresult)) {
2639 error(ERR_NONFATAL, "non-constant value given to `%%rotate'");
2640 return DIRECTIVE_FOUND;
2642 mmac = istk->mstk;
2643 while (mmac && !mmac->name) /* avoid mistaking %reps for macros */
2644 mmac = mmac->next_active;
2645 if (!mmac) {
2646 error(ERR_NONFATAL, "`%%rotate' invoked outside a macro call");
2647 } else if (mmac->nparam == 0) {
2648 error(ERR_NONFATAL,
2649 "`%%rotate' invoked within macro without parameters");
2650 } else {
2651 int rotate = mmac->rotate + reloc_value(evalresult);
2653 rotate %= (int)mmac->nparam;
2654 if (rotate < 0)
2655 rotate += mmac->nparam;
2657 mmac->rotate = rotate;
2659 return DIRECTIVE_FOUND;
2661 case PP_REP:
2662 nolist = false;
2663 do {
2664 tline = tline->next;
2665 } while (tok_type_(tline, TOK_WHITESPACE));
2667 if (tok_type_(tline, TOK_ID) &&
2668 nasm_stricmp(tline->text, ".nolist") == 0) {
2669 nolist = true;
2670 do {
2671 tline = tline->next;
2672 } while (tok_type_(tline, TOK_WHITESPACE));
2675 if (tline) {
2676 t = expand_smacro(tline);
2677 tptr = &t;
2678 tokval.t_type = TOKEN_INVALID;
2679 evalresult =
2680 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
2681 if (!evalresult) {
2682 free_tlist(origline);
2683 return DIRECTIVE_FOUND;
2685 if (tokval.t_type)
2686 error(ERR_WARNING|ERR_PASS1,
2687 "trailing garbage after expression ignored");
2688 if (!is_simple(evalresult)) {
2689 error(ERR_NONFATAL, "non-constant value given to `%%rep'");
2690 return DIRECTIVE_FOUND;
2692 count = reloc_value(evalresult) + 1;
2693 } else {
2694 error(ERR_NONFATAL, "`%%rep' expects a repeat count");
2695 count = 0;
2697 free_tlist(origline);
2699 tmp_defining = defining;
2700 defining = nasm_malloc(sizeof(MMacro));
2701 defining->name = NULL; /* flags this macro as a %rep block */
2702 defining->casesense = false;
2703 defining->plus = false;
2704 defining->nolist = nolist;
2705 defining->in_progress = count;
2706 defining->nparam_min = defining->nparam_max = 0;
2707 defining->defaults = NULL;
2708 defining->dlist = NULL;
2709 defining->expansion = NULL;
2710 defining->next_active = istk->mstk;
2711 defining->rep_nest = tmp_defining;
2712 return DIRECTIVE_FOUND;
2714 case PP_ENDREP:
2715 if (!defining || defining->name) {
2716 error(ERR_NONFATAL, "`%%endrep': no matching `%%rep'");
2717 return DIRECTIVE_FOUND;
2721 * Now we have a "macro" defined - although it has no name
2722 * and we won't be entering it in the hash tables - we must
2723 * push a macro-end marker for it on to istk->expansion.
2724 * After that, it will take care of propagating itself (a
2725 * macro-end marker line for a macro which is really a %rep
2726 * block will cause the macro to be re-expanded, complete
2727 * with another macro-end marker to ensure the process
2728 * continues) until the whole expansion is forcibly removed
2729 * from istk->expansion by a %exitrep.
2731 l = nasm_malloc(sizeof(Line));
2732 l->next = istk->expansion;
2733 l->finishes = defining;
2734 l->first = NULL;
2735 istk->expansion = l;
2737 istk->mstk = defining;
2739 list->uplevel(defining->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
2740 tmp_defining = defining;
2741 defining = defining->rep_nest;
2742 free_tlist(origline);
2743 return DIRECTIVE_FOUND;
2745 case PP_EXITREP:
2747 * We must search along istk->expansion until we hit a
2748 * macro-end marker for a macro with no name. Then we set
2749 * its `in_progress' flag to 0.
2751 for (l = istk->expansion; l; l = l->next)
2752 if (l->finishes && !l->finishes->name)
2753 break;
2755 if (l)
2756 l->finishes->in_progress = 1;
2757 else
2758 error(ERR_NONFATAL, "`%%exitrep' not within `%%rep' block");
2759 free_tlist(origline);
2760 return DIRECTIVE_FOUND;
2762 case PP_XDEFINE:
2763 case PP_IXDEFINE:
2764 case PP_DEFINE:
2765 case PP_IDEFINE:
2766 casesense = (i == PP_DEFINE || i == PP_XDEFINE);
2768 tline = tline->next;
2769 skip_white_(tline);
2770 tline = expand_id(tline);
2771 if (!tline || (tline->type != TOK_ID &&
2772 (tline->type != TOK_PREPROC_ID ||
2773 tline->text[1] != '$'))) {
2774 error(ERR_NONFATAL, "`%s' expects a macro identifier",
2775 pp_directives[i]);
2776 free_tlist(origline);
2777 return DIRECTIVE_FOUND;
2780 ctx = get_ctx(tline->text, &mname, false);
2781 last = tline;
2782 param_start = tline = tline->next;
2783 nparam = 0;
2785 /* Expand the macro definition now for %xdefine and %ixdefine */
2786 if ((i == PP_XDEFINE) || (i == PP_IXDEFINE))
2787 tline = expand_smacro(tline);
2789 if (tok_is_(tline, "(")) {
2791 * This macro has parameters.
2794 tline = tline->next;
2795 while (1) {
2796 skip_white_(tline);
2797 if (!tline) {
2798 error(ERR_NONFATAL, "parameter identifier expected");
2799 free_tlist(origline);
2800 return DIRECTIVE_FOUND;
2802 if (tline->type != TOK_ID) {
2803 error(ERR_NONFATAL,
2804 "`%s': parameter identifier expected",
2805 tline->text);
2806 free_tlist(origline);
2807 return DIRECTIVE_FOUND;
2809 tline->type = TOK_SMAC_PARAM + nparam++;
2810 tline = tline->next;
2811 skip_white_(tline);
2812 if (tok_is_(tline, ",")) {
2813 tline = tline->next;
2814 } else {
2815 if (!tok_is_(tline, ")")) {
2816 error(ERR_NONFATAL,
2817 "`)' expected to terminate macro template");
2818 free_tlist(origline);
2819 return DIRECTIVE_FOUND;
2821 break;
2824 last = tline;
2825 tline = tline->next;
2827 if (tok_type_(tline, TOK_WHITESPACE))
2828 last = tline, tline = tline->next;
2829 macro_start = NULL;
2830 last->next = NULL;
2831 t = tline;
2832 while (t) {
2833 if (t->type == TOK_ID) {
2834 for (tt = param_start; tt; tt = tt->next)
2835 if (tt->type >= TOK_SMAC_PARAM &&
2836 !strcmp(tt->text, t->text))
2837 t->type = tt->type;
2839 tt = t->next;
2840 t->next = macro_start;
2841 macro_start = t;
2842 t = tt;
2845 * Good. We now have a macro name, a parameter count, and a
2846 * token list (in reverse order) for an expansion. We ought
2847 * to be OK just to create an SMacro, store it, and let
2848 * free_tlist have the rest of the line (which we have
2849 * carefully re-terminated after chopping off the expansion
2850 * from the end).
2852 define_smacro(ctx, mname, casesense, nparam, macro_start);
2853 free_tlist(origline);
2854 return DIRECTIVE_FOUND;
2856 case PP_UNDEF:
2857 tline = tline->next;
2858 skip_white_(tline);
2859 tline = expand_id(tline);
2860 if (!tline || (tline->type != TOK_ID &&
2861 (tline->type != TOK_PREPROC_ID ||
2862 tline->text[1] != '$'))) {
2863 error(ERR_NONFATAL, "`%%undef' expects a macro identifier");
2864 free_tlist(origline);
2865 return DIRECTIVE_FOUND;
2867 if (tline->next) {
2868 error(ERR_WARNING|ERR_PASS1,
2869 "trailing garbage after macro name ignored");
2872 /* Find the context that symbol belongs to */
2873 ctx = get_ctx(tline->text, &mname, false);
2874 undef_smacro(ctx, mname);
2875 free_tlist(origline);
2876 return DIRECTIVE_FOUND;
2878 case PP_DEFSTR:
2879 case PP_IDEFSTR:
2880 casesense = (i == PP_DEFSTR);
2882 tline = tline->next;
2883 skip_white_(tline);
2884 tline = expand_id(tline);
2885 if (!tline || (tline->type != TOK_ID &&
2886 (tline->type != TOK_PREPROC_ID ||
2887 tline->text[1] != '$'))) {
2888 error(ERR_NONFATAL, "`%s' expects a macro identifier",
2889 pp_directives[i]);
2890 free_tlist(origline);
2891 return DIRECTIVE_FOUND;
2894 ctx = get_ctx(tline->text, &mname, false);
2895 last = tline;
2896 tline = expand_smacro(tline->next);
2897 last->next = NULL;
2899 while (tok_type_(tline, TOK_WHITESPACE))
2900 tline = delete_Token(tline);
2902 p = detoken(tline, false);
2903 macro_start = nasm_malloc(sizeof(*macro_start));
2904 macro_start->next = NULL;
2905 macro_start->text = nasm_quote(p, strlen(p));
2906 macro_start->type = TOK_STRING;
2907 macro_start->a.mac = NULL;
2908 nasm_free(p);
2911 * We now have a macro name, an implicit parameter count of
2912 * zero, and a string token to use as an expansion. Create
2913 * and store an SMacro.
2915 define_smacro(ctx, mname, casesense, 0, macro_start);
2916 free_tlist(origline);
2917 return DIRECTIVE_FOUND;
2919 case PP_PATHSEARCH:
2921 FILE *fp;
2922 StrList *xsl = NULL;
2923 StrList **xst = &xsl;
2925 casesense = true;
2927 tline = tline->next;
2928 skip_white_(tline);
2929 tline = expand_id(tline);
2930 if (!tline || (tline->type != TOK_ID &&
2931 (tline->type != TOK_PREPROC_ID ||
2932 tline->text[1] != '$'))) {
2933 error(ERR_NONFATAL,
2934 "`%%pathsearch' expects a macro identifier as first parameter");
2935 free_tlist(origline);
2936 return DIRECTIVE_FOUND;
2938 ctx = get_ctx(tline->text, &mname, false);
2939 last = tline;
2940 tline = expand_smacro(tline->next);
2941 last->next = NULL;
2943 t = tline;
2944 while (tok_type_(t, TOK_WHITESPACE))
2945 t = t->next;
2947 if (!t || (t->type != TOK_STRING &&
2948 t->type != TOK_INTERNAL_STRING)) {
2949 error(ERR_NONFATAL, "`%%pathsearch' expects a file name");
2950 free_tlist(tline);
2951 free_tlist(origline);
2952 return DIRECTIVE_FOUND; /* but we did _something_ */
2954 if (t->next)
2955 error(ERR_WARNING|ERR_PASS1,
2956 "trailing garbage after `%%pathsearch' ignored");
2957 p = t->text;
2958 if (t->type != TOK_INTERNAL_STRING)
2959 nasm_unquote(p, NULL);
2961 fp = inc_fopen(p, &xsl, &xst, true);
2962 if (fp) {
2963 p = xsl->str;
2964 fclose(fp); /* Don't actually care about the file */
2966 macro_start = nasm_malloc(sizeof(*macro_start));
2967 macro_start->next = NULL;
2968 macro_start->text = nasm_quote(p, strlen(p));
2969 macro_start->type = TOK_STRING;
2970 macro_start->a.mac = NULL;
2971 if (xsl)
2972 nasm_free(xsl);
2975 * We now have a macro name, an implicit parameter count of
2976 * zero, and a string token to use as an expansion. Create
2977 * and store an SMacro.
2979 define_smacro(ctx, mname, casesense, 0, macro_start);
2980 free_tlist(tline);
2981 free_tlist(origline);
2982 return DIRECTIVE_FOUND;
2985 case PP_STRLEN:
2986 casesense = true;
2988 tline = tline->next;
2989 skip_white_(tline);
2990 tline = expand_id(tline);
2991 if (!tline || (tline->type != TOK_ID &&
2992 (tline->type != TOK_PREPROC_ID ||
2993 tline->text[1] != '$'))) {
2994 error(ERR_NONFATAL,
2995 "`%%strlen' expects a macro identifier as first parameter");
2996 free_tlist(origline);
2997 return DIRECTIVE_FOUND;
2999 ctx = get_ctx(tline->text, &mname, false);
3000 last = tline;
3001 tline = expand_smacro(tline->next);
3002 last->next = NULL;
3004 t = tline;
3005 while (tok_type_(t, TOK_WHITESPACE))
3006 t = t->next;
3007 /* t should now point to the string */
3008 if (t->type != TOK_STRING) {
3009 error(ERR_NONFATAL,
3010 "`%%strlen` requires string as second parameter");
3011 free_tlist(tline);
3012 free_tlist(origline);
3013 return DIRECTIVE_FOUND;
3016 macro_start = nasm_malloc(sizeof(*macro_start));
3017 macro_start->next = NULL;
3018 make_tok_num(macro_start, nasm_unquote(t->text, NULL));
3019 macro_start->a.mac = NULL;
3022 * We now have a macro name, an implicit parameter count of
3023 * zero, and a numeric token to use as an expansion. Create
3024 * and store an SMacro.
3026 define_smacro(ctx, mname, casesense, 0, macro_start);
3027 free_tlist(tline);
3028 free_tlist(origline);
3029 return DIRECTIVE_FOUND;
3031 case PP_STRCAT:
3032 casesense = true;
3034 tline = tline->next;
3035 skip_white_(tline);
3036 tline = expand_id(tline);
3037 if (!tline || (tline->type != TOK_ID &&
3038 (tline->type != TOK_PREPROC_ID ||
3039 tline->text[1] != '$'))) {
3040 error(ERR_NONFATAL,
3041 "`%%strcat' expects a macro identifier as first parameter");
3042 free_tlist(origline);
3043 return DIRECTIVE_FOUND;
3045 ctx = get_ctx(tline->text, &mname, false);
3046 last = tline;
3047 tline = expand_smacro(tline->next);
3048 last->next = NULL;
3050 len = 0;
3051 for (t = tline; t; t = t->next) {
3052 switch (t->type) {
3053 case TOK_WHITESPACE:
3054 break;
3055 case TOK_STRING:
3056 len += t->a.len = nasm_unquote(t->text, NULL);
3057 break;
3058 case TOK_OTHER:
3059 if (!strcmp(t->text, ",")) /* permit comma separators */
3060 break;
3061 /* else fall through */
3062 default:
3063 error(ERR_NONFATAL,
3064 "non-string passed to `%%strcat' (%d)", t->type);
3065 free_tlist(tline);
3066 free_tlist(origline);
3067 return DIRECTIVE_FOUND;
3071 p = pp = nasm_malloc(len);
3072 t = tline;
3073 for (t = tline; t; t = t->next) {
3074 if (t->type == TOK_STRING) {
3075 memcpy(p, t->text, t->a.len);
3076 p += t->a.len;
3081 * We now have a macro name, an implicit parameter count of
3082 * zero, and a numeric token to use as an expansion. Create
3083 * and store an SMacro.
3085 macro_start = new_Token(NULL, TOK_STRING, NULL, 0);
3086 macro_start->text = nasm_quote(pp, len);
3087 nasm_free(pp);
3088 define_smacro(ctx, mname, casesense, 0, macro_start);
3089 free_tlist(tline);
3090 free_tlist(origline);
3091 return DIRECTIVE_FOUND;
3093 case PP_SUBSTR:
3095 int64_t a1, a2;
3096 size_t len;
3098 casesense = true;
3100 tline = tline->next;
3101 skip_white_(tline);
3102 tline = expand_id(tline);
3103 if (!tline || (tline->type != TOK_ID &&
3104 (tline->type != TOK_PREPROC_ID ||
3105 tline->text[1] != '$'))) {
3106 error(ERR_NONFATAL,
3107 "`%%substr' expects a macro identifier as first parameter");
3108 free_tlist(origline);
3109 return DIRECTIVE_FOUND;
3111 ctx = get_ctx(tline->text, &mname, false);
3112 last = tline;
3113 tline = expand_smacro(tline->next);
3114 last->next = NULL;
3116 t = tline->next;
3117 while (tok_type_(t, TOK_WHITESPACE))
3118 t = t->next;
3120 /* t should now point to the string */
3121 if (t->type != TOK_STRING) {
3122 error(ERR_NONFATAL,
3123 "`%%substr` requires string as second parameter");
3124 free_tlist(tline);
3125 free_tlist(origline);
3126 return DIRECTIVE_FOUND;
3129 tt = t->next;
3130 tptr = &tt;
3131 tokval.t_type = TOKEN_INVALID;
3132 evalresult = evaluate(ppscan, tptr, &tokval, NULL,
3133 pass, error, NULL);
3134 if (!evalresult) {
3135 free_tlist(tline);
3136 free_tlist(origline);
3137 return DIRECTIVE_FOUND;
3138 } else if (!is_simple(evalresult)) {
3139 error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3140 free_tlist(tline);
3141 free_tlist(origline);
3142 return DIRECTIVE_FOUND;
3144 a1 = evalresult->value-1;
3146 while (tok_type_(tt, TOK_WHITESPACE))
3147 tt = tt->next;
3148 if (!tt) {
3149 a2 = 1; /* Backwards compatibility: one character */
3150 } else {
3151 tokval.t_type = TOKEN_INVALID;
3152 evalresult = evaluate(ppscan, tptr, &tokval, NULL,
3153 pass, error, NULL);
3154 if (!evalresult) {
3155 free_tlist(tline);
3156 free_tlist(origline);
3157 return DIRECTIVE_FOUND;
3158 } else if (!is_simple(evalresult)) {
3159 error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3160 free_tlist(tline);
3161 free_tlist(origline);
3162 return DIRECTIVE_FOUND;
3164 a2 = evalresult->value;
3167 len = nasm_unquote(t->text, NULL);
3168 if (a2 < 0)
3169 a2 = a2+1+len-a1;
3170 if (a1+a2 > (int64_t)len)
3171 a2 = len-a1;
3173 macro_start = nasm_malloc(sizeof(*macro_start));
3174 macro_start->next = NULL;
3175 macro_start->text = nasm_quote((a1 < 0) ? "" : t->text+a1, a2);
3176 macro_start->type = TOK_STRING;
3177 macro_start->a.mac = NULL;
3180 * We now have a macro name, an implicit parameter count of
3181 * zero, and a numeric token to use as an expansion. Create
3182 * and store an SMacro.
3184 define_smacro(ctx, mname, casesense, 0, macro_start);
3185 free_tlist(tline);
3186 free_tlist(origline);
3187 return DIRECTIVE_FOUND;
3190 case PP_ASSIGN:
3191 case PP_IASSIGN:
3192 casesense = (i == PP_ASSIGN);
3194 tline = tline->next;
3195 skip_white_(tline);
3196 tline = expand_id(tline);
3197 if (!tline || (tline->type != TOK_ID &&
3198 (tline->type != TOK_PREPROC_ID ||
3199 tline->text[1] != '$'))) {
3200 error(ERR_NONFATAL,
3201 "`%%%sassign' expects a macro identifier",
3202 (i == PP_IASSIGN ? "i" : ""));
3203 free_tlist(origline);
3204 return DIRECTIVE_FOUND;
3206 ctx = get_ctx(tline->text, &mname, false);
3207 last = tline;
3208 tline = expand_smacro(tline->next);
3209 last->next = NULL;
3211 t = tline;
3212 tptr = &t;
3213 tokval.t_type = TOKEN_INVALID;
3214 evalresult =
3215 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
3216 free_tlist(tline);
3217 if (!evalresult) {
3218 free_tlist(origline);
3219 return DIRECTIVE_FOUND;
3222 if (tokval.t_type)
3223 error(ERR_WARNING|ERR_PASS1,
3224 "trailing garbage after expression ignored");
3226 if (!is_simple(evalresult)) {
3227 error(ERR_NONFATAL,
3228 "non-constant value given to `%%%sassign'",
3229 (i == PP_IASSIGN ? "i" : ""));
3230 free_tlist(origline);
3231 return DIRECTIVE_FOUND;
3234 macro_start = nasm_malloc(sizeof(*macro_start));
3235 macro_start->next = NULL;
3236 make_tok_num(macro_start, reloc_value(evalresult));
3237 macro_start->a.mac = NULL;
3240 * We now have a macro name, an implicit parameter count of
3241 * zero, and a numeric token to use as an expansion. Create
3242 * and store an SMacro.
3244 define_smacro(ctx, mname, casesense, 0, macro_start);
3245 free_tlist(origline);
3246 return DIRECTIVE_FOUND;
3248 case PP_LINE:
3250 * Syntax is `%line nnn[+mmm] [filename]'
3252 tline = tline->next;
3253 skip_white_(tline);
3254 if (!tok_type_(tline, TOK_NUMBER)) {
3255 error(ERR_NONFATAL, "`%%line' expects line number");
3256 free_tlist(origline);
3257 return DIRECTIVE_FOUND;
3259 k = readnum(tline->text, &err);
3260 m = 1;
3261 tline = tline->next;
3262 if (tok_is_(tline, "+")) {
3263 tline = tline->next;
3264 if (!tok_type_(tline, TOK_NUMBER)) {
3265 error(ERR_NONFATAL, "`%%line' expects line increment");
3266 free_tlist(origline);
3267 return DIRECTIVE_FOUND;
3269 m = readnum(tline->text, &err);
3270 tline = tline->next;
3272 skip_white_(tline);
3273 src_set_linnum(k);
3274 istk->lineinc = m;
3275 if (tline) {
3276 nasm_free(src_set_fname(detoken(tline, false)));
3278 free_tlist(origline);
3279 return DIRECTIVE_FOUND;
3281 default:
3282 error(ERR_FATAL,
3283 "preprocessor directive `%s' not yet implemented",
3284 pp_directives[i]);
3285 return DIRECTIVE_FOUND;
3290 * Ensure that a macro parameter contains a condition code and
3291 * nothing else. Return the condition code index if so, or -1
3292 * otherwise.
3294 static int find_cc(Token * t)
3296 Token *tt;
3297 int i, j, k, m;
3299 if (!t)
3300 return -1; /* Probably a %+ without a space */
3302 skip_white_(t);
3303 if (t->type != TOK_ID)
3304 return -1;
3305 tt = t->next;
3306 skip_white_(tt);
3307 if (tt && (tt->type != TOK_OTHER || strcmp(tt->text, ",")))
3308 return -1;
3310 i = -1;
3311 j = elements(conditions);
3312 while (j - i > 1) {
3313 k = (j + i) / 2;
3314 m = nasm_stricmp(t->text, conditions[k]);
3315 if (m == 0) {
3316 i = k;
3317 j = -2;
3318 break;
3319 } else if (m < 0) {
3320 j = k;
3321 } else
3322 i = k;
3324 if (j != -2)
3325 return -1;
3326 return i;
3330 * Expand MMacro-local things: parameter references (%0, %n, %+n,
3331 * %-n) and MMacro-local identifiers (%%foo) as well as
3332 * macro indirection (%[...]).
3334 static Token *expand_mmac_params(Token * tline)
3336 Token *t, *tt, **tail, *thead;
3337 bool changed = false;
3339 tail = &thead;
3340 thead = NULL;
3342 while (tline) {
3343 if (tline->type == TOK_PREPROC_ID &&
3344 (((tline->text[1] == '+' || tline->text[1] == '-')
3345 && tline->text[2]) || tline->text[1] == '%'
3346 || (tline->text[1] >= '0' && tline->text[1] <= '9'))) {
3347 char *text = NULL;
3348 int type = 0, cc; /* type = 0 to placate optimisers */
3349 char tmpbuf[30];
3350 unsigned int n;
3351 int i;
3352 MMacro *mac;
3354 t = tline;
3355 tline = tline->next;
3357 mac = istk->mstk;
3358 while (mac && !mac->name) /* avoid mistaking %reps for macros */
3359 mac = mac->next_active;
3360 if (!mac)
3361 error(ERR_NONFATAL, "`%s': not in a macro call", t->text);
3362 else
3363 switch (t->text[1]) {
3365 * We have to make a substitution of one of the
3366 * forms %1, %-1, %+1, %%foo, %0.
3368 case '0':
3369 type = TOK_NUMBER;
3370 snprintf(tmpbuf, sizeof(tmpbuf), "%d", mac->nparam);
3371 text = nasm_strdup(tmpbuf);
3372 break;
3373 case '%':
3374 type = TOK_ID;
3375 snprintf(tmpbuf, sizeof(tmpbuf), "..@%"PRIu64".",
3376 mac->unique);
3377 text = nasm_strcat(tmpbuf, t->text + 2);
3378 break;
3379 case '-':
3380 n = atoi(t->text + 2) - 1;
3381 if (n >= mac->nparam)
3382 tt = NULL;
3383 else {
3384 if (mac->nparam > 1)
3385 n = (n + mac->rotate) % mac->nparam;
3386 tt = mac->params[n];
3388 cc = find_cc(tt);
3389 if (cc == -1) {
3390 error(ERR_NONFATAL,
3391 "macro parameter %d is not a condition code",
3392 n + 1);
3393 text = NULL;
3394 } else {
3395 type = TOK_ID;
3396 if (inverse_ccs[cc] == -1) {
3397 error(ERR_NONFATAL,
3398 "condition code `%s' is not invertible",
3399 conditions[cc]);
3400 text = NULL;
3401 } else
3402 text = nasm_strdup(conditions[inverse_ccs[cc]]);
3404 break;
3405 case '+':
3406 n = atoi(t->text + 2) - 1;
3407 if (n >= mac->nparam)
3408 tt = NULL;
3409 else {
3410 if (mac->nparam > 1)
3411 n = (n + mac->rotate) % mac->nparam;
3412 tt = mac->params[n];
3414 cc = find_cc(tt);
3415 if (cc == -1) {
3416 error(ERR_NONFATAL,
3417 "macro parameter %d is not a condition code",
3418 n + 1);
3419 text = NULL;
3420 } else {
3421 type = TOK_ID;
3422 text = nasm_strdup(conditions[cc]);
3424 break;
3425 default:
3426 n = atoi(t->text + 1) - 1;
3427 if (n >= mac->nparam)
3428 tt = NULL;
3429 else {
3430 if (mac->nparam > 1)
3431 n = (n + mac->rotate) % mac->nparam;
3432 tt = mac->params[n];
3434 if (tt) {
3435 for (i = 0; i < mac->paramlen[n]; i++) {
3436 *tail = new_Token(NULL, tt->type, tt->text, 0);
3437 tail = &(*tail)->next;
3438 tt = tt->next;
3441 text = NULL; /* we've done it here */
3442 break;
3444 if (!text) {
3445 delete_Token(t);
3446 } else {
3447 *tail = t;
3448 tail = &t->next;
3449 t->type = type;
3450 nasm_free(t->text);
3451 t->text = text;
3452 t->a.mac = NULL;
3454 changed = true;
3455 continue;
3456 } else if (tline->type == TOK_INDIRECT) {
3457 t = tline;
3458 tline = tline->next;
3459 tt = tokenize(t->text);
3460 tt = expand_mmac_params(tt);
3461 tt = expand_smacro(tt);
3462 *tail = tt;
3463 while (tt) {
3464 tt->a.mac = NULL; /* Necessary? */
3465 tail = &tt->next;
3466 tt = tt->next;
3468 delete_Token(t);
3469 changed = true;
3470 } else {
3471 t = *tail = tline;
3472 tline = tline->next;
3473 t->a.mac = NULL;
3474 tail = &t->next;
3477 *tail = NULL;
3479 if (!changed)
3480 return thead;
3482 /* Now handle token pasting... */
3483 tail = &thead;
3484 while ((t = *tail) && (tt = t->next)) {
3485 switch (t->type) {
3486 case TOK_WHITESPACE:
3487 if (tt->type == TOK_WHITESPACE) {
3488 t->next = delete_Token(tt);
3489 } else {
3490 tail = &t->next;
3492 break;
3493 case TOK_ID:
3494 case TOK_NUMBER:
3495 case TOK_FLOAT:
3497 size_t len = 0;
3498 char *tmp, *p;
3500 while (tt &&
3501 (tt->type == TOK_ID || tt->type == TOK_NUMBER ||
3502 tt->type == TOK_FLOAT || tt->type == TOK_OTHER)) {
3503 len += strlen(tt->text);
3504 tt = tt->next;
3507 /* Now tt points to the first token after the potential
3508 paste area... */
3509 if (tt != t->next) {
3510 /* We have at least two tokens... */
3511 len += strlen(t->text);
3512 p = tmp = nasm_malloc(len+1);
3514 while (t != tt) {
3515 strcpy(p, t->text);
3516 p = strchr(p, '\0');
3517 t = delete_Token(t);
3520 t = *tail = tokenize(tmp);
3521 nasm_free(tmp);
3523 while (t->next)
3524 t = t->next;
3526 t->next = tt; /* Attach the remaining token chain */
3528 tail = &t->next;
3529 break;
3531 default:
3532 tail = &t->next;
3533 break;
3536 return thead;
3540 * Expand all single-line macro calls made in the given line.
3541 * Return the expanded version of the line. The original is deemed
3542 * to be destroyed in the process. (In reality we'll just move
3543 * Tokens from input to output a lot of the time, rather than
3544 * actually bothering to destroy and replicate.)
3546 #define DEADMAN_LIMIT (1 << 20)
3548 static Token *expand_smacro(Token * tline)
3550 Token *t, *tt, *mstart, **tail, *thead;
3551 struct hash_table *smtbl;
3552 SMacro *head = NULL, *m;
3553 Token **params;
3554 int *paramsize;
3555 unsigned int nparam, sparam;
3556 int brackets, rescan;
3557 Token *org_tline = tline;
3558 Context *ctx;
3559 const char *mname;
3560 int deadman = DEADMAN_LIMIT;
3563 * Trick: we should avoid changing the start token pointer since it can
3564 * be contained in "next" field of other token. Because of this
3565 * we allocate a copy of first token and work with it; at the end of
3566 * routine we copy it back
3568 if (org_tline) {
3569 tline =
3570 new_Token(org_tline->next, org_tline->type, org_tline->text,
3572 tline->a.mac = org_tline->a.mac;
3573 nasm_free(org_tline->text);
3574 org_tline->text = NULL;
3577 again:
3578 tail = &thead;
3579 thead = NULL;
3581 while (tline) { /* main token loop */
3582 if (!--deadman) {
3583 error(ERR_NONFATAL, "interminable macro recursion");
3584 break;
3587 if ((mname = tline->text)) {
3588 /* if this token is a local macro, look in local context */
3589 if (tline->type == TOK_ID || tline->type == TOK_PREPROC_ID)
3590 ctx = get_ctx(mname, &mname, true);
3591 else
3592 ctx = NULL;
3593 smtbl = ctx ? &ctx->localmac : &smacros;
3594 head = (SMacro *) hash_findix(smtbl, mname);
3597 * We've hit an identifier. As in is_mmacro below, we first
3598 * check whether the identifier is a single-line macro at
3599 * all, then think about checking for parameters if
3600 * necessary.
3602 for (m = head; m; m = m->next)
3603 if (!mstrcmp(m->name, mname, m->casesense))
3604 break;
3605 if (m) {
3606 mstart = tline;
3607 params = NULL;
3608 paramsize = NULL;
3609 if (m->nparam == 0) {
3611 * Simple case: the macro is parameterless. Discard the
3612 * one token that the macro call took, and push the
3613 * expansion back on the to-do stack.
3615 if (!m->expansion) {
3616 if (!strcmp("__FILE__", m->name)) {
3617 int32_t num = 0;
3618 char *file = NULL;
3619 src_get(&num, &file);
3620 tline->text = nasm_quote(file, strlen(file));
3621 tline->type = TOK_STRING;
3622 nasm_free(file);
3623 continue;
3625 if (!strcmp("__LINE__", m->name)) {
3626 nasm_free(tline->text);
3627 make_tok_num(tline, src_get_linnum());
3628 continue;
3630 if (!strcmp("__BITS__", m->name)) {
3631 nasm_free(tline->text);
3632 make_tok_num(tline, globalbits);
3633 continue;
3635 tline = delete_Token(tline);
3636 continue;
3638 } else {
3640 * Complicated case: at least one macro with this name
3641 * exists and takes parameters. We must find the
3642 * parameters in the call, count them, find the SMacro
3643 * that corresponds to that form of the macro call, and
3644 * substitute for the parameters when we expand. What a
3645 * pain.
3647 /*tline = tline->next;
3648 skip_white_(tline); */
3649 do {
3650 t = tline->next;
3651 while (tok_type_(t, TOK_SMAC_END)) {
3652 t->a.mac->in_progress = false;
3653 t->text = NULL;
3654 t = tline->next = delete_Token(t);
3656 tline = t;
3657 } while (tok_type_(tline, TOK_WHITESPACE));
3658 if (!tok_is_(tline, "(")) {
3660 * This macro wasn't called with parameters: ignore
3661 * the call. (Behaviour borrowed from gnu cpp.)
3663 tline = mstart;
3664 m = NULL;
3665 } else {
3666 int paren = 0;
3667 int white = 0;
3668 brackets = 0;
3669 nparam = 0;
3670 sparam = PARAM_DELTA;
3671 params = nasm_malloc(sparam * sizeof(Token *));
3672 params[0] = tline->next;
3673 paramsize = nasm_malloc(sparam * sizeof(int));
3674 paramsize[0] = 0;
3675 while (true) { /* parameter loop */
3677 * For some unusual expansions
3678 * which concatenates function call
3680 t = tline->next;
3681 while (tok_type_(t, TOK_SMAC_END)) {
3682 t->a.mac->in_progress = false;
3683 t->text = NULL;
3684 t = tline->next = delete_Token(t);
3686 tline = t;
3688 if (!tline) {
3689 error(ERR_NONFATAL,
3690 "macro call expects terminating `)'");
3691 break;
3693 if (tline->type == TOK_WHITESPACE
3694 && brackets <= 0) {
3695 if (paramsize[nparam])
3696 white++;
3697 else
3698 params[nparam] = tline->next;
3699 continue; /* parameter loop */
3701 if (tline->type == TOK_OTHER
3702 && tline->text[1] == 0) {
3703 char ch = tline->text[0];
3704 if (ch == ',' && !paren && brackets <= 0) {
3705 if (++nparam >= sparam) {
3706 sparam += PARAM_DELTA;
3707 params = nasm_realloc(params,
3708 sparam *
3709 sizeof(Token
3710 *));
3711 paramsize =
3712 nasm_realloc(paramsize,
3713 sparam *
3714 sizeof(int));
3716 params[nparam] = tline->next;
3717 paramsize[nparam] = 0;
3718 white = 0;
3719 continue; /* parameter loop */
3721 if (ch == '{' &&
3722 (brackets > 0 || (brackets == 0 &&
3723 !paramsize[nparam])))
3725 if (!(brackets++)) {
3726 params[nparam] = tline->next;
3727 continue; /* parameter loop */
3730 if (ch == '}' && brackets > 0)
3731 if (--brackets == 0) {
3732 brackets = -1;
3733 continue; /* parameter loop */
3735 if (ch == '(' && !brackets)
3736 paren++;
3737 if (ch == ')' && brackets <= 0)
3738 if (--paren < 0)
3739 break;
3741 if (brackets < 0) {
3742 brackets = 0;
3743 error(ERR_NONFATAL, "braces do not "
3744 "enclose all of macro parameter");
3746 paramsize[nparam] += white + 1;
3747 white = 0;
3748 } /* parameter loop */
3749 nparam++;
3750 while (m && (m->nparam != nparam ||
3751 mstrcmp(m->name, mname,
3752 m->casesense)))
3753 m = m->next;
3754 if (!m)
3755 error(ERR_WARNING|ERR_PASS1|ERR_WARN_MNP,
3756 "macro `%s' exists, "
3757 "but not taking %d parameters",
3758 mstart->text, nparam);
3761 if (m && m->in_progress)
3762 m = NULL;
3763 if (!m) { /* in progess or didn't find '(' or wrong nparam */
3765 * Design question: should we handle !tline, which
3766 * indicates missing ')' here, or expand those
3767 * macros anyway, which requires the (t) test a few
3768 * lines down?
3770 nasm_free(params);
3771 nasm_free(paramsize);
3772 tline = mstart;
3773 } else {
3775 * Expand the macro: we are placed on the last token of the
3776 * call, so that we can easily split the call from the
3777 * following tokens. We also start by pushing an SMAC_END
3778 * token for the cycle removal.
3780 t = tline;
3781 if (t) {
3782 tline = t->next;
3783 t->next = NULL;
3785 tt = new_Token(tline, TOK_SMAC_END, NULL, 0);
3786 tt->a.mac = m;
3787 m->in_progress = true;
3788 tline = tt;
3789 for (t = m->expansion; t; t = t->next) {
3790 if (t->type >= TOK_SMAC_PARAM) {
3791 Token *pcopy = tline, **ptail = &pcopy;
3792 Token *ttt, *pt;
3793 int i;
3795 ttt = params[t->type - TOK_SMAC_PARAM];
3796 for (i = paramsize[t->type - TOK_SMAC_PARAM];
3797 --i >= 0;) {
3798 pt = *ptail =
3799 new_Token(tline, ttt->type, ttt->text,
3801 ptail = &pt->next;
3802 ttt = ttt->next;
3804 tline = pcopy;
3805 } else if (t->type == TOK_PREPROC_Q) {
3806 tt = new_Token(tline, TOK_ID, mname, 0);
3807 tline = tt;
3808 } else if (t->type == TOK_PREPROC_QQ) {
3809 tt = new_Token(tline, TOK_ID, m->name, 0);
3810 tline = tt;
3811 } else {
3812 tt = new_Token(tline, t->type, t->text, 0);
3813 tline = tt;
3818 * Having done that, get rid of the macro call, and clean
3819 * up the parameters.
3821 nasm_free(params);
3822 nasm_free(paramsize);
3823 free_tlist(mstart);
3824 continue; /* main token loop */
3829 if (tline->type == TOK_SMAC_END) {
3830 tline->a.mac->in_progress = false;
3831 tline = delete_Token(tline);
3832 } else {
3833 t = *tail = tline;
3834 tline = tline->next;
3835 t->a.mac = NULL;
3836 t->next = NULL;
3837 tail = &t->next;
3842 * Now scan the entire line and look for successive TOK_IDs that resulted
3843 * after expansion (they can't be produced by tokenize()). The successive
3844 * TOK_IDs should be concatenated.
3845 * Also we look for %+ tokens and concatenate the tokens before and after
3846 * them (without white spaces in between).
3848 t = thead;
3849 rescan = 0;
3850 while (t) {
3851 while (t && t->type != TOK_ID && t->type != TOK_PREPROC_ID)
3852 t = t->next;
3853 if (!t || !t->next)
3854 break;
3855 if (t->next->type == TOK_ID ||
3856 t->next->type == TOK_PREPROC_ID ||
3857 t->next->type == TOK_NUMBER) {
3858 char *p = nasm_strcat(t->text, t->next->text);
3859 nasm_free(t->text);
3860 t->next = delete_Token(t->next);
3861 t->text = p;
3862 rescan = 1;
3863 } else if (t->next->type == TOK_WHITESPACE && t->next->next &&
3864 t->next->next->type == TOK_PREPROC_ID &&
3865 strcmp(t->next->next->text, "%+") == 0) {
3866 /* free the next whitespace, the %+ token and next whitespace */
3867 int i;
3868 for (i = 1; i <= 3; i++) {
3869 if (!t->next
3870 || (i != 2 && t->next->type != TOK_WHITESPACE))
3871 break;
3872 t->next = delete_Token(t->next);
3873 } /* endfor */
3874 } else
3875 t = t->next;
3877 /* If we concatenaded something, re-scan the line for macros */
3878 if (rescan) {
3879 tline = thead;
3880 goto again;
3883 if (org_tline) {
3884 if (thead) {
3885 *org_tline = *thead;
3886 /* since we just gave text to org_line, don't free it */
3887 thead->text = NULL;
3888 delete_Token(thead);
3889 } else {
3890 /* the expression expanded to empty line;
3891 we can't return NULL for some reasons
3892 we just set the line to a single WHITESPACE token. */
3893 memset(org_tline, 0, sizeof(*org_tline));
3894 org_tline->text = NULL;
3895 org_tline->type = TOK_WHITESPACE;
3897 thead = org_tline;
3900 return thead;
3904 * Similar to expand_smacro but used exclusively with macro identifiers
3905 * right before they are fetched in. The reason is that there can be
3906 * identifiers consisting of several subparts. We consider that if there
3907 * are more than one element forming the name, user wants a expansion,
3908 * otherwise it will be left as-is. Example:
3910 * %define %$abc cde
3912 * the identifier %$abc will be left as-is so that the handler for %define
3913 * will suck it and define the corresponding value. Other case:
3915 * %define _%$abc cde
3917 * In this case user wants name to be expanded *before* %define starts
3918 * working, so we'll expand %$abc into something (if it has a value;
3919 * otherwise it will be left as-is) then concatenate all successive
3920 * PP_IDs into one.
3922 static Token *expand_id(Token * tline)
3924 Token *cur, *oldnext = NULL;
3926 if (!tline || !tline->next)
3927 return tline;
3929 cur = tline;
3930 while (cur->next &&
3931 (cur->next->type == TOK_ID ||
3932 cur->next->type == TOK_PREPROC_ID
3933 || cur->next->type == TOK_NUMBER))
3934 cur = cur->next;
3936 /* If identifier consists of just one token, don't expand */
3937 if (cur == tline)
3938 return tline;
3940 if (cur) {
3941 oldnext = cur->next; /* Detach the tail past identifier */
3942 cur->next = NULL; /* so that expand_smacro stops here */
3945 tline = expand_smacro(tline);
3947 if (cur) {
3948 /* expand_smacro possibly changhed tline; re-scan for EOL */
3949 cur = tline;
3950 while (cur && cur->next)
3951 cur = cur->next;
3952 if (cur)
3953 cur->next = oldnext;
3956 return tline;
3960 * Determine whether the given line constitutes a multi-line macro
3961 * call, and return the MMacro structure called if so. Doesn't have
3962 * to check for an initial label - that's taken care of in
3963 * expand_mmacro - but must check numbers of parameters. Guaranteed
3964 * to be called with tline->type == TOK_ID, so the putative macro
3965 * name is easy to find.
3967 static MMacro *is_mmacro(Token * tline, Token *** params_array)
3969 MMacro *head, *m;
3970 Token **params;
3971 int nparam;
3973 head = (MMacro *) hash_findix(&mmacros, tline->text);
3976 * Efficiency: first we see if any macro exists with the given
3977 * name. If not, we can return NULL immediately. _Then_ we
3978 * count the parameters, and then we look further along the
3979 * list if necessary to find the proper MMacro.
3981 for (m = head; m; m = m->next)
3982 if (!mstrcmp(m->name, tline->text, m->casesense))
3983 break;
3984 if (!m)
3985 return NULL;
3988 * OK, we have a potential macro. Count and demarcate the
3989 * parameters.
3991 count_mmac_params(tline->next, &nparam, &params);
3994 * So we know how many parameters we've got. Find the MMacro
3995 * structure that handles this number.
3997 while (m) {
3998 if (m->nparam_min <= nparam
3999 && (m->plus || nparam <= m->nparam_max)) {
4001 * This one is right. Just check if cycle removal
4002 * prohibits us using it before we actually celebrate...
4004 if (m->in_progress) {
4005 #if 0
4006 error(ERR_NONFATAL,
4007 "self-reference in multi-line macro `%s'", m->name);
4008 #endif
4009 nasm_free(params);
4010 return NULL;
4013 * It's right, and we can use it. Add its default
4014 * parameters to the end of our list if necessary.
4016 if (m->defaults && nparam < m->nparam_min + m->ndefs) {
4017 params =
4018 nasm_realloc(params,
4019 ((m->nparam_min + m->ndefs +
4020 1) * sizeof(*params)));
4021 while (nparam < m->nparam_min + m->ndefs) {
4022 params[nparam] = m->defaults[nparam - m->nparam_min];
4023 nparam++;
4027 * If we've gone over the maximum parameter count (and
4028 * we're in Plus mode), ignore parameters beyond
4029 * nparam_max.
4031 if (m->plus && nparam > m->nparam_max)
4032 nparam = m->nparam_max;
4034 * Then terminate the parameter list, and leave.
4036 if (!params) { /* need this special case */
4037 params = nasm_malloc(sizeof(*params));
4038 nparam = 0;
4040 params[nparam] = NULL;
4041 *params_array = params;
4042 return m;
4045 * This one wasn't right: look for the next one with the
4046 * same name.
4048 for (m = m->next; m; m = m->next)
4049 if (!mstrcmp(m->name, tline->text, m->casesense))
4050 break;
4054 * After all that, we didn't find one with the right number of
4055 * parameters. Issue a warning, and fail to expand the macro.
4057 error(ERR_WARNING|ERR_PASS1|ERR_WARN_MNP,
4058 "macro `%s' exists, but not taking %d parameters",
4059 tline->text, nparam);
4060 nasm_free(params);
4061 return NULL;
4065 * Expand the multi-line macro call made by the given line, if
4066 * there is one to be expanded. If there is, push the expansion on
4067 * istk->expansion and return 1. Otherwise return 0.
4069 static int expand_mmacro(Token * tline)
4071 Token *startline = tline;
4072 Token *label = NULL;
4073 int dont_prepend = 0;
4074 Token **params, *t, *mtok, *tt;
4075 MMacro *m;
4076 Line *l, *ll;
4077 int i, nparam, *paramlen;
4078 const char *mname;
4080 t = tline;
4081 skip_white_(t);
4082 /* if (!tok_type_(t, TOK_ID)) Lino 02/25/02 */
4083 if (!tok_type_(t, TOK_ID) && !tok_type_(t, TOK_PREPROC_ID))
4084 return 0;
4085 mtok = t;
4086 m = is_mmacro(t, &params);
4087 if (m) {
4088 mname = t->text;
4089 } else {
4090 Token *last;
4092 * We have an id which isn't a macro call. We'll assume
4093 * it might be a label; we'll also check to see if a
4094 * colon follows it. Then, if there's another id after
4095 * that lot, we'll check it again for macro-hood.
4097 label = last = t;
4098 t = t->next;
4099 if (tok_type_(t, TOK_WHITESPACE))
4100 last = t, t = t->next;
4101 if (tok_is_(t, ":")) {
4102 dont_prepend = 1;
4103 last = t, t = t->next;
4104 if (tok_type_(t, TOK_WHITESPACE))
4105 last = t, t = t->next;
4107 if (!tok_type_(t, TOK_ID) || (m = is_mmacro(t, &params)) == NULL)
4108 return 0;
4109 last->next = NULL;
4110 mname = t->text;
4111 tline = t;
4115 * Fix up the parameters: this involves stripping leading and
4116 * trailing whitespace, then stripping braces if they are
4117 * present.
4119 for (nparam = 0; params[nparam]; nparam++) ;
4120 paramlen = nparam ? nasm_malloc(nparam * sizeof(*paramlen)) : NULL;
4122 for (i = 0; params[i]; i++) {
4123 int brace = false;
4124 int comma = (!m->plus || i < nparam - 1);
4126 t = params[i];
4127 skip_white_(t);
4128 if (tok_is_(t, "{"))
4129 t = t->next, brace = true, comma = false;
4130 params[i] = t;
4131 paramlen[i] = 0;
4132 while (t) {
4133 if (comma && t->type == TOK_OTHER && !strcmp(t->text, ","))
4134 break; /* ... because we have hit a comma */
4135 if (comma && t->type == TOK_WHITESPACE
4136 && tok_is_(t->next, ","))
4137 break; /* ... or a space then a comma */
4138 if (brace && t->type == TOK_OTHER && !strcmp(t->text, "}"))
4139 break; /* ... or a brace */
4140 t = t->next;
4141 paramlen[i]++;
4146 * OK, we have a MMacro structure together with a set of
4147 * parameters. We must now go through the expansion and push
4148 * copies of each Line on to istk->expansion. Substitution of
4149 * parameter tokens and macro-local tokens doesn't get done
4150 * until the single-line macro substitution process; this is
4151 * because delaying them allows us to change the semantics
4152 * later through %rotate.
4154 * First, push an end marker on to istk->expansion, mark this
4155 * macro as in progress, and set up its invocation-specific
4156 * variables.
4158 ll = nasm_malloc(sizeof(Line));
4159 ll->next = istk->expansion;
4160 ll->finishes = m;
4161 ll->first = NULL;
4162 istk->expansion = ll;
4164 m->in_progress = true;
4165 m->params = params;
4166 m->iline = tline;
4167 m->nparam = nparam;
4168 m->rotate = 0;
4169 m->paramlen = paramlen;
4170 m->unique = unique++;
4171 m->lineno = 0;
4173 m->next_active = istk->mstk;
4174 istk->mstk = m;
4176 for (l = m->expansion; l; l = l->next) {
4177 Token **tail;
4179 ll = nasm_malloc(sizeof(Line));
4180 ll->finishes = NULL;
4181 ll->next = istk->expansion;
4182 istk->expansion = ll;
4183 tail = &ll->first;
4185 for (t = l->first; t; t = t->next) {
4186 Token *x = t;
4187 switch (t->type) {
4188 case TOK_PREPROC_Q:
4189 tt = *tail = new_Token(NULL, TOK_ID, mname, 0);
4190 break;
4191 case TOK_PREPROC_QQ:
4192 tt = *tail = new_Token(NULL, TOK_ID, m->name, 0);
4193 break;
4194 case TOK_PREPROC_ID:
4195 if (t->text[1] == '0' && t->text[2] == '0') {
4196 dont_prepend = -1;
4197 x = label;
4198 if (!x)
4199 continue;
4201 /* fall through */
4202 default:
4203 tt = *tail = new_Token(NULL, x->type, x->text, 0);
4204 break;
4206 tail = &tt->next;
4208 *tail = NULL;
4212 * If we had a label, push it on as the first line of
4213 * the macro expansion.
4215 if (label) {
4216 if (dont_prepend < 0)
4217 free_tlist(startline);
4218 else {
4219 ll = nasm_malloc(sizeof(Line));
4220 ll->finishes = NULL;
4221 ll->next = istk->expansion;
4222 istk->expansion = ll;
4223 ll->first = startline;
4224 if (!dont_prepend) {
4225 while (label->next)
4226 label = label->next;
4227 label->next = tt = new_Token(NULL, TOK_OTHER, ":", 0);
4232 list->uplevel(m->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
4234 return 1;
4237 /* The function that actually does the error reporting */
4238 static void verror(int severity, const char *fmt, va_list arg)
4240 char buff[1024];
4242 vsnprintf(buff, sizeof(buff), fmt, arg);
4244 if (istk && istk->mstk && istk->mstk->name)
4245 _error(severity, "(%s:%d) %s", istk->mstk->name,
4246 istk->mstk->lineno, buff);
4247 else
4248 _error(severity, "%s", buff);
4252 * Since preprocessor always operate only on the line that didn't
4253 * arrived yet, we should always use ERR_OFFBY1.
4255 static void error(int severity, const char *fmt, ...)
4257 va_list arg;
4259 /* If we're in a dead branch of IF or something like it, ignore the error */
4260 if (istk && istk->conds && !emitting(istk->conds->state))
4261 return;
4263 va_start(arg, fmt);
4264 verror(severity, fmt, arg);
4265 va_end(arg);
4269 * Because %else etc are evaluated in the state context
4270 * of the previous branch, errors might get lost with error():
4271 * %if 0 ... %else trailing garbage ... %endif
4272 * So %else etc should report errors with this function.
4274 static void error_precond(int severity, const char *fmt, ...)
4276 va_list arg;
4278 /* Only ignore the error if it's really in a dead branch */
4279 if (istk && istk->conds && istk->conds->state == COND_NEVER)
4280 return;
4282 va_start(arg, fmt);
4283 verror(severity, fmt, arg);
4284 va_end(arg);
4287 static void
4288 pp_reset(char *file, int apass, efunc errfunc, evalfunc eval,
4289 ListGen * listgen, StrList **deplist)
4291 Token *t;
4293 _error = errfunc;
4294 cstk = NULL;
4295 istk = nasm_malloc(sizeof(Include));
4296 istk->next = NULL;
4297 istk->conds = NULL;
4298 istk->expansion = NULL;
4299 istk->mstk = NULL;
4300 istk->fp = fopen(file, "r");
4301 istk->fname = NULL;
4302 src_set_fname(nasm_strdup(file));
4303 src_set_linnum(0);
4304 istk->lineinc = 1;
4305 if (!istk->fp)
4306 error(ERR_FATAL|ERR_NOFILE, "unable to open input file `%s'",
4307 file);
4308 defining = NULL;
4309 nested_mac_count = 0;
4310 nested_rep_count = 0;
4311 init_macros();
4312 unique = 0;
4313 if (tasm_compatible_mode) {
4314 stdmacpos = nasm_stdmac;
4315 } else {
4316 stdmacpos = nasm_stdmac_after_tasm;
4318 any_extrastdmac = extrastdmac && *extrastdmac;
4319 do_predef = true;
4320 list = listgen;
4321 evaluate = eval;
4324 * 0 for dependencies, 1 for preparatory passes, 2 for final pass.
4325 * The caller, however, will also pass in 3 for preprocess-only so
4326 * we can set __PASS__ accordingly.
4328 pass = apass > 2 ? 2 : apass;
4330 dephead = deptail = deplist;
4331 if (deplist) {
4332 StrList *sl = nasm_malloc(strlen(file)+1+sizeof sl->next);
4333 sl->next = NULL;
4334 strcpy(sl->str, file);
4335 *deptail = sl;
4336 deptail = &sl->next;
4340 * Define the __PASS__ macro. This is defined here unlike
4341 * all the other builtins, because it is special -- it varies between
4342 * passes.
4344 t = nasm_malloc(sizeof(*t));
4345 t->next = NULL;
4346 make_tok_num(t, apass);
4347 t->a.mac = NULL;
4348 define_smacro(NULL, "__PASS__", true, 0, t);
4351 static char *pp_getline(void)
4353 char *line;
4354 Token *tline;
4356 while (1) {
4358 * Fetch a tokenized line, either from the macro-expansion
4359 * buffer or from the input file.
4361 tline = NULL;
4362 while (istk->expansion && istk->expansion->finishes) {
4363 Line *l = istk->expansion;
4364 if (!l->finishes->name && l->finishes->in_progress > 1) {
4365 Line *ll;
4368 * This is a macro-end marker for a macro with no
4369 * name, which means it's not really a macro at all
4370 * but a %rep block, and the `in_progress' field is
4371 * more than 1, meaning that we still need to
4372 * repeat. (1 means the natural last repetition; 0
4373 * means termination by %exitrep.) We have
4374 * therefore expanded up to the %endrep, and must
4375 * push the whole block on to the expansion buffer
4376 * again. We don't bother to remove the macro-end
4377 * marker: we'd only have to generate another one
4378 * if we did.
4380 l->finishes->in_progress--;
4381 for (l = l->finishes->expansion; l; l = l->next) {
4382 Token *t, *tt, **tail;
4384 ll = nasm_malloc(sizeof(Line));
4385 ll->next = istk->expansion;
4386 ll->finishes = NULL;
4387 ll->first = NULL;
4388 tail = &ll->first;
4390 for (t = l->first; t; t = t->next) {
4391 if (t->text || t->type == TOK_WHITESPACE) {
4392 tt = *tail =
4393 new_Token(NULL, t->type, t->text, 0);
4394 tail = &tt->next;
4398 istk->expansion = ll;
4400 } else {
4402 * Check whether a `%rep' was started and not ended
4403 * within this macro expansion. This can happen and
4404 * should be detected. It's a fatal error because
4405 * I'm too confused to work out how to recover
4406 * sensibly from it.
4408 if (defining) {
4409 if (defining->name)
4410 error(ERR_PANIC,
4411 "defining with name in expansion");
4412 else if (istk->mstk->name)
4413 error(ERR_FATAL,
4414 "`%%rep' without `%%endrep' within"
4415 " expansion of macro `%s'",
4416 istk->mstk->name);
4420 * FIXME: investigate the relationship at this point between
4421 * istk->mstk and l->finishes
4424 MMacro *m = istk->mstk;
4425 istk->mstk = m->next_active;
4426 if (m->name) {
4428 * This was a real macro call, not a %rep, and
4429 * therefore the parameter information needs to
4430 * be freed.
4432 nasm_free(m->params);
4433 free_tlist(m->iline);
4434 nasm_free(m->paramlen);
4435 l->finishes->in_progress = false;
4436 } else
4437 free_mmacro(m);
4439 istk->expansion = l->next;
4440 nasm_free(l);
4441 list->downlevel(LIST_MACRO);
4444 while (1) { /* until we get a line we can use */
4446 if (istk->expansion) { /* from a macro expansion */
4447 char *p;
4448 Line *l = istk->expansion;
4449 if (istk->mstk)
4450 istk->mstk->lineno++;
4451 tline = l->first;
4452 istk->expansion = l->next;
4453 nasm_free(l);
4454 p = detoken(tline, false);
4455 list->line(LIST_MACRO, p);
4456 nasm_free(p);
4457 break;
4459 line = read_line();
4460 if (line) { /* from the current input file */
4461 line = prepreproc(line);
4462 tline = tokenize(line);
4463 nasm_free(line);
4464 break;
4467 * The current file has ended; work down the istk
4470 Include *i = istk;
4471 fclose(i->fp);
4472 if (i->conds)
4473 error(ERR_FATAL,
4474 "expected `%%endif' before end of file");
4475 /* only set line and file name if there's a next node */
4476 if (i->next) {
4477 src_set_linnum(i->lineno);
4478 nasm_free(src_set_fname(i->fname));
4480 istk = i->next;
4481 list->downlevel(LIST_INCLUDE);
4482 nasm_free(i);
4483 if (!istk)
4484 return NULL;
4485 if (istk->expansion && istk->expansion->finishes)
4486 break;
4491 * We must expand MMacro parameters and MMacro-local labels
4492 * _before_ we plunge into directive processing, to cope
4493 * with things like `%define something %1' such as STRUC
4494 * uses. Unless we're _defining_ a MMacro, in which case
4495 * those tokens should be left alone to go into the
4496 * definition; and unless we're in a non-emitting
4497 * condition, in which case we don't want to meddle with
4498 * anything.
4500 if (!defining && !(istk->conds && !emitting(istk->conds->state))
4501 && !(istk->mstk && !istk->mstk->in_progress)) {
4502 tline = expand_mmac_params(tline);
4506 * Check the line to see if it's a preprocessor directive.
4508 if (do_directive(tline) == DIRECTIVE_FOUND) {
4509 continue;
4510 } else if (defining) {
4512 * We're defining a multi-line macro. We emit nothing
4513 * at all, and just
4514 * shove the tokenized line on to the macro definition.
4516 Line *l = nasm_malloc(sizeof(Line));
4517 l->next = defining->expansion;
4518 l->first = tline;
4519 l->finishes = NULL;
4520 defining->expansion = l;
4521 continue;
4522 } else if (istk->conds && !emitting(istk->conds->state)) {
4524 * We're in a non-emitting branch of a condition block.
4525 * Emit nothing at all, not even a blank line: when we
4526 * emerge from the condition we'll give a line-number
4527 * directive so we keep our place correctly.
4529 free_tlist(tline);
4530 continue;
4531 } else if (istk->mstk && !istk->mstk->in_progress) {
4533 * We're in a %rep block which has been terminated, so
4534 * we're walking through to the %endrep without
4535 * emitting anything. Emit nothing at all, not even a
4536 * blank line: when we emerge from the %rep block we'll
4537 * give a line-number directive so we keep our place
4538 * correctly.
4540 free_tlist(tline);
4541 continue;
4542 } else {
4543 tline = expand_smacro(tline);
4544 if (!expand_mmacro(tline)) {
4546 * De-tokenize the line again, and emit it.
4548 line = detoken(tline, true);
4549 free_tlist(tline);
4550 break;
4551 } else {
4552 continue; /* expand_mmacro calls free_tlist */
4557 return line;
4560 static void pp_cleanup(int pass)
4562 if (defining) {
4563 if(defining->name) {
4564 error(ERR_NONFATAL,
4565 "end of file while still defining macro `%s'",
4566 defining->name);
4567 } else {
4568 error(ERR_NONFATAL, "end of file while still in %%rep");
4571 free_mmacro(defining);
4573 while (cstk)
4574 ctx_pop();
4575 free_macros();
4576 while (istk) {
4577 Include *i = istk;
4578 istk = istk->next;
4579 fclose(i->fp);
4580 nasm_free(i->fname);
4581 nasm_free(i);
4583 while (cstk)
4584 ctx_pop();
4585 nasm_free(src_set_fname(NULL));
4586 if (pass == 0) {
4587 IncPath *i;
4588 free_llist(predef);
4589 delete_Blocks();
4590 while ((i = ipath)) {
4591 ipath = i->next;
4592 if (i->path)
4593 nasm_free(i->path);
4594 nasm_free(i);
4599 void pp_include_path(char *path)
4601 IncPath *i;
4603 i = nasm_malloc(sizeof(IncPath));
4604 i->path = path ? nasm_strdup(path) : NULL;
4605 i->next = NULL;
4607 if (ipath != NULL) {
4608 IncPath *j = ipath;
4609 while (j->next != NULL)
4610 j = j->next;
4611 j->next = i;
4612 } else {
4613 ipath = i;
4617 void pp_pre_include(char *fname)
4619 Token *inc, *space, *name;
4620 Line *l;
4622 name = new_Token(NULL, TOK_INTERNAL_STRING, fname, 0);
4623 space = new_Token(name, TOK_WHITESPACE, NULL, 0);
4624 inc = new_Token(space, TOK_PREPROC_ID, "%include", 0);
4626 l = nasm_malloc(sizeof(Line));
4627 l->next = predef;
4628 l->first = inc;
4629 l->finishes = NULL;
4630 predef = l;
4633 void pp_pre_define(char *definition)
4635 Token *def, *space;
4636 Line *l;
4637 char *equals;
4639 equals = strchr(definition, '=');
4640 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
4641 def = new_Token(space, TOK_PREPROC_ID, "%define", 0);
4642 if (equals)
4643 *equals = ' ';
4644 space->next = tokenize(definition);
4645 if (equals)
4646 *equals = '=';
4648 l = nasm_malloc(sizeof(Line));
4649 l->next = predef;
4650 l->first = def;
4651 l->finishes = NULL;
4652 predef = l;
4655 void pp_pre_undefine(char *definition)
4657 Token *def, *space;
4658 Line *l;
4660 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
4661 def = new_Token(space, TOK_PREPROC_ID, "%undef", 0);
4662 space->next = tokenize(definition);
4664 l = nasm_malloc(sizeof(Line));
4665 l->next = predef;
4666 l->first = def;
4667 l->finishes = NULL;
4668 predef = l;
4672 * Added by Keith Kanios:
4674 * This function is used to assist with "runtime" preprocessor
4675 * directives. (e.g. pp_runtime("%define __BITS__ 64");)
4677 * ERRORS ARE IGNORED HERE, SO MAKE COMPLETELY SURE THAT YOU
4678 * PASS A VALID STRING TO THIS FUNCTION!!!!!
4681 void pp_runtime(char *definition)
4683 Token *def;
4685 def = tokenize(definition);
4686 if(do_directive(def) == NO_DIRECTIVE_FOUND)
4687 free_tlist(def);
4691 void pp_extra_stdmac(macros_t *macros)
4693 extrastdmac = macros;
4696 static void make_tok_num(Token * tok, int64_t val)
4698 char numbuf[20];
4699 snprintf(numbuf, sizeof(numbuf), "%"PRId64"", val);
4700 tok->text = nasm_strdup(numbuf);
4701 tok->type = TOK_NUMBER;
4704 Preproc nasmpp = {
4705 pp_reset,
4706 pp_getline,
4707 pp_cleanup