doc/changes.src: something closer to final release notes for 2.04
[nasm/perl-rewrite.git] / preproc.c
blobde67e0649dbdaf55822eff6ade4fa2b88c9e7c33
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_SMAC_PARAM, /* MUST BE LAST IN THE LIST!!! */
163 TOK_MAX = INT_MAX /* Keep compiler from reducing the range */
166 struct Token {
167 Token *next;
168 char *text;
169 union {
170 SMacro *mac; /* associated macro for TOK_SMAC_END */
171 size_t len; /* scratch length field */
172 } a; /* Auxiliary data */
173 enum pp_token_type type;
177 * Multi-line macro definitions are stored as a linked list of
178 * these, which is essentially a container to allow several linked
179 * lists of Tokens.
181 * Note that in this module, linked lists are treated as stacks
182 * wherever possible. For this reason, Lines are _pushed_ on to the
183 * `expansion' field in MMacro structures, so that the linked list,
184 * if walked, would give the macro lines in reverse order; this
185 * means that we can walk the list when expanding a macro, and thus
186 * push the lines on to the `expansion' field in _istk_ in reverse
187 * order (so that when popped back off they are in the right
188 * order). It may seem cockeyed, and it relies on my design having
189 * an even number of steps in, but it works...
191 * Some of these structures, rather than being actual lines, are
192 * markers delimiting the end of the expansion of a given macro.
193 * This is for use in the cycle-tracking and %rep-handling code.
194 * Such structures have `finishes' non-NULL, and `first' NULL. All
195 * others have `finishes' NULL, but `first' may still be NULL if
196 * the line is blank.
198 struct Line {
199 Line *next;
200 MMacro *finishes;
201 Token *first;
205 * To handle an arbitrary level of file inclusion, we maintain a
206 * stack (ie linked list) of these things.
208 struct Include {
209 Include *next;
210 FILE *fp;
211 Cond *conds;
212 Line *expansion;
213 char *fname;
214 int lineno, lineinc;
215 MMacro *mstk; /* stack of active macros/reps */
219 * Include search path. This is simply a list of strings which get
220 * prepended, in turn, to the name of an include file, in an
221 * attempt to find the file if it's not in the current directory.
223 struct IncPath {
224 IncPath *next;
225 char *path;
229 * Conditional assembly: we maintain a separate stack of these for
230 * each level of file inclusion. (The only reason we keep the
231 * stacks separate is to ensure that a stray `%endif' in a file
232 * included from within the true branch of a `%if' won't terminate
233 * it and cause confusion: instead, rightly, it'll cause an error.)
235 struct Cond {
236 Cond *next;
237 int state;
239 enum {
241 * These states are for use just after %if or %elif: IF_TRUE
242 * means the condition has evaluated to truth so we are
243 * currently emitting, whereas IF_FALSE means we are not
244 * currently emitting but will start doing so if a %else comes
245 * up. In these states, all directives are admissible: %elif,
246 * %else and %endif. (And of course %if.)
248 COND_IF_TRUE, COND_IF_FALSE,
250 * These states come up after a %else: ELSE_TRUE means we're
251 * emitting, and ELSE_FALSE means we're not. In ELSE_* states,
252 * any %elif or %else will cause an error.
254 COND_ELSE_TRUE, COND_ELSE_FALSE,
256 * These states mean that we're not emitting now, and also that
257 * nothing until %endif will be emitted at all. COND_DONE is
258 * used when we've had our moment of emission
259 * and have now started seeing %elifs. COND_NEVER is used when
260 * the condition construct in question is contained within a
261 * non-emitting branch of a larger condition construct,
262 * or if there is an error.
264 COND_DONE, COND_NEVER
266 #define emitting(x) ( (x) == COND_IF_TRUE || (x) == COND_ELSE_TRUE )
269 * These defines are used as the possible return values for do_directive
271 #define NO_DIRECTIVE_FOUND 0
272 #define DIRECTIVE_FOUND 1
275 * Condition codes. Note that we use c_ prefix not C_ because C_ is
276 * used in nasm.h for the "real" condition codes. At _this_ level,
277 * we treat CXZ and ECXZ as condition codes, albeit non-invertible
278 * ones, so we need a different enum...
280 static const char * const conditions[] = {
281 "a", "ae", "b", "be", "c", "cxz", "e", "ecxz", "g", "ge", "l", "le",
282 "na", "nae", "nb", "nbe", "nc", "ne", "ng", "nge", "nl", "nle", "no",
283 "np", "ns", "nz", "o", "p", "pe", "po", "rcxz", "s", "z"
285 enum pp_conds {
286 c_A, c_AE, c_B, c_BE, c_C, c_CXZ, c_E, c_ECXZ, c_G, c_GE, c_L, c_LE,
287 c_NA, c_NAE, c_NB, c_NBE, c_NC, c_NE, c_NG, c_NGE, c_NL, c_NLE, c_NO,
288 c_NP, c_NS, c_NZ, c_O, c_P, c_PE, c_PO, c_RCXZ, c_S, c_Z,
289 c_none = -1
291 static const enum pp_conds inverse_ccs[] = {
292 c_NA, c_NAE, c_NB, c_NBE, c_NC, -1, c_NE, -1, c_NG, c_NGE, c_NL, c_NLE,
293 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,
294 c_Z, c_NO, c_NP, c_PO, c_PE, -1, c_NS, c_NZ
298 * Directive names.
300 /* If this is a an IF, ELIF, ELSE or ENDIF keyword */
301 static int is_condition(enum preproc_token arg)
303 return PP_IS_COND(arg) || (arg == PP_ELSE) || (arg == PP_ENDIF);
306 /* For TASM compatibility we need to be able to recognise TASM compatible
307 * conditional compilation directives. Using the NASM pre-processor does
308 * not work, so we look for them specifically from the following list and
309 * then jam in the equivalent NASM directive into the input stream.
312 enum {
313 TM_ARG, TM_ELIF, TM_ELSE, TM_ENDIF, TM_IF, TM_IFDEF, TM_IFDIFI,
314 TM_IFNDEF, TM_INCLUDE, TM_LOCAL
317 static const char * const tasm_directives[] = {
318 "arg", "elif", "else", "endif", "if", "ifdef", "ifdifi",
319 "ifndef", "include", "local"
322 static int StackSize = 4;
323 static char *StackPointer = "ebp";
324 static int ArgOffset = 8;
325 static int LocalOffset = 0;
327 static Context *cstk;
328 static Include *istk;
329 static IncPath *ipath = NULL;
331 static efunc _error; /* Pointer to client-provided error reporting function */
332 static evalfunc evaluate;
334 static int pass; /* HACK: pass 0 = generate dependencies only */
335 static StrList **dephead, **deptail; /* Dependency list */
337 static uint64_t unique; /* unique identifier numbers */
339 static Line *predef = NULL;
340 static bool do_predef;
342 static ListGen *list;
345 * The current set of multi-line macros we have defined.
347 static struct hash_table mmacros;
350 * The current set of single-line macros we have defined.
352 static struct hash_table smacros;
355 * The multi-line macro we are currently defining, or the %rep
356 * block we are currently reading, if any.
358 static MMacro *defining;
360 static uint64_t nested_mac_count;
361 static uint64_t nested_rep_count;
364 * The number of macro parameters to allocate space for at a time.
366 #define PARAM_DELTA 16
369 * The standard macro set: defined in macros.c in the array nasm_stdmac.
370 * This gives our position in the macro set, when we're processing it.
372 static macros_t *stdmacpos;
375 * The extra standard macros that come from the object format, if
376 * any.
378 static macros_t *extrastdmac = NULL;
379 static bool any_extrastdmac;
382 * Tokens are allocated in blocks to improve speed
384 #define TOKEN_BLOCKSIZE 4096
385 static Token *freeTokens = NULL;
386 struct Blocks {
387 Blocks *next;
388 void *chunk;
391 static Blocks blocks = { NULL, NULL };
394 * Forward declarations.
396 static Token *expand_mmac_params(Token * tline);
397 static Token *expand_smacro(Token * tline);
398 static Token *expand_id(Token * tline);
399 static Context *get_ctx(const char *name, bool all_contexts);
400 static void make_tok_num(Token * tok, int64_t val);
401 static void error(int severity, const char *fmt, ...);
402 static void error_precond(int severity, const char *fmt, ...);
403 static void *new_Block(size_t size);
404 static void delete_Blocks(void);
405 static Token *new_Token(Token * next, enum pp_token_type type,
406 const char *text, int txtlen);
407 static Token *delete_Token(Token * t);
410 * Macros for safe checking of token pointers, avoid *(NULL)
412 #define tok_type_(x,t) ((x) && (x)->type == (t))
413 #define skip_white_(x) if (tok_type_((x), TOK_WHITESPACE)) (x)=(x)->next
414 #define tok_is_(x,v) (tok_type_((x), TOK_OTHER) && !strcmp((x)->text,(v)))
415 #define tok_isnt_(x,v) ((x) && ((x)->type!=TOK_OTHER || strcmp((x)->text,(v))))
417 /* Handle TASM specific directives, which do not contain a % in
418 * front of them. We do it here because I could not find any other
419 * place to do it for the moment, and it is a hack (ideally it would
420 * be nice to be able to use the NASM pre-processor to do it).
422 static char *check_tasm_directive(char *line)
424 int32_t i, j, k, m, len;
425 char *p = line, *oldline, oldchar;
427 /* Skip whitespace */
428 while (nasm_isspace(*p) && *p != 0)
429 p++;
431 /* Binary search for the directive name */
432 i = -1;
433 j = elements(tasm_directives);
434 len = 0;
435 while (!nasm_isspace(p[len]) && p[len] != 0)
436 len++;
437 if (len) {
438 oldchar = p[len];
439 p[len] = 0;
440 while (j - i > 1) {
441 k = (j + i) / 2;
442 m = nasm_stricmp(p, tasm_directives[k]);
443 if (m == 0) {
444 /* We have found a directive, so jam a % in front of it
445 * so that NASM will then recognise it as one if it's own.
447 p[len] = oldchar;
448 len = strlen(p);
449 oldline = line;
450 line = nasm_malloc(len + 2);
451 line[0] = '%';
452 if (k == TM_IFDIFI) {
453 /* NASM does not recognise IFDIFI, so we convert it to
454 * %ifdef BOGUS. This is not used in NASM comaptible
455 * code, but does need to parse for the TASM macro
456 * package.
458 strcpy(line + 1, "ifdef BOGUS");
459 } else {
460 memcpy(line + 1, p, len + 1);
462 nasm_free(oldline);
463 return line;
464 } else if (m < 0) {
465 j = k;
466 } else
467 i = k;
469 p[len] = oldchar;
471 return line;
475 * The pre-preprocessing stage... This function translates line
476 * number indications as they emerge from GNU cpp (`# lineno "file"
477 * flags') into NASM preprocessor line number indications (`%line
478 * lineno file').
480 static char *prepreproc(char *line)
482 int lineno, fnlen;
483 char *fname, *oldline;
485 if (line[0] == '#' && line[1] == ' ') {
486 oldline = line;
487 fname = oldline + 2;
488 lineno = atoi(fname);
489 fname += strspn(fname, "0123456789 ");
490 if (*fname == '"')
491 fname++;
492 fnlen = strcspn(fname, "\"");
493 line = nasm_malloc(20 + fnlen);
494 snprintf(line, 20 + fnlen, "%%line %d %.*s", lineno, fnlen, fname);
495 nasm_free(oldline);
497 if (tasm_compatible_mode)
498 return check_tasm_directive(line);
499 return line;
503 * Free a linked list of tokens.
505 static void free_tlist(Token * list)
507 while (list) {
508 list = delete_Token(list);
513 * Free a linked list of lines.
515 static void free_llist(Line * list)
517 Line *l;
518 while (list) {
519 l = list;
520 list = list->next;
521 free_tlist(l->first);
522 nasm_free(l);
527 * Free an MMacro
529 static void free_mmacro(MMacro * m)
531 nasm_free(m->name);
532 free_tlist(m->dlist);
533 nasm_free(m->defaults);
534 free_llist(m->expansion);
535 nasm_free(m);
539 * Free all currently defined macros, and free the hash tables
541 static void free_smacro_table(struct hash_table *smt)
543 SMacro *s;
544 const char *key;
545 struct hash_tbl_node *it = NULL;
547 while ((s = hash_iterate(smt, &it, &key)) != NULL) {
548 nasm_free((void *)key);
549 while (s) {
550 SMacro *ns = s->next;
551 nasm_free(s->name);
552 free_tlist(s->expansion);
553 nasm_free(s);
554 s = ns;
557 hash_free(smt);
560 static void free_mmacro_table(struct hash_table *mmt)
562 MMacro *m;
563 const char *key;
564 struct hash_tbl_node *it = NULL;
566 it = NULL;
567 while ((m = hash_iterate(mmt, &it, &key)) != NULL) {
568 nasm_free((void *)key);
569 while (m) {
570 MMacro *nm = m->next;
571 free_mmacro(m);
572 m = nm;
575 hash_free(mmt);
578 static void free_macros(void)
580 free_smacro_table(&smacros);
581 free_mmacro_table(&mmacros);
585 * Initialize the hash tables
587 static void init_macros(void)
589 hash_init(&smacros, HASH_LARGE);
590 hash_init(&mmacros, HASH_LARGE);
594 * Pop the context stack.
596 static void ctx_pop(void)
598 Context *c = cstk;
600 cstk = cstk->next;
601 free_smacro_table(&c->localmac);
602 nasm_free(c->name);
603 nasm_free(c);
607 * Search for a key in the hash index; adding it if necessary
608 * (in which case we initialize the data pointer to NULL.)
610 static void **
611 hash_findi_add(struct hash_table *hash, const char *str)
613 struct hash_insert hi;
614 void **r;
615 char *strx;
617 r = hash_findi(hash, str, &hi);
618 if (r)
619 return r;
621 strx = nasm_strdup(str); /* Use a more efficient allocator here? */
622 return hash_add(&hi, strx, NULL);
626 * Like hash_findi, but returns the data element rather than a pointer
627 * to it. Used only when not adding a new element, hence no third
628 * argument.
630 static void *
631 hash_findix(struct hash_table *hash, const char *str)
633 void **p;
635 p = hash_findi(hash, str, NULL);
636 return p ? *p : NULL;
639 #define BUF_DELTA 512
641 * Read a line from the top file in istk, handling multiple CR/LFs
642 * at the end of the line read, and handling spurious ^Zs. Will
643 * return lines from the standard macro set if this has not already
644 * been done.
646 static char *read_line(void)
648 char *buffer, *p, *q;
649 int bufsize, continued_count;
651 if (stdmacpos) {
652 unsigned char c;
653 const unsigned char *p = stdmacpos;
654 char *ret, *q;
655 size_t len = 0;
656 while ((c = *p++)) {
657 if (c >= 0x80)
658 len += pp_directives_len[c-0x80]+1;
659 else
660 len++;
662 ret = nasm_malloc(len+1);
663 q = ret;
664 while ((c = *stdmacpos++)) {
665 if (c >= 0x80) {
666 memcpy(q, pp_directives[c-0x80], pp_directives_len[c-0x80]);
667 q += pp_directives_len[c-0x80];
668 *q++ = ' ';
669 } else {
670 *q++ = c;
673 stdmacpos = p;
674 *q = '\0';
676 if (!*stdmacpos) {
677 /* This was the last of the standard macro chain... */
678 stdmacpos = NULL;
679 if (any_extrastdmac) {
680 stdmacpos = extrastdmac;
681 any_extrastdmac = false;
682 } else if (do_predef) {
683 Line *pd, *l;
684 Token *head, **tail, *t;
687 * Nasty hack: here we push the contents of
688 * `predef' on to the top-level expansion stack,
689 * since this is the most convenient way to
690 * implement the pre-include and pre-define
691 * features.
693 for (pd = predef; pd; pd = pd->next) {
694 head = NULL;
695 tail = &head;
696 for (t = pd->first; t; t = t->next) {
697 *tail = new_Token(NULL, t->type, t->text, 0);
698 tail = &(*tail)->next;
700 l = nasm_malloc(sizeof(Line));
701 l->next = istk->expansion;
702 l->first = head;
703 l->finishes = NULL;
704 istk->expansion = l;
706 do_predef = false;
709 return ret;
712 bufsize = BUF_DELTA;
713 buffer = nasm_malloc(BUF_DELTA);
714 p = buffer;
715 continued_count = 0;
716 while (1) {
717 q = fgets(p, bufsize - (p - buffer), istk->fp);
718 if (!q)
719 break;
720 p += strlen(p);
721 if (p > buffer && p[-1] == '\n') {
722 /* Convert backslash-CRLF line continuation sequences into
723 nothing at all (for DOS and Windows) */
724 if (((p - 2) > buffer) && (p[-3] == '\\') && (p[-2] == '\r')) {
725 p -= 3;
726 *p = 0;
727 continued_count++;
729 /* Also convert backslash-LF line continuation sequences into
730 nothing at all (for Unix) */
731 else if (((p - 1) > buffer) && (p[-2] == '\\')) {
732 p -= 2;
733 *p = 0;
734 continued_count++;
735 } else {
736 break;
739 if (p - buffer > bufsize - 10) {
740 int32_t offset = p - buffer;
741 bufsize += BUF_DELTA;
742 buffer = nasm_realloc(buffer, bufsize);
743 p = buffer + offset; /* prevent stale-pointer problems */
747 if (!q && p == buffer) {
748 nasm_free(buffer);
749 return NULL;
752 src_set_linnum(src_get_linnum() + istk->lineinc +
753 (continued_count * istk->lineinc));
756 * Play safe: remove CRs as well as LFs, if any of either are
757 * present at the end of the line.
759 while (--p >= buffer && (*p == '\n' || *p == '\r'))
760 *p = '\0';
763 * Handle spurious ^Z, which may be inserted into source files
764 * by some file transfer utilities.
766 buffer[strcspn(buffer, "\032")] = '\0';
768 list->line(LIST_READ, buffer);
770 return buffer;
774 * Tokenize a line of text. This is a very simple process since we
775 * don't need to parse the value out of e.g. numeric tokens: we
776 * simply split one string into many.
778 static Token *tokenize(char *line)
780 char *p = line;
781 enum pp_token_type type;
782 Token *list = NULL;
783 Token *t, **tail = &list;
785 while (*line) {
786 p = line;
787 if (*p == '%') {
788 p++;
789 if (nasm_isdigit(*p) ||
790 ((*p == '-' || *p == '+') && nasm_isdigit(p[1])) ||
791 ((*p == '+') && (nasm_isspace(p[1]) || !p[1]))) {
792 do {
793 p++;
795 while (nasm_isdigit(*p));
796 type = TOK_PREPROC_ID;
797 } else if (*p == '{') {
798 p++;
799 while (*p && *p != '}') {
800 p[-1] = *p;
801 p++;
803 p[-1] = '\0';
804 if (*p)
805 p++;
806 type = TOK_PREPROC_ID;
807 } else if (*p == '?') {
808 type = TOK_PREPROC_Q; /* %? */
809 p++;
810 if (*p == '?') {
811 type = TOK_PREPROC_QQ; /* %?? */
812 p++;
814 } else if (isidchar(*p) ||
815 ((*p == '!' || *p == '%' || *p == '$') &&
816 isidchar(p[1]))) {
817 do {
818 p++;
820 while (isidchar(*p));
821 type = TOK_PREPROC_ID;
822 } else {
823 type = TOK_OTHER;
824 if (*p == '%')
825 p++;
827 } else if (isidstart(*p) || (*p == '$' && isidstart(p[1]))) {
828 type = TOK_ID;
829 p++;
830 while (*p && isidchar(*p))
831 p++;
832 } else if (*p == '\'' || *p == '"' || *p == '`') {
834 * A string token.
836 type = TOK_STRING;
837 p = nasm_skip_string(p);
839 if (*p) {
840 p++;
841 } else {
842 error(ERR_WARNING, "unterminated string");
843 /* Handling unterminated strings by UNV */
844 /* type = -1; */
846 } else if (isnumstart(*p)) {
847 bool is_hex = false;
848 bool is_float = false;
849 bool has_e = false;
850 char c, *r;
853 * A numeric token.
856 if (*p == '$') {
857 p++;
858 is_hex = true;
861 for (;;) {
862 c = *p++;
864 if (!is_hex && (c == 'e' || c == 'E')) {
865 has_e = true;
866 if (*p == '+' || *p == '-') {
867 /* e can only be followed by +/- if it is either a
868 prefixed hex number or a floating-point number */
869 p++;
870 is_float = true;
872 } else if (c == 'H' || c == 'h' || c == 'X' || c == 'x') {
873 is_hex = true;
874 } else if (c == 'P' || c == 'p') {
875 is_float = true;
876 if (*p == '+' || *p == '-')
877 p++;
878 } else if (isnumchar(c) || c == '_')
879 ; /* just advance */
880 else if (c == '.') {
881 /* we need to deal with consequences of the legacy
882 parser, like "1.nolist" being two tokens
883 (TOK_NUMBER, TOK_ID) here; at least give it
884 a shot for now. In the future, we probably need
885 a flex-based scanner with proper pattern matching
886 to do it as well as it can be done. Nothing in
887 the world is going to help the person who wants
888 0x123.p16 interpreted as two tokens, though. */
889 r = p;
890 while (*r == '_')
891 r++;
893 if (nasm_isdigit(*r) || (is_hex && nasm_isxdigit(*r)) ||
894 (!is_hex && (*r == 'e' || *r == 'E')) ||
895 (*r == 'p' || *r == 'P')) {
896 p = r;
897 is_float = true;
898 } else
899 break; /* Terminate the token */
900 } else
901 break;
903 p--; /* Point to first character beyond number */
905 if (has_e && !is_hex) {
906 /* 1e13 is floating-point, but 1e13h is not */
907 is_float = true;
910 type = is_float ? TOK_FLOAT : TOK_NUMBER;
911 } else if (nasm_isspace(*p)) {
912 type = TOK_WHITESPACE;
913 p++;
914 while (*p && nasm_isspace(*p))
915 p++;
917 * Whitespace just before end-of-line is discarded by
918 * pretending it's a comment; whitespace just before a
919 * comment gets lumped into the comment.
921 if (!*p || *p == ';') {
922 type = TOK_COMMENT;
923 while (*p)
924 p++;
926 } else if (*p == ';') {
927 type = TOK_COMMENT;
928 while (*p)
929 p++;
930 } else {
932 * Anything else is an operator of some kind. We check
933 * for all the double-character operators (>>, <<, //,
934 * %%, <=, >=, ==, !=, <>, &&, ||, ^^), but anything
935 * else is a single-character operator.
937 type = TOK_OTHER;
938 if ((p[0] == '>' && p[1] == '>') ||
939 (p[0] == '<' && p[1] == '<') ||
940 (p[0] == '/' && p[1] == '/') ||
941 (p[0] == '<' && p[1] == '=') ||
942 (p[0] == '>' && p[1] == '=') ||
943 (p[0] == '=' && p[1] == '=') ||
944 (p[0] == '!' && p[1] == '=') ||
945 (p[0] == '<' && p[1] == '>') ||
946 (p[0] == '&' && p[1] == '&') ||
947 (p[0] == '|' && p[1] == '|') ||
948 (p[0] == '^' && p[1] == '^')) {
949 p++;
951 p++;
954 /* Handling unterminated string by UNV */
955 /*if (type == -1)
957 *tail = t = new_Token(NULL, TOK_STRING, line, p-line+1);
958 t->text[p-line] = *line;
959 tail = &t->next;
961 else */
962 if (type != TOK_COMMENT) {
963 *tail = t = new_Token(NULL, type, line, p - line);
964 tail = &t->next;
966 line = p;
968 return list;
972 * this function allocates a new managed block of memory and
973 * returns a pointer to the block. The managed blocks are
974 * deleted only all at once by the delete_Blocks function.
976 static void *new_Block(size_t size)
978 Blocks *b = &blocks;
980 /* first, get to the end of the linked list */
981 while (b->next)
982 b = b->next;
983 /* now allocate the requested chunk */
984 b->chunk = nasm_malloc(size);
986 /* now allocate a new block for the next request */
987 b->next = nasm_malloc(sizeof(Blocks));
988 /* and initialize the contents of the new block */
989 b->next->next = NULL;
990 b->next->chunk = NULL;
991 return b->chunk;
995 * this function deletes all managed blocks of memory
997 static void delete_Blocks(void)
999 Blocks *a, *b = &blocks;
1002 * keep in mind that the first block, pointed to by blocks
1003 * is a static and not dynamically allocated, so we don't
1004 * free it.
1006 while (b) {
1007 if (b->chunk)
1008 nasm_free(b->chunk);
1009 a = b;
1010 b = b->next;
1011 if (a != &blocks)
1012 nasm_free(a);
1017 * this function creates a new Token and passes a pointer to it
1018 * back to the caller. It sets the type and text elements, and
1019 * also the a.mac and next elements to NULL.
1021 static Token *new_Token(Token * next, enum pp_token_type type,
1022 const char *text, int txtlen)
1024 Token *t;
1025 int i;
1027 if (freeTokens == NULL) {
1028 freeTokens = (Token *) new_Block(TOKEN_BLOCKSIZE * sizeof(Token));
1029 for (i = 0; i < TOKEN_BLOCKSIZE - 1; i++)
1030 freeTokens[i].next = &freeTokens[i + 1];
1031 freeTokens[i].next = NULL;
1033 t = freeTokens;
1034 freeTokens = t->next;
1035 t->next = next;
1036 t->a.mac = NULL;
1037 t->type = type;
1038 if (type == TOK_WHITESPACE || text == NULL) {
1039 t->text = NULL;
1040 } else {
1041 if (txtlen == 0)
1042 txtlen = strlen(text);
1043 t->text = nasm_malloc(txtlen+1);
1044 memcpy(t->text, text, txtlen);
1045 t->text[txtlen] = '\0';
1047 return t;
1050 static Token *delete_Token(Token * t)
1052 Token *next = t->next;
1053 nasm_free(t->text);
1054 t->next = freeTokens;
1055 freeTokens = t;
1056 return next;
1060 * Convert a line of tokens back into text.
1061 * If expand_locals is not zero, identifiers of the form "%$*xxx"
1062 * will be transformed into ..@ctxnum.xxx
1064 static char *detoken(Token * tlist, bool expand_locals)
1066 Token *t;
1067 int len;
1068 char *line, *p;
1069 const char *q;
1071 len = 0;
1072 for (t = tlist; t; t = t->next) {
1073 if (t->type == TOK_PREPROC_ID && t->text[1] == '!') {
1074 char *p = getenv(t->text + 2);
1075 nasm_free(t->text);
1076 if (p)
1077 t->text = nasm_strdup(p);
1078 else
1079 t->text = NULL;
1081 /* Expand local macros here and not during preprocessing */
1082 if (expand_locals &&
1083 t->type == TOK_PREPROC_ID && t->text &&
1084 t->text[0] == '%' && t->text[1] == '$') {
1085 Context *ctx = get_ctx(t->text, false);
1086 if (ctx) {
1087 char buffer[40];
1088 char *p, *q = t->text + 2;
1090 q += strspn(q, "$");
1091 snprintf(buffer, sizeof(buffer), "..@%"PRIu32".", ctx->number);
1092 p = nasm_strcat(buffer, q);
1093 nasm_free(t->text);
1094 t->text = p;
1097 if (t->type == TOK_WHITESPACE) {
1098 len++;
1099 } else if (t->text) {
1100 len += strlen(t->text);
1103 p = line = nasm_malloc(len + 1);
1104 for (t = tlist; t; t = t->next) {
1105 if (t->type == TOK_WHITESPACE) {
1106 *p++ = ' ';
1107 } else if (t->text) {
1108 q = t->text;
1109 while (*q)
1110 *p++ = *q++;
1113 *p = '\0';
1114 return line;
1118 * A scanner, suitable for use by the expression evaluator, which
1119 * operates on a line of Tokens. Expects a pointer to a pointer to
1120 * the first token in the line to be passed in as its private_data
1121 * field.
1123 * FIX: This really needs to be unified with stdscan.
1125 static int ppscan(void *private_data, struct tokenval *tokval)
1127 Token **tlineptr = private_data;
1128 Token *tline;
1129 char ourcopy[MAX_KEYWORD+1], *p, *r, *s;
1131 do {
1132 tline = *tlineptr;
1133 *tlineptr = tline ? tline->next : NULL;
1135 while (tline && (tline->type == TOK_WHITESPACE ||
1136 tline->type == TOK_COMMENT));
1138 if (!tline)
1139 return tokval->t_type = TOKEN_EOS;
1141 tokval->t_charptr = tline->text;
1143 if (tline->text[0] == '$' && !tline->text[1])
1144 return tokval->t_type = TOKEN_HERE;
1145 if (tline->text[0] == '$' && tline->text[1] == '$' && !tline->text[2])
1146 return tokval->t_type = TOKEN_BASE;
1148 if (tline->type == TOK_ID) {
1149 p = tokval->t_charptr = tline->text;
1150 if (p[0] == '$') {
1151 tokval->t_charptr++;
1152 return tokval->t_type = TOKEN_ID;
1155 for (r = p, s = ourcopy; *r; r++) {
1156 if (r >= p+MAX_KEYWORD)
1157 return tokval->t_type = TOKEN_ID; /* Not a keyword */
1158 *s++ = nasm_tolower(*r);
1160 *s = '\0';
1161 /* right, so we have an identifier sitting in temp storage. now,
1162 * is it actually a register or instruction name, or what? */
1163 return nasm_token_hash(ourcopy, tokval);
1166 if (tline->type == TOK_NUMBER) {
1167 bool rn_error;
1168 tokval->t_integer = readnum(tline->text, &rn_error);
1169 tokval->t_charptr = tline->text;
1170 if (rn_error)
1171 return tokval->t_type = TOKEN_ERRNUM;
1172 else
1173 return tokval->t_type = TOKEN_NUM;
1176 if (tline->type == TOK_FLOAT) {
1177 return tokval->t_type = TOKEN_FLOAT;
1180 if (tline->type == TOK_STRING) {
1181 char bq, *ep;
1183 bq = tline->text[0];
1184 tokval->t_charptr = tline->text;
1185 tokval->t_inttwo = nasm_unquote(tline->text, &ep);
1187 if (ep[0] != bq || ep[1] != '\0')
1188 return tokval->t_type = TOKEN_ERRSTR;
1189 else
1190 return tokval->t_type = TOKEN_STR;
1193 if (tline->type == TOK_OTHER) {
1194 if (!strcmp(tline->text, "<<"))
1195 return tokval->t_type = TOKEN_SHL;
1196 if (!strcmp(tline->text, ">>"))
1197 return tokval->t_type = TOKEN_SHR;
1198 if (!strcmp(tline->text, "//"))
1199 return tokval->t_type = TOKEN_SDIV;
1200 if (!strcmp(tline->text, "%%"))
1201 return tokval->t_type = TOKEN_SMOD;
1202 if (!strcmp(tline->text, "=="))
1203 return tokval->t_type = TOKEN_EQ;
1204 if (!strcmp(tline->text, "<>"))
1205 return tokval->t_type = TOKEN_NE;
1206 if (!strcmp(tline->text, "!="))
1207 return tokval->t_type = TOKEN_NE;
1208 if (!strcmp(tline->text, "<="))
1209 return tokval->t_type = TOKEN_LE;
1210 if (!strcmp(tline->text, ">="))
1211 return tokval->t_type = TOKEN_GE;
1212 if (!strcmp(tline->text, "&&"))
1213 return tokval->t_type = TOKEN_DBL_AND;
1214 if (!strcmp(tline->text, "^^"))
1215 return tokval->t_type = TOKEN_DBL_XOR;
1216 if (!strcmp(tline->text, "||"))
1217 return tokval->t_type = TOKEN_DBL_OR;
1221 * We have no other options: just return the first character of
1222 * the token text.
1224 return tokval->t_type = tline->text[0];
1228 * Compare a string to the name of an existing macro; this is a
1229 * simple wrapper which calls either strcmp or nasm_stricmp
1230 * depending on the value of the `casesense' parameter.
1232 static int mstrcmp(const char *p, const char *q, bool casesense)
1234 return casesense ? strcmp(p, q) : nasm_stricmp(p, q);
1238 * Compare a string to the name of an existing macro; this is a
1239 * simple wrapper which calls either strcmp or nasm_stricmp
1240 * depending on the value of the `casesense' parameter.
1242 static int mmemcmp(const char *p, const char *q, size_t l, bool casesense)
1244 return casesense ? memcmp(p, q, l) : nasm_memicmp(p, q, l);
1248 * Return the Context structure associated with a %$ token. Return
1249 * NULL, having _already_ reported an error condition, if the
1250 * context stack isn't deep enough for the supplied number of $
1251 * signs.
1252 * If all_contexts == true, contexts that enclose current are
1253 * also scanned for such smacro, until it is found; if not -
1254 * only the context that directly results from the number of $'s
1255 * in variable's name.
1257 static Context *get_ctx(const char *name, bool all_contexts)
1259 Context *ctx;
1260 SMacro *m;
1261 int i;
1263 if (!name || name[0] != '%' || name[1] != '$')
1264 return NULL;
1266 if (!cstk) {
1267 error(ERR_NONFATAL, "`%s': context stack is empty", name);
1268 return NULL;
1271 for (i = strspn(name + 2, "$"), ctx = cstk; (i > 0) && ctx; i--) {
1272 ctx = ctx->next;
1273 /* i--; Lino - 02/25/02 */
1275 if (!ctx) {
1276 error(ERR_NONFATAL, "`%s': context stack is only"
1277 " %d level%s deep", name, i - 1, (i == 2 ? "" : "s"));
1278 return NULL;
1280 if (!all_contexts)
1281 return ctx;
1283 do {
1284 /* Search for this smacro in found context */
1285 m = hash_findix(&ctx->localmac, name);
1286 while (m) {
1287 if (!mstrcmp(m->name, name, m->casesense))
1288 return ctx;
1289 m = m->next;
1291 ctx = ctx->next;
1293 while (ctx);
1294 return NULL;
1298 * Check to see if a file is already in a string list
1300 static bool in_list(const StrList *list, const char *str)
1302 while (list) {
1303 if (!strcmp(list->str, str))
1304 return true;
1305 list = list->next;
1307 return false;
1311 * Open an include file. This routine must always return a valid
1312 * file pointer if it returns - it's responsible for throwing an
1313 * ERR_FATAL and bombing out completely if not. It should also try
1314 * the include path one by one until it finds the file or reaches
1315 * the end of the path.
1317 static FILE *inc_fopen(const char *file, StrList **dhead, StrList ***dtail,
1318 bool missing_ok)
1320 FILE *fp;
1321 char *prefix = "";
1322 IncPath *ip = ipath;
1323 int len = strlen(file);
1324 size_t prefix_len = 0;
1325 StrList *sl;
1327 while (1) {
1328 sl = nasm_malloc(prefix_len+len+1+sizeof sl->next);
1329 memcpy(sl->str, prefix, prefix_len);
1330 memcpy(sl->str+prefix_len, file, len+1);
1331 fp = fopen(sl->str, "r");
1332 if (fp && dhead && !in_list(*dhead, sl->str)) {
1333 sl->next = NULL;
1334 **dtail = sl;
1335 *dtail = &sl->next;
1336 } else {
1337 nasm_free(sl);
1339 if (fp)
1340 return fp;
1341 if (!ip) {
1342 if (!missing_ok)
1343 break;
1344 prefix = NULL;
1345 } else {
1346 prefix = ip->path;
1347 ip = ip->next;
1349 if (prefix) {
1350 prefix_len = strlen(prefix);
1351 } else {
1352 /* -MG given and file not found */
1353 if (dhead && !in_list(*dhead, file)) {
1354 sl = nasm_malloc(len+1+sizeof sl->next);
1355 sl->next = NULL;
1356 strcpy(sl->str, file);
1357 **dtail = sl;
1358 *dtail = &sl->next;
1360 return NULL;
1364 error(ERR_FATAL, "unable to open include file `%s'", file);
1365 return NULL; /* never reached - placate compilers */
1369 * Determine if we should warn on defining a single-line macro of
1370 * name `name', with `nparam' parameters. If nparam is 0 or -1, will
1371 * return true if _any_ single-line macro of that name is defined.
1372 * Otherwise, will return true if a single-line macro with either
1373 * `nparam' or no parameters is defined.
1375 * If a macro with precisely the right number of parameters is
1376 * defined, or nparam is -1, the address of the definition structure
1377 * will be returned in `defn'; otherwise NULL will be returned. If `defn'
1378 * is NULL, no action will be taken regarding its contents, and no
1379 * error will occur.
1381 * Note that this is also called with nparam zero to resolve
1382 * `ifdef'.
1384 * If you already know which context macro belongs to, you can pass
1385 * the context pointer as first parameter; if you won't but name begins
1386 * with %$ the context will be automatically computed. If all_contexts
1387 * is true, macro will be searched in outer contexts as well.
1389 static bool
1390 smacro_defined(Context * ctx, const char *name, int nparam, SMacro ** defn,
1391 bool nocase)
1393 struct hash_table *smtbl;
1394 SMacro *m;
1396 if (ctx) {
1397 smtbl = &ctx->localmac;
1398 } else if (name[0] == '%' && name[1] == '$') {
1399 if (cstk)
1400 ctx = get_ctx(name, false);
1401 if (!ctx)
1402 return false; /* got to return _something_ */
1403 smtbl = &ctx->localmac;
1404 } else {
1405 smtbl = &smacros;
1407 m = (SMacro *) hash_findix(smtbl, name);
1409 while (m) {
1410 if (!mstrcmp(m->name, name, m->casesense && nocase) &&
1411 (nparam <= 0 || m->nparam == 0 || nparam == (int) m->nparam)) {
1412 if (defn) {
1413 if (nparam == (int) m->nparam || nparam == -1)
1414 *defn = m;
1415 else
1416 *defn = NULL;
1418 return true;
1420 m = m->next;
1423 return false;
1427 * Count and mark off the parameters in a multi-line macro call.
1428 * This is called both from within the multi-line macro expansion
1429 * code, and also to mark off the default parameters when provided
1430 * in a %macro definition line.
1432 static void count_mmac_params(Token * t, int *nparam, Token *** params)
1434 int paramsize, brace;
1436 *nparam = paramsize = 0;
1437 *params = NULL;
1438 while (t) {
1439 /* +1: we need space for the final NULL */
1440 if (*nparam+1 >= paramsize) {
1441 paramsize += PARAM_DELTA;
1442 *params = nasm_realloc(*params, sizeof(**params) * paramsize);
1444 skip_white_(t);
1445 brace = false;
1446 if (tok_is_(t, "{"))
1447 brace = true;
1448 (*params)[(*nparam)++] = t;
1449 while (tok_isnt_(t, brace ? "}" : ","))
1450 t = t->next;
1451 if (t) { /* got a comma/brace */
1452 t = t->next;
1453 if (brace) {
1455 * Now we've found the closing brace, look further
1456 * for the comma.
1458 skip_white_(t);
1459 if (tok_isnt_(t, ",")) {
1460 error(ERR_NONFATAL,
1461 "braces do not enclose all of macro parameter");
1462 while (tok_isnt_(t, ","))
1463 t = t->next;
1465 if (t)
1466 t = t->next; /* eat the comma */
1473 * Determine whether one of the various `if' conditions is true or
1474 * not.
1476 * We must free the tline we get passed.
1478 static bool if_condition(Token * tline, enum preproc_token ct)
1480 enum pp_conditional i = PP_COND(ct);
1481 bool j;
1482 Token *t, *tt, **tptr, *origline;
1483 struct tokenval tokval;
1484 expr *evalresult;
1485 enum pp_token_type needtype;
1487 origline = tline;
1489 switch (i) {
1490 case PPC_IFCTX:
1491 j = false; /* have we matched yet? */
1492 while (true) {
1493 skip_white_(tline);
1494 if (!tline)
1495 break;
1496 if (tline->type != TOK_ID) {
1497 error(ERR_NONFATAL,
1498 "`%s' expects context identifiers", pp_directives[ct]);
1499 free_tlist(origline);
1500 return -1;
1502 if (cstk && cstk->name && !nasm_stricmp(tline->text, cstk->name))
1503 j = true;
1504 tline = tline->next;
1506 break;
1508 case PPC_IFDEF:
1509 j = false; /* have we matched yet? */
1510 while (tline) {
1511 skip_white_(tline);
1512 if (!tline || (tline->type != TOK_ID &&
1513 (tline->type != TOK_PREPROC_ID ||
1514 tline->text[1] != '$'))) {
1515 error(ERR_NONFATAL,
1516 "`%s' expects macro identifiers", pp_directives[ct]);
1517 goto fail;
1519 if (smacro_defined(NULL, tline->text, 0, NULL, true))
1520 j = true;
1521 tline = tline->next;
1523 break;
1525 case PPC_IFIDN:
1526 case PPC_IFIDNI:
1527 tline = expand_smacro(tline);
1528 t = tt = tline;
1529 while (tok_isnt_(tt, ","))
1530 tt = tt->next;
1531 if (!tt) {
1532 error(ERR_NONFATAL,
1533 "`%s' expects two comma-separated arguments",
1534 pp_directives[ct]);
1535 goto fail;
1537 tt = tt->next;
1538 j = true; /* assume equality unless proved not */
1539 while ((t->type != TOK_OTHER || strcmp(t->text, ",")) && tt) {
1540 if (tt->type == TOK_OTHER && !strcmp(tt->text, ",")) {
1541 error(ERR_NONFATAL, "`%s': more than one comma on line",
1542 pp_directives[ct]);
1543 goto fail;
1545 if (t->type == TOK_WHITESPACE) {
1546 t = t->next;
1547 continue;
1549 if (tt->type == TOK_WHITESPACE) {
1550 tt = tt->next;
1551 continue;
1553 if (tt->type != t->type) {
1554 j = false; /* found mismatching tokens */
1555 break;
1557 /* When comparing strings, need to unquote them first */
1558 if (t->type == TOK_STRING) {
1559 size_t l1 = nasm_unquote(t->text, NULL);
1560 size_t l2 = nasm_unquote(tt->text, NULL);
1562 if (l1 != l2) {
1563 j = false;
1564 break;
1566 if (mmemcmp(t->text, tt->text, l1, i == PPC_IFIDN)) {
1567 j = false;
1568 break;
1570 } else if (mstrcmp(tt->text, t->text, i == PPC_IFIDN) != 0) {
1571 j = false; /* found mismatching tokens */
1572 break;
1575 t = t->next;
1576 tt = tt->next;
1578 if ((t->type != TOK_OTHER || strcmp(t->text, ",")) || tt)
1579 j = false; /* trailing gunk on one end or other */
1580 break;
1582 case PPC_IFMACRO:
1584 bool found = false;
1585 MMacro searching, *mmac;
1587 tline = tline->next;
1588 skip_white_(tline);
1589 tline = expand_id(tline);
1590 if (!tok_type_(tline, TOK_ID)) {
1591 error(ERR_NONFATAL,
1592 "`%s' expects a macro name", pp_directives[ct]);
1593 goto fail;
1595 searching.name = nasm_strdup(tline->text);
1596 searching.casesense = true;
1597 searching.plus = false;
1598 searching.nolist = false;
1599 searching.in_progress = 0;
1600 searching.rep_nest = NULL;
1601 searching.nparam_min = 0;
1602 searching.nparam_max = INT_MAX;
1603 tline = expand_smacro(tline->next);
1604 skip_white_(tline);
1605 if (!tline) {
1606 } else if (!tok_type_(tline, TOK_NUMBER)) {
1607 error(ERR_NONFATAL,
1608 "`%s' expects a parameter count or nothing",
1609 pp_directives[ct]);
1610 } else {
1611 searching.nparam_min = searching.nparam_max =
1612 readnum(tline->text, &j);
1613 if (j)
1614 error(ERR_NONFATAL,
1615 "unable to parse parameter count `%s'",
1616 tline->text);
1618 if (tline && tok_is_(tline->next, "-")) {
1619 tline = tline->next->next;
1620 if (tok_is_(tline, "*"))
1621 searching.nparam_max = INT_MAX;
1622 else if (!tok_type_(tline, TOK_NUMBER))
1623 error(ERR_NONFATAL,
1624 "`%s' expects a parameter count after `-'",
1625 pp_directives[ct]);
1626 else {
1627 searching.nparam_max = readnum(tline->text, &j);
1628 if (j)
1629 error(ERR_NONFATAL,
1630 "unable to parse parameter count `%s'",
1631 tline->text);
1632 if (searching.nparam_min > searching.nparam_max)
1633 error(ERR_NONFATAL,
1634 "minimum parameter count exceeds maximum");
1637 if (tline && tok_is_(tline->next, "+")) {
1638 tline = tline->next;
1639 searching.plus = true;
1641 mmac = (MMacro *) hash_findix(&mmacros, searching.name);
1642 while (mmac) {
1643 if (!strcmp(mmac->name, searching.name) &&
1644 (mmac->nparam_min <= searching.nparam_max
1645 || searching.plus)
1646 && (searching.nparam_min <= mmac->nparam_max
1647 || mmac->plus)) {
1648 found = true;
1649 break;
1651 mmac = mmac->next;
1653 if(tline && tline->next)
1654 error(ERR_WARNING, "trailing garbage after %%ifmacro ignored");
1655 nasm_free(searching.name);
1656 j = found;
1657 break;
1660 case PPC_IFID:
1661 needtype = TOK_ID;
1662 goto iftype;
1663 case PPC_IFNUM:
1664 needtype = TOK_NUMBER;
1665 goto iftype;
1666 case PPC_IFSTR:
1667 needtype = TOK_STRING;
1668 goto iftype;
1670 iftype:
1671 t = tline = expand_smacro(tline);
1673 while (tok_type_(t, TOK_WHITESPACE) ||
1674 (needtype == TOK_NUMBER &&
1675 tok_type_(t, TOK_OTHER) &&
1676 (t->text[0] == '-' || t->text[0] == '+') &&
1677 !t->text[1]))
1678 t = t->next;
1680 j = tok_type_(t, needtype);
1681 break;
1683 case PPC_IFTOKEN:
1684 t = tline = expand_smacro(tline);
1685 while (tok_type_(t, TOK_WHITESPACE))
1686 t = t->next;
1688 j = false;
1689 if (t) {
1690 t = t->next; /* Skip the actual token */
1691 while (tok_type_(t, TOK_WHITESPACE))
1692 t = t->next;
1693 j = !t; /* Should be nothing left */
1695 break;
1697 case PPC_IFEMPTY:
1698 t = tline = expand_smacro(tline);
1699 while (tok_type_(t, TOK_WHITESPACE))
1700 t = t->next;
1702 j = !t; /* Should be empty */
1703 break;
1705 case PPC_IF:
1706 t = tline = expand_smacro(tline);
1707 tptr = &t;
1708 tokval.t_type = TOKEN_INVALID;
1709 evalresult = evaluate(ppscan, tptr, &tokval,
1710 NULL, pass | CRITICAL, error, NULL);
1711 if (!evalresult)
1712 return -1;
1713 if (tokval.t_type)
1714 error(ERR_WARNING,
1715 "trailing garbage after expression ignored");
1716 if (!is_simple(evalresult)) {
1717 error(ERR_NONFATAL,
1718 "non-constant value given to `%s'", pp_directives[ct]);
1719 goto fail;
1721 j = reloc_value(evalresult) != 0;
1722 break;
1724 default:
1725 error(ERR_FATAL,
1726 "preprocessor directive `%s' not yet implemented",
1727 pp_directives[ct]);
1728 goto fail;
1731 free_tlist(origline);
1732 return j ^ PP_NEGATIVE(ct);
1734 fail:
1735 free_tlist(origline);
1736 return -1;
1740 * Common code for defining an smacro
1742 static bool define_smacro(Context *ctx, char *mname, bool casesense,
1743 int nparam, Token *expansion)
1745 SMacro *smac, **smhead;
1746 struct hash_table *smtbl;
1748 if (smacro_defined(ctx, mname, nparam, &smac, casesense)) {
1749 if (!smac) {
1750 error(ERR_WARNING,
1751 "single-line macro `%s' defined both with and"
1752 " without parameters", mname);
1754 /* Some instances of the old code considered this a failure,
1755 some others didn't. What is the right thing to do here? */
1756 free_tlist(expansion);
1757 return false; /* Failure */
1758 } else {
1760 * We're redefining, so we have to take over an
1761 * existing SMacro structure. This means freeing
1762 * what was already in it.
1764 nasm_free(smac->name);
1765 free_tlist(smac->expansion);
1767 } else {
1768 smtbl = ctx ? &ctx->localmac : &smacros;
1769 smhead = (SMacro **) hash_findi_add(smtbl, mname);
1770 smac = nasm_malloc(sizeof(SMacro));
1771 smac->next = *smhead;
1772 *smhead = smac;
1774 smac->name = nasm_strdup(mname);
1775 smac->casesense = casesense;
1776 smac->nparam = nparam;
1777 smac->expansion = expansion;
1778 smac->in_progress = false;
1779 return true; /* Success */
1783 * Undefine an smacro
1785 static void undef_smacro(Context *ctx, const char *mname)
1787 SMacro **smhead, *s, **sp;
1788 struct hash_table *smtbl;
1790 smtbl = ctx ? &ctx->localmac : &smacros;
1791 smhead = (SMacro **)hash_findi(smtbl, mname, NULL);
1793 if (smhead) {
1795 * We now have a macro name... go hunt for it.
1797 sp = smhead;
1798 while ((s = *sp) != NULL) {
1799 if (!mstrcmp(s->name, mname, s->casesense)) {
1800 *sp = s->next;
1801 nasm_free(s->name);
1802 free_tlist(s->expansion);
1803 nasm_free(s);
1804 } else {
1805 sp = &s->next;
1812 * Parse a mmacro specification.
1814 static bool parse_mmacro_spec(Token *tline, MMacro *def, const char *directive)
1816 bool err;
1818 tline = tline->next;
1819 skip_white_(tline);
1820 tline = expand_id(tline);
1821 if (!tok_type_(tline, TOK_ID)) {
1822 error(ERR_NONFATAL, "`%s' expects a macro name", directive);
1823 return false;
1826 def->name = nasm_strdup(tline->text);
1827 def->plus = false;
1828 def->nolist = false;
1829 def->in_progress = 0;
1830 def->rep_nest = NULL;
1831 def->nparam_min = 0;
1832 def->nparam_max = 0;
1834 tline = expand_smacro(tline->next);
1835 skip_white_(tline);
1836 if (!tok_type_(tline, TOK_NUMBER)) {
1837 error(ERR_NONFATAL, "`%s' expects a parameter count", directive);
1838 } else {
1839 def->nparam_min = def->nparam_max =
1840 readnum(tline->text, &err);
1841 if (err)
1842 error(ERR_NONFATAL,
1843 "unable to parse parameter count `%s'", tline->text);
1845 if (tline && tok_is_(tline->next, "-")) {
1846 tline = tline->next->next;
1847 if (tok_is_(tline, "*")) {
1848 def->nparam_max = INT_MAX;
1849 } else if (!tok_type_(tline, TOK_NUMBER)) {
1850 error(ERR_NONFATAL,
1851 "`%s' expects a parameter count after `-'", directive);
1852 } else {
1853 def->nparam_max = readnum(tline->text, &err);
1854 if (err) {
1855 error(ERR_NONFATAL, "unable to parse parameter count `%s'",
1856 tline->text);
1858 if (def->nparam_min > def->nparam_max) {
1859 error(ERR_NONFATAL, "minimum parameter count exceeds maximum");
1863 if (tline && tok_is_(tline->next, "+")) {
1864 tline = tline->next;
1865 def->plus = true;
1867 if (tline && tok_type_(tline->next, TOK_ID) &&
1868 !nasm_stricmp(tline->next->text, ".nolist")) {
1869 tline = tline->next;
1870 def->nolist = true;
1874 * Handle default parameters.
1876 if (tline && tline->next) {
1877 def->dlist = tline->next;
1878 tline->next = NULL;
1879 count_mmac_params(def->dlist, &def->ndefs, &def->defaults);
1880 } else {
1881 def->dlist = NULL;
1882 def->defaults = NULL;
1884 def->expansion = NULL;
1886 if(def->defaults &&
1887 def->ndefs > def->nparam_max - def->nparam_min &&
1888 !def->plus)
1889 error(ERR_WARNING | ERR_WARN_MDP, "too many default macro parameters");
1891 return true;
1896 * Decode a size directive
1898 static int parse_size(const char *str) {
1899 static const char *size_names[] =
1900 { "byte", "dword", "oword", "qword", "tword", "word", "yword" };
1901 static const int sizes[] =
1902 { 0, 1, 4, 16, 8, 10, 2, 32 };
1904 return sizes[bsii(str, size_names, elements(size_names))+1];
1908 * find and process preprocessor directive in passed line
1909 * Find out if a line contains a preprocessor directive, and deal
1910 * with it if so.
1912 * If a directive _is_ found, it is the responsibility of this routine
1913 * (and not the caller) to free_tlist() the line.
1915 * @param tline a pointer to the current tokeninzed line linked list
1916 * @return DIRECTIVE_FOUND or NO_DIRECTIVE_FOUND
1919 static int do_directive(Token * tline)
1921 enum preproc_token i;
1922 int j;
1923 bool err;
1924 int nparam;
1925 bool nolist;
1926 bool casesense;
1927 int k, m;
1928 int offset;
1929 char *p, *pp, *mname;
1930 Include *inc;
1931 Context *ctx;
1932 Cond *cond;
1933 MMacro *mmac, **mmhead;
1934 Token *t, *tt, *param_start, *macro_start, *last, **tptr, *origline;
1935 Line *l;
1936 struct tokenval tokval;
1937 expr *evalresult;
1938 MMacro *tmp_defining; /* Used when manipulating rep_nest */
1939 int64_t count;
1940 size_t len;
1941 int severity;
1943 origline = tline;
1945 skip_white_(tline);
1946 if (!tline || !tok_type_(tline, TOK_PREPROC_ID) ||
1947 (tline->text[1] == '%' || tline->text[1] == '$'
1948 || tline->text[1] == '!'))
1949 return NO_DIRECTIVE_FOUND;
1951 i = pp_token_hash(tline->text);
1954 * If we're in a non-emitting branch of a condition construct,
1955 * or walking to the end of an already terminated %rep block,
1956 * we should ignore all directives except for condition
1957 * directives.
1959 if (((istk->conds && !emitting(istk->conds->state)) ||
1960 (istk->mstk && !istk->mstk->in_progress)) && !is_condition(i)) {
1961 return NO_DIRECTIVE_FOUND;
1965 * If we're defining a macro or reading a %rep block, we should
1966 * ignore all directives except for %macro/%imacro (which nest),
1967 * %endm/%endmacro, and (only if we're in a %rep block) %endrep.
1968 * If we're in a %rep block, another %rep nests, so should be let through.
1970 if (defining && i != PP_MACRO && i != PP_IMACRO &&
1971 i != PP_ENDMACRO && i != PP_ENDM &&
1972 (defining->name || (i != PP_ENDREP && i != PP_REP))) {
1973 return NO_DIRECTIVE_FOUND;
1976 if (defining) {
1977 if (i == PP_MACRO || i == PP_IMACRO) {
1978 nested_mac_count++;
1979 return NO_DIRECTIVE_FOUND;
1980 } else if (nested_mac_count > 0) {
1981 if (i == PP_ENDMACRO) {
1982 nested_mac_count--;
1983 return NO_DIRECTIVE_FOUND;
1986 if (!defining->name) {
1987 if (i == PP_REP) {
1988 nested_rep_count++;
1989 return NO_DIRECTIVE_FOUND;
1990 } else if (nested_rep_count > 0) {
1991 if (i == PP_ENDREP) {
1992 nested_rep_count--;
1993 return NO_DIRECTIVE_FOUND;
1999 switch (i) {
2000 case PP_INVALID:
2001 error(ERR_NONFATAL, "unknown preprocessor directive `%s'",
2002 tline->text);
2003 return NO_DIRECTIVE_FOUND; /* didn't get it */
2005 case PP_STACKSIZE:
2006 /* Directive to tell NASM what the default stack size is. The
2007 * default is for a 16-bit stack, and this can be overriden with
2008 * %stacksize large.
2009 * the following form:
2011 * ARG arg1:WORD, arg2:DWORD, arg4:QWORD
2013 tline = tline->next;
2014 if (tline && tline->type == TOK_WHITESPACE)
2015 tline = tline->next;
2016 if (!tline || tline->type != TOK_ID) {
2017 error(ERR_NONFATAL, "`%%stacksize' missing size parameter");
2018 free_tlist(origline);
2019 return DIRECTIVE_FOUND;
2021 if (nasm_stricmp(tline->text, "flat") == 0) {
2022 /* All subsequent ARG directives are for a 32-bit stack */
2023 StackSize = 4;
2024 StackPointer = "ebp";
2025 ArgOffset = 8;
2026 LocalOffset = 0;
2027 } else if (nasm_stricmp(tline->text, "flat64") == 0) {
2028 /* All subsequent ARG directives are for a 64-bit stack */
2029 StackSize = 8;
2030 StackPointer = "rbp";
2031 ArgOffset = 8;
2032 LocalOffset = 0;
2033 } else if (nasm_stricmp(tline->text, "large") == 0) {
2034 /* All subsequent ARG directives are for a 16-bit stack,
2035 * far function call.
2037 StackSize = 2;
2038 StackPointer = "bp";
2039 ArgOffset = 4;
2040 LocalOffset = 0;
2041 } else if (nasm_stricmp(tline->text, "small") == 0) {
2042 /* All subsequent ARG directives are for a 16-bit stack,
2043 * far function call. We don't support near functions.
2045 StackSize = 2;
2046 StackPointer = "bp";
2047 ArgOffset = 6;
2048 LocalOffset = 0;
2049 } else {
2050 error(ERR_NONFATAL, "`%%stacksize' invalid size type");
2051 free_tlist(origline);
2052 return DIRECTIVE_FOUND;
2054 free_tlist(origline);
2055 return DIRECTIVE_FOUND;
2057 case PP_ARG:
2058 /* TASM like ARG directive to define arguments to functions, in
2059 * the following form:
2061 * ARG arg1:WORD, arg2:DWORD, arg4:QWORD
2063 offset = ArgOffset;
2064 do {
2065 char *arg, directive[256];
2066 int size = StackSize;
2068 /* Find the argument name */
2069 tline = tline->next;
2070 if (tline && tline->type == TOK_WHITESPACE)
2071 tline = tline->next;
2072 if (!tline || tline->type != TOK_ID) {
2073 error(ERR_NONFATAL, "`%%arg' missing argument parameter");
2074 free_tlist(origline);
2075 return DIRECTIVE_FOUND;
2077 arg = tline->text;
2079 /* Find the argument size type */
2080 tline = tline->next;
2081 if (!tline || tline->type != TOK_OTHER
2082 || tline->text[0] != ':') {
2083 error(ERR_NONFATAL,
2084 "Syntax error processing `%%arg' directive");
2085 free_tlist(origline);
2086 return DIRECTIVE_FOUND;
2088 tline = tline->next;
2089 if (!tline || tline->type != TOK_ID) {
2090 error(ERR_NONFATAL, "`%%arg' missing size type parameter");
2091 free_tlist(origline);
2092 return DIRECTIVE_FOUND;
2095 /* Allow macro expansion of type parameter */
2096 tt = tokenize(tline->text);
2097 tt = expand_smacro(tt);
2098 size = parse_size(tt->text);
2099 if (!size) {
2100 error(ERR_NONFATAL,
2101 "Invalid size type for `%%arg' missing directive");
2102 free_tlist(tt);
2103 free_tlist(origline);
2104 return DIRECTIVE_FOUND;
2106 free_tlist(tt);
2108 /* Round up to even stack slots */
2109 size = (size+StackSize-1) & ~(StackSize-1);
2111 /* Now define the macro for the argument */
2112 snprintf(directive, sizeof(directive), "%%define %s (%s+%d)",
2113 arg, StackPointer, offset);
2114 do_directive(tokenize(directive));
2115 offset += size;
2117 /* Move to the next argument in the list */
2118 tline = tline->next;
2119 if (tline && tline->type == TOK_WHITESPACE)
2120 tline = tline->next;
2121 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2122 ArgOffset = offset;
2123 free_tlist(origline);
2124 return DIRECTIVE_FOUND;
2126 case PP_LOCAL:
2127 /* TASM like LOCAL directive to define local variables for a
2128 * function, in the following form:
2130 * LOCAL local1:WORD, local2:DWORD, local4:QWORD = LocalSize
2132 * The '= LocalSize' at the end is ignored by NASM, but is
2133 * required by TASM to define the local parameter size (and used
2134 * by the TASM macro package).
2136 offset = LocalOffset;
2137 do {
2138 char *local, directive[256];
2139 int size = StackSize;
2141 /* Find the argument name */
2142 tline = tline->next;
2143 if (tline && tline->type == TOK_WHITESPACE)
2144 tline = tline->next;
2145 if (!tline || tline->type != TOK_ID) {
2146 error(ERR_NONFATAL,
2147 "`%%local' missing argument parameter");
2148 free_tlist(origline);
2149 return DIRECTIVE_FOUND;
2151 local = tline->text;
2153 /* Find the argument size type */
2154 tline = tline->next;
2155 if (!tline || tline->type != TOK_OTHER
2156 || tline->text[0] != ':') {
2157 error(ERR_NONFATAL,
2158 "Syntax error processing `%%local' directive");
2159 free_tlist(origline);
2160 return DIRECTIVE_FOUND;
2162 tline = tline->next;
2163 if (!tline || tline->type != TOK_ID) {
2164 error(ERR_NONFATAL,
2165 "`%%local' missing size type parameter");
2166 free_tlist(origline);
2167 return DIRECTIVE_FOUND;
2170 /* Allow macro expansion of type parameter */
2171 tt = tokenize(tline->text);
2172 tt = expand_smacro(tt);
2173 size = parse_size(tt->text);
2174 if (!size) {
2175 error(ERR_NONFATAL,
2176 "Invalid size type for `%%local' missing directive");
2177 free_tlist(tt);
2178 free_tlist(origline);
2179 return DIRECTIVE_FOUND;
2181 free_tlist(tt);
2183 /* Round up to even stack slots */
2184 size = (size+StackSize-1) & ~(StackSize-1);
2186 offset += size; /* Negative offset, increment before */
2188 /* Now define the macro for the argument */
2189 snprintf(directive, sizeof(directive), "%%define %s (%s-%d)",
2190 local, StackPointer, offset);
2191 do_directive(tokenize(directive));
2193 /* Now define the assign to setup the enter_c macro correctly */
2194 snprintf(directive, sizeof(directive),
2195 "%%assign %%$localsize %%$localsize+%d", size);
2196 do_directive(tokenize(directive));
2198 /* Move to the next argument in the list */
2199 tline = tline->next;
2200 if (tline && tline->type == TOK_WHITESPACE)
2201 tline = tline->next;
2202 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2203 LocalOffset = offset;
2204 free_tlist(origline);
2205 return DIRECTIVE_FOUND;
2207 case PP_CLEAR:
2208 if (tline->next)
2209 error(ERR_WARNING, "trailing garbage after `%%clear' ignored");
2210 free_macros();
2211 init_macros();
2212 free_tlist(origline);
2213 return DIRECTIVE_FOUND;
2215 case PP_DEPEND:
2216 t = tline->next = expand_smacro(tline->next);
2217 skip_white_(t);
2218 if (!t || (t->type != TOK_STRING &&
2219 t->type != TOK_INTERNAL_STRING)) {
2220 error(ERR_NONFATAL, "`%%depend' expects a file name");
2221 free_tlist(origline);
2222 return DIRECTIVE_FOUND; /* but we did _something_ */
2224 if (t->next)
2225 error(ERR_WARNING,
2226 "trailing garbage after `%%depend' ignored");
2227 p = t->text;
2228 if (t->type != TOK_INTERNAL_STRING)
2229 nasm_unquote(p, NULL);
2230 if (dephead && !in_list(*dephead, p)) {
2231 StrList *sl = nasm_malloc(strlen(p)+1+sizeof sl->next);
2232 sl->next = NULL;
2233 strcpy(sl->str, p);
2234 *deptail = sl;
2235 deptail = &sl->next;
2237 free_tlist(origline);
2238 return DIRECTIVE_FOUND;
2240 case PP_INCLUDE:
2241 t = tline->next = expand_smacro(tline->next);
2242 skip_white_(t);
2244 if (!t || (t->type != TOK_STRING &&
2245 t->type != TOK_INTERNAL_STRING)) {
2246 error(ERR_NONFATAL, "`%%include' expects a file name");
2247 free_tlist(origline);
2248 return DIRECTIVE_FOUND; /* but we did _something_ */
2250 if (t->next)
2251 error(ERR_WARNING,
2252 "trailing garbage after `%%include' ignored");
2253 p = t->text;
2254 if (t->type != TOK_INTERNAL_STRING)
2255 nasm_unquote(p, NULL);
2256 inc = nasm_malloc(sizeof(Include));
2257 inc->next = istk;
2258 inc->conds = NULL;
2259 inc->fp = inc_fopen(p, dephead, &deptail, pass == 0);
2260 if (!inc->fp) {
2261 /* -MG given but file not found */
2262 nasm_free(inc);
2263 } else {
2264 inc->fname = src_set_fname(nasm_strdup(p));
2265 inc->lineno = src_set_linnum(0);
2266 inc->lineinc = 1;
2267 inc->expansion = NULL;
2268 inc->mstk = NULL;
2269 istk = inc;
2270 list->uplevel(LIST_INCLUDE);
2272 free_tlist(origline);
2273 return DIRECTIVE_FOUND;
2275 case PP_USE:
2277 static macros_t *use_pkg;
2278 const char *pkg_macro;
2280 t = tline->next = expand_smacro(tline->next);
2281 skip_white_(t);
2283 if (!t || (t->type != TOK_STRING &&
2284 t->type != TOK_INTERNAL_STRING &&
2285 t->type != TOK_ID)) {
2286 error(ERR_NONFATAL, "`%%use' expects a package name");
2287 free_tlist(origline);
2288 return DIRECTIVE_FOUND; /* but we did _something_ */
2290 if (t->next)
2291 error(ERR_WARNING,
2292 "trailing garbage after `%%use' ignored");
2293 if (t->type == TOK_STRING)
2294 nasm_unquote(t->text, NULL);
2295 use_pkg = nasm_stdmac_find_package(t->text);
2296 if (!use_pkg)
2297 error(ERR_NONFATAL, "unknown `%%use' package: %s", t->text);
2298 /* The first string will be <%define>__USE_*__ */
2299 pkg_macro = (char *)use_pkg + 1;
2300 if (!smacro_defined(NULL, pkg_macro, 0, NULL, true)) {
2301 /* Not already included, go ahead and include it */
2302 stdmacpos = use_pkg;
2304 free_tlist(origline);
2305 return DIRECTIVE_FOUND;
2307 case PP_PUSH:
2308 tline = tline->next;
2309 skip_white_(tline);
2310 tline = expand_id(tline);
2311 if (tline) {
2312 if (!tok_type_(tline, TOK_ID)) {
2313 error(ERR_NONFATAL, "`%%push' expects a context identifier");
2314 free_tlist(origline);
2315 return DIRECTIVE_FOUND; /* but we did _something_ */
2317 if (tline->next)
2318 error(ERR_WARNING, "trailing garbage after `%%push' ignored");
2319 p = nasm_strdup(tline->text);
2320 } else {
2321 p = NULL; /* Anonymous context */
2323 ctx = nasm_malloc(sizeof(Context));
2324 ctx->next = cstk;
2325 hash_init(&ctx->localmac, HASH_SMALL);
2326 ctx->name = p;
2327 ctx->number = unique++;
2328 cstk = ctx;
2329 free_tlist(origline);
2330 return DIRECTIVE_FOUND;
2332 case PP_REPL:
2333 tline = tline->next;
2334 skip_white_(tline);
2335 tline = expand_id(tline);
2336 if (tline) {
2337 if (!tok_type_(tline, TOK_ID)) {
2338 error(ERR_NONFATAL, "`%%repl' expects a context identifier");
2339 free_tlist(origline);
2340 return DIRECTIVE_FOUND; /* but we did _something_ */
2342 if (tline->next)
2343 error(ERR_WARNING, "trailing garbage after `%%repl' ignored");
2344 p = nasm_strdup(tline->text);
2345 } else {
2346 p = NULL;
2348 if (!cstk)
2349 error(ERR_NONFATAL, "`%%repl': context stack is empty");
2350 else {
2351 nasm_free(cstk->name);
2352 cstk->name = p;
2354 free_tlist(origline);
2355 return DIRECTIVE_FOUND;
2357 case PP_POP:
2358 if (tline->next)
2359 error(ERR_WARNING, "trailing garbage after `%%pop' ignored");
2360 if (!cstk)
2361 error(ERR_NONFATAL, "`%%pop': context stack is already empty");
2362 else
2363 ctx_pop();
2364 free_tlist(origline);
2365 return DIRECTIVE_FOUND;
2367 case PP_FATAL:
2368 severity = ERR_FATAL|ERR_NO_SEVERITY;
2369 goto issue_error;
2370 case PP_ERROR:
2371 severity = ERR_NONFATAL|ERR_NO_SEVERITY;
2372 goto issue_error;
2373 case PP_WARNING:
2374 severity = ERR_WARNING|ERR_NO_SEVERITY;
2375 goto issue_error;
2377 issue_error:
2379 /* Only error out if this is the final pass */
2380 if (pass != 2 && i != PP_FATAL)
2381 return DIRECTIVE_FOUND;
2383 tline->next = expand_smacro(tline->next);
2384 tline = tline->next;
2385 skip_white_(tline);
2386 t = tline ? tline->next : NULL;
2387 skip_white_(t);
2388 if (tok_type_(tline, TOK_STRING) && !t) {
2389 /* The line contains only a quoted string */
2390 p = tline->text;
2391 nasm_unquote(p, NULL);
2392 error(severity, "%s: %s", pp_directives[i], p);
2393 } else {
2394 /* Not a quoted string, or more than a quoted string */
2395 p = detoken(tline, false);
2396 error(severity, "%s: %s", pp_directives[i], p);
2397 nasm_free(p);
2399 free_tlist(origline);
2400 return DIRECTIVE_FOUND;
2403 CASE_PP_IF:
2404 if (istk->conds && !emitting(istk->conds->state))
2405 j = COND_NEVER;
2406 else {
2407 j = if_condition(tline->next, i);
2408 tline->next = NULL; /* it got freed */
2409 j = j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2411 cond = nasm_malloc(sizeof(Cond));
2412 cond->next = istk->conds;
2413 cond->state = j;
2414 istk->conds = cond;
2415 free_tlist(origline);
2416 return DIRECTIVE_FOUND;
2418 CASE_PP_ELIF:
2419 if (!istk->conds)
2420 error(ERR_FATAL, "`%s': no matching `%%if'", pp_directives[i]);
2421 switch(istk->conds->state) {
2422 case COND_IF_TRUE:
2423 istk->conds->state = COND_DONE;
2424 break;
2426 case COND_DONE:
2427 case COND_NEVER:
2428 break;
2430 case COND_ELSE_TRUE:
2431 case COND_ELSE_FALSE:
2432 error_precond(ERR_WARNING, "`%%elif' after `%%else' ignored");
2433 istk->conds->state = COND_NEVER;
2434 break;
2436 case COND_IF_FALSE:
2438 * IMPORTANT: In the case of %if, we will already have
2439 * called expand_mmac_params(); however, if we're
2440 * processing an %elif we must have been in a
2441 * non-emitting mode, which would have inhibited
2442 * the normal invocation of expand_mmac_params(). Therefore,
2443 * we have to do it explicitly here.
2445 j = if_condition(expand_mmac_params(tline->next), i);
2446 tline->next = NULL; /* it got freed */
2447 istk->conds->state =
2448 j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2449 break;
2451 free_tlist(origline);
2452 return DIRECTIVE_FOUND;
2454 case PP_ELSE:
2455 if (tline->next)
2456 error_precond(ERR_WARNING, "trailing garbage after `%%else' ignored");
2457 if (!istk->conds)
2458 error(ERR_FATAL, "`%%else': no matching `%%if'");
2459 switch(istk->conds->state) {
2460 case COND_IF_TRUE:
2461 case COND_DONE:
2462 istk->conds->state = COND_ELSE_FALSE;
2463 break;
2465 case COND_NEVER:
2466 break;
2468 case COND_IF_FALSE:
2469 istk->conds->state = COND_ELSE_TRUE;
2470 break;
2472 case COND_ELSE_TRUE:
2473 case COND_ELSE_FALSE:
2474 error_precond(ERR_WARNING, "`%%else' after `%%else' ignored.");
2475 istk->conds->state = COND_NEVER;
2476 break;
2478 free_tlist(origline);
2479 return DIRECTIVE_FOUND;
2481 case PP_ENDIF:
2482 if (tline->next)
2483 error_precond(ERR_WARNING, "trailing garbage after `%%endif' ignored");
2484 if (!istk->conds)
2485 error(ERR_FATAL, "`%%endif': no matching `%%if'");
2486 cond = istk->conds;
2487 istk->conds = cond->next;
2488 nasm_free(cond);
2489 free_tlist(origline);
2490 return DIRECTIVE_FOUND;
2492 case PP_MACRO:
2493 case PP_IMACRO:
2494 if (defining) {
2495 error(ERR_FATAL,
2496 "`%%%smacro': already defining a macro",
2497 (i == PP_IMACRO ? "i" : ""));
2498 return DIRECTIVE_FOUND;
2500 defining = nasm_malloc(sizeof(MMacro));
2501 defining->casesense = (i == PP_MACRO);
2502 if (!parse_mmacro_spec(tline, defining, pp_directives[i])) {
2503 nasm_free(defining);
2504 defining = NULL;
2505 return DIRECTIVE_FOUND;
2508 mmac = (MMacro *) hash_findix(&mmacros, defining->name);
2509 while (mmac) {
2510 if (!strcmp(mmac->name, defining->name) &&
2511 (mmac->nparam_min <= defining->nparam_max
2512 || defining->plus)
2513 && (defining->nparam_min <= mmac->nparam_max
2514 || mmac->plus)) {
2515 error(ERR_WARNING,
2516 "redefining multi-line macro `%s'", defining->name);
2517 return DIRECTIVE_FOUND;
2519 mmac = mmac->next;
2521 free_tlist(origline);
2522 return DIRECTIVE_FOUND;
2524 case PP_ENDM:
2525 case PP_ENDMACRO:
2526 if (! (defining && defining->name)) {
2527 error(ERR_NONFATAL, "`%s': not defining a macro", tline->text);
2528 return DIRECTIVE_FOUND;
2530 mmhead = (MMacro **) hash_findi_add(&mmacros, defining->name);
2531 defining->next = *mmhead;
2532 *mmhead = defining;
2533 defining = NULL;
2534 free_tlist(origline);
2535 return DIRECTIVE_FOUND;
2537 case PP_UNMACRO:
2538 case PP_UNIMACRO:
2540 MMacro **mmac_p;
2541 MMacro spec;
2543 spec.casesense = (i == PP_UNMACRO);
2544 if (!parse_mmacro_spec(tline, &spec, pp_directives[i])) {
2545 return DIRECTIVE_FOUND;
2547 mmac_p = (MMacro **) hash_findi(&mmacros, spec.name, NULL);
2548 while (mmac_p && *mmac_p) {
2549 mmac = *mmac_p;
2550 if (mmac->casesense == spec.casesense &&
2551 !mstrcmp(mmac->name, spec.name, spec.casesense) &&
2552 mmac->nparam_min == spec.nparam_min &&
2553 mmac->nparam_max == spec.nparam_max &&
2554 mmac->plus == spec.plus) {
2555 *mmac_p = mmac->next;
2556 free_mmacro(mmac);
2557 } else {
2558 mmac_p = &mmac->next;
2561 free_tlist(origline);
2562 free_tlist(spec.dlist);
2563 return DIRECTIVE_FOUND;
2566 case PP_ROTATE:
2567 if (tline->next && tline->next->type == TOK_WHITESPACE)
2568 tline = tline->next;
2569 if (tline->next == NULL) {
2570 free_tlist(origline);
2571 error(ERR_NONFATAL, "`%%rotate' missing rotate count");
2572 return DIRECTIVE_FOUND;
2574 t = expand_smacro(tline->next);
2575 tline->next = NULL;
2576 free_tlist(origline);
2577 tline = t;
2578 tptr = &t;
2579 tokval.t_type = TOKEN_INVALID;
2580 evalresult =
2581 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
2582 free_tlist(tline);
2583 if (!evalresult)
2584 return DIRECTIVE_FOUND;
2585 if (tokval.t_type)
2586 error(ERR_WARNING,
2587 "trailing garbage after expression ignored");
2588 if (!is_simple(evalresult)) {
2589 error(ERR_NONFATAL, "non-constant value given to `%%rotate'");
2590 return DIRECTIVE_FOUND;
2592 mmac = istk->mstk;
2593 while (mmac && !mmac->name) /* avoid mistaking %reps for macros */
2594 mmac = mmac->next_active;
2595 if (!mmac) {
2596 error(ERR_NONFATAL, "`%%rotate' invoked outside a macro call");
2597 } else if (mmac->nparam == 0) {
2598 error(ERR_NONFATAL,
2599 "`%%rotate' invoked within macro without parameters");
2600 } else {
2601 int rotate = mmac->rotate + reloc_value(evalresult);
2603 rotate %= (int)mmac->nparam;
2604 if (rotate < 0)
2605 rotate += mmac->nparam;
2607 mmac->rotate = rotate;
2609 return DIRECTIVE_FOUND;
2611 case PP_REP:
2612 nolist = false;
2613 do {
2614 tline = tline->next;
2615 } while (tok_type_(tline, TOK_WHITESPACE));
2617 if (tok_type_(tline, TOK_ID) &&
2618 nasm_stricmp(tline->text, ".nolist") == 0) {
2619 nolist = true;
2620 do {
2621 tline = tline->next;
2622 } while (tok_type_(tline, TOK_WHITESPACE));
2625 if (tline) {
2626 t = expand_smacro(tline);
2627 tptr = &t;
2628 tokval.t_type = TOKEN_INVALID;
2629 evalresult =
2630 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
2631 if (!evalresult) {
2632 free_tlist(origline);
2633 return DIRECTIVE_FOUND;
2635 if (tokval.t_type)
2636 error(ERR_WARNING,
2637 "trailing garbage after expression ignored");
2638 if (!is_simple(evalresult)) {
2639 error(ERR_NONFATAL, "non-constant value given to `%%rep'");
2640 return DIRECTIVE_FOUND;
2642 count = reloc_value(evalresult) + 1;
2643 } else {
2644 error(ERR_NONFATAL, "`%%rep' expects a repeat count");
2645 count = 0;
2647 free_tlist(origline);
2649 tmp_defining = defining;
2650 defining = nasm_malloc(sizeof(MMacro));
2651 defining->name = NULL; /* flags this macro as a %rep block */
2652 defining->casesense = false;
2653 defining->plus = false;
2654 defining->nolist = nolist;
2655 defining->in_progress = count;
2656 defining->nparam_min = defining->nparam_max = 0;
2657 defining->defaults = NULL;
2658 defining->dlist = NULL;
2659 defining->expansion = NULL;
2660 defining->next_active = istk->mstk;
2661 defining->rep_nest = tmp_defining;
2662 return DIRECTIVE_FOUND;
2664 case PP_ENDREP:
2665 if (!defining || defining->name) {
2666 error(ERR_NONFATAL, "`%%endrep': no matching `%%rep'");
2667 return DIRECTIVE_FOUND;
2671 * Now we have a "macro" defined - although it has no name
2672 * and we won't be entering it in the hash tables - we must
2673 * push a macro-end marker for it on to istk->expansion.
2674 * After that, it will take care of propagating itself (a
2675 * macro-end marker line for a macro which is really a %rep
2676 * block will cause the macro to be re-expanded, complete
2677 * with another macro-end marker to ensure the process
2678 * continues) until the whole expansion is forcibly removed
2679 * from istk->expansion by a %exitrep.
2681 l = nasm_malloc(sizeof(Line));
2682 l->next = istk->expansion;
2683 l->finishes = defining;
2684 l->first = NULL;
2685 istk->expansion = l;
2687 istk->mstk = defining;
2689 list->uplevel(defining->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
2690 tmp_defining = defining;
2691 defining = defining->rep_nest;
2692 free_tlist(origline);
2693 return DIRECTIVE_FOUND;
2695 case PP_EXITREP:
2697 * We must search along istk->expansion until we hit a
2698 * macro-end marker for a macro with no name. Then we set
2699 * its `in_progress' flag to 0.
2701 for (l = istk->expansion; l; l = l->next)
2702 if (l->finishes && !l->finishes->name)
2703 break;
2705 if (l)
2706 l->finishes->in_progress = 1;
2707 else
2708 error(ERR_NONFATAL, "`%%exitrep' not within `%%rep' block");
2709 free_tlist(origline);
2710 return DIRECTIVE_FOUND;
2712 case PP_XDEFINE:
2713 case PP_IXDEFINE:
2714 case PP_DEFINE:
2715 case PP_IDEFINE:
2716 casesense = (i == PP_DEFINE || i == PP_XDEFINE);
2718 tline = tline->next;
2719 skip_white_(tline);
2720 tline = expand_id(tline);
2721 if (!tline || (tline->type != TOK_ID &&
2722 (tline->type != TOK_PREPROC_ID ||
2723 tline->text[1] != '$'))) {
2724 error(ERR_NONFATAL, "`%s' expects a macro identifier",
2725 pp_directives[i]);
2726 free_tlist(origline);
2727 return DIRECTIVE_FOUND;
2730 ctx = get_ctx(tline->text, false);
2732 mname = tline->text;
2733 last = tline;
2734 param_start = tline = tline->next;
2735 nparam = 0;
2737 /* Expand the macro definition now for %xdefine and %ixdefine */
2738 if ((i == PP_XDEFINE) || (i == PP_IXDEFINE))
2739 tline = expand_smacro(tline);
2741 if (tok_is_(tline, "(")) {
2743 * This macro has parameters.
2746 tline = tline->next;
2747 while (1) {
2748 skip_white_(tline);
2749 if (!tline) {
2750 error(ERR_NONFATAL, "parameter identifier expected");
2751 free_tlist(origline);
2752 return DIRECTIVE_FOUND;
2754 if (tline->type != TOK_ID) {
2755 error(ERR_NONFATAL,
2756 "`%s': parameter identifier expected",
2757 tline->text);
2758 free_tlist(origline);
2759 return DIRECTIVE_FOUND;
2761 tline->type = TOK_SMAC_PARAM + nparam++;
2762 tline = tline->next;
2763 skip_white_(tline);
2764 if (tok_is_(tline, ",")) {
2765 tline = tline->next;
2766 } else {
2767 if (!tok_is_(tline, ")")) {
2768 error(ERR_NONFATAL,
2769 "`)' expected to terminate macro template");
2770 free_tlist(origline);
2771 return DIRECTIVE_FOUND;
2773 break;
2776 last = tline;
2777 tline = tline->next;
2779 if (tok_type_(tline, TOK_WHITESPACE))
2780 last = tline, tline = tline->next;
2781 macro_start = NULL;
2782 last->next = NULL;
2783 t = tline;
2784 while (t) {
2785 if (t->type == TOK_ID) {
2786 for (tt = param_start; tt; tt = tt->next)
2787 if (tt->type >= TOK_SMAC_PARAM &&
2788 !strcmp(tt->text, t->text))
2789 t->type = tt->type;
2791 tt = t->next;
2792 t->next = macro_start;
2793 macro_start = t;
2794 t = tt;
2797 * Good. We now have a macro name, a parameter count, and a
2798 * token list (in reverse order) for an expansion. We ought
2799 * to be OK just to create an SMacro, store it, and let
2800 * free_tlist have the rest of the line (which we have
2801 * carefully re-terminated after chopping off the expansion
2802 * from the end).
2804 define_smacro(ctx, mname, casesense, nparam, macro_start);
2805 free_tlist(origline);
2806 return DIRECTIVE_FOUND;
2808 case PP_UNDEF:
2809 tline = tline->next;
2810 skip_white_(tline);
2811 tline = expand_id(tline);
2812 if (!tline || (tline->type != TOK_ID &&
2813 (tline->type != TOK_PREPROC_ID ||
2814 tline->text[1] != '$'))) {
2815 error(ERR_NONFATAL, "`%%undef' expects a macro identifier");
2816 free_tlist(origline);
2817 return DIRECTIVE_FOUND;
2819 if (tline->next) {
2820 error(ERR_WARNING,
2821 "trailing garbage after macro name ignored");
2824 /* Find the context that symbol belongs to */
2825 ctx = get_ctx(tline->text, false);
2826 undef_smacro(ctx, tline->text);
2827 free_tlist(origline);
2828 return DIRECTIVE_FOUND;
2830 case PP_DEFSTR:
2831 case PP_IDEFSTR:
2832 casesense = (i == PP_DEFSTR);
2834 tline = tline->next;
2835 skip_white_(tline);
2836 tline = expand_id(tline);
2837 if (!tline || (tline->type != TOK_ID &&
2838 (tline->type != TOK_PREPROC_ID ||
2839 tline->text[1] != '$'))) {
2840 error(ERR_NONFATAL, "`%s' expects a macro identifier",
2841 pp_directives[i]);
2842 free_tlist(origline);
2843 return DIRECTIVE_FOUND;
2846 ctx = get_ctx(tline->text, false);
2848 mname = tline->text;
2849 last = tline;
2850 tline = expand_smacro(tline->next);
2851 last->next = NULL;
2853 while (tok_type_(tline, TOK_WHITESPACE))
2854 tline = delete_Token(tline);
2856 p = detoken(tline, false);
2857 macro_start = nasm_malloc(sizeof(*macro_start));
2858 macro_start->next = NULL;
2859 macro_start->text = nasm_quote(p, strlen(p));
2860 macro_start->type = TOK_STRING;
2861 macro_start->a.mac = NULL;
2862 nasm_free(p);
2865 * We now have a macro name, an implicit parameter count of
2866 * zero, and a string token to use as an expansion. Create
2867 * and store an SMacro.
2869 define_smacro(ctx, mname, casesense, 0, macro_start);
2870 free_tlist(origline);
2871 return DIRECTIVE_FOUND;
2873 case PP_PATHSEARCH:
2875 FILE *fp;
2876 StrList *xsl = NULL;
2877 StrList **xst = &xsl;
2879 casesense = true;
2881 tline = tline->next;
2882 skip_white_(tline);
2883 tline = expand_id(tline);
2884 if (!tline || (tline->type != TOK_ID &&
2885 (tline->type != TOK_PREPROC_ID ||
2886 tline->text[1] != '$'))) {
2887 error(ERR_NONFATAL,
2888 "`%%pathsearch' expects a macro identifier as first parameter");
2889 free_tlist(origline);
2890 return DIRECTIVE_FOUND;
2892 ctx = get_ctx(tline->text, false);
2894 mname = tline->text;
2895 last = tline;
2896 tline = expand_smacro(tline->next);
2897 last->next = NULL;
2899 t = tline;
2900 while (tok_type_(t, TOK_WHITESPACE))
2901 t = t->next;
2903 if (!t || (t->type != TOK_STRING &&
2904 t->type != TOK_INTERNAL_STRING)) {
2905 error(ERR_NONFATAL, "`%%pathsearch' expects a file name");
2906 free_tlist(tline);
2907 free_tlist(origline);
2908 return DIRECTIVE_FOUND; /* but we did _something_ */
2910 if (t->next)
2911 error(ERR_WARNING,
2912 "trailing garbage after `%%pathsearch' ignored");
2913 p = t->text;
2914 if (t->type != TOK_INTERNAL_STRING)
2915 nasm_unquote(p, NULL);
2917 fp = inc_fopen(p, &xsl, &xst, true);
2918 if (fp) {
2919 p = xsl->str;
2920 fclose(fp); /* Don't actually care about the file */
2922 macro_start = nasm_malloc(sizeof(*macro_start));
2923 macro_start->next = NULL;
2924 macro_start->text = nasm_quote(p, strlen(p));
2925 macro_start->type = TOK_STRING;
2926 macro_start->a.mac = NULL;
2927 if (xsl)
2928 nasm_free(xsl);
2931 * We now have a macro name, an implicit parameter count of
2932 * zero, and a string token to use as an expansion. Create
2933 * and store an SMacro.
2935 define_smacro(ctx, mname, casesense, 0, macro_start);
2936 free_tlist(tline);
2937 free_tlist(origline);
2938 return DIRECTIVE_FOUND;
2941 case PP_STRLEN:
2942 casesense = true;
2944 tline = tline->next;
2945 skip_white_(tline);
2946 tline = expand_id(tline);
2947 if (!tline || (tline->type != TOK_ID &&
2948 (tline->type != TOK_PREPROC_ID ||
2949 tline->text[1] != '$'))) {
2950 error(ERR_NONFATAL,
2951 "`%%strlen' expects a macro identifier as first parameter");
2952 free_tlist(origline);
2953 return DIRECTIVE_FOUND;
2955 ctx = get_ctx(tline->text, false);
2957 mname = tline->text;
2958 last = tline;
2959 tline = expand_smacro(tline->next);
2960 last->next = NULL;
2962 t = tline;
2963 while (tok_type_(t, TOK_WHITESPACE))
2964 t = t->next;
2965 /* t should now point to the string */
2966 if (t->type != TOK_STRING) {
2967 error(ERR_NONFATAL,
2968 "`%%strlen` requires string as second parameter");
2969 free_tlist(tline);
2970 free_tlist(origline);
2971 return DIRECTIVE_FOUND;
2974 macro_start = nasm_malloc(sizeof(*macro_start));
2975 macro_start->next = NULL;
2976 make_tok_num(macro_start, nasm_unquote(t->text, NULL));
2977 macro_start->a.mac = NULL;
2980 * We now have a macro name, an implicit parameter count of
2981 * zero, and a numeric token to use as an expansion. Create
2982 * and store an SMacro.
2984 define_smacro(ctx, mname, casesense, 0, macro_start);
2985 free_tlist(tline);
2986 free_tlist(origline);
2987 return DIRECTIVE_FOUND;
2989 case PP_STRCAT:
2990 casesense = true;
2992 tline = tline->next;
2993 skip_white_(tline);
2994 tline = expand_id(tline);
2995 if (!tline || (tline->type != TOK_ID &&
2996 (tline->type != TOK_PREPROC_ID ||
2997 tline->text[1] != '$'))) {
2998 error(ERR_NONFATAL,
2999 "`%%strcat' expects a macro identifier as first parameter");
3000 free_tlist(origline);
3001 return DIRECTIVE_FOUND;
3003 ctx = get_ctx(tline->text, false);
3005 mname = tline->text;
3006 last = tline;
3007 tline = expand_smacro(tline->next);
3008 last->next = NULL;
3010 len = 0;
3011 for (t = tline; t; t = t->next) {
3012 switch (t->type) {
3013 case TOK_WHITESPACE:
3014 break;
3015 case TOK_STRING:
3016 len += t->a.len = nasm_unquote(t->text, NULL);
3017 break;
3018 case TOK_OTHER:
3019 if (!strcmp(t->text, ",")) /* permit comma separators */
3020 break;
3021 /* else fall through */
3022 default:
3023 error(ERR_NONFATAL,
3024 "non-string passed to `%%strcat' (%d)", t->type);
3025 free_tlist(tline);
3026 free_tlist(origline);
3027 return DIRECTIVE_FOUND;
3031 p = pp = nasm_malloc(len);
3032 t = tline;
3033 for (t = tline; t; t = t->next) {
3034 if (t->type == TOK_STRING) {
3035 memcpy(p, t->text, t->a.len);
3036 p += t->a.len;
3041 * We now have a macro name, an implicit parameter count of
3042 * zero, and a numeric token to use as an expansion. Create
3043 * and store an SMacro.
3045 macro_start = new_Token(NULL, TOK_STRING, NULL, 0);
3046 macro_start->text = nasm_quote(pp, len);
3047 nasm_free(pp);
3048 define_smacro(ctx, mname, casesense, 0, macro_start);
3049 free_tlist(tline);
3050 free_tlist(origline);
3051 return DIRECTIVE_FOUND;
3053 case PP_SUBSTR:
3055 int64_t a1, a2;
3056 size_t len;
3058 casesense = true;
3060 tline = tline->next;
3061 skip_white_(tline);
3062 tline = expand_id(tline);
3063 if (!tline || (tline->type != TOK_ID &&
3064 (tline->type != TOK_PREPROC_ID ||
3065 tline->text[1] != '$'))) {
3066 error(ERR_NONFATAL,
3067 "`%%substr' expects a macro identifier as first parameter");
3068 free_tlist(origline);
3069 return DIRECTIVE_FOUND;
3071 ctx = get_ctx(tline->text, false);
3073 mname = tline->text;
3074 last = tline;
3075 tline = expand_smacro(tline->next);
3076 last->next = NULL;
3078 t = tline->next;
3079 while (tok_type_(t, TOK_WHITESPACE))
3080 t = t->next;
3082 /* t should now point to the string */
3083 if (t->type != TOK_STRING) {
3084 error(ERR_NONFATAL,
3085 "`%%substr` requires string as second parameter");
3086 free_tlist(tline);
3087 free_tlist(origline);
3088 return DIRECTIVE_FOUND;
3091 tt = t->next;
3092 tptr = &tt;
3093 tokval.t_type = TOKEN_INVALID;
3094 evalresult = evaluate(ppscan, tptr, &tokval, NULL,
3095 pass, error, NULL);
3096 if (!evalresult) {
3097 free_tlist(tline);
3098 free_tlist(origline);
3099 return DIRECTIVE_FOUND;
3100 } else if (!is_simple(evalresult)) {
3101 error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3102 free_tlist(tline);
3103 free_tlist(origline);
3104 return DIRECTIVE_FOUND;
3106 a1 = evalresult->value-1;
3108 while (tok_type_(tt, TOK_WHITESPACE))
3109 tt = tt->next;
3110 if (!tt) {
3111 a2 = 1; /* Backwards compatibility: one character */
3112 } else {
3113 tokval.t_type = TOKEN_INVALID;
3114 evalresult = evaluate(ppscan, tptr, &tokval, NULL,
3115 pass, error, NULL);
3116 if (!evalresult) {
3117 free_tlist(tline);
3118 free_tlist(origline);
3119 return DIRECTIVE_FOUND;
3120 } else if (!is_simple(evalresult)) {
3121 error(ERR_NONFATAL, "non-constant value given to `%%substr`");
3122 free_tlist(tline);
3123 free_tlist(origline);
3124 return DIRECTIVE_FOUND;
3126 a2 = evalresult->value;
3129 len = nasm_unquote(t->text, NULL);
3130 if (a2 < 0)
3131 a2 = a2+1+len-a1;
3132 if (a1+a2 > (int64_t)len)
3133 a2 = len-a1;
3135 macro_start = nasm_malloc(sizeof(*macro_start));
3136 macro_start->next = NULL;
3137 macro_start->text = nasm_quote((a1 < 0) ? "" : t->text+a1, a2);
3138 macro_start->type = TOK_STRING;
3139 macro_start->a.mac = NULL;
3142 * We now have a macro name, an implicit parameter count of
3143 * zero, and a numeric token to use as an expansion. Create
3144 * and store an SMacro.
3146 define_smacro(ctx, mname, casesense, 0, macro_start);
3147 free_tlist(tline);
3148 free_tlist(origline);
3149 return DIRECTIVE_FOUND;
3152 case PP_ASSIGN:
3153 case PP_IASSIGN:
3154 casesense = (i == PP_ASSIGN);
3156 tline = tline->next;
3157 skip_white_(tline);
3158 tline = expand_id(tline);
3159 if (!tline || (tline->type != TOK_ID &&
3160 (tline->type != TOK_PREPROC_ID ||
3161 tline->text[1] != '$'))) {
3162 error(ERR_NONFATAL,
3163 "`%%%sassign' expects a macro identifier",
3164 (i == PP_IASSIGN ? "i" : ""));
3165 free_tlist(origline);
3166 return DIRECTIVE_FOUND;
3168 ctx = get_ctx(tline->text, false);
3170 mname = tline->text;
3171 last = tline;
3172 tline = expand_smacro(tline->next);
3173 last->next = NULL;
3175 t = tline;
3176 tptr = &t;
3177 tokval.t_type = TOKEN_INVALID;
3178 evalresult =
3179 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
3180 free_tlist(tline);
3181 if (!evalresult) {
3182 free_tlist(origline);
3183 return DIRECTIVE_FOUND;
3186 if (tokval.t_type)
3187 error(ERR_WARNING,
3188 "trailing garbage after expression ignored");
3190 if (!is_simple(evalresult)) {
3191 error(ERR_NONFATAL,
3192 "non-constant value given to `%%%sassign'",
3193 (i == PP_IASSIGN ? "i" : ""));
3194 free_tlist(origline);
3195 return DIRECTIVE_FOUND;
3198 macro_start = nasm_malloc(sizeof(*macro_start));
3199 macro_start->next = NULL;
3200 make_tok_num(macro_start, reloc_value(evalresult));
3201 macro_start->a.mac = NULL;
3204 * We now have a macro name, an implicit parameter count of
3205 * zero, and a numeric token to use as an expansion. Create
3206 * and store an SMacro.
3208 define_smacro(ctx, mname, casesense, 0, macro_start);
3209 free_tlist(origline);
3210 return DIRECTIVE_FOUND;
3212 case PP_LINE:
3214 * Syntax is `%line nnn[+mmm] [filename]'
3216 tline = tline->next;
3217 skip_white_(tline);
3218 if (!tok_type_(tline, TOK_NUMBER)) {
3219 error(ERR_NONFATAL, "`%%line' expects line number");
3220 free_tlist(origline);
3221 return DIRECTIVE_FOUND;
3223 k = readnum(tline->text, &err);
3224 m = 1;
3225 tline = tline->next;
3226 if (tok_is_(tline, "+")) {
3227 tline = tline->next;
3228 if (!tok_type_(tline, TOK_NUMBER)) {
3229 error(ERR_NONFATAL, "`%%line' expects line increment");
3230 free_tlist(origline);
3231 return DIRECTIVE_FOUND;
3233 m = readnum(tline->text, &err);
3234 tline = tline->next;
3236 skip_white_(tline);
3237 src_set_linnum(k);
3238 istk->lineinc = m;
3239 if (tline) {
3240 nasm_free(src_set_fname(detoken(tline, false)));
3242 free_tlist(origline);
3243 return DIRECTIVE_FOUND;
3245 default:
3246 error(ERR_FATAL,
3247 "preprocessor directive `%s' not yet implemented",
3248 pp_directives[i]);
3249 return DIRECTIVE_FOUND;
3254 * Ensure that a macro parameter contains a condition code and
3255 * nothing else. Return the condition code index if so, or -1
3256 * otherwise.
3258 static int find_cc(Token * t)
3260 Token *tt;
3261 int i, j, k, m;
3263 if (!t)
3264 return -1; /* Probably a %+ without a space */
3266 skip_white_(t);
3267 if (t->type != TOK_ID)
3268 return -1;
3269 tt = t->next;
3270 skip_white_(tt);
3271 if (tt && (tt->type != TOK_OTHER || strcmp(tt->text, ",")))
3272 return -1;
3274 i = -1;
3275 j = elements(conditions);
3276 while (j - i > 1) {
3277 k = (j + i) / 2;
3278 m = nasm_stricmp(t->text, conditions[k]);
3279 if (m == 0) {
3280 i = k;
3281 j = -2;
3282 break;
3283 } else if (m < 0) {
3284 j = k;
3285 } else
3286 i = k;
3288 if (j != -2)
3289 return -1;
3290 return i;
3294 * Expand MMacro-local things: parameter references (%0, %n, %+n,
3295 * %-n) and MMacro-local identifiers (%%foo).
3297 static Token *expand_mmac_params(Token * tline)
3299 Token *t, *tt, **tail, *thead;
3301 tail = &thead;
3302 thead = NULL;
3304 while (tline) {
3305 if (tline->type == TOK_PREPROC_ID &&
3306 (((tline->text[1] == '+' || tline->text[1] == '-')
3307 && tline->text[2]) || tline->text[1] == '%'
3308 || (tline->text[1] >= '0' && tline->text[1] <= '9'))) {
3309 char *text = NULL;
3310 int type = 0, cc; /* type = 0 to placate optimisers */
3311 char tmpbuf[30];
3312 unsigned int n;
3313 int i;
3314 MMacro *mac;
3316 t = tline;
3317 tline = tline->next;
3319 mac = istk->mstk;
3320 while (mac && !mac->name) /* avoid mistaking %reps for macros */
3321 mac = mac->next_active;
3322 if (!mac)
3323 error(ERR_NONFATAL, "`%s': not in a macro call", t->text);
3324 else
3325 switch (t->text[1]) {
3327 * We have to make a substitution of one of the
3328 * forms %1, %-1, %+1, %%foo, %0.
3330 case '0':
3331 type = TOK_NUMBER;
3332 snprintf(tmpbuf, sizeof(tmpbuf), "%d", mac->nparam);
3333 text = nasm_strdup(tmpbuf);
3334 break;
3335 case '%':
3336 type = TOK_ID;
3337 snprintf(tmpbuf, sizeof(tmpbuf), "..@%"PRIu64".",
3338 mac->unique);
3339 text = nasm_strcat(tmpbuf, t->text + 2);
3340 break;
3341 case '-':
3342 n = atoi(t->text + 2) - 1;
3343 if (n >= mac->nparam)
3344 tt = NULL;
3345 else {
3346 if (mac->nparam > 1)
3347 n = (n + mac->rotate) % mac->nparam;
3348 tt = mac->params[n];
3350 cc = find_cc(tt);
3351 if (cc == -1) {
3352 error(ERR_NONFATAL,
3353 "macro parameter %d is not a condition code",
3354 n + 1);
3355 text = NULL;
3356 } else {
3357 type = TOK_ID;
3358 if (inverse_ccs[cc] == -1) {
3359 error(ERR_NONFATAL,
3360 "condition code `%s' is not invertible",
3361 conditions[cc]);
3362 text = NULL;
3363 } else
3364 text =
3365 nasm_strdup(conditions[inverse_ccs[cc]]);
3367 break;
3368 case '+':
3369 n = atoi(t->text + 2) - 1;
3370 if (n >= mac->nparam)
3371 tt = NULL;
3372 else {
3373 if (mac->nparam > 1)
3374 n = (n + mac->rotate) % mac->nparam;
3375 tt = mac->params[n];
3377 cc = find_cc(tt);
3378 if (cc == -1) {
3379 error(ERR_NONFATAL,
3380 "macro parameter %d is not a condition code",
3381 n + 1);
3382 text = NULL;
3383 } else {
3384 type = TOK_ID;
3385 text = nasm_strdup(conditions[cc]);
3387 break;
3388 default:
3389 n = atoi(t->text + 1) - 1;
3390 if (n >= mac->nparam)
3391 tt = NULL;
3392 else {
3393 if (mac->nparam > 1)
3394 n = (n + mac->rotate) % mac->nparam;
3395 tt = mac->params[n];
3397 if (tt) {
3398 for (i = 0; i < mac->paramlen[n]; i++) {
3399 *tail = new_Token(NULL, tt->type, tt->text, 0);
3400 tail = &(*tail)->next;
3401 tt = tt->next;
3404 text = NULL; /* we've done it here */
3405 break;
3407 if (!text) {
3408 delete_Token(t);
3409 } else {
3410 *tail = t;
3411 tail = &t->next;
3412 t->type = type;
3413 nasm_free(t->text);
3414 t->text = text;
3415 t->a.mac = NULL;
3417 continue;
3418 } else {
3419 t = *tail = tline;
3420 tline = tline->next;
3421 t->a.mac = NULL;
3422 tail = &t->next;
3425 *tail = NULL;
3426 t = thead;
3427 for (; t && (tt = t->next) != NULL; t = t->next)
3428 switch (t->type) {
3429 case TOK_WHITESPACE:
3430 if (tt->type == TOK_WHITESPACE) {
3431 t->next = delete_Token(tt);
3433 break;
3434 case TOK_ID:
3435 if (tt->type == TOK_ID || tt->type == TOK_NUMBER) {
3436 char *tmp = nasm_strcat(t->text, tt->text);
3437 nasm_free(t->text);
3438 t->text = tmp;
3439 t->next = delete_Token(tt);
3441 break;
3442 case TOK_NUMBER:
3443 if (tt->type == TOK_NUMBER) {
3444 char *tmp = nasm_strcat(t->text, tt->text);
3445 nasm_free(t->text);
3446 t->text = tmp;
3447 t->next = delete_Token(tt);
3449 break;
3450 default:
3451 break;
3454 return thead;
3458 * Expand all single-line macro calls made in the given line.
3459 * Return the expanded version of the line. The original is deemed
3460 * to be destroyed in the process. (In reality we'll just move
3461 * Tokens from input to output a lot of the time, rather than
3462 * actually bothering to destroy and replicate.)
3464 #define DEADMAN_LIMIT (1 << 20)
3466 static Token *expand_smacro(Token * tline)
3468 Token *t, *tt, *mstart, **tail, *thead;
3469 struct hash_table *smtbl;
3470 SMacro *head = NULL, *m;
3471 Token **params;
3472 int *paramsize;
3473 unsigned int nparam, sparam;
3474 int brackets, rescan;
3475 Token *org_tline = tline;
3476 Context *ctx;
3477 char *mname;
3478 int deadman = DEADMAN_LIMIT;
3481 * Trick: we should avoid changing the start token pointer since it can
3482 * be contained in "next" field of other token. Because of this
3483 * we allocate a copy of first token and work with it; at the end of
3484 * routine we copy it back
3486 if (org_tline) {
3487 tline =
3488 new_Token(org_tline->next, org_tline->type, org_tline->text,
3490 tline->a.mac = org_tline->a.mac;
3491 nasm_free(org_tline->text);
3492 org_tline->text = NULL;
3495 again:
3496 tail = &thead;
3497 thead = NULL;
3499 while (tline) { /* main token loop */
3500 if (!--deadman) {
3501 error(ERR_NONFATAL, "interminable macro recursion");
3502 break;
3505 if ((mname = tline->text)) {
3506 /* if this token is a local macro, look in local context */
3507 ctx = NULL;
3508 smtbl = &smacros;
3509 if (tline->type == TOK_ID || tline->type == TOK_PREPROC_ID) {
3510 ctx = get_ctx(mname, true);
3511 if (ctx)
3512 smtbl = &ctx->localmac;
3514 head = (SMacro *) hash_findix(smtbl, mname);
3517 * We've hit an identifier. As in is_mmacro below, we first
3518 * check whether the identifier is a single-line macro at
3519 * all, then think about checking for parameters if
3520 * necessary.
3522 for (m = head; m; m = m->next)
3523 if (!mstrcmp(m->name, mname, m->casesense))
3524 break;
3525 if (m) {
3526 mstart = tline;
3527 params = NULL;
3528 paramsize = NULL;
3529 if (m->nparam == 0) {
3531 * Simple case: the macro is parameterless. Discard the
3532 * one token that the macro call took, and push the
3533 * expansion back on the to-do stack.
3535 if (!m->expansion) {
3536 if (!strcmp("__FILE__", m->name)) {
3537 int32_t num = 0;
3538 char *file = NULL;
3539 src_get(&num, &file);
3540 tline->text = nasm_quote(file, strlen(file));
3541 tline->type = TOK_STRING;
3542 nasm_free(file);
3543 continue;
3545 if (!strcmp("__LINE__", m->name)) {
3546 nasm_free(tline->text);
3547 make_tok_num(tline, src_get_linnum());
3548 continue;
3550 if (!strcmp("__BITS__", m->name)) {
3551 nasm_free(tline->text);
3552 make_tok_num(tline, globalbits);
3553 continue;
3555 tline = delete_Token(tline);
3556 continue;
3558 } else {
3560 * Complicated case: at least one macro with this name
3561 * exists and takes parameters. We must find the
3562 * parameters in the call, count them, find the SMacro
3563 * that corresponds to that form of the macro call, and
3564 * substitute for the parameters when we expand. What a
3565 * pain.
3567 /*tline = tline->next;
3568 skip_white_(tline); */
3569 do {
3570 t = tline->next;
3571 while (tok_type_(t, TOK_SMAC_END)) {
3572 t->a.mac->in_progress = false;
3573 t->text = NULL;
3574 t = tline->next = delete_Token(t);
3576 tline = t;
3577 } while (tok_type_(tline, TOK_WHITESPACE));
3578 if (!tok_is_(tline, "(")) {
3580 * This macro wasn't called with parameters: ignore
3581 * the call. (Behaviour borrowed from gnu cpp.)
3583 tline = mstart;
3584 m = NULL;
3585 } else {
3586 int paren = 0;
3587 int white = 0;
3588 brackets = 0;
3589 nparam = 0;
3590 sparam = PARAM_DELTA;
3591 params = nasm_malloc(sparam * sizeof(Token *));
3592 params[0] = tline->next;
3593 paramsize = nasm_malloc(sparam * sizeof(int));
3594 paramsize[0] = 0;
3595 while (true) { /* parameter loop */
3597 * For some unusual expansions
3598 * which concatenates function call
3600 t = tline->next;
3601 while (tok_type_(t, TOK_SMAC_END)) {
3602 t->a.mac->in_progress = false;
3603 t->text = NULL;
3604 t = tline->next = delete_Token(t);
3606 tline = t;
3608 if (!tline) {
3609 error(ERR_NONFATAL,
3610 "macro call expects terminating `)'");
3611 break;
3613 if (tline->type == TOK_WHITESPACE
3614 && brackets <= 0) {
3615 if (paramsize[nparam])
3616 white++;
3617 else
3618 params[nparam] = tline->next;
3619 continue; /* parameter loop */
3621 if (tline->type == TOK_OTHER
3622 && tline->text[1] == 0) {
3623 char ch = tline->text[0];
3624 if (ch == ',' && !paren && brackets <= 0) {
3625 if (++nparam >= sparam) {
3626 sparam += PARAM_DELTA;
3627 params = nasm_realloc(params,
3628 sparam *
3629 sizeof(Token
3630 *));
3631 paramsize =
3632 nasm_realloc(paramsize,
3633 sparam *
3634 sizeof(int));
3636 params[nparam] = tline->next;
3637 paramsize[nparam] = 0;
3638 white = 0;
3639 continue; /* parameter loop */
3641 if (ch == '{' &&
3642 (brackets > 0 || (brackets == 0 &&
3643 !paramsize[nparam])))
3645 if (!(brackets++)) {
3646 params[nparam] = tline->next;
3647 continue; /* parameter loop */
3650 if (ch == '}' && brackets > 0)
3651 if (--brackets == 0) {
3652 brackets = -1;
3653 continue; /* parameter loop */
3655 if (ch == '(' && !brackets)
3656 paren++;
3657 if (ch == ')' && brackets <= 0)
3658 if (--paren < 0)
3659 break;
3661 if (brackets < 0) {
3662 brackets = 0;
3663 error(ERR_NONFATAL, "braces do not "
3664 "enclose all of macro parameter");
3666 paramsize[nparam] += white + 1;
3667 white = 0;
3668 } /* parameter loop */
3669 nparam++;
3670 while (m && (m->nparam != nparam ||
3671 mstrcmp(m->name, mname,
3672 m->casesense)))
3673 m = m->next;
3674 if (!m)
3675 error(ERR_WARNING | ERR_WARN_MNP,
3676 "macro `%s' exists, "
3677 "but not taking %d parameters",
3678 mstart->text, nparam);
3681 if (m && m->in_progress)
3682 m = NULL;
3683 if (!m) { /* in progess or didn't find '(' or wrong nparam */
3685 * Design question: should we handle !tline, which
3686 * indicates missing ')' here, or expand those
3687 * macros anyway, which requires the (t) test a few
3688 * lines down?
3690 nasm_free(params);
3691 nasm_free(paramsize);
3692 tline = mstart;
3693 } else {
3695 * Expand the macro: we are placed on the last token of the
3696 * call, so that we can easily split the call from the
3697 * following tokens. We also start by pushing an SMAC_END
3698 * token for the cycle removal.
3700 t = tline;
3701 if (t) {
3702 tline = t->next;
3703 t->next = NULL;
3705 tt = new_Token(tline, TOK_SMAC_END, NULL, 0);
3706 tt->a.mac = m;
3707 m->in_progress = true;
3708 tline = tt;
3709 for (t = m->expansion; t; t = t->next) {
3710 if (t->type >= TOK_SMAC_PARAM) {
3711 Token *pcopy = tline, **ptail = &pcopy;
3712 Token *ttt, *pt;
3713 int i;
3715 ttt = params[t->type - TOK_SMAC_PARAM];
3716 for (i = paramsize[t->type - TOK_SMAC_PARAM];
3717 --i >= 0;) {
3718 pt = *ptail =
3719 new_Token(tline, ttt->type, ttt->text,
3721 ptail = &pt->next;
3722 ttt = ttt->next;
3724 tline = pcopy;
3725 } else if (t->type == TOK_PREPROC_Q) {
3726 tt = new_Token(tline, TOK_ID, mname, 0);
3727 tline = tt;
3728 } else if (t->type == TOK_PREPROC_QQ) {
3729 tt = new_Token(tline, TOK_ID, m->name, 0);
3730 tline = tt;
3731 } else {
3732 tt = new_Token(tline, t->type, t->text, 0);
3733 tline = tt;
3738 * Having done that, get rid of the macro call, and clean
3739 * up the parameters.
3741 nasm_free(params);
3742 nasm_free(paramsize);
3743 free_tlist(mstart);
3744 continue; /* main token loop */
3749 if (tline->type == TOK_SMAC_END) {
3750 tline->a.mac->in_progress = false;
3751 tline = delete_Token(tline);
3752 } else {
3753 t = *tail = tline;
3754 tline = tline->next;
3755 t->a.mac = NULL;
3756 t->next = NULL;
3757 tail = &t->next;
3762 * Now scan the entire line and look for successive TOK_IDs that resulted
3763 * after expansion (they can't be produced by tokenize()). The successive
3764 * TOK_IDs should be concatenated.
3765 * Also we look for %+ tokens and concatenate the tokens before and after
3766 * them (without white spaces in between).
3768 t = thead;
3769 rescan = 0;
3770 while (t) {
3771 while (t && t->type != TOK_ID && t->type != TOK_PREPROC_ID)
3772 t = t->next;
3773 if (!t || !t->next)
3774 break;
3775 if (t->next->type == TOK_ID ||
3776 t->next->type == TOK_PREPROC_ID ||
3777 t->next->type == TOK_NUMBER) {
3778 char *p = nasm_strcat(t->text, t->next->text);
3779 nasm_free(t->text);
3780 t->next = delete_Token(t->next);
3781 t->text = p;
3782 rescan = 1;
3783 } else if (t->next->type == TOK_WHITESPACE && t->next->next &&
3784 t->next->next->type == TOK_PREPROC_ID &&
3785 strcmp(t->next->next->text, "%+") == 0) {
3786 /* free the next whitespace, the %+ token and next whitespace */
3787 int i;
3788 for (i = 1; i <= 3; i++) {
3789 if (!t->next
3790 || (i != 2 && t->next->type != TOK_WHITESPACE))
3791 break;
3792 t->next = delete_Token(t->next);
3793 } /* endfor */
3794 } else
3795 t = t->next;
3797 /* If we concatenaded something, re-scan the line for macros */
3798 if (rescan) {
3799 tline = thead;
3800 goto again;
3803 if (org_tline) {
3804 if (thead) {
3805 *org_tline = *thead;
3806 /* since we just gave text to org_line, don't free it */
3807 thead->text = NULL;
3808 delete_Token(thead);
3809 } else {
3810 /* the expression expanded to empty line;
3811 we can't return NULL for some reasons
3812 we just set the line to a single WHITESPACE token. */
3813 memset(org_tline, 0, sizeof(*org_tline));
3814 org_tline->text = NULL;
3815 org_tline->type = TOK_WHITESPACE;
3817 thead = org_tline;
3820 return thead;
3824 * Similar to expand_smacro but used exclusively with macro identifiers
3825 * right before they are fetched in. The reason is that there can be
3826 * identifiers consisting of several subparts. We consider that if there
3827 * are more than one element forming the name, user wants a expansion,
3828 * otherwise it will be left as-is. Example:
3830 * %define %$abc cde
3832 * the identifier %$abc will be left as-is so that the handler for %define
3833 * will suck it and define the corresponding value. Other case:
3835 * %define _%$abc cde
3837 * In this case user wants name to be expanded *before* %define starts
3838 * working, so we'll expand %$abc into something (if it has a value;
3839 * otherwise it will be left as-is) then concatenate all successive
3840 * PP_IDs into one.
3842 static Token *expand_id(Token * tline)
3844 Token *cur, *oldnext = NULL;
3846 if (!tline || !tline->next)
3847 return tline;
3849 cur = tline;
3850 while (cur->next &&
3851 (cur->next->type == TOK_ID ||
3852 cur->next->type == TOK_PREPROC_ID
3853 || cur->next->type == TOK_NUMBER))
3854 cur = cur->next;
3856 /* If identifier consists of just one token, don't expand */
3857 if (cur == tline)
3858 return tline;
3860 if (cur) {
3861 oldnext = cur->next; /* Detach the tail past identifier */
3862 cur->next = NULL; /* so that expand_smacro stops here */
3865 tline = expand_smacro(tline);
3867 if (cur) {
3868 /* expand_smacro possibly changhed tline; re-scan for EOL */
3869 cur = tline;
3870 while (cur && cur->next)
3871 cur = cur->next;
3872 if (cur)
3873 cur->next = oldnext;
3876 return tline;
3880 * Determine whether the given line constitutes a multi-line macro
3881 * call, and return the MMacro structure called if so. Doesn't have
3882 * to check for an initial label - that's taken care of in
3883 * expand_mmacro - but must check numbers of parameters. Guaranteed
3884 * to be called with tline->type == TOK_ID, so the putative macro
3885 * name is easy to find.
3887 static MMacro *is_mmacro(Token * tline, Token *** params_array)
3889 MMacro *head, *m;
3890 Token **params;
3891 int nparam;
3893 head = (MMacro *) hash_findix(&mmacros, tline->text);
3896 * Efficiency: first we see if any macro exists with the given
3897 * name. If not, we can return NULL immediately. _Then_ we
3898 * count the parameters, and then we look further along the
3899 * list if necessary to find the proper MMacro.
3901 for (m = head; m; m = m->next)
3902 if (!mstrcmp(m->name, tline->text, m->casesense))
3903 break;
3904 if (!m)
3905 return NULL;
3908 * OK, we have a potential macro. Count and demarcate the
3909 * parameters.
3911 count_mmac_params(tline->next, &nparam, &params);
3914 * So we know how many parameters we've got. Find the MMacro
3915 * structure that handles this number.
3917 while (m) {
3918 if (m->nparam_min <= nparam
3919 && (m->plus || nparam <= m->nparam_max)) {
3921 * This one is right. Just check if cycle removal
3922 * prohibits us using it before we actually celebrate...
3924 if (m->in_progress) {
3925 #if 0
3926 error(ERR_NONFATAL,
3927 "self-reference in multi-line macro `%s'", m->name);
3928 #endif
3929 nasm_free(params);
3930 return NULL;
3933 * It's right, and we can use it. Add its default
3934 * parameters to the end of our list if necessary.
3936 if (m->defaults && nparam < m->nparam_min + m->ndefs) {
3937 params =
3938 nasm_realloc(params,
3939 ((m->nparam_min + m->ndefs +
3940 1) * sizeof(*params)));
3941 while (nparam < m->nparam_min + m->ndefs) {
3942 params[nparam] = m->defaults[nparam - m->nparam_min];
3943 nparam++;
3947 * If we've gone over the maximum parameter count (and
3948 * we're in Plus mode), ignore parameters beyond
3949 * nparam_max.
3951 if (m->plus && nparam > m->nparam_max)
3952 nparam = m->nparam_max;
3954 * Then terminate the parameter list, and leave.
3956 if (!params) { /* need this special case */
3957 params = nasm_malloc(sizeof(*params));
3958 nparam = 0;
3960 params[nparam] = NULL;
3961 *params_array = params;
3962 return m;
3965 * This one wasn't right: look for the next one with the
3966 * same name.
3968 for (m = m->next; m; m = m->next)
3969 if (!mstrcmp(m->name, tline->text, m->casesense))
3970 break;
3974 * After all that, we didn't find one with the right number of
3975 * parameters. Issue a warning, and fail to expand the macro.
3977 error(ERR_WARNING | ERR_WARN_MNP,
3978 "macro `%s' exists, but not taking %d parameters",
3979 tline->text, nparam);
3980 nasm_free(params);
3981 return NULL;
3985 * Expand the multi-line macro call made by the given line, if
3986 * there is one to be expanded. If there is, push the expansion on
3987 * istk->expansion and return 1. Otherwise return 0.
3989 static int expand_mmacro(Token * tline)
3991 Token *startline = tline;
3992 Token *label = NULL;
3993 int dont_prepend = 0;
3994 Token **params, *t, *mtok, *tt;
3995 MMacro *m;
3996 Line *l, *ll;
3997 int i, nparam, *paramlen;
3998 const char *mname;
4000 t = tline;
4001 skip_white_(t);
4002 /* if (!tok_type_(t, TOK_ID)) Lino 02/25/02 */
4003 if (!tok_type_(t, TOK_ID) && !tok_type_(t, TOK_PREPROC_ID))
4004 return 0;
4005 mtok = t;
4006 m = is_mmacro(t, &params);
4007 if (m) {
4008 mname = t->text;
4009 } else {
4010 Token *last;
4012 * We have an id which isn't a macro call. We'll assume
4013 * it might be a label; we'll also check to see if a
4014 * colon follows it. Then, if there's another id after
4015 * that lot, we'll check it again for macro-hood.
4017 label = last = t;
4018 t = t->next;
4019 if (tok_type_(t, TOK_WHITESPACE))
4020 last = t, t = t->next;
4021 if (tok_is_(t, ":")) {
4022 dont_prepend = 1;
4023 last = t, t = t->next;
4024 if (tok_type_(t, TOK_WHITESPACE))
4025 last = t, t = t->next;
4027 if (!tok_type_(t, TOK_ID) || (m = is_mmacro(t, &params)) == NULL)
4028 return 0;
4029 last->next = NULL;
4030 mname = t->text;
4031 tline = t;
4035 * Fix up the parameters: this involves stripping leading and
4036 * trailing whitespace, then stripping braces if they are
4037 * present.
4039 for (nparam = 0; params[nparam]; nparam++) ;
4040 paramlen = nparam ? nasm_malloc(nparam * sizeof(*paramlen)) : NULL;
4042 for (i = 0; params[i]; i++) {
4043 int brace = false;
4044 int comma = (!m->plus || i < nparam - 1);
4046 t = params[i];
4047 skip_white_(t);
4048 if (tok_is_(t, "{"))
4049 t = t->next, brace = true, comma = false;
4050 params[i] = t;
4051 paramlen[i] = 0;
4052 while (t) {
4053 if (comma && t->type == TOK_OTHER && !strcmp(t->text, ","))
4054 break; /* ... because we have hit a comma */
4055 if (comma && t->type == TOK_WHITESPACE
4056 && tok_is_(t->next, ","))
4057 break; /* ... or a space then a comma */
4058 if (brace && t->type == TOK_OTHER && !strcmp(t->text, "}"))
4059 break; /* ... or a brace */
4060 t = t->next;
4061 paramlen[i]++;
4066 * OK, we have a MMacro structure together with a set of
4067 * parameters. We must now go through the expansion and push
4068 * copies of each Line on to istk->expansion. Substitution of
4069 * parameter tokens and macro-local tokens doesn't get done
4070 * until the single-line macro substitution process; this is
4071 * because delaying them allows us to change the semantics
4072 * later through %rotate.
4074 * First, push an end marker on to istk->expansion, mark this
4075 * macro as in progress, and set up its invocation-specific
4076 * variables.
4078 ll = nasm_malloc(sizeof(Line));
4079 ll->next = istk->expansion;
4080 ll->finishes = m;
4081 ll->first = NULL;
4082 istk->expansion = ll;
4084 m->in_progress = true;
4085 m->params = params;
4086 m->iline = tline;
4087 m->nparam = nparam;
4088 m->rotate = 0;
4089 m->paramlen = paramlen;
4090 m->unique = unique++;
4091 m->lineno = 0;
4093 m->next_active = istk->mstk;
4094 istk->mstk = m;
4096 for (l = m->expansion; l; l = l->next) {
4097 Token **tail;
4099 ll = nasm_malloc(sizeof(Line));
4100 ll->finishes = NULL;
4101 ll->next = istk->expansion;
4102 istk->expansion = ll;
4103 tail = &ll->first;
4105 for (t = l->first; t; t = t->next) {
4106 Token *x = t;
4107 switch (t->type) {
4108 case TOK_PREPROC_Q:
4109 tt = *tail = new_Token(NULL, TOK_ID, mname, 0);
4110 break;
4111 case TOK_PREPROC_QQ:
4112 tt = *tail = new_Token(NULL, TOK_ID, m->name, 0);
4113 break;
4114 case TOK_PREPROC_ID:
4115 if (t->text[1] == '0' && t->text[2] == '0') {
4116 dont_prepend = -1;
4117 x = label;
4118 if (!x)
4119 continue;
4121 /* fall through */
4122 default:
4123 tt = *tail = new_Token(NULL, x->type, x->text, 0);
4124 break;
4126 tail = &tt->next;
4128 *tail = NULL;
4132 * If we had a label, push it on as the first line of
4133 * the macro expansion.
4135 if (label) {
4136 if (dont_prepend < 0)
4137 free_tlist(startline);
4138 else {
4139 ll = nasm_malloc(sizeof(Line));
4140 ll->finishes = NULL;
4141 ll->next = istk->expansion;
4142 istk->expansion = ll;
4143 ll->first = startline;
4144 if (!dont_prepend) {
4145 while (label->next)
4146 label = label->next;
4147 label->next = tt = new_Token(NULL, TOK_OTHER, ":", 0);
4152 list->uplevel(m->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
4154 return 1;
4157 /* The function that actually does the error reporting */
4158 static void verror(int severity, const char *fmt, va_list arg)
4160 char buff[1024];
4162 vsnprintf(buff, sizeof(buff), fmt, arg);
4164 if (istk && istk->mstk && istk->mstk->name)
4165 _error(severity | ERR_PASS1, "(%s:%d) %s", istk->mstk->name,
4166 istk->mstk->lineno, buff);
4167 else
4168 _error(severity | ERR_PASS1, "%s", buff);
4172 * Since preprocessor always operate only on the line that didn't
4173 * arrived yet, we should always use ERR_OFFBY1. Also since user
4174 * won't want to see same error twice (preprocessing is done once
4175 * per pass) we will want to show errors only during pass one.
4177 static void error(int severity, const char *fmt, ...)
4179 va_list arg;
4181 /* If we're in a dead branch of IF or something like it, ignore the error */
4182 if (istk && istk->conds && !emitting(istk->conds->state))
4183 return;
4185 va_start(arg, fmt);
4186 verror(severity, fmt, arg);
4187 va_end(arg);
4191 * Because %else etc are evaluated in the state context
4192 * of the previous branch, errors might get lost with error():
4193 * %if 0 ... %else trailing garbage ... %endif
4194 * So %else etc should report errors with this function.
4196 static void error_precond(int severity, const char *fmt, ...)
4198 va_list arg;
4200 /* Only ignore the error if it's really in a dead branch */
4201 if (istk && istk->conds && istk->conds->state == COND_NEVER)
4202 return;
4204 va_start(arg, fmt);
4205 verror(severity, fmt, arg);
4206 va_end(arg);
4209 static void
4210 pp_reset(char *file, int apass, efunc errfunc, evalfunc eval,
4211 ListGen * listgen, StrList **deplist)
4213 _error = errfunc;
4214 cstk = NULL;
4215 istk = nasm_malloc(sizeof(Include));
4216 istk->next = NULL;
4217 istk->conds = NULL;
4218 istk->expansion = NULL;
4219 istk->mstk = NULL;
4220 istk->fp = fopen(file, "r");
4221 istk->fname = NULL;
4222 src_set_fname(nasm_strdup(file));
4223 src_set_linnum(0);
4224 istk->lineinc = 1;
4225 if (!istk->fp)
4226 error(ERR_FATAL | ERR_NOFILE, "unable to open input file `%s'",
4227 file);
4228 defining = NULL;
4229 nested_mac_count = 0;
4230 nested_rep_count = 0;
4231 init_macros();
4232 unique = 0;
4233 if (tasm_compatible_mode) {
4234 stdmacpos = nasm_stdmac;
4235 } else {
4236 stdmacpos = nasm_stdmac_after_tasm;
4238 any_extrastdmac = extrastdmac && *extrastdmac;
4239 do_predef = true;
4240 list = listgen;
4241 evaluate = eval;
4242 pass = apass;
4243 dephead = deptail = deplist;
4244 if (deplist) {
4245 StrList *sl = nasm_malloc(strlen(file)+1+sizeof sl->next);
4246 sl->next = NULL;
4247 strcpy(sl->str, file);
4248 *deptail = sl;
4249 deptail = &sl->next;
4253 static char *pp_getline(void)
4255 char *line;
4256 Token *tline;
4258 while (1) {
4260 * Fetch a tokenized line, either from the macro-expansion
4261 * buffer or from the input file.
4263 tline = NULL;
4264 while (istk->expansion && istk->expansion->finishes) {
4265 Line *l = istk->expansion;
4266 if (!l->finishes->name && l->finishes->in_progress > 1) {
4267 Line *ll;
4270 * This is a macro-end marker for a macro with no
4271 * name, which means it's not really a macro at all
4272 * but a %rep block, and the `in_progress' field is
4273 * more than 1, meaning that we still need to
4274 * repeat. (1 means the natural last repetition; 0
4275 * means termination by %exitrep.) We have
4276 * therefore expanded up to the %endrep, and must
4277 * push the whole block on to the expansion buffer
4278 * again. We don't bother to remove the macro-end
4279 * marker: we'd only have to generate another one
4280 * if we did.
4282 l->finishes->in_progress--;
4283 for (l = l->finishes->expansion; l; l = l->next) {
4284 Token *t, *tt, **tail;
4286 ll = nasm_malloc(sizeof(Line));
4287 ll->next = istk->expansion;
4288 ll->finishes = NULL;
4289 ll->first = NULL;
4290 tail = &ll->first;
4292 for (t = l->first; t; t = t->next) {
4293 if (t->text || t->type == TOK_WHITESPACE) {
4294 tt = *tail =
4295 new_Token(NULL, t->type, t->text, 0);
4296 tail = &tt->next;
4300 istk->expansion = ll;
4302 } else {
4304 * Check whether a `%rep' was started and not ended
4305 * within this macro expansion. This can happen and
4306 * should be detected. It's a fatal error because
4307 * I'm too confused to work out how to recover
4308 * sensibly from it.
4310 if (defining) {
4311 if (defining->name)
4312 error(ERR_PANIC,
4313 "defining with name in expansion");
4314 else if (istk->mstk->name)
4315 error(ERR_FATAL,
4316 "`%%rep' without `%%endrep' within"
4317 " expansion of macro `%s'",
4318 istk->mstk->name);
4322 * FIXME: investigate the relationship at this point between
4323 * istk->mstk and l->finishes
4326 MMacro *m = istk->mstk;
4327 istk->mstk = m->next_active;
4328 if (m->name) {
4330 * This was a real macro call, not a %rep, and
4331 * therefore the parameter information needs to
4332 * be freed.
4334 nasm_free(m->params);
4335 free_tlist(m->iline);
4336 nasm_free(m->paramlen);
4337 l->finishes->in_progress = false;
4338 } else
4339 free_mmacro(m);
4341 istk->expansion = l->next;
4342 nasm_free(l);
4343 list->downlevel(LIST_MACRO);
4346 while (1) { /* until we get a line we can use */
4348 if (istk->expansion) { /* from a macro expansion */
4349 char *p;
4350 Line *l = istk->expansion;
4351 if (istk->mstk)
4352 istk->mstk->lineno++;
4353 tline = l->first;
4354 istk->expansion = l->next;
4355 nasm_free(l);
4356 p = detoken(tline, false);
4357 list->line(LIST_MACRO, p);
4358 nasm_free(p);
4359 break;
4361 line = read_line();
4362 if (line) { /* from the current input file */
4363 line = prepreproc(line);
4364 tline = tokenize(line);
4365 nasm_free(line);
4366 break;
4369 * The current file has ended; work down the istk
4372 Include *i = istk;
4373 fclose(i->fp);
4374 if (i->conds)
4375 error(ERR_FATAL,
4376 "expected `%%endif' before end of file");
4377 /* only set line and file name if there's a next node */
4378 if (i->next) {
4379 src_set_linnum(i->lineno);
4380 nasm_free(src_set_fname(i->fname));
4382 istk = i->next;
4383 list->downlevel(LIST_INCLUDE);
4384 nasm_free(i);
4385 if (!istk)
4386 return NULL;
4391 * We must expand MMacro parameters and MMacro-local labels
4392 * _before_ we plunge into directive processing, to cope
4393 * with things like `%define something %1' such as STRUC
4394 * uses. Unless we're _defining_ a MMacro, in which case
4395 * those tokens should be left alone to go into the
4396 * definition; and unless we're in a non-emitting
4397 * condition, in which case we don't want to meddle with
4398 * anything.
4400 if (!defining && !(istk->conds && !emitting(istk->conds->state))
4401 && !(istk->mstk && !istk->mstk->in_progress))
4402 tline = expand_mmac_params(tline);
4405 * Check the line to see if it's a preprocessor directive.
4407 if (do_directive(tline) == DIRECTIVE_FOUND) {
4408 continue;
4409 } else if (defining) {
4411 * We're defining a multi-line macro. We emit nothing
4412 * at all, and just
4413 * shove the tokenized line on to the macro definition.
4415 Line *l = nasm_malloc(sizeof(Line));
4416 l->next = defining->expansion;
4417 l->first = tline;
4418 l->finishes = NULL;
4419 defining->expansion = l;
4420 continue;
4421 } else if (istk->conds && !emitting(istk->conds->state)) {
4423 * We're in a non-emitting branch of a condition block.
4424 * Emit nothing at all, not even a blank line: when we
4425 * emerge from the condition we'll give a line-number
4426 * directive so we keep our place correctly.
4428 free_tlist(tline);
4429 continue;
4430 } else if (istk->mstk && !istk->mstk->in_progress) {
4432 * We're in a %rep block which has been terminated, so
4433 * we're walking through to the %endrep without
4434 * emitting anything. Emit nothing at all, not even a
4435 * blank line: when we emerge from the %rep block we'll
4436 * give a line-number directive so we keep our place
4437 * correctly.
4439 free_tlist(tline);
4440 continue;
4441 } else {
4442 tline = expand_smacro(tline);
4443 if (!expand_mmacro(tline)) {
4445 * De-tokenize the line again, and emit it.
4447 line = detoken(tline, true);
4448 free_tlist(tline);
4449 break;
4450 } else {
4451 continue; /* expand_mmacro calls free_tlist */
4456 return line;
4459 static void pp_cleanup(int pass)
4461 if (defining) {
4462 if(defining->name) {
4463 error(ERR_NONFATAL,
4464 "end of file while still defining macro `%s'",
4465 defining->name);
4466 } else {
4467 error(ERR_NONFATAL, "end of file while still in %%rep");
4470 free_mmacro(defining);
4472 while (cstk)
4473 ctx_pop();
4474 free_macros();
4475 while (istk) {
4476 Include *i = istk;
4477 istk = istk->next;
4478 fclose(i->fp);
4479 nasm_free(i->fname);
4480 nasm_free(i);
4482 while (cstk)
4483 ctx_pop();
4484 nasm_free(src_set_fname(NULL));
4485 if (pass == 0) {
4486 IncPath *i;
4487 free_llist(predef);
4488 delete_Blocks();
4489 while ((i = ipath)) {
4490 ipath = i->next;
4491 if (i->path)
4492 nasm_free(i->path);
4493 nasm_free(i);
4498 void pp_include_path(char *path)
4500 IncPath *i;
4502 i = nasm_malloc(sizeof(IncPath));
4503 i->path = path ? nasm_strdup(path) : NULL;
4504 i->next = NULL;
4506 if (ipath != NULL) {
4507 IncPath *j = ipath;
4508 while (j->next != NULL)
4509 j = j->next;
4510 j->next = i;
4511 } else {
4512 ipath = i;
4516 void pp_pre_include(char *fname)
4518 Token *inc, *space, *name;
4519 Line *l;
4521 name = new_Token(NULL, TOK_INTERNAL_STRING, fname, 0);
4522 space = new_Token(name, TOK_WHITESPACE, NULL, 0);
4523 inc = new_Token(space, TOK_PREPROC_ID, "%include", 0);
4525 l = nasm_malloc(sizeof(Line));
4526 l->next = predef;
4527 l->first = inc;
4528 l->finishes = NULL;
4529 predef = l;
4532 void pp_pre_define(char *definition)
4534 Token *def, *space;
4535 Line *l;
4536 char *equals;
4538 equals = strchr(definition, '=');
4539 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
4540 def = new_Token(space, TOK_PREPROC_ID, "%define", 0);
4541 if (equals)
4542 *equals = ' ';
4543 space->next = tokenize(definition);
4544 if (equals)
4545 *equals = '=';
4547 l = nasm_malloc(sizeof(Line));
4548 l->next = predef;
4549 l->first = def;
4550 l->finishes = NULL;
4551 predef = l;
4554 void pp_pre_undefine(char *definition)
4556 Token *def, *space;
4557 Line *l;
4559 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
4560 def = new_Token(space, TOK_PREPROC_ID, "%undef", 0);
4561 space->next = tokenize(definition);
4563 l = nasm_malloc(sizeof(Line));
4564 l->next = predef;
4565 l->first = def;
4566 l->finishes = NULL;
4567 predef = l;
4571 * Added by Keith Kanios:
4573 * This function is used to assist with "runtime" preprocessor
4574 * directives. (e.g. pp_runtime("%define __BITS__ 64");)
4576 * ERRORS ARE IGNORED HERE, SO MAKE COMPLETELY SURE THAT YOU
4577 * PASS A VALID STRING TO THIS FUNCTION!!!!!
4580 void pp_runtime(char *definition)
4582 Token *def;
4584 def = tokenize(definition);
4585 if(do_directive(def) == NO_DIRECTIVE_FOUND)
4586 free_tlist(def);
4590 void pp_extra_stdmac(macros_t *macros)
4592 extrastdmac = macros;
4595 static void make_tok_num(Token * tok, int64_t val)
4597 char numbuf[20];
4598 snprintf(numbuf, sizeof(numbuf), "%"PRId64"", val);
4599 tok->text = nasm_strdup(numbuf);
4600 tok->type = TOK_NUMBER;
4603 Preproc nasmpp = {
4604 pp_reset,
4605 pp_getline,
4606 pp_cleanup