Documentation facelift backported to 2.03.x
[nasm.git] / preproc.c
blob8859005bfe7f75955a060ccc0f06418201c5ff25
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 SMacro *mac; /* associated macro for TOK_SMAC_END */
170 enum pp_token_type type;
174 * Multi-line macro definitions are stored as a linked list of
175 * these, which is essentially a container to allow several linked
176 * lists of Tokens.
178 * Note that in this module, linked lists are treated as stacks
179 * wherever possible. For this reason, Lines are _pushed_ on to the
180 * `expansion' field in MMacro structures, so that the linked list,
181 * if walked, would give the macro lines in reverse order; this
182 * means that we can walk the list when expanding a macro, and thus
183 * push the lines on to the `expansion' field in _istk_ in reverse
184 * order (so that when popped back off they are in the right
185 * order). It may seem cockeyed, and it relies on my design having
186 * an even number of steps in, but it works...
188 * Some of these structures, rather than being actual lines, are
189 * markers delimiting the end of the expansion of a given macro.
190 * This is for use in the cycle-tracking and %rep-handling code.
191 * Such structures have `finishes' non-NULL, and `first' NULL. All
192 * others have `finishes' NULL, but `first' may still be NULL if
193 * the line is blank.
195 struct Line {
196 Line *next;
197 MMacro *finishes;
198 Token *first;
202 * To handle an arbitrary level of file inclusion, we maintain a
203 * stack (ie linked list) of these things.
205 struct Include {
206 Include *next;
207 FILE *fp;
208 Cond *conds;
209 Line *expansion;
210 char *fname;
211 int lineno, lineinc;
212 MMacro *mstk; /* stack of active macros/reps */
216 * Include search path. This is simply a list of strings which get
217 * prepended, in turn, to the name of an include file, in an
218 * attempt to find the file if it's not in the current directory.
220 struct IncPath {
221 IncPath *next;
222 char *path;
226 * Conditional assembly: we maintain a separate stack of these for
227 * each level of file inclusion. (The only reason we keep the
228 * stacks separate is to ensure that a stray `%endif' in a file
229 * included from within the true branch of a `%if' won't terminate
230 * it and cause confusion: instead, rightly, it'll cause an error.)
232 struct Cond {
233 Cond *next;
234 int state;
236 enum {
238 * These states are for use just after %if or %elif: IF_TRUE
239 * means the condition has evaluated to truth so we are
240 * currently emitting, whereas IF_FALSE means we are not
241 * currently emitting but will start doing so if a %else comes
242 * up. In these states, all directives are admissible: %elif,
243 * %else and %endif. (And of course %if.)
245 COND_IF_TRUE, COND_IF_FALSE,
247 * These states come up after a %else: ELSE_TRUE means we're
248 * emitting, and ELSE_FALSE means we're not. In ELSE_* states,
249 * any %elif or %else will cause an error.
251 COND_ELSE_TRUE, COND_ELSE_FALSE,
253 * This state means that we're not emitting now, and also that
254 * nothing until %endif will be emitted at all. It's for use in
255 * two circumstances: (i) when we've had our moment of emission
256 * and have now started seeing %elifs, and (ii) when the
257 * condition construct in question is contained within a
258 * non-emitting branch of a larger condition construct.
260 COND_NEVER
262 #define emitting(x) ( (x) == COND_IF_TRUE || (x) == COND_ELSE_TRUE )
265 * These defines are used as the possible return values for do_directive
267 #define NO_DIRECTIVE_FOUND 0
268 #define DIRECTIVE_FOUND 1
271 * Condition codes. Note that we use c_ prefix not C_ because C_ is
272 * used in nasm.h for the "real" condition codes. At _this_ level,
273 * we treat CXZ and ECXZ as condition codes, albeit non-invertible
274 * ones, so we need a different enum...
276 static const char * const conditions[] = {
277 "a", "ae", "b", "be", "c", "cxz", "e", "ecxz", "g", "ge", "l", "le",
278 "na", "nae", "nb", "nbe", "nc", "ne", "ng", "nge", "nl", "nle", "no",
279 "np", "ns", "nz", "o", "p", "pe", "po", "rcxz", "s", "z"
281 enum pp_conds {
282 c_A, c_AE, c_B, c_BE, c_C, c_CXZ, c_E, c_ECXZ, c_G, c_GE, c_L, c_LE,
283 c_NA, c_NAE, c_NB, c_NBE, c_NC, c_NE, c_NG, c_NGE, c_NL, c_NLE, c_NO,
284 c_NP, c_NS, c_NZ, c_O, c_P, c_PE, c_PO, c_RCXZ, c_S, c_Z,
285 c_none = -1
287 static const enum pp_conds inverse_ccs[] = {
288 c_NA, c_NAE, c_NB, c_NBE, c_NC, -1, c_NE, -1, c_NG, c_NGE, c_NL, c_NLE,
289 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,
290 c_Z, c_NO, c_NP, c_PO, c_PE, -1, c_NS, c_NZ
294 * Directive names.
296 /* If this is a an IF, ELIF, ELSE or ENDIF keyword */
297 static int is_condition(enum preproc_token arg)
299 return PP_IS_COND(arg) || (arg == PP_ELSE) || (arg == PP_ENDIF);
302 /* For TASM compatibility we need to be able to recognise TASM compatible
303 * conditional compilation directives. Using the NASM pre-processor does
304 * not work, so we look for them specifically from the following list and
305 * then jam in the equivalent NASM directive into the input stream.
308 enum {
309 TM_ARG, TM_ELIF, TM_ELSE, TM_ENDIF, TM_IF, TM_IFDEF, TM_IFDIFI,
310 TM_IFNDEF, TM_INCLUDE, TM_LOCAL
313 static const char * const tasm_directives[] = {
314 "arg", "elif", "else", "endif", "if", "ifdef", "ifdifi",
315 "ifndef", "include", "local"
318 static int StackSize = 4;
319 static char *StackPointer = "ebp";
320 static int ArgOffset = 8;
321 static int LocalOffset = 0;
323 static Context *cstk;
324 static Include *istk;
325 static IncPath *ipath = NULL;
327 static efunc _error; /* Pointer to client-provided error reporting function */
328 static evalfunc evaluate;
330 static int pass; /* HACK: pass 0 = generate dependencies only */
331 static StrList **dephead, **deptail; /* Dependency list */
333 static uint64_t unique; /* unique identifier numbers */
335 static Line *predef = NULL;
337 static ListGen *list;
340 * The current set of multi-line macros we have defined.
342 static struct hash_table mmacros;
345 * The current set of single-line macros we have defined.
347 static struct hash_table smacros;
350 * The multi-line macro we are currently defining, or the %rep
351 * block we are currently reading, if any.
353 static MMacro *defining;
356 * The number of macro parameters to allocate space for at a time.
358 #define PARAM_DELTA 16
361 * The standard macro set: defined in macros.c in the array nasm_stdmac.
362 * This gives our position in the macro set, when we're processing it.
364 static const char * const *stdmacpos;
367 * The extra standard macros that come from the object format, if
368 * any.
370 static const char * const *extrastdmac = NULL;
371 bool any_extrastdmac;
374 * Tokens are allocated in blocks to improve speed
376 #define TOKEN_BLOCKSIZE 4096
377 static Token *freeTokens = NULL;
378 struct Blocks {
379 Blocks *next;
380 void *chunk;
383 static Blocks blocks = { NULL, NULL };
386 * Forward declarations.
388 static Token *expand_mmac_params(Token * tline);
389 static Token *expand_smacro(Token * tline);
390 static Token *expand_id(Token * tline);
391 static Context *get_ctx(char *name, bool all_contexts);
392 static void make_tok_num(Token * tok, int64_t val);
393 static void error(int severity, const char *fmt, ...);
394 static void *new_Block(size_t size);
395 static void delete_Blocks(void);
396 static Token *new_Token(Token * next, enum pp_token_type type,
397 const char *text, int txtlen);
398 static Token *delete_Token(Token * t);
401 * Macros for safe checking of token pointers, avoid *(NULL)
403 #define tok_type_(x,t) ((x) && (x)->type == (t))
404 #define skip_white_(x) if (tok_type_((x), TOK_WHITESPACE)) (x)=(x)->next
405 #define tok_is_(x,v) (tok_type_((x), TOK_OTHER) && !strcmp((x)->text,(v)))
406 #define tok_isnt_(x,v) ((x) && ((x)->type!=TOK_OTHER || strcmp((x)->text,(v))))
408 /* Handle TASM specific directives, which do not contain a % in
409 * front of them. We do it here because I could not find any other
410 * place to do it for the moment, and it is a hack (ideally it would
411 * be nice to be able to use the NASM pre-processor to do it).
413 static char *check_tasm_directive(char *line)
415 int32_t i, j, k, m, len;
416 char *p = line, *oldline, oldchar;
418 /* Skip whitespace */
419 while (isspace(*p) && *p != 0)
420 p++;
422 /* Binary search for the directive name */
423 i = -1;
424 j = elements(tasm_directives);
425 len = 0;
426 while (!isspace(p[len]) && p[len] != 0)
427 len++;
428 if (len) {
429 oldchar = p[len];
430 p[len] = 0;
431 while (j - i > 1) {
432 k = (j + i) / 2;
433 m = nasm_stricmp(p, tasm_directives[k]);
434 if (m == 0) {
435 /* We have found a directive, so jam a % in front of it
436 * so that NASM will then recognise it as one if it's own.
438 p[len] = oldchar;
439 len = strlen(p);
440 oldline = line;
441 line = nasm_malloc(len + 2);
442 line[0] = '%';
443 if (k == TM_IFDIFI) {
444 /* NASM does not recognise IFDIFI, so we convert it to
445 * %ifdef BOGUS. This is not used in NASM comaptible
446 * code, but does need to parse for the TASM macro
447 * package.
449 strcpy(line + 1, "ifdef BOGUS");
450 } else {
451 memcpy(line + 1, p, len + 1);
453 nasm_free(oldline);
454 return line;
455 } else if (m < 0) {
456 j = k;
457 } else
458 i = k;
460 p[len] = oldchar;
462 return line;
466 * The pre-preprocessing stage... This function translates line
467 * number indications as they emerge from GNU cpp (`# lineno "file"
468 * flags') into NASM preprocessor line number indications (`%line
469 * lineno file').
471 static char *prepreproc(char *line)
473 int lineno, fnlen;
474 char *fname, *oldline;
476 if (line[0] == '#' && line[1] == ' ') {
477 oldline = line;
478 fname = oldline + 2;
479 lineno = atoi(fname);
480 fname += strspn(fname, "0123456789 ");
481 if (*fname == '"')
482 fname++;
483 fnlen = strcspn(fname, "\"");
484 line = nasm_malloc(20 + fnlen);
485 snprintf(line, 20 + fnlen, "%%line %d %.*s", lineno, fnlen, fname);
486 nasm_free(oldline);
488 if (tasm_compatible_mode)
489 return check_tasm_directive(line);
490 return line;
494 * Free a linked list of tokens.
496 static void free_tlist(Token * list)
498 while (list) {
499 list = delete_Token(list);
504 * Free a linked list of lines.
506 static void free_llist(Line * list)
508 Line *l;
509 while (list) {
510 l = list;
511 list = list->next;
512 free_tlist(l->first);
513 nasm_free(l);
518 * Free an MMacro
520 static void free_mmacro(MMacro * m)
522 nasm_free(m->name);
523 free_tlist(m->dlist);
524 nasm_free(m->defaults);
525 free_llist(m->expansion);
526 nasm_free(m);
530 * Free all currently defined macros, and free the hash tables
532 static void free_smacro_table(struct hash_table *smt)
534 SMacro *s;
535 const char *key;
536 struct hash_tbl_node *it = NULL;
538 while ((s = hash_iterate(smt, &it, &key)) != NULL) {
539 nasm_free((void *)key);
540 while (s) {
541 SMacro *ns = s->next;
542 nasm_free(s->name);
543 free_tlist(s->expansion);
544 nasm_free(s);
545 s = ns;
548 hash_free(smt);
551 static void free_mmacro_table(struct hash_table *mmt)
553 MMacro *m;
554 const char *key;
555 struct hash_tbl_node *it = NULL;
557 it = NULL;
558 while ((m = hash_iterate(mmt, &it, &key)) != NULL) {
559 nasm_free((void *)key);
560 while (m) {
561 MMacro *nm = m->next;
562 free_mmacro(m);
563 m = nm;
566 hash_free(mmt);
569 static void free_macros(void)
571 free_smacro_table(&smacros);
572 free_mmacro_table(&mmacros);
576 * Initialize the hash tables
578 static void init_macros(void)
580 hash_init(&smacros, HASH_LARGE);
581 hash_init(&mmacros, HASH_LARGE);
585 * Pop the context stack.
587 static void ctx_pop(void)
589 Context *c = cstk;
591 cstk = cstk->next;
592 free_smacro_table(&c->localmac);
593 nasm_free(c->name);
594 nasm_free(c);
598 * Search for a key in the hash index; adding it if necessary
599 * (in which case we initialize the data pointer to NULL.)
601 static void **
602 hash_findi_add(struct hash_table *hash, const char *str)
604 struct hash_insert hi;
605 void **r;
606 char *strx;
608 r = hash_findi(hash, str, &hi);
609 if (r)
610 return r;
612 strx = nasm_strdup(str); /* Use a more efficient allocator here? */
613 return hash_add(&hi, strx, NULL);
617 * Like hash_findi, but returns the data element rather than a pointer
618 * to it. Used only when not adding a new element, hence no third
619 * argument.
621 static void *
622 hash_findix(struct hash_table *hash, const char *str)
624 void **p;
626 p = hash_findi(hash, str, NULL);
627 return p ? *p : NULL;
630 #define BUF_DELTA 512
632 * Read a line from the top file in istk, handling multiple CR/LFs
633 * at the end of the line read, and handling spurious ^Zs. Will
634 * return lines from the standard macro set if this has not already
635 * been done.
637 static char *read_line(void)
639 char *buffer, *p, *q;
640 int bufsize, continued_count;
642 if (stdmacpos) {
643 if (*stdmacpos) {
644 char *ret = nasm_strdup(*stdmacpos++);
645 if (!*stdmacpos && any_extrastdmac) {
646 stdmacpos = extrastdmac;
647 any_extrastdmac = false;
648 return ret;
651 * Nasty hack: here we push the contents of `predef' on
652 * to the top-level expansion stack, since this is the
653 * most convenient way to implement the pre-include and
654 * pre-define features.
656 if (!*stdmacpos) {
657 Line *pd, *l;
658 Token *head, **tail, *t;
660 for (pd = predef; pd; pd = pd->next) {
661 head = NULL;
662 tail = &head;
663 for (t = pd->first; t; t = t->next) {
664 *tail = new_Token(NULL, t->type, t->text, 0);
665 tail = &(*tail)->next;
667 l = nasm_malloc(sizeof(Line));
668 l->next = istk->expansion;
669 l->first = head;
670 l->finishes = false;
671 istk->expansion = l;
674 return ret;
675 } else {
676 stdmacpos = NULL;
680 bufsize = BUF_DELTA;
681 buffer = nasm_malloc(BUF_DELTA);
682 p = buffer;
683 continued_count = 0;
684 while (1) {
685 q = fgets(p, bufsize - (p - buffer), istk->fp);
686 if (!q)
687 break;
688 p += strlen(p);
689 if (p > buffer && p[-1] == '\n') {
690 /* Convert backslash-CRLF line continuation sequences into
691 nothing at all (for DOS and Windows) */
692 if (((p - 2) > buffer) && (p[-3] == '\\') && (p[-2] == '\r')) {
693 p -= 3;
694 *p = 0;
695 continued_count++;
697 /* Also convert backslash-LF line continuation sequences into
698 nothing at all (for Unix) */
699 else if (((p - 1) > buffer) && (p[-2] == '\\')) {
700 p -= 2;
701 *p = 0;
702 continued_count++;
703 } else {
704 break;
707 if (p - buffer > bufsize - 10) {
708 int32_t offset = p - buffer;
709 bufsize += BUF_DELTA;
710 buffer = nasm_realloc(buffer, bufsize);
711 p = buffer + offset; /* prevent stale-pointer problems */
715 if (!q && p == buffer) {
716 nasm_free(buffer);
717 return NULL;
720 src_set_linnum(src_get_linnum() + istk->lineinc +
721 (continued_count * istk->lineinc));
724 * Play safe: remove CRs as well as LFs, if any of either are
725 * present at the end of the line.
727 while (--p >= buffer && (*p == '\n' || *p == '\r'))
728 *p = '\0';
731 * Handle spurious ^Z, which may be inserted into source files
732 * by some file transfer utilities.
734 buffer[strcspn(buffer, "\032")] = '\0';
736 list->line(LIST_READ, buffer);
738 return buffer;
742 * Tokenize a line of text. This is a very simple process since we
743 * don't need to parse the value out of e.g. numeric tokens: we
744 * simply split one string into many.
746 static Token *tokenize(char *line)
748 char *p = line;
749 enum pp_token_type type;
750 Token *list = NULL;
751 Token *t, **tail = &list;
753 while (*line) {
754 p = line;
755 if (*p == '%') {
756 p++;
757 if (isdigit(*p) ||
758 ((*p == '-' || *p == '+') && isdigit(p[1])) ||
759 ((*p == '+') && (isspace(p[1]) || !p[1]))) {
760 do {
761 p++;
763 while (isdigit(*p));
764 type = TOK_PREPROC_ID;
765 } else if (*p == '{') {
766 p++;
767 while (*p && *p != '}') {
768 p[-1] = *p;
769 p++;
771 p[-1] = '\0';
772 if (*p)
773 p++;
774 type = TOK_PREPROC_ID;
775 } else if (*p == '?') {
776 type = TOK_PREPROC_Q; /* %? */
777 p++;
778 if (*p == '?') {
779 type = TOK_PREPROC_QQ; /* %?? */
780 p++;
782 } else if (isidchar(*p) ||
783 ((*p == '!' || *p == '%' || *p == '$') &&
784 isidchar(p[1]))) {
785 do {
786 p++;
788 while (isidchar(*p));
789 type = TOK_PREPROC_ID;
790 } else {
791 type = TOK_OTHER;
792 if (*p == '%')
793 p++;
795 } else if (isidstart(*p) || (*p == '$' && isidstart(p[1]))) {
796 type = TOK_ID;
797 p++;
798 while (*p && isidchar(*p))
799 p++;
800 } else if (*p == '\'' || *p == '"' || *p == '`') {
802 * A string token.
804 type = TOK_STRING;
805 p = nasm_skip_string(p);
807 if (*p) {
808 p++;
809 } else {
810 error(ERR_WARNING, "unterminated string");
811 /* Handling unterminated strings by UNV */
812 /* type = -1; */
814 } else if (isnumstart(*p)) {
815 bool is_hex = false;
816 bool is_float = false;
817 bool has_e = false;
818 char c, *r;
821 * A numeric token.
824 if (*p == '$') {
825 p++;
826 is_hex = true;
829 for (;;) {
830 c = *p++;
832 if (!is_hex && (c == 'e' || c == 'E')) {
833 has_e = true;
834 if (*p == '+' || *p == '-') {
835 /* e can only be followed by +/- if it is either a
836 prefixed hex number or a floating-point number */
837 p++;
838 is_float = true;
840 } else if (c == 'H' || c == 'h' || c == 'X' || c == 'x') {
841 is_hex = true;
842 } else if (c == 'P' || c == 'p') {
843 is_float = true;
844 if (*p == '+' || *p == '-')
845 p++;
846 } else if (isnumchar(c) || c == '_')
847 ; /* just advance */
848 else if (c == '.') {
849 /* we need to deal with consequences of the legacy
850 parser, like "1.nolist" being two tokens
851 (TOK_NUMBER, TOK_ID) here; at least give it
852 a shot for now. In the future, we probably need
853 a flex-based scanner with proper pattern matching
854 to do it as well as it can be done. Nothing in
855 the world is going to help the person who wants
856 0x123.p16 interpreted as two tokens, though. */
857 r = p;
858 while (*r == '_')
859 r++;
861 if (isdigit(*r) || (is_hex && isxdigit(*r)) ||
862 (!is_hex && (*r == 'e' || *r == 'E')) ||
863 (*r == 'p' || *r == 'P')) {
864 p = r;
865 is_float = true;
866 } else
867 break; /* Terminate the token */
868 } else
869 break;
871 p--; /* Point to first character beyond number */
873 if (has_e && !is_hex) {
874 /* 1e13 is floating-point, but 1e13h is not */
875 is_float = true;
878 type = is_float ? TOK_FLOAT : TOK_NUMBER;
879 } else if (isspace(*p)) {
880 type = TOK_WHITESPACE;
881 p++;
882 while (*p && isspace(*p))
883 p++;
885 * Whitespace just before end-of-line is discarded by
886 * pretending it's a comment; whitespace just before a
887 * comment gets lumped into the comment.
889 if (!*p || *p == ';') {
890 type = TOK_COMMENT;
891 while (*p)
892 p++;
894 } else if (*p == ';') {
895 type = TOK_COMMENT;
896 while (*p)
897 p++;
898 } else {
900 * Anything else is an operator of some kind. We check
901 * for all the double-character operators (>>, <<, //,
902 * %%, <=, >=, ==, !=, <>, &&, ||, ^^), but anything
903 * else is a single-character operator.
905 type = TOK_OTHER;
906 if ((p[0] == '>' && p[1] == '>') ||
907 (p[0] == '<' && p[1] == '<') ||
908 (p[0] == '/' && p[1] == '/') ||
909 (p[0] == '<' && p[1] == '=') ||
910 (p[0] == '>' && p[1] == '=') ||
911 (p[0] == '=' && p[1] == '=') ||
912 (p[0] == '!' && p[1] == '=') ||
913 (p[0] == '<' && p[1] == '>') ||
914 (p[0] == '&' && p[1] == '&') ||
915 (p[0] == '|' && p[1] == '|') ||
916 (p[0] == '^' && p[1] == '^')) {
917 p++;
919 p++;
922 /* Handling unterminated string by UNV */
923 /*if (type == -1)
925 *tail = t = new_Token(NULL, TOK_STRING, line, p-line+1);
926 t->text[p-line] = *line;
927 tail = &t->next;
929 else */
930 if (type != TOK_COMMENT) {
931 *tail = t = new_Token(NULL, type, line, p - line);
932 tail = &t->next;
934 line = p;
936 return list;
940 * this function allocates a new managed block of memory and
941 * returns a pointer to the block. The managed blocks are
942 * deleted only all at once by the delete_Blocks function.
944 static void *new_Block(size_t size)
946 Blocks *b = &blocks;
948 /* first, get to the end of the linked list */
949 while (b->next)
950 b = b->next;
951 /* now allocate the requested chunk */
952 b->chunk = nasm_malloc(size);
954 /* now allocate a new block for the next request */
955 b->next = nasm_malloc(sizeof(Blocks));
956 /* and initialize the contents of the new block */
957 b->next->next = NULL;
958 b->next->chunk = NULL;
959 return b->chunk;
963 * this function deletes all managed blocks of memory
965 static void delete_Blocks(void)
967 Blocks *a, *b = &blocks;
970 * keep in mind that the first block, pointed to by blocks
971 * is a static and not dynamically allocated, so we don't
972 * free it.
974 while (b) {
975 if (b->chunk)
976 nasm_free(b->chunk);
977 a = b;
978 b = b->next;
979 if (a != &blocks)
980 nasm_free(a);
985 * this function creates a new Token and passes a pointer to it
986 * back to the caller. It sets the type and text elements, and
987 * also the mac and next elements to NULL.
989 static Token *new_Token(Token * next, enum pp_token_type type,
990 const char *text, int txtlen)
992 Token *t;
993 int i;
995 if (freeTokens == NULL) {
996 freeTokens = (Token *) new_Block(TOKEN_BLOCKSIZE * sizeof(Token));
997 for (i = 0; i < TOKEN_BLOCKSIZE - 1; i++)
998 freeTokens[i].next = &freeTokens[i + 1];
999 freeTokens[i].next = NULL;
1001 t = freeTokens;
1002 freeTokens = t->next;
1003 t->next = next;
1004 t->mac = NULL;
1005 t->type = type;
1006 if (type == TOK_WHITESPACE || text == NULL) {
1007 t->text = NULL;
1008 } else {
1009 if (txtlen == 0)
1010 txtlen = strlen(text);
1011 t->text = nasm_malloc(txtlen+1);
1012 memcpy(t->text, text, txtlen);
1013 t->text[txtlen] = '\0';
1015 return t;
1018 static Token *delete_Token(Token * t)
1020 Token *next = t->next;
1021 nasm_free(t->text);
1022 t->next = freeTokens;
1023 freeTokens = t;
1024 return next;
1028 * Convert a line of tokens back into text.
1029 * If expand_locals is not zero, identifiers of the form "%$*xxx"
1030 * will be transformed into ..@ctxnum.xxx
1032 static char *detoken(Token * tlist, bool expand_locals)
1034 Token *t;
1035 int len;
1036 char *line, *p;
1037 const char *q;
1039 len = 0;
1040 for (t = tlist; t; t = t->next) {
1041 if (t->type == TOK_PREPROC_ID && t->text[1] == '!') {
1042 char *p = getenv(t->text + 2);
1043 nasm_free(t->text);
1044 if (p)
1045 t->text = nasm_strdup(p);
1046 else
1047 t->text = NULL;
1049 /* Expand local macros here and not during preprocessing */
1050 if (expand_locals &&
1051 t->type == TOK_PREPROC_ID && t->text &&
1052 t->text[0] == '%' && t->text[1] == '$') {
1053 Context *ctx = get_ctx(t->text, false);
1054 if (ctx) {
1055 char buffer[40];
1056 char *p, *q = t->text + 2;
1058 q += strspn(q, "$");
1059 snprintf(buffer, sizeof(buffer), "..@%"PRIu32".", ctx->number);
1060 p = nasm_strcat(buffer, q);
1061 nasm_free(t->text);
1062 t->text = p;
1065 if (t->type == TOK_WHITESPACE) {
1066 len++;
1067 } else if (t->text) {
1068 len += strlen(t->text);
1071 p = line = nasm_malloc(len + 1);
1072 for (t = tlist; t; t = t->next) {
1073 if (t->type == TOK_WHITESPACE) {
1074 *p++ = ' ';
1075 } else if (t->text) {
1076 q = t->text;
1077 while (*q)
1078 *p++ = *q++;
1081 *p = '\0';
1082 return line;
1086 * A scanner, suitable for use by the expression evaluator, which
1087 * operates on a line of Tokens. Expects a pointer to a pointer to
1088 * the first token in the line to be passed in as its private_data
1089 * field.
1091 * FIX: This really needs to be unified with stdscan.
1093 static int ppscan(void *private_data, struct tokenval *tokval)
1095 Token **tlineptr = private_data;
1096 Token *tline;
1097 char ourcopy[MAX_KEYWORD+1], *p, *r, *s;
1099 do {
1100 tline = *tlineptr;
1101 *tlineptr = tline ? tline->next : NULL;
1103 while (tline && (tline->type == TOK_WHITESPACE ||
1104 tline->type == TOK_COMMENT));
1106 if (!tline)
1107 return tokval->t_type = TOKEN_EOS;
1109 tokval->t_charptr = tline->text;
1111 if (tline->text[0] == '$' && !tline->text[1])
1112 return tokval->t_type = TOKEN_HERE;
1113 if (tline->text[0] == '$' && tline->text[1] == '$' && !tline->text[2])
1114 return tokval->t_type = TOKEN_BASE;
1116 if (tline->type == TOK_ID) {
1117 p = tokval->t_charptr = tline->text;
1118 if (p[0] == '$') {
1119 tokval->t_charptr++;
1120 return tokval->t_type = TOKEN_ID;
1123 for (r = p, s = ourcopy; *r; r++) {
1124 if (r >= p+MAX_KEYWORD)
1125 return tokval->t_type = TOKEN_ID; /* Not a keyword */
1126 *s++ = nasm_tolower(*r);
1128 *s = '\0';
1129 /* right, so we have an identifier sitting in temp storage. now,
1130 * is it actually a register or instruction name, or what? */
1131 return nasm_token_hash(ourcopy, tokval);
1134 if (tline->type == TOK_NUMBER) {
1135 bool rn_error;
1136 tokval->t_integer = readnum(tline->text, &rn_error);
1137 tokval->t_charptr = tline->text;
1138 if (rn_error)
1139 return tokval->t_type = TOKEN_ERRNUM;
1140 else
1141 return tokval->t_type = TOKEN_NUM;
1144 if (tline->type == TOK_FLOAT) {
1145 return tokval->t_type = TOKEN_FLOAT;
1148 if (tline->type == TOK_STRING) {
1149 char bq, *ep;
1151 bq = tline->text[0];
1152 tokval->t_charptr = tline->text;
1153 tokval->t_inttwo = nasm_unquote(tline->text, &ep);
1155 if (ep[0] != bq || ep[1] != '\0')
1156 return tokval->t_type = TOKEN_ERRSTR;
1157 else
1158 return tokval->t_type = TOKEN_STR;
1161 if (tline->type == TOK_OTHER) {
1162 if (!strcmp(tline->text, "<<"))
1163 return tokval->t_type = TOKEN_SHL;
1164 if (!strcmp(tline->text, ">>"))
1165 return tokval->t_type = TOKEN_SHR;
1166 if (!strcmp(tline->text, "//"))
1167 return tokval->t_type = TOKEN_SDIV;
1168 if (!strcmp(tline->text, "%%"))
1169 return tokval->t_type = TOKEN_SMOD;
1170 if (!strcmp(tline->text, "=="))
1171 return tokval->t_type = TOKEN_EQ;
1172 if (!strcmp(tline->text, "<>"))
1173 return tokval->t_type = TOKEN_NE;
1174 if (!strcmp(tline->text, "!="))
1175 return tokval->t_type = TOKEN_NE;
1176 if (!strcmp(tline->text, "<="))
1177 return tokval->t_type = TOKEN_LE;
1178 if (!strcmp(tline->text, ">="))
1179 return tokval->t_type = TOKEN_GE;
1180 if (!strcmp(tline->text, "&&"))
1181 return tokval->t_type = TOKEN_DBL_AND;
1182 if (!strcmp(tline->text, "^^"))
1183 return tokval->t_type = TOKEN_DBL_XOR;
1184 if (!strcmp(tline->text, "||"))
1185 return tokval->t_type = TOKEN_DBL_OR;
1189 * We have no other options: just return the first character of
1190 * the token text.
1192 return tokval->t_type = tline->text[0];
1196 * Compare a string to the name of an existing macro; this is a
1197 * simple wrapper which calls either strcmp or nasm_stricmp
1198 * depending on the value of the `casesense' parameter.
1200 static int mstrcmp(const char *p, const char *q, bool casesense)
1202 return casesense ? strcmp(p, q) : nasm_stricmp(p, q);
1206 * Compare a string to the name of an existing macro; this is a
1207 * simple wrapper which calls either strcmp or nasm_stricmp
1208 * depending on the value of the `casesense' parameter.
1210 static int mmemcmp(const char *p, const char *q, size_t l, bool casesense)
1212 return casesense ? memcmp(p, q, l) : nasm_memicmp(p, q, l);
1216 * Return the Context structure associated with a %$ token. Return
1217 * NULL, having _already_ reported an error condition, if the
1218 * context stack isn't deep enough for the supplied number of $
1219 * signs.
1220 * If all_contexts == true, contexts that enclose current are
1221 * also scanned for such smacro, until it is found; if not -
1222 * only the context that directly results from the number of $'s
1223 * in variable's name.
1225 static Context *get_ctx(char *name, bool all_contexts)
1227 Context *ctx;
1228 SMacro *m;
1229 int i;
1231 if (!name || name[0] != '%' || name[1] != '$')
1232 return NULL;
1234 if (!cstk) {
1235 error(ERR_NONFATAL, "`%s': context stack is empty", name);
1236 return NULL;
1239 for (i = strspn(name + 2, "$"), ctx = cstk; (i > 0) && ctx; i--) {
1240 ctx = ctx->next;
1241 /* i--; Lino - 02/25/02 */
1243 if (!ctx) {
1244 error(ERR_NONFATAL, "`%s': context stack is only"
1245 " %d level%s deep", name, i - 1, (i == 2 ? "" : "s"));
1246 return NULL;
1248 if (!all_contexts)
1249 return ctx;
1251 do {
1252 /* Search for this smacro in found context */
1253 m = hash_findix(&ctx->localmac, name);
1254 while (m) {
1255 if (!mstrcmp(m->name, name, m->casesense))
1256 return ctx;
1257 m = m->next;
1259 ctx = ctx->next;
1261 while (ctx);
1262 return NULL;
1266 * Check to see if a file is already in a string list
1268 static bool in_list(const StrList *list, const char *str)
1270 while (list) {
1271 if (!strcmp(list->str, str))
1272 return true;
1273 list = list->next;
1275 return false;
1279 * Open an include file. This routine must always return a valid
1280 * file pointer if it returns - it's responsible for throwing an
1281 * ERR_FATAL and bombing out completely if not. It should also try
1282 * the include path one by one until it finds the file or reaches
1283 * the end of the path.
1285 static FILE *inc_fopen(const char *file, StrList **dhead, StrList ***dtail,
1286 bool missing_ok)
1288 FILE *fp;
1289 char *prefix = "";
1290 IncPath *ip = ipath;
1291 int len = strlen(file);
1292 size_t prefix_len = 0;
1293 StrList *sl;
1295 while (1) {
1296 sl = nasm_malloc(prefix_len+len+1+sizeof sl->next);
1297 memcpy(sl->str, prefix, prefix_len);
1298 memcpy(sl->str+prefix_len, file, len+1);
1299 fp = fopen(sl->str, "r");
1300 if (fp && dhead && !in_list(*dhead, sl->str)) {
1301 sl->next = NULL;
1302 **dtail = sl;
1303 *dtail = &sl->next;
1304 } else {
1305 nasm_free(sl);
1307 if (fp)
1308 return fp;
1309 if (!ip) {
1310 if (!missing_ok)
1311 break;
1312 prefix = NULL;
1313 } else {
1314 prefix = ip->path;
1315 ip = ip->next;
1317 if (prefix) {
1318 prefix_len = strlen(prefix);
1319 } else {
1320 /* -MG given and file not found */
1321 if (dhead && !in_list(*dhead, file)) {
1322 sl = nasm_malloc(len+1+sizeof sl->next);
1323 sl->next = NULL;
1324 strcpy(sl->str, file);
1325 **dtail = sl;
1326 *dtail = &sl->next;
1328 return NULL;
1332 error(ERR_FATAL, "unable to open include file `%s'", file);
1333 return NULL; /* never reached - placate compilers */
1337 * Determine if we should warn on defining a single-line macro of
1338 * name `name', with `nparam' parameters. If nparam is 0 or -1, will
1339 * return true if _any_ single-line macro of that name is defined.
1340 * Otherwise, will return true if a single-line macro with either
1341 * `nparam' or no parameters is defined.
1343 * If a macro with precisely the right number of parameters is
1344 * defined, or nparam is -1, the address of the definition structure
1345 * will be returned in `defn'; otherwise NULL will be returned. If `defn'
1346 * is NULL, no action will be taken regarding its contents, and no
1347 * error will occur.
1349 * Note that this is also called with nparam zero to resolve
1350 * `ifdef'.
1352 * If you already know which context macro belongs to, you can pass
1353 * the context pointer as first parameter; if you won't but name begins
1354 * with %$ the context will be automatically computed. If all_contexts
1355 * is true, macro will be searched in outer contexts as well.
1357 static bool
1358 smacro_defined(Context * ctx, char *name, int nparam, SMacro ** defn,
1359 bool nocase)
1361 struct hash_table *smtbl;
1362 SMacro *m;
1364 if (ctx) {
1365 smtbl = &ctx->localmac;
1366 } else if (name[0] == '%' && name[1] == '$') {
1367 if (cstk)
1368 ctx = get_ctx(name, false);
1369 if (!ctx)
1370 return false; /* got to return _something_ */
1371 smtbl = &ctx->localmac;
1372 } else {
1373 smtbl = &smacros;
1375 m = (SMacro *) hash_findix(smtbl, name);
1377 while (m) {
1378 if (!mstrcmp(m->name, name, m->casesense && nocase) &&
1379 (nparam <= 0 || m->nparam == 0 || nparam == (int) m->nparam)) {
1380 if (defn) {
1381 if (nparam == (int) m->nparam || nparam == -1)
1382 *defn = m;
1383 else
1384 *defn = NULL;
1386 return true;
1388 m = m->next;
1391 return false;
1395 * Count and mark off the parameters in a multi-line macro call.
1396 * This is called both from within the multi-line macro expansion
1397 * code, and also to mark off the default parameters when provided
1398 * in a %macro definition line.
1400 static void count_mmac_params(Token * t, int *nparam, Token *** params)
1402 int paramsize, brace;
1404 *nparam = paramsize = 0;
1405 *params = NULL;
1406 while (t) {
1407 if (*nparam >= paramsize) {
1408 paramsize += PARAM_DELTA;
1409 *params = nasm_realloc(*params, sizeof(**params) * paramsize);
1411 skip_white_(t);
1412 brace = false;
1413 if (tok_is_(t, "{"))
1414 brace = true;
1415 (*params)[(*nparam)++] = t;
1416 while (tok_isnt_(t, brace ? "}" : ","))
1417 t = t->next;
1418 if (t) { /* got a comma/brace */
1419 t = t->next;
1420 if (brace) {
1422 * Now we've found the closing brace, look further
1423 * for the comma.
1425 skip_white_(t);
1426 if (tok_isnt_(t, ",")) {
1427 error(ERR_NONFATAL,
1428 "braces do not enclose all of macro parameter");
1429 while (tok_isnt_(t, ","))
1430 t = t->next;
1432 if (t)
1433 t = t->next; /* eat the comma */
1440 * Determine whether one of the various `if' conditions is true or
1441 * not.
1443 * We must free the tline we get passed.
1445 static bool if_condition(Token * tline, enum preproc_token ct)
1447 enum pp_conditional i = PP_COND(ct);
1448 bool j;
1449 Token *t, *tt, **tptr, *origline;
1450 struct tokenval tokval;
1451 expr *evalresult;
1452 enum pp_token_type needtype;
1454 origline = tline;
1456 switch (i) {
1457 case PPC_IFCTX:
1458 j = false; /* have we matched yet? */
1459 while (cstk && tline) {
1460 skip_white_(tline);
1461 if (!tline || tline->type != TOK_ID) {
1462 error(ERR_NONFATAL,
1463 "`%s' expects context identifiers", pp_directives[ct]);
1464 free_tlist(origline);
1465 return -1;
1467 if (!nasm_stricmp(tline->text, cstk->name))
1468 j = true;
1469 tline = tline->next;
1471 break;
1473 case PPC_IFDEF:
1474 j = false; /* have we matched yet? */
1475 while (tline) {
1476 skip_white_(tline);
1477 if (!tline || (tline->type != TOK_ID &&
1478 (tline->type != TOK_PREPROC_ID ||
1479 tline->text[1] != '$'))) {
1480 error(ERR_NONFATAL,
1481 "`%s' expects macro identifiers", pp_directives[ct]);
1482 goto fail;
1484 if (smacro_defined(NULL, tline->text, 0, NULL, true))
1485 j = true;
1486 tline = tline->next;
1488 break;
1490 case PPC_IFIDN:
1491 case PPC_IFIDNI:
1492 tline = expand_smacro(tline);
1493 t = tt = tline;
1494 while (tok_isnt_(tt, ","))
1495 tt = tt->next;
1496 if (!tt) {
1497 error(ERR_NONFATAL,
1498 "`%s' expects two comma-separated arguments",
1499 pp_directives[ct]);
1500 goto fail;
1502 tt = tt->next;
1503 j = true; /* assume equality unless proved not */
1504 while ((t->type != TOK_OTHER || strcmp(t->text, ",")) && tt) {
1505 if (tt->type == TOK_OTHER && !strcmp(tt->text, ",")) {
1506 error(ERR_NONFATAL, "`%s': more than one comma on line",
1507 pp_directives[ct]);
1508 goto fail;
1510 if (t->type == TOK_WHITESPACE) {
1511 t = t->next;
1512 continue;
1514 if (tt->type == TOK_WHITESPACE) {
1515 tt = tt->next;
1516 continue;
1518 if (tt->type != t->type) {
1519 j = false; /* found mismatching tokens */
1520 break;
1522 /* When comparing strings, need to unquote them first */
1523 if (t->type == TOK_STRING) {
1524 size_t l1 = nasm_unquote(t->text, NULL);
1525 size_t l2 = nasm_unquote(tt->text, NULL);
1527 if (l1 != l2) {
1528 j = false;
1529 break;
1531 if (mmemcmp(t->text, tt->text, l1, i == PPC_IFIDN)) {
1532 j = false;
1533 break;
1535 } else if (mstrcmp(tt->text, t->text, i == PPC_IFIDN) != 0) {
1536 j = false; /* found mismatching tokens */
1537 break;
1540 t = t->next;
1541 tt = tt->next;
1543 if ((t->type != TOK_OTHER || strcmp(t->text, ",")) || tt)
1544 j = false; /* trailing gunk on one end or other */
1545 break;
1547 case PPC_IFMACRO:
1549 bool found = false;
1550 MMacro searching, *mmac;
1552 tline = tline->next;
1553 skip_white_(tline);
1554 tline = expand_id(tline);
1555 if (!tok_type_(tline, TOK_ID)) {
1556 error(ERR_NONFATAL,
1557 "`%s' expects a macro name", pp_directives[ct]);
1558 goto fail;
1560 searching.name = nasm_strdup(tline->text);
1561 searching.casesense = true;
1562 searching.plus = false;
1563 searching.nolist = false;
1564 searching.in_progress = 0;
1565 searching.rep_nest = NULL;
1566 searching.nparam_min = 0;
1567 searching.nparam_max = INT_MAX;
1568 tline = expand_smacro(tline->next);
1569 skip_white_(tline);
1570 if (!tline) {
1571 } else if (!tok_type_(tline, TOK_NUMBER)) {
1572 error(ERR_NONFATAL,
1573 "`%s' expects a parameter count or nothing",
1574 pp_directives[ct]);
1575 } else {
1576 searching.nparam_min = searching.nparam_max =
1577 readnum(tline->text, &j);
1578 if (j)
1579 error(ERR_NONFATAL,
1580 "unable to parse parameter count `%s'",
1581 tline->text);
1583 if (tline && tok_is_(tline->next, "-")) {
1584 tline = tline->next->next;
1585 if (tok_is_(tline, "*"))
1586 searching.nparam_max = INT_MAX;
1587 else if (!tok_type_(tline, TOK_NUMBER))
1588 error(ERR_NONFATAL,
1589 "`%s' expects a parameter count after `-'",
1590 pp_directives[ct]);
1591 else {
1592 searching.nparam_max = readnum(tline->text, &j);
1593 if (j)
1594 error(ERR_NONFATAL,
1595 "unable to parse parameter count `%s'",
1596 tline->text);
1597 if (searching.nparam_min > searching.nparam_max)
1598 error(ERR_NONFATAL,
1599 "minimum parameter count exceeds maximum");
1602 if (tline && tok_is_(tline->next, "+")) {
1603 tline = tline->next;
1604 searching.plus = true;
1606 mmac = (MMacro *) hash_findix(&mmacros, searching.name);
1607 while (mmac) {
1608 if (!strcmp(mmac->name, searching.name) &&
1609 (mmac->nparam_min <= searching.nparam_max
1610 || searching.plus)
1611 && (searching.nparam_min <= mmac->nparam_max
1612 || mmac->plus)) {
1613 found = true;
1614 break;
1616 mmac = mmac->next;
1618 nasm_free(searching.name);
1619 j = found;
1620 break;
1623 case PPC_IFID:
1624 needtype = TOK_ID;
1625 goto iftype;
1626 case PPC_IFNUM:
1627 needtype = TOK_NUMBER;
1628 goto iftype;
1629 case PPC_IFSTR:
1630 needtype = TOK_STRING;
1631 goto iftype;
1633 iftype:
1634 t = tline = expand_smacro(tline);
1636 while (tok_type_(t, TOK_WHITESPACE) ||
1637 (needtype == TOK_NUMBER &&
1638 tok_type_(t, TOK_OTHER) &&
1639 (t->text[0] == '-' || t->text[0] == '+') &&
1640 !t->text[1]))
1641 t = t->next;
1643 j = tok_type_(t, needtype);
1644 break;
1646 case PPC_IFTOKEN:
1647 t = tline = expand_smacro(tline);
1648 while (tok_type_(t, TOK_WHITESPACE))
1649 t = t->next;
1651 j = false;
1652 if (t) {
1653 t = t->next; /* Skip the actual token */
1654 while (tok_type_(t, TOK_WHITESPACE))
1655 t = t->next;
1656 j = !t; /* Should be nothing left */
1658 break;
1660 case PPC_IFEMPTY:
1661 t = tline = expand_smacro(tline);
1662 while (tok_type_(t, TOK_WHITESPACE))
1663 t = t->next;
1665 j = !t; /* Should be empty */
1666 break;
1668 case PPC_IF:
1669 t = tline = expand_smacro(tline);
1670 tptr = &t;
1671 tokval.t_type = TOKEN_INVALID;
1672 evalresult = evaluate(ppscan, tptr, &tokval,
1673 NULL, pass | CRITICAL, error, NULL);
1674 if (!evalresult)
1675 return -1;
1676 if (tokval.t_type)
1677 error(ERR_WARNING,
1678 "trailing garbage after expression ignored");
1679 if (!is_simple(evalresult)) {
1680 error(ERR_NONFATAL,
1681 "non-constant value given to `%s'", pp_directives[ct]);
1682 goto fail;
1684 j = reloc_value(evalresult) != 0;
1685 return j;
1687 default:
1688 error(ERR_FATAL,
1689 "preprocessor directive `%s' not yet implemented",
1690 pp_directives[ct]);
1691 goto fail;
1694 free_tlist(origline);
1695 return j ^ PP_NEGATIVE(ct);
1697 fail:
1698 free_tlist(origline);
1699 return -1;
1703 * Common code for defining an smacro
1705 static bool define_smacro(Context *ctx, char *mname, bool casesense,
1706 int nparam, Token *expansion)
1708 SMacro *smac, **smhead;
1709 struct hash_table *smtbl;
1711 if (smacro_defined(ctx, mname, nparam, &smac, casesense)) {
1712 if (!smac) {
1713 error(ERR_WARNING,
1714 "single-line macro `%s' defined both with and"
1715 " without parameters", mname);
1717 /* Some instances of the old code considered this a failure,
1718 some others didn't. What is the right thing to do here? */
1719 free_tlist(expansion);
1720 return false; /* Failure */
1721 } else {
1723 * We're redefining, so we have to take over an
1724 * existing SMacro structure. This means freeing
1725 * what was already in it.
1727 nasm_free(smac->name);
1728 free_tlist(smac->expansion);
1730 } else {
1731 smtbl = ctx ? &ctx->localmac : &smacros;
1732 smhead = (SMacro **) hash_findi_add(smtbl, mname);
1733 smac = nasm_malloc(sizeof(SMacro));
1734 smac->next = *smhead;
1735 *smhead = smac;
1737 smac->name = nasm_strdup(mname);
1738 smac->casesense = casesense;
1739 smac->nparam = nparam;
1740 smac->expansion = expansion;
1741 smac->in_progress = false;
1742 return true; /* Success */
1746 * Undefine an smacro
1748 static void undef_smacro(Context *ctx, const char *mname)
1750 SMacro **smhead, *s, **sp;
1751 struct hash_table *smtbl;
1753 smtbl = ctx ? &ctx->localmac : &smacros;
1754 smhead = (SMacro **)hash_findi(smtbl, mname, NULL);
1756 if (smhead) {
1758 * We now have a macro name... go hunt for it.
1760 sp = smhead;
1761 while ((s = *sp) != NULL) {
1762 if (!mstrcmp(s->name, mname, s->casesense)) {
1763 *sp = s->next;
1764 nasm_free(s->name);
1765 free_tlist(s->expansion);
1766 nasm_free(s);
1767 } else {
1768 sp = &s->next;
1775 * Decode a size directive
1777 static int parse_size(const char *str) {
1778 static const char *size_names[] =
1779 { "byte", "dword", "oword", "qword", "tword", "word", "yword" };
1780 static const int sizes[] =
1781 { 0, 1, 4, 16, 8, 10, 2, 32 };
1783 return sizes[bsii(str, size_names, elements(size_names))+1];
1787 * find and process preprocessor directive in passed line
1788 * Find out if a line contains a preprocessor directive, and deal
1789 * with it if so.
1791 * If a directive _is_ found, it is the responsibility of this routine
1792 * (and not the caller) to free_tlist() the line.
1794 * @param tline a pointer to the current tokeninzed line linked list
1795 * @return DIRECTIVE_FOUND or NO_DIRECTIVE_FOUND
1798 static int do_directive(Token * tline)
1800 enum preproc_token i;
1801 int j;
1802 bool err;
1803 int nparam;
1804 bool nolist;
1805 bool casesense;
1806 int k, m;
1807 int offset;
1808 char *p, *mname;
1809 Include *inc;
1810 Context *ctx;
1811 Cond *cond;
1812 MMacro *mmac, **mmhead;
1813 Token *t, *tt, *param_start, *macro_start, *last, **tptr, *origline;
1814 Line *l;
1815 struct tokenval tokval;
1816 expr *evalresult;
1817 MMacro *tmp_defining; /* Used when manipulating rep_nest */
1818 int64_t count;
1820 origline = tline;
1822 skip_white_(tline);
1823 if (!tline || !tok_type_(tline, TOK_PREPROC_ID) ||
1824 (tline->text[1] == '%' || tline->text[1] == '$'
1825 || tline->text[1] == '!'))
1826 return NO_DIRECTIVE_FOUND;
1828 i = pp_token_hash(tline->text);
1831 * If we're in a non-emitting branch of a condition construct,
1832 * or walking to the end of an already terminated %rep block,
1833 * we should ignore all directives except for condition
1834 * directives.
1836 if (((istk->conds && !emitting(istk->conds->state)) ||
1837 (istk->mstk && !istk->mstk->in_progress)) && !is_condition(i)) {
1838 return NO_DIRECTIVE_FOUND;
1842 * If we're defining a macro or reading a %rep block, we should
1843 * ignore all directives except for %macro/%imacro (which
1844 * generate an error), %endm/%endmacro, and (only if we're in a
1845 * %rep block) %endrep. If we're in a %rep block, another %rep
1846 * causes an error, so should be let through.
1848 if (defining && i != PP_MACRO && i != PP_IMACRO &&
1849 i != PP_ENDMACRO && i != PP_ENDM &&
1850 (defining->name || (i != PP_ENDREP && i != PP_REP))) {
1851 return NO_DIRECTIVE_FOUND;
1854 switch (i) {
1855 case PP_INVALID:
1856 error(ERR_NONFATAL, "unknown preprocessor directive `%s'",
1857 tline->text);
1858 return NO_DIRECTIVE_FOUND; /* didn't get it */
1860 case PP_STACKSIZE:
1861 /* Directive to tell NASM what the default stack size is. The
1862 * default is for a 16-bit stack, and this can be overriden with
1863 * %stacksize large.
1864 * the following form:
1866 * ARG arg1:WORD, arg2:DWORD, arg4:QWORD
1868 tline = tline->next;
1869 if (tline && tline->type == TOK_WHITESPACE)
1870 tline = tline->next;
1871 if (!tline || tline->type != TOK_ID) {
1872 error(ERR_NONFATAL, "`%%stacksize' missing size parameter");
1873 free_tlist(origline);
1874 return DIRECTIVE_FOUND;
1876 if (nasm_stricmp(tline->text, "flat") == 0) {
1877 /* All subsequent ARG directives are for a 32-bit stack */
1878 StackSize = 4;
1879 StackPointer = "ebp";
1880 ArgOffset = 8;
1881 LocalOffset = 0;
1882 } else if (nasm_stricmp(tline->text, "flat64") == 0) {
1883 /* All subsequent ARG directives are for a 64-bit stack */
1884 StackSize = 8;
1885 StackPointer = "rbp";
1886 ArgOffset = 8;
1887 LocalOffset = 0;
1888 } else if (nasm_stricmp(tline->text, "large") == 0) {
1889 /* All subsequent ARG directives are for a 16-bit stack,
1890 * far function call.
1892 StackSize = 2;
1893 StackPointer = "bp";
1894 ArgOffset = 4;
1895 LocalOffset = 0;
1896 } else if (nasm_stricmp(tline->text, "small") == 0) {
1897 /* All subsequent ARG directives are for a 16-bit stack,
1898 * far function call. We don't support near functions.
1900 StackSize = 2;
1901 StackPointer = "bp";
1902 ArgOffset = 6;
1903 LocalOffset = 0;
1904 } else {
1905 error(ERR_NONFATAL, "`%%stacksize' invalid size type");
1906 free_tlist(origline);
1907 return DIRECTIVE_FOUND;
1909 free_tlist(origline);
1910 return DIRECTIVE_FOUND;
1912 case PP_ARG:
1913 /* TASM like ARG directive to define arguments to functions, in
1914 * the following form:
1916 * ARG arg1:WORD, arg2:DWORD, arg4:QWORD
1918 offset = ArgOffset;
1919 do {
1920 char *arg, directive[256];
1921 int size = StackSize;
1923 /* Find the argument name */
1924 tline = tline->next;
1925 if (tline && tline->type == TOK_WHITESPACE)
1926 tline = tline->next;
1927 if (!tline || tline->type != TOK_ID) {
1928 error(ERR_NONFATAL, "`%%arg' missing argument parameter");
1929 free_tlist(origline);
1930 return DIRECTIVE_FOUND;
1932 arg = tline->text;
1934 /* Find the argument size type */
1935 tline = tline->next;
1936 if (!tline || tline->type != TOK_OTHER
1937 || tline->text[0] != ':') {
1938 error(ERR_NONFATAL,
1939 "Syntax error processing `%%arg' directive");
1940 free_tlist(origline);
1941 return DIRECTIVE_FOUND;
1943 tline = tline->next;
1944 if (!tline || tline->type != TOK_ID) {
1945 error(ERR_NONFATAL, "`%%arg' missing size type parameter");
1946 free_tlist(origline);
1947 return DIRECTIVE_FOUND;
1950 /* Allow macro expansion of type parameter */
1951 tt = tokenize(tline->text);
1952 tt = expand_smacro(tt);
1953 size = parse_size(tt->text);
1954 if (!size) {
1955 error(ERR_NONFATAL,
1956 "Invalid size type for `%%arg' missing directive");
1957 free_tlist(tt);
1958 free_tlist(origline);
1959 return DIRECTIVE_FOUND;
1961 free_tlist(tt);
1963 /* Round up to even stack slots */
1964 size = (size+StackSize-1) & ~(StackSize-1);
1966 /* Now define the macro for the argument */
1967 snprintf(directive, sizeof(directive), "%%define %s (%s+%d)",
1968 arg, StackPointer, offset);
1969 do_directive(tokenize(directive));
1970 offset += size;
1972 /* Move to the next argument in the list */
1973 tline = tline->next;
1974 if (tline && tline->type == TOK_WHITESPACE)
1975 tline = tline->next;
1976 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
1977 ArgOffset = offset;
1978 free_tlist(origline);
1979 return DIRECTIVE_FOUND;
1981 case PP_LOCAL:
1982 /* TASM like LOCAL directive to define local variables for a
1983 * function, in the following form:
1985 * LOCAL local1:WORD, local2:DWORD, local4:QWORD = LocalSize
1987 * The '= LocalSize' at the end is ignored by NASM, but is
1988 * required by TASM to define the local parameter size (and used
1989 * by the TASM macro package).
1991 offset = LocalOffset;
1992 do {
1993 char *local, directive[256];
1994 int size = StackSize;
1996 /* Find the argument name */
1997 tline = tline->next;
1998 if (tline && tline->type == TOK_WHITESPACE)
1999 tline = tline->next;
2000 if (!tline || tline->type != TOK_ID) {
2001 error(ERR_NONFATAL,
2002 "`%%local' missing argument parameter");
2003 free_tlist(origline);
2004 return DIRECTIVE_FOUND;
2006 local = tline->text;
2008 /* Find the argument size type */
2009 tline = tline->next;
2010 if (!tline || tline->type != TOK_OTHER
2011 || tline->text[0] != ':') {
2012 error(ERR_NONFATAL,
2013 "Syntax error processing `%%local' directive");
2014 free_tlist(origline);
2015 return DIRECTIVE_FOUND;
2017 tline = tline->next;
2018 if (!tline || tline->type != TOK_ID) {
2019 error(ERR_NONFATAL,
2020 "`%%local' missing size type parameter");
2021 free_tlist(origline);
2022 return DIRECTIVE_FOUND;
2025 /* Allow macro expansion of type parameter */
2026 tt = tokenize(tline->text);
2027 tt = expand_smacro(tt);
2028 size = parse_size(tt->text);
2029 if (!size) {
2030 error(ERR_NONFATAL,
2031 "Invalid size type for `%%local' missing directive");
2032 free_tlist(tt);
2033 free_tlist(origline);
2034 return DIRECTIVE_FOUND;
2036 free_tlist(tt);
2038 /* Round up to even stack slots */
2039 size = (size+StackSize-1) & ~(StackSize-1);
2041 offset += size; /* Negative offset, increment before */
2043 /* Now define the macro for the argument */
2044 snprintf(directive, sizeof(directive), "%%define %s (%s-%d)",
2045 local, StackPointer, offset);
2046 do_directive(tokenize(directive));
2048 /* Now define the assign to setup the enter_c macro correctly */
2049 snprintf(directive, sizeof(directive),
2050 "%%assign %%$localsize %%$localsize+%d", size);
2051 do_directive(tokenize(directive));
2053 /* Move to the next argument in the list */
2054 tline = tline->next;
2055 if (tline && tline->type == TOK_WHITESPACE)
2056 tline = tline->next;
2057 } while (tline && tline->type == TOK_OTHER && tline->text[0] == ',');
2058 LocalOffset = offset;
2059 free_tlist(origline);
2060 return DIRECTIVE_FOUND;
2062 case PP_CLEAR:
2063 if (tline->next)
2064 error(ERR_WARNING, "trailing garbage after `%%clear' ignored");
2065 free_macros();
2066 init_macros();
2067 free_tlist(origline);
2068 return DIRECTIVE_FOUND;
2070 case PP_DEPEND:
2071 t = tline->next = expand_smacro(tline->next);
2072 skip_white_(t);
2073 if (!t || (t->type != TOK_STRING &&
2074 t->type != TOK_INTERNAL_STRING)) {
2075 error(ERR_NONFATAL, "`%%depend' expects a file name");
2076 free_tlist(origline);
2077 return DIRECTIVE_FOUND; /* but we did _something_ */
2079 if (t->next)
2080 error(ERR_WARNING,
2081 "trailing garbage after `%%depend' ignored");
2082 p = t->text;
2083 if (t->type != TOK_INTERNAL_STRING)
2084 nasm_unquote(p, NULL);
2085 if (dephead && !in_list(*dephead, p)) {
2086 StrList *sl = nasm_malloc(strlen(p)+1+sizeof sl->next);
2087 sl->next = NULL;
2088 strcpy(sl->str, p);
2089 *deptail = sl;
2090 deptail = &sl->next;
2092 free_tlist(origline);
2093 return DIRECTIVE_FOUND;
2095 case PP_INCLUDE:
2096 t = tline->next = expand_smacro(tline->next);
2097 skip_white_(t);
2099 if (!t || (t->type != TOK_STRING &&
2100 t->type != TOK_INTERNAL_STRING)) {
2101 error(ERR_NONFATAL, "`%%include' expects a file name");
2102 free_tlist(origline);
2103 return DIRECTIVE_FOUND; /* but we did _something_ */
2105 if (t->next)
2106 error(ERR_WARNING,
2107 "trailing garbage after `%%include' ignored");
2108 p = t->text;
2109 if (t->type != TOK_INTERNAL_STRING)
2110 nasm_unquote(p, NULL);
2111 inc = nasm_malloc(sizeof(Include));
2112 inc->next = istk;
2113 inc->conds = NULL;
2114 inc->fp = inc_fopen(p, dephead, &deptail, pass == 0);
2115 if (!inc->fp) {
2116 /* -MG given but file not found */
2117 nasm_free(inc);
2118 } else {
2119 inc->fname = src_set_fname(nasm_strdup(p));
2120 inc->lineno = src_set_linnum(0);
2121 inc->lineinc = 1;
2122 inc->expansion = NULL;
2123 inc->mstk = NULL;
2124 istk = inc;
2125 list->uplevel(LIST_INCLUDE);
2127 free_tlist(origline);
2128 return DIRECTIVE_FOUND;
2130 case PP_PUSH:
2131 tline = tline->next;
2132 skip_white_(tline);
2133 tline = expand_id(tline);
2134 if (!tok_type_(tline, TOK_ID)) {
2135 error(ERR_NONFATAL, "`%%push' expects a context identifier");
2136 free_tlist(origline);
2137 return DIRECTIVE_FOUND; /* but we did _something_ */
2139 if (tline->next)
2140 error(ERR_WARNING, "trailing garbage after `%%push' ignored");
2141 ctx = nasm_malloc(sizeof(Context));
2142 ctx->next = cstk;
2143 hash_init(&ctx->localmac, HASH_SMALL);
2144 ctx->name = nasm_strdup(tline->text);
2145 ctx->number = unique++;
2146 cstk = ctx;
2147 free_tlist(origline);
2148 break;
2150 case PP_REPL:
2151 tline = tline->next;
2152 skip_white_(tline);
2153 tline = expand_id(tline);
2154 if (!tok_type_(tline, TOK_ID)) {
2155 error(ERR_NONFATAL, "`%%repl' expects a context identifier");
2156 free_tlist(origline);
2157 return DIRECTIVE_FOUND; /* but we did _something_ */
2159 if (tline->next)
2160 error(ERR_WARNING, "trailing garbage after `%%repl' ignored");
2161 if (!cstk)
2162 error(ERR_NONFATAL, "`%%repl': context stack is empty");
2163 else {
2164 nasm_free(cstk->name);
2165 cstk->name = nasm_strdup(tline->text);
2167 free_tlist(origline);
2168 break;
2170 case PP_POP:
2171 if (tline->next)
2172 error(ERR_WARNING, "trailing garbage after `%%pop' ignored");
2173 if (!cstk)
2174 error(ERR_NONFATAL, "`%%pop': context stack is already empty");
2175 else
2176 ctx_pop();
2177 free_tlist(origline);
2178 break;
2180 case PP_ERROR:
2181 case PP_WARNING:
2183 int severity = PP_ERROR ? ERR_NONFATAL|ERR_NO_SEVERITY :
2184 ERR_WARNING|ERR_NO_SEVERITY;
2186 tline->next = expand_smacro(tline->next);
2187 tline = tline->next;
2188 skip_white_(tline);
2189 t = tline ? tline->next : NULL;
2190 skip_white_(t);
2191 if (tok_type_(tline, TOK_STRING) && !t) {
2192 /* The line contains only a quoted string */
2193 p = tline->text;
2194 nasm_unquote(p, NULL);
2195 error(severity, "%s: %s", pp_directives[i], p);
2196 } else {
2197 /* Not a quoted string, or more than a quoted string */
2198 p = detoken(tline, false);
2199 error(severity, "%s: %s", pp_directives[i], p);
2200 nasm_free(p);
2202 free_tlist(origline);
2203 break;
2206 CASE_PP_IF:
2207 if (istk->conds && !emitting(istk->conds->state))
2208 j = COND_NEVER;
2209 else {
2210 j = if_condition(tline->next, i);
2211 tline->next = NULL; /* it got freed */
2212 j = j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2214 cond = nasm_malloc(sizeof(Cond));
2215 cond->next = istk->conds;
2216 cond->state = j;
2217 istk->conds = cond;
2218 free_tlist(origline);
2219 return DIRECTIVE_FOUND;
2221 CASE_PP_ELIF:
2222 if (!istk->conds)
2223 error(ERR_FATAL, "`%s': no matching `%%if'", pp_directives[i]);
2224 if (emitting(istk->conds->state)
2225 || istk->conds->state == COND_NEVER)
2226 istk->conds->state = COND_NEVER;
2227 else {
2229 * IMPORTANT: In the case of %if, we will already have
2230 * called expand_mmac_params(); however, if we're
2231 * processing an %elif we must have been in a
2232 * non-emitting mode, which would have inhibited
2233 * the normal invocation of expand_mmac_params(). Therefore,
2234 * we have to do it explicitly here.
2236 j = if_condition(expand_mmac_params(tline->next), i);
2237 tline->next = NULL; /* it got freed */
2238 istk->conds->state =
2239 j < 0 ? COND_NEVER : j ? COND_IF_TRUE : COND_IF_FALSE;
2241 free_tlist(origline);
2242 return DIRECTIVE_FOUND;
2244 case PP_ELSE:
2245 if (tline->next)
2246 error(ERR_WARNING, "trailing garbage after `%%else' ignored");
2247 if (!istk->conds)
2248 error(ERR_FATAL, "`%%else': no matching `%%if'");
2249 if (emitting(istk->conds->state)
2250 || istk->conds->state == COND_NEVER)
2251 istk->conds->state = COND_ELSE_FALSE;
2252 else
2253 istk->conds->state = COND_ELSE_TRUE;
2254 free_tlist(origline);
2255 return DIRECTIVE_FOUND;
2257 case PP_ENDIF:
2258 if (tline->next)
2259 error(ERR_WARNING, "trailing garbage after `%%endif' ignored");
2260 if (!istk->conds)
2261 error(ERR_FATAL, "`%%endif': no matching `%%if'");
2262 cond = istk->conds;
2263 istk->conds = cond->next;
2264 nasm_free(cond);
2265 free_tlist(origline);
2266 return DIRECTIVE_FOUND;
2268 case PP_MACRO:
2269 case PP_IMACRO:
2270 if (defining)
2271 error(ERR_FATAL,
2272 "`%%%smacro': already defining a macro",
2273 (i == PP_IMACRO ? "i" : ""));
2274 tline = tline->next;
2275 skip_white_(tline);
2276 tline = expand_id(tline);
2277 if (!tok_type_(tline, TOK_ID)) {
2278 error(ERR_NONFATAL,
2279 "`%%%smacro' expects a macro name",
2280 (i == PP_IMACRO ? "i" : ""));
2281 return DIRECTIVE_FOUND;
2283 defining = nasm_malloc(sizeof(MMacro));
2284 defining->name = nasm_strdup(tline->text);
2285 defining->casesense = (i == PP_MACRO);
2286 defining->plus = false;
2287 defining->nolist = false;
2288 defining->in_progress = 0;
2289 defining->rep_nest = NULL;
2290 tline = expand_smacro(tline->next);
2291 skip_white_(tline);
2292 if (!tok_type_(tline, TOK_NUMBER)) {
2293 error(ERR_NONFATAL,
2294 "`%%%smacro' expects a parameter count",
2295 (i == PP_IMACRO ? "i" : ""));
2296 defining->nparam_min = defining->nparam_max = 0;
2297 } else {
2298 defining->nparam_min = defining->nparam_max =
2299 readnum(tline->text, &err);
2300 if (err)
2301 error(ERR_NONFATAL,
2302 "unable to parse parameter count `%s'", tline->text);
2304 if (tline && tok_is_(tline->next, "-")) {
2305 tline = tline->next->next;
2306 if (tok_is_(tline, "*"))
2307 defining->nparam_max = INT_MAX;
2308 else if (!tok_type_(tline, TOK_NUMBER))
2309 error(ERR_NONFATAL,
2310 "`%%%smacro' expects a parameter count after `-'",
2311 (i == PP_IMACRO ? "i" : ""));
2312 else {
2313 defining->nparam_max = readnum(tline->text, &err);
2314 if (err)
2315 error(ERR_NONFATAL,
2316 "unable to parse parameter count `%s'",
2317 tline->text);
2318 if (defining->nparam_min > defining->nparam_max)
2319 error(ERR_NONFATAL,
2320 "minimum parameter count exceeds maximum");
2323 if (tline && tok_is_(tline->next, "+")) {
2324 tline = tline->next;
2325 defining->plus = true;
2327 if (tline && tok_type_(tline->next, TOK_ID) &&
2328 !nasm_stricmp(tline->next->text, ".nolist")) {
2329 tline = tline->next;
2330 defining->nolist = true;
2332 mmac = (MMacro *) hash_findix(&mmacros, defining->name);
2333 while (mmac) {
2334 if (!strcmp(mmac->name, defining->name) &&
2335 (mmac->nparam_min <= defining->nparam_max
2336 || defining->plus)
2337 && (defining->nparam_min <= mmac->nparam_max
2338 || mmac->plus)) {
2339 error(ERR_WARNING,
2340 "redefining multi-line macro `%s'", defining->name);
2341 break;
2343 mmac = mmac->next;
2346 * Handle default parameters.
2348 if (tline && tline->next) {
2349 defining->dlist = tline->next;
2350 tline->next = NULL;
2351 count_mmac_params(defining->dlist, &defining->ndefs,
2352 &defining->defaults);
2353 } else {
2354 defining->dlist = NULL;
2355 defining->defaults = NULL;
2357 defining->expansion = NULL;
2358 free_tlist(origline);
2359 return DIRECTIVE_FOUND;
2361 case PP_ENDM:
2362 case PP_ENDMACRO:
2363 if (!defining) {
2364 error(ERR_NONFATAL, "`%s': not defining a macro", tline->text);
2365 return DIRECTIVE_FOUND;
2367 mmhead = (MMacro **) hash_findi_add(&mmacros, defining->name);
2368 defining->next = *mmhead;
2369 *mmhead = defining;
2370 defining = NULL;
2371 free_tlist(origline);
2372 return DIRECTIVE_FOUND;
2374 case PP_ROTATE:
2375 if (tline->next && tline->next->type == TOK_WHITESPACE)
2376 tline = tline->next;
2377 if (tline->next == NULL) {
2378 free_tlist(origline);
2379 error(ERR_NONFATAL, "`%%rotate' missing rotate count");
2380 return DIRECTIVE_FOUND;
2382 t = expand_smacro(tline->next);
2383 tline->next = NULL;
2384 free_tlist(origline);
2385 tline = t;
2386 tptr = &t;
2387 tokval.t_type = TOKEN_INVALID;
2388 evalresult =
2389 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
2390 free_tlist(tline);
2391 if (!evalresult)
2392 return DIRECTIVE_FOUND;
2393 if (tokval.t_type)
2394 error(ERR_WARNING,
2395 "trailing garbage after expression ignored");
2396 if (!is_simple(evalresult)) {
2397 error(ERR_NONFATAL, "non-constant value given to `%%rotate'");
2398 return DIRECTIVE_FOUND;
2400 mmac = istk->mstk;
2401 while (mmac && !mmac->name) /* avoid mistaking %reps for macros */
2402 mmac = mmac->next_active;
2403 if (!mmac) {
2404 error(ERR_NONFATAL, "`%%rotate' invoked outside a macro call");
2405 } else if (mmac->nparam == 0) {
2406 error(ERR_NONFATAL,
2407 "`%%rotate' invoked within macro without parameters");
2408 } else {
2409 int rotate = mmac->rotate + reloc_value(evalresult);
2411 rotate %= (int)mmac->nparam;
2412 if (rotate < 0)
2413 rotate += mmac->nparam;
2415 mmac->rotate = rotate;
2417 return DIRECTIVE_FOUND;
2419 case PP_REP:
2420 nolist = false;
2421 do {
2422 tline = tline->next;
2423 } while (tok_type_(tline, TOK_WHITESPACE));
2425 if (tok_type_(tline, TOK_ID) &&
2426 nasm_stricmp(tline->text, ".nolist") == 0) {
2427 nolist = true;
2428 do {
2429 tline = tline->next;
2430 } while (tok_type_(tline, TOK_WHITESPACE));
2433 if (tline) {
2434 t = expand_smacro(tline);
2435 tptr = &t;
2436 tokval.t_type = TOKEN_INVALID;
2437 evalresult =
2438 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
2439 if (!evalresult) {
2440 free_tlist(origline);
2441 return DIRECTIVE_FOUND;
2443 if (tokval.t_type)
2444 error(ERR_WARNING,
2445 "trailing garbage after expression ignored");
2446 if (!is_simple(evalresult)) {
2447 error(ERR_NONFATAL, "non-constant value given to `%%rep'");
2448 return DIRECTIVE_FOUND;
2450 count = reloc_value(evalresult) + 1;
2451 } else {
2452 error(ERR_NONFATAL, "`%%rep' expects a repeat count");
2453 count = 0;
2455 free_tlist(origline);
2457 tmp_defining = defining;
2458 defining = nasm_malloc(sizeof(MMacro));
2459 defining->name = NULL; /* flags this macro as a %rep block */
2460 defining->casesense = false;
2461 defining->plus = false;
2462 defining->nolist = nolist;
2463 defining->in_progress = count;
2464 defining->nparam_min = defining->nparam_max = 0;
2465 defining->defaults = NULL;
2466 defining->dlist = NULL;
2467 defining->expansion = NULL;
2468 defining->next_active = istk->mstk;
2469 defining->rep_nest = tmp_defining;
2470 return DIRECTIVE_FOUND;
2472 case PP_ENDREP:
2473 if (!defining || defining->name) {
2474 error(ERR_NONFATAL, "`%%endrep': no matching `%%rep'");
2475 return DIRECTIVE_FOUND;
2479 * Now we have a "macro" defined - although it has no name
2480 * and we won't be entering it in the hash tables - we must
2481 * push a macro-end marker for it on to istk->expansion.
2482 * After that, it will take care of propagating itself (a
2483 * macro-end marker line for a macro which is really a %rep
2484 * block will cause the macro to be re-expanded, complete
2485 * with another macro-end marker to ensure the process
2486 * continues) until the whole expansion is forcibly removed
2487 * from istk->expansion by a %exitrep.
2489 l = nasm_malloc(sizeof(Line));
2490 l->next = istk->expansion;
2491 l->finishes = defining;
2492 l->first = NULL;
2493 istk->expansion = l;
2495 istk->mstk = defining;
2497 list->uplevel(defining->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
2498 tmp_defining = defining;
2499 defining = defining->rep_nest;
2500 free_tlist(origline);
2501 return DIRECTIVE_FOUND;
2503 case PP_EXITREP:
2505 * We must search along istk->expansion until we hit a
2506 * macro-end marker for a macro with no name. Then we set
2507 * its `in_progress' flag to 0.
2509 for (l = istk->expansion; l; l = l->next)
2510 if (l->finishes && !l->finishes->name)
2511 break;
2513 if (l)
2514 l->finishes->in_progress = 0;
2515 else
2516 error(ERR_NONFATAL, "`%%exitrep' not within `%%rep' block");
2517 free_tlist(origline);
2518 return DIRECTIVE_FOUND;
2520 case PP_XDEFINE:
2521 case PP_IXDEFINE:
2522 case PP_DEFINE:
2523 case PP_IDEFINE:
2524 casesense = (i == PP_DEFINE || i == PP_XDEFINE);
2526 tline = tline->next;
2527 skip_white_(tline);
2528 tline = expand_id(tline);
2529 if (!tline || (tline->type != TOK_ID &&
2530 (tline->type != TOK_PREPROC_ID ||
2531 tline->text[1] != '$'))) {
2532 error(ERR_NONFATAL, "`%s' expects a macro identifier",
2533 pp_directives[i]);
2534 free_tlist(origline);
2535 return DIRECTIVE_FOUND;
2538 ctx = get_ctx(tline->text, false);
2540 mname = tline->text;
2541 last = tline;
2542 param_start = tline = tline->next;
2543 nparam = 0;
2545 /* Expand the macro definition now for %xdefine and %ixdefine */
2546 if ((i == PP_XDEFINE) || (i == PP_IXDEFINE))
2547 tline = expand_smacro(tline);
2549 if (tok_is_(tline, "(")) {
2551 * This macro has parameters.
2554 tline = tline->next;
2555 while (1) {
2556 skip_white_(tline);
2557 if (!tline) {
2558 error(ERR_NONFATAL, "parameter identifier expected");
2559 free_tlist(origline);
2560 return DIRECTIVE_FOUND;
2562 if (tline->type != TOK_ID) {
2563 error(ERR_NONFATAL,
2564 "`%s': parameter identifier expected",
2565 tline->text);
2566 free_tlist(origline);
2567 return DIRECTIVE_FOUND;
2569 tline->type = TOK_SMAC_PARAM + nparam++;
2570 tline = tline->next;
2571 skip_white_(tline);
2572 if (tok_is_(tline, ",")) {
2573 tline = tline->next;
2574 continue;
2576 if (!tok_is_(tline, ")")) {
2577 error(ERR_NONFATAL,
2578 "`)' expected to terminate macro template");
2579 free_tlist(origline);
2580 return DIRECTIVE_FOUND;
2582 break;
2584 last = tline;
2585 tline = tline->next;
2587 if (tok_type_(tline, TOK_WHITESPACE))
2588 last = tline, tline = tline->next;
2589 macro_start = NULL;
2590 last->next = NULL;
2591 t = tline;
2592 while (t) {
2593 if (t->type == TOK_ID) {
2594 for (tt = param_start; tt; tt = tt->next)
2595 if (tt->type >= TOK_SMAC_PARAM &&
2596 !strcmp(tt->text, t->text))
2597 t->type = tt->type;
2599 tt = t->next;
2600 t->next = macro_start;
2601 macro_start = t;
2602 t = tt;
2605 * Good. We now have a macro name, a parameter count, and a
2606 * token list (in reverse order) for an expansion. We ought
2607 * to be OK just to create an SMacro, store it, and let
2608 * free_tlist have the rest of the line (which we have
2609 * carefully re-terminated after chopping off the expansion
2610 * from the end).
2612 define_smacro(ctx, mname, casesense, nparam, macro_start);
2613 free_tlist(origline);
2614 return DIRECTIVE_FOUND;
2616 case PP_UNDEF:
2617 tline = tline->next;
2618 skip_white_(tline);
2619 tline = expand_id(tline);
2620 if (!tline || (tline->type != TOK_ID &&
2621 (tline->type != TOK_PREPROC_ID ||
2622 tline->text[1] != '$'))) {
2623 error(ERR_NONFATAL, "`%%undef' expects a macro identifier");
2624 free_tlist(origline);
2625 return DIRECTIVE_FOUND;
2627 if (tline->next) {
2628 error(ERR_WARNING,
2629 "trailing garbage after macro name ignored");
2632 /* Find the context that symbol belongs to */
2633 ctx = get_ctx(tline->text, false);
2634 undef_smacro(ctx, tline->text);
2635 free_tlist(origline);
2636 return DIRECTIVE_FOUND;
2638 case PP_DEFSTR:
2639 case PP_IDEFSTR:
2640 casesense = (i == PP_DEFSTR);
2642 tline = tline->next;
2643 skip_white_(tline);
2644 tline = expand_id(tline);
2645 if (!tline || (tline->type != TOK_ID &&
2646 (tline->type != TOK_PREPROC_ID ||
2647 tline->text[1] != '$'))) {
2648 error(ERR_NONFATAL, "`%s' expects a macro identifier",
2649 pp_directives[i]);
2650 free_tlist(origline);
2651 return DIRECTIVE_FOUND;
2654 ctx = get_ctx(tline->text, false);
2656 mname = tline->text;
2657 last = tline;
2658 tline = expand_smacro(tline->next);
2659 last->next = NULL;
2661 while (tok_type_(tline, TOK_WHITESPACE))
2662 tline = delete_Token(tline);
2664 p = detoken(tline, false);
2665 macro_start = nasm_malloc(sizeof(*macro_start));
2666 macro_start->next = NULL;
2667 macro_start->text = nasm_quote(p, strlen(p));
2668 macro_start->type = TOK_STRING;
2669 macro_start->mac = NULL;
2670 nasm_free(p);
2673 * We now have a macro name, an implicit parameter count of
2674 * zero, and a string token to use as an expansion. Create
2675 * and store an SMacro.
2677 define_smacro(ctx, mname, casesense, 0, macro_start);
2678 free_tlist(origline);
2679 return DIRECTIVE_FOUND;
2681 case PP_PATHSEARCH:
2683 FILE *fp;
2684 StrList *xsl = NULL;
2685 StrList **xst = &xsl;
2687 casesense = true;
2689 tline = tline->next;
2690 skip_white_(tline);
2691 tline = expand_id(tline);
2692 if (!tline || (tline->type != TOK_ID &&
2693 (tline->type != TOK_PREPROC_ID ||
2694 tline->text[1] != '$'))) {
2695 error(ERR_NONFATAL,
2696 "`%%pathsearch' expects a macro identifier as first parameter");
2697 free_tlist(origline);
2698 return DIRECTIVE_FOUND;
2700 ctx = get_ctx(tline->text, false);
2702 mname = tline->text;
2703 last = tline;
2704 tline = expand_smacro(tline->next);
2705 last->next = NULL;
2707 t = tline;
2708 while (tok_type_(t, TOK_WHITESPACE))
2709 t = t->next;
2711 if (!t || (t->type != TOK_STRING &&
2712 t->type != TOK_INTERNAL_STRING)) {
2713 error(ERR_NONFATAL, "`%%pathsearch' expects a file name");
2714 free_tlist(tline);
2715 free_tlist(origline);
2716 return DIRECTIVE_FOUND; /* but we did _something_ */
2718 if (t->next)
2719 error(ERR_WARNING,
2720 "trailing garbage after `%%pathsearch' ignored");
2721 p = t->text;
2722 if (t->type != TOK_INTERNAL_STRING)
2723 nasm_unquote(p, NULL);
2725 fp = inc_fopen(p, &xsl, &xst, true);
2726 if (fp) {
2727 p = xsl->str;
2728 fclose(fp); /* Don't actually care about the file */
2730 macro_start = nasm_malloc(sizeof(*macro_start));
2731 macro_start->next = NULL;
2732 macro_start->text = nasm_quote(p, strlen(p));
2733 macro_start->type = TOK_STRING;
2734 macro_start->mac = NULL;
2735 if (xsl)
2736 nasm_free(xsl);
2739 * We now have a macro name, an implicit parameter count of
2740 * zero, and a string token to use as an expansion. Create
2741 * and store an SMacro.
2743 define_smacro(ctx, mname, casesense, 0, macro_start);
2744 free_tlist(tline);
2745 free_tlist(origline);
2746 return DIRECTIVE_FOUND;
2749 case PP_STRLEN:
2750 casesense = true;
2752 tline = tline->next;
2753 skip_white_(tline);
2754 tline = expand_id(tline);
2755 if (!tline || (tline->type != TOK_ID &&
2756 (tline->type != TOK_PREPROC_ID ||
2757 tline->text[1] != '$'))) {
2758 error(ERR_NONFATAL,
2759 "`%%strlen' expects a macro identifier as first parameter");
2760 free_tlist(origline);
2761 return DIRECTIVE_FOUND;
2763 ctx = get_ctx(tline->text, false);
2765 mname = tline->text;
2766 last = tline;
2767 tline = expand_smacro(tline->next);
2768 last->next = NULL;
2770 t = tline;
2771 while (tok_type_(t, TOK_WHITESPACE))
2772 t = t->next;
2773 /* t should now point to the string */
2774 if (t->type != TOK_STRING) {
2775 error(ERR_NONFATAL,
2776 "`%%strlen` requires string as second parameter");
2777 free_tlist(tline);
2778 free_tlist(origline);
2779 return DIRECTIVE_FOUND;
2782 macro_start = nasm_malloc(sizeof(*macro_start));
2783 macro_start->next = NULL;
2784 make_tok_num(macro_start, nasm_unquote(t->text, NULL));
2785 macro_start->mac = NULL;
2788 * We now have a macro name, an implicit parameter count of
2789 * zero, and a numeric token to use as an expansion. Create
2790 * and store an SMacro.
2792 define_smacro(ctx, mname, casesense, 0, macro_start);
2793 free_tlist(tline);
2794 free_tlist(origline);
2795 return DIRECTIVE_FOUND;
2797 case PP_SUBSTR:
2799 int64_t a1, a2;
2800 size_t len;
2802 casesense = true;
2804 tline = tline->next;
2805 skip_white_(tline);
2806 tline = expand_id(tline);
2807 if (!tline || (tline->type != TOK_ID &&
2808 (tline->type != TOK_PREPROC_ID ||
2809 tline->text[1] != '$'))) {
2810 error(ERR_NONFATAL,
2811 "`%%substr' expects a macro identifier as first parameter");
2812 free_tlist(origline);
2813 return DIRECTIVE_FOUND;
2815 ctx = get_ctx(tline->text, false);
2817 mname = tline->text;
2818 last = tline;
2819 tline = expand_smacro(tline->next);
2820 last->next = NULL;
2822 t = tline->next;
2823 while (tok_type_(t, TOK_WHITESPACE))
2824 t = t->next;
2826 /* t should now point to the string */
2827 if (t->type != TOK_STRING) {
2828 error(ERR_NONFATAL,
2829 "`%%substr` requires string as second parameter");
2830 free_tlist(tline);
2831 free_tlist(origline);
2832 return DIRECTIVE_FOUND;
2835 tt = t->next;
2836 tptr = &tt;
2837 tokval.t_type = TOKEN_INVALID;
2838 evalresult = evaluate(ppscan, tptr, &tokval, NULL,
2839 pass, error, NULL);
2840 if (!evalresult) {
2841 free_tlist(tline);
2842 free_tlist(origline);
2843 return DIRECTIVE_FOUND;
2844 } else if (!is_simple(evalresult)) {
2845 error(ERR_NONFATAL, "non-constant value given to `%%substr`");
2846 free_tlist(tline);
2847 free_tlist(origline);
2848 return DIRECTIVE_FOUND;
2850 a1 = evalresult->value-1;
2852 while (tok_type_(tt, TOK_WHITESPACE))
2853 tt = tt->next;
2854 if (!tt) {
2855 a2 = 1; /* Backwards compatibility: one character */
2856 } else {
2857 tokval.t_type = TOKEN_INVALID;
2858 evalresult = evaluate(ppscan, tptr, &tokval, NULL,
2859 pass, error, NULL);
2860 if (!evalresult) {
2861 free_tlist(tline);
2862 free_tlist(origline);
2863 return DIRECTIVE_FOUND;
2864 } else if (!is_simple(evalresult)) {
2865 error(ERR_NONFATAL, "non-constant value given to `%%substr`");
2866 free_tlist(tline);
2867 free_tlist(origline);
2868 return DIRECTIVE_FOUND;
2870 a2 = evalresult->value;
2873 len = nasm_unquote(t->text, NULL);
2874 if (a2 < 0)
2875 a2 = a2+1+len-a1;
2876 if (a1+a2 > (int64_t)len)
2877 a2 = len-a1;
2879 macro_start = nasm_malloc(sizeof(*macro_start));
2880 macro_start->next = NULL;
2881 macro_start->text = nasm_quote((a1 < 0) ? "" : t->text+a1, a2);
2882 macro_start->type = TOK_STRING;
2883 macro_start->mac = NULL;
2886 * We now have a macro name, an implicit parameter count of
2887 * zero, and a numeric token to use as an expansion. Create
2888 * and store an SMacro.
2890 define_smacro(ctx, mname, casesense, 0, macro_start);
2891 free_tlist(tline);
2892 free_tlist(origline);
2893 return DIRECTIVE_FOUND;
2896 case PP_ASSIGN:
2897 case PP_IASSIGN:
2898 casesense = (i == PP_ASSIGN);
2900 tline = tline->next;
2901 skip_white_(tline);
2902 tline = expand_id(tline);
2903 if (!tline || (tline->type != TOK_ID &&
2904 (tline->type != TOK_PREPROC_ID ||
2905 tline->text[1] != '$'))) {
2906 error(ERR_NONFATAL,
2907 "`%%%sassign' expects a macro identifier",
2908 (i == PP_IASSIGN ? "i" : ""));
2909 free_tlist(origline);
2910 return DIRECTIVE_FOUND;
2912 ctx = get_ctx(tline->text, false);
2914 mname = tline->text;
2915 last = tline;
2916 tline = expand_smacro(tline->next);
2917 last->next = NULL;
2919 t = tline;
2920 tptr = &t;
2921 tokval.t_type = TOKEN_INVALID;
2922 evalresult =
2923 evaluate(ppscan, tptr, &tokval, NULL, pass, error, NULL);
2924 free_tlist(tline);
2925 if (!evalresult) {
2926 free_tlist(origline);
2927 return DIRECTIVE_FOUND;
2930 if (tokval.t_type)
2931 error(ERR_WARNING,
2932 "trailing garbage after expression ignored");
2934 if (!is_simple(evalresult)) {
2935 error(ERR_NONFATAL,
2936 "non-constant value given to `%%%sassign'",
2937 (i == PP_IASSIGN ? "i" : ""));
2938 free_tlist(origline);
2939 return DIRECTIVE_FOUND;
2942 macro_start = nasm_malloc(sizeof(*macro_start));
2943 macro_start->next = NULL;
2944 make_tok_num(macro_start, reloc_value(evalresult));
2945 macro_start->mac = NULL;
2948 * We now have a macro name, an implicit parameter count of
2949 * zero, and a numeric token to use as an expansion. Create
2950 * and store an SMacro.
2952 define_smacro(ctx, mname, casesense, 0, macro_start);
2953 free_tlist(origline);
2954 return DIRECTIVE_FOUND;
2956 case PP_LINE:
2958 * Syntax is `%line nnn[+mmm] [filename]'
2960 tline = tline->next;
2961 skip_white_(tline);
2962 if (!tok_type_(tline, TOK_NUMBER)) {
2963 error(ERR_NONFATAL, "`%%line' expects line number");
2964 free_tlist(origline);
2965 return DIRECTIVE_FOUND;
2967 k = readnum(tline->text, &err);
2968 m = 1;
2969 tline = tline->next;
2970 if (tok_is_(tline, "+")) {
2971 tline = tline->next;
2972 if (!tok_type_(tline, TOK_NUMBER)) {
2973 error(ERR_NONFATAL, "`%%line' expects line increment");
2974 free_tlist(origline);
2975 return DIRECTIVE_FOUND;
2977 m = readnum(tline->text, &err);
2978 tline = tline->next;
2980 skip_white_(tline);
2981 src_set_linnum(k);
2982 istk->lineinc = m;
2983 if (tline) {
2984 nasm_free(src_set_fname(detoken(tline, false)));
2986 free_tlist(origline);
2987 return DIRECTIVE_FOUND;
2989 default:
2990 error(ERR_FATAL,
2991 "preprocessor directive `%s' not yet implemented",
2992 pp_directives[i]);
2993 break;
2995 return DIRECTIVE_FOUND;
2999 * Ensure that a macro parameter contains a condition code and
3000 * nothing else. Return the condition code index if so, or -1
3001 * otherwise.
3003 static int find_cc(Token * t)
3005 Token *tt;
3006 int i, j, k, m;
3008 if (!t)
3009 return -1; /* Probably a %+ without a space */
3011 skip_white_(t);
3012 if (t->type != TOK_ID)
3013 return -1;
3014 tt = t->next;
3015 skip_white_(tt);
3016 if (tt && (tt->type != TOK_OTHER || strcmp(tt->text, ",")))
3017 return -1;
3019 i = -1;
3020 j = elements(conditions);
3021 while (j - i > 1) {
3022 k = (j + i) / 2;
3023 m = nasm_stricmp(t->text, conditions[k]);
3024 if (m == 0) {
3025 i = k;
3026 j = -2;
3027 break;
3028 } else if (m < 0) {
3029 j = k;
3030 } else
3031 i = k;
3033 if (j != -2)
3034 return -1;
3035 return i;
3039 * Expand MMacro-local things: parameter references (%0, %n, %+n,
3040 * %-n) and MMacro-local identifiers (%%foo).
3042 static Token *expand_mmac_params(Token * tline)
3044 Token *t, *tt, **tail, *thead;
3046 tail = &thead;
3047 thead = NULL;
3049 while (tline) {
3050 if (tline->type == TOK_PREPROC_ID &&
3051 (((tline->text[1] == '+' || tline->text[1] == '-')
3052 && tline->text[2]) || tline->text[1] == '%'
3053 || (tline->text[1] >= '0' && tline->text[1] <= '9'))) {
3054 char *text = NULL;
3055 int type = 0, cc; /* type = 0 to placate optimisers */
3056 char tmpbuf[30];
3057 unsigned int n;
3058 int i;
3059 MMacro *mac;
3061 t = tline;
3062 tline = tline->next;
3064 mac = istk->mstk;
3065 while (mac && !mac->name) /* avoid mistaking %reps for macros */
3066 mac = mac->next_active;
3067 if (!mac)
3068 error(ERR_NONFATAL, "`%s': not in a macro call", t->text);
3069 else
3070 switch (t->text[1]) {
3072 * We have to make a substitution of one of the
3073 * forms %1, %-1, %+1, %%foo, %0.
3075 case '0':
3076 type = TOK_NUMBER;
3077 snprintf(tmpbuf, sizeof(tmpbuf), "%d", mac->nparam);
3078 text = nasm_strdup(tmpbuf);
3079 break;
3080 case '%':
3081 type = TOK_ID;
3082 snprintf(tmpbuf, sizeof(tmpbuf), "..@%"PRIu64".",
3083 mac->unique);
3084 text = nasm_strcat(tmpbuf, t->text + 2);
3085 break;
3086 case '-':
3087 n = atoi(t->text + 2) - 1;
3088 if (n >= mac->nparam)
3089 tt = NULL;
3090 else {
3091 if (mac->nparam > 1)
3092 n = (n + mac->rotate) % mac->nparam;
3093 tt = mac->params[n];
3095 cc = find_cc(tt);
3096 if (cc == -1) {
3097 error(ERR_NONFATAL,
3098 "macro parameter %d is not a condition code",
3099 n + 1);
3100 text = NULL;
3101 } else {
3102 type = TOK_ID;
3103 if (inverse_ccs[cc] == -1) {
3104 error(ERR_NONFATAL,
3105 "condition code `%s' is not invertible",
3106 conditions[cc]);
3107 text = NULL;
3108 } else
3109 text =
3110 nasm_strdup(conditions[inverse_ccs[cc]]);
3112 break;
3113 case '+':
3114 n = atoi(t->text + 2) - 1;
3115 if (n >= mac->nparam)
3116 tt = NULL;
3117 else {
3118 if (mac->nparam > 1)
3119 n = (n + mac->rotate) % mac->nparam;
3120 tt = mac->params[n];
3122 cc = find_cc(tt);
3123 if (cc == -1) {
3124 error(ERR_NONFATAL,
3125 "macro parameter %d is not a condition code",
3126 n + 1);
3127 text = NULL;
3128 } else {
3129 type = TOK_ID;
3130 text = nasm_strdup(conditions[cc]);
3132 break;
3133 default:
3134 n = atoi(t->text + 1) - 1;
3135 if (n >= mac->nparam)
3136 tt = NULL;
3137 else {
3138 if (mac->nparam > 1)
3139 n = (n + mac->rotate) % mac->nparam;
3140 tt = mac->params[n];
3142 if (tt) {
3143 for (i = 0; i < mac->paramlen[n]; i++) {
3144 *tail = new_Token(NULL, tt->type, tt->text, 0);
3145 tail = &(*tail)->next;
3146 tt = tt->next;
3149 text = NULL; /* we've done it here */
3150 break;
3152 if (!text) {
3153 delete_Token(t);
3154 } else {
3155 *tail = t;
3156 tail = &t->next;
3157 t->type = type;
3158 nasm_free(t->text);
3159 t->text = text;
3160 t->mac = NULL;
3162 continue;
3163 } else {
3164 t = *tail = tline;
3165 tline = tline->next;
3166 t->mac = NULL;
3167 tail = &t->next;
3170 *tail = NULL;
3171 t = thead;
3172 for (; t && (tt = t->next) != NULL; t = t->next)
3173 switch (t->type) {
3174 case TOK_WHITESPACE:
3175 if (tt->type == TOK_WHITESPACE) {
3176 t->next = delete_Token(tt);
3178 break;
3179 case TOK_ID:
3180 if (tt->type == TOK_ID || tt->type == TOK_NUMBER) {
3181 char *tmp = nasm_strcat(t->text, tt->text);
3182 nasm_free(t->text);
3183 t->text = tmp;
3184 t->next = delete_Token(tt);
3186 break;
3187 case TOK_NUMBER:
3188 if (tt->type == TOK_NUMBER) {
3189 char *tmp = nasm_strcat(t->text, tt->text);
3190 nasm_free(t->text);
3191 t->text = tmp;
3192 t->next = delete_Token(tt);
3194 break;
3195 default:
3196 break;
3199 return thead;
3203 * Expand all single-line macro calls made in the given line.
3204 * Return the expanded version of the line. The original is deemed
3205 * to be destroyed in the process. (In reality we'll just move
3206 * Tokens from input to output a lot of the time, rather than
3207 * actually bothering to destroy and replicate.)
3209 #define DEADMAN_LIMIT (1 << 20)
3211 static Token *expand_smacro(Token * tline)
3213 Token *t, *tt, *mstart, **tail, *thead;
3214 struct hash_table *smtbl;
3215 SMacro *head = NULL, *m;
3216 Token **params;
3217 int *paramsize;
3218 unsigned int nparam, sparam;
3219 int brackets, rescan;
3220 Token *org_tline = tline;
3221 Context *ctx;
3222 char *mname;
3223 int deadman = DEADMAN_LIMIT;
3226 * Trick: we should avoid changing the start token pointer since it can
3227 * be contained in "next" field of other token. Because of this
3228 * we allocate a copy of first token and work with it; at the end of
3229 * routine we copy it back
3231 if (org_tline) {
3232 tline =
3233 new_Token(org_tline->next, org_tline->type, org_tline->text,
3235 tline->mac = org_tline->mac;
3236 nasm_free(org_tline->text);
3237 org_tline->text = NULL;
3240 again:
3241 tail = &thead;
3242 thead = NULL;
3244 while (tline) { /* main token loop */
3245 if (!--deadman) {
3246 error(ERR_NONFATAL, "interminable macro recursion");
3247 break;
3250 if ((mname = tline->text)) {
3251 /* if this token is a local macro, look in local context */
3252 ctx = NULL;
3253 smtbl = &smacros;
3254 if (tline->type == TOK_ID || tline->type == TOK_PREPROC_ID) {
3255 ctx = get_ctx(mname, true);
3256 if (ctx)
3257 smtbl = &ctx->localmac;
3259 head = (SMacro *) hash_findix(smtbl, mname);
3262 * We've hit an identifier. As in is_mmacro below, we first
3263 * check whether the identifier is a single-line macro at
3264 * all, then think about checking for parameters if
3265 * necessary.
3267 for (m = head; m; m = m->next)
3268 if (!mstrcmp(m->name, mname, m->casesense))
3269 break;
3270 if (m) {
3271 mstart = tline;
3272 params = NULL;
3273 paramsize = NULL;
3274 if (m->nparam == 0) {
3276 * Simple case: the macro is parameterless. Discard the
3277 * one token that the macro call took, and push the
3278 * expansion back on the to-do stack.
3280 if (!m->expansion) {
3281 if (!strcmp("__FILE__", m->name)) {
3282 int32_t num = 0;
3283 char *file;
3284 src_get(&num, &file);
3285 tline->text = nasm_quote(file, strlen(file));
3286 tline->type = TOK_STRING;
3287 nasm_free(file);
3288 continue;
3290 if (!strcmp("__LINE__", m->name)) {
3291 nasm_free(tline->text);
3292 make_tok_num(tline, src_get_linnum());
3293 continue;
3295 if (!strcmp("__BITS__", m->name)) {
3296 nasm_free(tline->text);
3297 make_tok_num(tline, globalbits);
3298 continue;
3300 tline = delete_Token(tline);
3301 continue;
3303 } else {
3305 * Complicated case: at least one macro with this name
3306 * exists and takes parameters. We must find the
3307 * parameters in the call, count them, find the SMacro
3308 * that corresponds to that form of the macro call, and
3309 * substitute for the parameters when we expand. What a
3310 * pain.
3312 /*tline = tline->next;
3313 skip_white_(tline); */
3314 do {
3315 t = tline->next;
3316 while (tok_type_(t, TOK_SMAC_END)) {
3317 t->mac->in_progress = false;
3318 t->text = NULL;
3319 t = tline->next = delete_Token(t);
3321 tline = t;
3322 } while (tok_type_(tline, TOK_WHITESPACE));
3323 if (!tok_is_(tline, "(")) {
3325 * This macro wasn't called with parameters: ignore
3326 * the call. (Behaviour borrowed from gnu cpp.)
3328 tline = mstart;
3329 m = NULL;
3330 } else {
3331 int paren = 0;
3332 int white = 0;
3333 brackets = 0;
3334 nparam = 0;
3335 sparam = PARAM_DELTA;
3336 params = nasm_malloc(sparam * sizeof(Token *));
3337 params[0] = tline->next;
3338 paramsize = nasm_malloc(sparam * sizeof(int));
3339 paramsize[0] = 0;
3340 while (true) { /* parameter loop */
3342 * For some unusual expansions
3343 * which concatenates function call
3345 t = tline->next;
3346 while (tok_type_(t, TOK_SMAC_END)) {
3347 t->mac->in_progress = false;
3348 t->text = NULL;
3349 t = tline->next = delete_Token(t);
3351 tline = t;
3353 if (!tline) {
3354 error(ERR_NONFATAL,
3355 "macro call expects terminating `)'");
3356 break;
3358 if (tline->type == TOK_WHITESPACE
3359 && brackets <= 0) {
3360 if (paramsize[nparam])
3361 white++;
3362 else
3363 params[nparam] = tline->next;
3364 continue; /* parameter loop */
3366 if (tline->type == TOK_OTHER
3367 && tline->text[1] == 0) {
3368 char ch = tline->text[0];
3369 if (ch == ',' && !paren && brackets <= 0) {
3370 if (++nparam >= sparam) {
3371 sparam += PARAM_DELTA;
3372 params = nasm_realloc(params,
3373 sparam *
3374 sizeof(Token
3375 *));
3376 paramsize =
3377 nasm_realloc(paramsize,
3378 sparam *
3379 sizeof(int));
3381 params[nparam] = tline->next;
3382 paramsize[nparam] = 0;
3383 white = 0;
3384 continue; /* parameter loop */
3386 if (ch == '{' &&
3387 (brackets > 0 || (brackets == 0 &&
3388 !paramsize[nparam])))
3390 if (!(brackets++)) {
3391 params[nparam] = tline->next;
3392 continue; /* parameter loop */
3395 if (ch == '}' && brackets > 0)
3396 if (--brackets == 0) {
3397 brackets = -1;
3398 continue; /* parameter loop */
3400 if (ch == '(' && !brackets)
3401 paren++;
3402 if (ch == ')' && brackets <= 0)
3403 if (--paren < 0)
3404 break;
3406 if (brackets < 0) {
3407 brackets = 0;
3408 error(ERR_NONFATAL, "braces do not "
3409 "enclose all of macro parameter");
3411 paramsize[nparam] += white + 1;
3412 white = 0;
3413 } /* parameter loop */
3414 nparam++;
3415 while (m && (m->nparam != nparam ||
3416 mstrcmp(m->name, mname,
3417 m->casesense)))
3418 m = m->next;
3419 if (!m)
3420 error(ERR_WARNING | ERR_WARN_MNP,
3421 "macro `%s' exists, "
3422 "but not taking %d parameters",
3423 mstart->text, nparam);
3426 if (m && m->in_progress)
3427 m = NULL;
3428 if (!m) { /* in progess or didn't find '(' or wrong nparam */
3430 * Design question: should we handle !tline, which
3431 * indicates missing ')' here, or expand those
3432 * macros anyway, which requires the (t) test a few
3433 * lines down?
3435 nasm_free(params);
3436 nasm_free(paramsize);
3437 tline = mstart;
3438 } else {
3440 * Expand the macro: we are placed on the last token of the
3441 * call, so that we can easily split the call from the
3442 * following tokens. We also start by pushing an SMAC_END
3443 * token for the cycle removal.
3445 t = tline;
3446 if (t) {
3447 tline = t->next;
3448 t->next = NULL;
3450 tt = new_Token(tline, TOK_SMAC_END, NULL, 0);
3451 tt->mac = m;
3452 m->in_progress = true;
3453 tline = tt;
3454 for (t = m->expansion; t; t = t->next) {
3455 if (t->type >= TOK_SMAC_PARAM) {
3456 Token *pcopy = tline, **ptail = &pcopy;
3457 Token *ttt, *pt;
3458 int i;
3460 ttt = params[t->type - TOK_SMAC_PARAM];
3461 for (i = paramsize[t->type - TOK_SMAC_PARAM];
3462 --i >= 0;) {
3463 pt = *ptail =
3464 new_Token(tline, ttt->type, ttt->text,
3466 ptail = &pt->next;
3467 ttt = ttt->next;
3469 tline = pcopy;
3470 } else if (t->type == TOK_PREPROC_Q) {
3471 tt = new_Token(tline, TOK_ID, mname, 0);
3472 tline = tt;
3473 } else if (t->type == TOK_PREPROC_QQ) {
3474 tt = new_Token(tline, TOK_ID, m->name, 0);
3475 tline = tt;
3476 } else {
3477 tt = new_Token(tline, t->type, t->text, 0);
3478 tline = tt;
3483 * Having done that, get rid of the macro call, and clean
3484 * up the parameters.
3486 nasm_free(params);
3487 nasm_free(paramsize);
3488 free_tlist(mstart);
3489 continue; /* main token loop */
3494 if (tline->type == TOK_SMAC_END) {
3495 tline->mac->in_progress = false;
3496 tline = delete_Token(tline);
3497 } else {
3498 t = *tail = tline;
3499 tline = tline->next;
3500 t->mac = NULL;
3501 t->next = NULL;
3502 tail = &t->next;
3507 * Now scan the entire line and look for successive TOK_IDs that resulted
3508 * after expansion (they can't be produced by tokenize()). The successive
3509 * TOK_IDs should be concatenated.
3510 * Also we look for %+ tokens and concatenate the tokens before and after
3511 * them (without white spaces in between).
3513 t = thead;
3514 rescan = 0;
3515 while (t) {
3516 while (t && t->type != TOK_ID && t->type != TOK_PREPROC_ID)
3517 t = t->next;
3518 if (!t || !t->next)
3519 break;
3520 if (t->next->type == TOK_ID ||
3521 t->next->type == TOK_PREPROC_ID ||
3522 t->next->type == TOK_NUMBER) {
3523 char *p = nasm_strcat(t->text, t->next->text);
3524 nasm_free(t->text);
3525 t->next = delete_Token(t->next);
3526 t->text = p;
3527 rescan = 1;
3528 } else if (t->next->type == TOK_WHITESPACE && t->next->next &&
3529 t->next->next->type == TOK_PREPROC_ID &&
3530 strcmp(t->next->next->text, "%+") == 0) {
3531 /* free the next whitespace, the %+ token and next whitespace */
3532 int i;
3533 for (i = 1; i <= 3; i++) {
3534 if (!t->next
3535 || (i != 2 && t->next->type != TOK_WHITESPACE))
3536 break;
3537 t->next = delete_Token(t->next);
3538 } /* endfor */
3539 } else
3540 t = t->next;
3542 /* If we concatenaded something, re-scan the line for macros */
3543 if (rescan) {
3544 tline = thead;
3545 goto again;
3548 if (org_tline) {
3549 if (thead) {
3550 *org_tline = *thead;
3551 /* since we just gave text to org_line, don't free it */
3552 thead->text = NULL;
3553 delete_Token(thead);
3554 } else {
3555 /* the expression expanded to empty line;
3556 we can't return NULL for some reasons
3557 we just set the line to a single WHITESPACE token. */
3558 memset(org_tline, 0, sizeof(*org_tline));
3559 org_tline->text = NULL;
3560 org_tline->type = TOK_WHITESPACE;
3562 thead = org_tline;
3565 return thead;
3569 * Similar to expand_smacro but used exclusively with macro identifiers
3570 * right before they are fetched in. The reason is that there can be
3571 * identifiers consisting of several subparts. We consider that if there
3572 * are more than one element forming the name, user wants a expansion,
3573 * otherwise it will be left as-is. Example:
3575 * %define %$abc cde
3577 * the identifier %$abc will be left as-is so that the handler for %define
3578 * will suck it and define the corresponding value. Other case:
3580 * %define _%$abc cde
3582 * In this case user wants name to be expanded *before* %define starts
3583 * working, so we'll expand %$abc into something (if it has a value;
3584 * otherwise it will be left as-is) then concatenate all successive
3585 * PP_IDs into one.
3587 static Token *expand_id(Token * tline)
3589 Token *cur, *oldnext = NULL;
3591 if (!tline || !tline->next)
3592 return tline;
3594 cur = tline;
3595 while (cur->next &&
3596 (cur->next->type == TOK_ID ||
3597 cur->next->type == TOK_PREPROC_ID
3598 || cur->next->type == TOK_NUMBER))
3599 cur = cur->next;
3601 /* If identifier consists of just one token, don't expand */
3602 if (cur == tline)
3603 return tline;
3605 if (cur) {
3606 oldnext = cur->next; /* Detach the tail past identifier */
3607 cur->next = NULL; /* so that expand_smacro stops here */
3610 tline = expand_smacro(tline);
3612 if (cur) {
3613 /* expand_smacro possibly changhed tline; re-scan for EOL */
3614 cur = tline;
3615 while (cur && cur->next)
3616 cur = cur->next;
3617 if (cur)
3618 cur->next = oldnext;
3621 return tline;
3625 * Determine whether the given line constitutes a multi-line macro
3626 * call, and return the MMacro structure called if so. Doesn't have
3627 * to check for an initial label - that's taken care of in
3628 * expand_mmacro - but must check numbers of parameters. Guaranteed
3629 * to be called with tline->type == TOK_ID, so the putative macro
3630 * name is easy to find.
3632 static MMacro *is_mmacro(Token * tline, Token *** params_array)
3634 MMacro *head, *m;
3635 Token **params;
3636 int nparam;
3638 head = (MMacro *) hash_findix(&mmacros, tline->text);
3641 * Efficiency: first we see if any macro exists with the given
3642 * name. If not, we can return NULL immediately. _Then_ we
3643 * count the parameters, and then we look further along the
3644 * list if necessary to find the proper MMacro.
3646 for (m = head; m; m = m->next)
3647 if (!mstrcmp(m->name, tline->text, m->casesense))
3648 break;
3649 if (!m)
3650 return NULL;
3653 * OK, we have a potential macro. Count and demarcate the
3654 * parameters.
3656 count_mmac_params(tline->next, &nparam, &params);
3659 * So we know how many parameters we've got. Find the MMacro
3660 * structure that handles this number.
3662 while (m) {
3663 if (m->nparam_min <= nparam
3664 && (m->plus || nparam <= m->nparam_max)) {
3666 * This one is right. Just check if cycle removal
3667 * prohibits us using it before we actually celebrate...
3669 if (m->in_progress) {
3670 #if 0
3671 error(ERR_NONFATAL,
3672 "self-reference in multi-line macro `%s'", m->name);
3673 #endif
3674 nasm_free(params);
3675 return NULL;
3678 * It's right, and we can use it. Add its default
3679 * parameters to the end of our list if necessary.
3681 if (m->defaults && nparam < m->nparam_min + m->ndefs) {
3682 params =
3683 nasm_realloc(params,
3684 ((m->nparam_min + m->ndefs +
3685 1) * sizeof(*params)));
3686 while (nparam < m->nparam_min + m->ndefs) {
3687 params[nparam] = m->defaults[nparam - m->nparam_min];
3688 nparam++;
3692 * If we've gone over the maximum parameter count (and
3693 * we're in Plus mode), ignore parameters beyond
3694 * nparam_max.
3696 if (m->plus && nparam > m->nparam_max)
3697 nparam = m->nparam_max;
3699 * Then terminate the parameter list, and leave.
3701 if (!params) { /* need this special case */
3702 params = nasm_malloc(sizeof(*params));
3703 nparam = 0;
3705 params[nparam] = NULL;
3706 *params_array = params;
3707 return m;
3710 * This one wasn't right: look for the next one with the
3711 * same name.
3713 for (m = m->next; m; m = m->next)
3714 if (!mstrcmp(m->name, tline->text, m->casesense))
3715 break;
3719 * After all that, we didn't find one with the right number of
3720 * parameters. Issue a warning, and fail to expand the macro.
3722 error(ERR_WARNING | ERR_WARN_MNP,
3723 "macro `%s' exists, but not taking %d parameters",
3724 tline->text, nparam);
3725 nasm_free(params);
3726 return NULL;
3730 * Expand the multi-line macro call made by the given line, if
3731 * there is one to be expanded. If there is, push the expansion on
3732 * istk->expansion and return 1. Otherwise return 0.
3734 static int expand_mmacro(Token * tline)
3736 Token *startline = tline;
3737 Token *label = NULL;
3738 int dont_prepend = 0;
3739 Token **params, *t, *mtok, *tt;
3740 MMacro *m;
3741 Line *l, *ll;
3742 int i, nparam, *paramlen;
3743 const char *mname;
3745 t = tline;
3746 skip_white_(t);
3747 /* if (!tok_type_(t, TOK_ID)) Lino 02/25/02 */
3748 if (!tok_type_(t, TOK_ID) && !tok_type_(t, TOK_PREPROC_ID))
3749 return 0;
3750 mtok = t;
3751 m = is_mmacro(t, &params);
3752 if (m) {
3753 mname = t->text;
3754 } else {
3755 Token *last;
3757 * We have an id which isn't a macro call. We'll assume
3758 * it might be a label; we'll also check to see if a
3759 * colon follows it. Then, if there's another id after
3760 * that lot, we'll check it again for macro-hood.
3762 label = last = t;
3763 t = t->next;
3764 if (tok_type_(t, TOK_WHITESPACE))
3765 last = t, t = t->next;
3766 if (tok_is_(t, ":")) {
3767 dont_prepend = 1;
3768 last = t, t = t->next;
3769 if (tok_type_(t, TOK_WHITESPACE))
3770 last = t, t = t->next;
3772 if (!tok_type_(t, TOK_ID) || (m = is_mmacro(t, &params)) == NULL)
3773 return 0;
3774 last->next = NULL;
3775 mname = t->text;
3776 tline = t;
3780 * Fix up the parameters: this involves stripping leading and
3781 * trailing whitespace, then stripping braces if they are
3782 * present.
3784 for (nparam = 0; params[nparam]; nparam++) ;
3785 paramlen = nparam ? nasm_malloc(nparam * sizeof(*paramlen)) : NULL;
3787 for (i = 0; params[i]; i++) {
3788 int brace = false;
3789 int comma = (!m->plus || i < nparam - 1);
3791 t = params[i];
3792 skip_white_(t);
3793 if (tok_is_(t, "{"))
3794 t = t->next, brace = true, comma = false;
3795 params[i] = t;
3796 paramlen[i] = 0;
3797 while (t) {
3798 if (comma && t->type == TOK_OTHER && !strcmp(t->text, ","))
3799 break; /* ... because we have hit a comma */
3800 if (comma && t->type == TOK_WHITESPACE
3801 && tok_is_(t->next, ","))
3802 break; /* ... or a space then a comma */
3803 if (brace && t->type == TOK_OTHER && !strcmp(t->text, "}"))
3804 break; /* ... or a brace */
3805 t = t->next;
3806 paramlen[i]++;
3811 * OK, we have a MMacro structure together with a set of
3812 * parameters. We must now go through the expansion and push
3813 * copies of each Line on to istk->expansion. Substitution of
3814 * parameter tokens and macro-local tokens doesn't get done
3815 * until the single-line macro substitution process; this is
3816 * because delaying them allows us to change the semantics
3817 * later through %rotate.
3819 * First, push an end marker on to istk->expansion, mark this
3820 * macro as in progress, and set up its invocation-specific
3821 * variables.
3823 ll = nasm_malloc(sizeof(Line));
3824 ll->next = istk->expansion;
3825 ll->finishes = m;
3826 ll->first = NULL;
3827 istk->expansion = ll;
3829 m->in_progress = true;
3830 m->params = params;
3831 m->iline = tline;
3832 m->nparam = nparam;
3833 m->rotate = 0;
3834 m->paramlen = paramlen;
3835 m->unique = unique++;
3836 m->lineno = 0;
3838 m->next_active = istk->mstk;
3839 istk->mstk = m;
3841 for (l = m->expansion; l; l = l->next) {
3842 Token **tail;
3844 ll = nasm_malloc(sizeof(Line));
3845 ll->finishes = NULL;
3846 ll->next = istk->expansion;
3847 istk->expansion = ll;
3848 tail = &ll->first;
3850 for (t = l->first; t; t = t->next) {
3851 Token *x = t;
3852 switch (t->type) {
3853 case TOK_PREPROC_Q:
3854 tt = *tail = new_Token(NULL, TOK_ID, mname, 0);
3855 break;
3856 case TOK_PREPROC_QQ:
3857 tt = *tail = new_Token(NULL, TOK_ID, m->name, 0);
3858 break;
3859 case TOK_PREPROC_ID:
3860 if (t->text[1] == '0' && t->text[2] == '0') {
3861 dont_prepend = -1;
3862 x = label;
3863 if (!x)
3864 continue;
3866 /* fall through */
3867 default:
3868 tt = *tail = new_Token(NULL, x->type, x->text, 0);
3869 break;
3871 tail = &tt->next;
3873 *tail = NULL;
3877 * If we had a label, push it on as the first line of
3878 * the macro expansion.
3880 if (label) {
3881 if (dont_prepend < 0)
3882 free_tlist(startline);
3883 else {
3884 ll = nasm_malloc(sizeof(Line));
3885 ll->finishes = NULL;
3886 ll->next = istk->expansion;
3887 istk->expansion = ll;
3888 ll->first = startline;
3889 if (!dont_prepend) {
3890 while (label->next)
3891 label = label->next;
3892 label->next = tt = new_Token(NULL, TOK_OTHER, ":", 0);
3897 list->uplevel(m->nolist ? LIST_MACRO_NOLIST : LIST_MACRO);
3899 return 1;
3903 * Since preprocessor always operate only on the line that didn't
3904 * arrived yet, we should always use ERR_OFFBY1. Also since user
3905 * won't want to see same error twice (preprocessing is done once
3906 * per pass) we will want to show errors only during pass one.
3908 static void error(int severity, const char *fmt, ...)
3910 va_list arg;
3911 char buff[1024];
3913 /* If we're in a dead branch of IF or something like it, ignore the error */
3914 if (istk && istk->conds && !emitting(istk->conds->state))
3915 return;
3917 va_start(arg, fmt);
3918 vsnprintf(buff, sizeof(buff), fmt, arg);
3919 va_end(arg);
3921 if (istk && istk->mstk && istk->mstk->name)
3922 _error(severity | ERR_PASS1, "(%s:%d) %s", istk->mstk->name,
3923 istk->mstk->lineno, buff);
3924 else
3925 _error(severity | ERR_PASS1, "%s", buff);
3928 static void
3929 pp_reset(char *file, int apass, efunc errfunc, evalfunc eval,
3930 ListGen * listgen, StrList **deplist)
3932 _error = errfunc;
3933 cstk = NULL;
3934 istk = nasm_malloc(sizeof(Include));
3935 istk->next = NULL;
3936 istk->conds = NULL;
3937 istk->expansion = NULL;
3938 istk->mstk = NULL;
3939 istk->fp = fopen(file, "r");
3940 istk->fname = NULL;
3941 src_set_fname(nasm_strdup(file));
3942 src_set_linnum(0);
3943 istk->lineinc = 1;
3944 if (!istk->fp)
3945 error(ERR_FATAL | ERR_NOFILE, "unable to open input file `%s'",
3946 file);
3947 defining = NULL;
3948 init_macros();
3949 unique = 0;
3950 if (tasm_compatible_mode) {
3951 stdmacpos = nasm_stdmac;
3952 } else {
3953 stdmacpos = nasm_stdmac_after_tasm;
3955 any_extrastdmac = (extrastdmac != NULL);
3956 list = listgen;
3957 evaluate = eval;
3958 pass = apass;
3959 dephead = deptail = deplist;
3960 if (deplist) {
3961 StrList *sl = nasm_malloc(strlen(file)+1+sizeof sl->next);
3962 sl->next = NULL;
3963 strcpy(sl->str, file);
3964 *deptail = sl;
3965 deptail = &sl->next;
3969 static char *pp_getline(void)
3971 char *line;
3972 Token *tline;
3974 while (1) {
3976 * Fetch a tokenized line, either from the macro-expansion
3977 * buffer or from the input file.
3979 tline = NULL;
3980 while (istk->expansion && istk->expansion->finishes) {
3981 Line *l = istk->expansion;
3982 if (!l->finishes->name && l->finishes->in_progress > 1) {
3983 Line *ll;
3986 * This is a macro-end marker for a macro with no
3987 * name, which means it's not really a macro at all
3988 * but a %rep block, and the `in_progress' field is
3989 * more than 1, meaning that we still need to
3990 * repeat. (1 means the natural last repetition; 0
3991 * means termination by %exitrep.) We have
3992 * therefore expanded up to the %endrep, and must
3993 * push the whole block on to the expansion buffer
3994 * again. We don't bother to remove the macro-end
3995 * marker: we'd only have to generate another one
3996 * if we did.
3998 l->finishes->in_progress--;
3999 for (l = l->finishes->expansion; l; l = l->next) {
4000 Token *t, *tt, **tail;
4002 ll = nasm_malloc(sizeof(Line));
4003 ll->next = istk->expansion;
4004 ll->finishes = NULL;
4005 ll->first = NULL;
4006 tail = &ll->first;
4008 for (t = l->first; t; t = t->next) {
4009 if (t->text || t->type == TOK_WHITESPACE) {
4010 tt = *tail =
4011 new_Token(NULL, t->type, t->text, 0);
4012 tail = &tt->next;
4016 istk->expansion = ll;
4018 } else {
4020 * Check whether a `%rep' was started and not ended
4021 * within this macro expansion. This can happen and
4022 * should be detected. It's a fatal error because
4023 * I'm too confused to work out how to recover
4024 * sensibly from it.
4026 if (defining) {
4027 if (defining->name)
4028 error(ERR_PANIC,
4029 "defining with name in expansion");
4030 else if (istk->mstk->name)
4031 error(ERR_FATAL,
4032 "`%%rep' without `%%endrep' within"
4033 " expansion of macro `%s'",
4034 istk->mstk->name);
4038 * FIXME: investigate the relationship at this point between
4039 * istk->mstk and l->finishes
4042 MMacro *m = istk->mstk;
4043 istk->mstk = m->next_active;
4044 if (m->name) {
4046 * This was a real macro call, not a %rep, and
4047 * therefore the parameter information needs to
4048 * be freed.
4050 nasm_free(m->params);
4051 free_tlist(m->iline);
4052 nasm_free(m->paramlen);
4053 l->finishes->in_progress = false;
4054 } else
4055 free_mmacro(m);
4057 istk->expansion = l->next;
4058 nasm_free(l);
4059 list->downlevel(LIST_MACRO);
4062 while (1) { /* until we get a line we can use */
4064 if (istk->expansion) { /* from a macro expansion */
4065 char *p;
4066 Line *l = istk->expansion;
4067 if (istk->mstk)
4068 istk->mstk->lineno++;
4069 tline = l->first;
4070 istk->expansion = l->next;
4071 nasm_free(l);
4072 p = detoken(tline, false);
4073 list->line(LIST_MACRO, p);
4074 nasm_free(p);
4075 break;
4077 line = read_line();
4078 if (line) { /* from the current input file */
4079 line = prepreproc(line);
4080 tline = tokenize(line);
4081 nasm_free(line);
4082 break;
4085 * The current file has ended; work down the istk
4088 Include *i = istk;
4089 fclose(i->fp);
4090 if (i->conds)
4091 error(ERR_FATAL,
4092 "expected `%%endif' before end of file");
4093 /* only set line and file name if there's a next node */
4094 if (i->next) {
4095 src_set_linnum(i->lineno);
4096 nasm_free(src_set_fname(i->fname));
4098 istk = i->next;
4099 list->downlevel(LIST_INCLUDE);
4100 nasm_free(i);
4101 if (!istk)
4102 return NULL;
4107 * We must expand MMacro parameters and MMacro-local labels
4108 * _before_ we plunge into directive processing, to cope
4109 * with things like `%define something %1' such as STRUC
4110 * uses. Unless we're _defining_ a MMacro, in which case
4111 * those tokens should be left alone to go into the
4112 * definition; and unless we're in a non-emitting
4113 * condition, in which case we don't want to meddle with
4114 * anything.
4116 if (!defining && !(istk->conds && !emitting(istk->conds->state)))
4117 tline = expand_mmac_params(tline);
4120 * Check the line to see if it's a preprocessor directive.
4122 if (do_directive(tline) == DIRECTIVE_FOUND) {
4123 continue;
4124 } else if (defining) {
4126 * We're defining a multi-line macro. We emit nothing
4127 * at all, and just
4128 * shove the tokenized line on to the macro definition.
4130 Line *l = nasm_malloc(sizeof(Line));
4131 l->next = defining->expansion;
4132 l->first = tline;
4133 l->finishes = false;
4134 defining->expansion = l;
4135 continue;
4136 } else if (istk->conds && !emitting(istk->conds->state)) {
4138 * We're in a non-emitting branch of a condition block.
4139 * Emit nothing at all, not even a blank line: when we
4140 * emerge from the condition we'll give a line-number
4141 * directive so we keep our place correctly.
4143 free_tlist(tline);
4144 continue;
4145 } else if (istk->mstk && !istk->mstk->in_progress) {
4147 * We're in a %rep block which has been terminated, so
4148 * we're walking through to the %endrep without
4149 * emitting anything. Emit nothing at all, not even a
4150 * blank line: when we emerge from the %rep block we'll
4151 * give a line-number directive so we keep our place
4152 * correctly.
4154 free_tlist(tline);
4155 continue;
4156 } else {
4157 tline = expand_smacro(tline);
4158 if (!expand_mmacro(tline)) {
4160 * De-tokenize the line again, and emit it.
4162 line = detoken(tline, true);
4163 free_tlist(tline);
4164 break;
4165 } else {
4166 continue; /* expand_mmacro calls free_tlist */
4171 return line;
4174 static void pp_cleanup(int pass)
4176 if (defining) {
4177 error(ERR_NONFATAL, "end of file while still defining macro `%s'",
4178 defining->name);
4179 free_mmacro(defining);
4181 while (cstk)
4182 ctx_pop();
4183 free_macros();
4184 while (istk) {
4185 Include *i = istk;
4186 istk = istk->next;
4187 fclose(i->fp);
4188 nasm_free(i->fname);
4189 nasm_free(i);
4191 while (cstk)
4192 ctx_pop();
4193 if (pass == 0) {
4194 free_llist(predef);
4195 delete_Blocks();
4199 void pp_include_path(char *path)
4201 IncPath *i;
4203 i = nasm_malloc(sizeof(IncPath));
4204 i->path = path ? nasm_strdup(path) : NULL;
4205 i->next = NULL;
4207 if (ipath != NULL) {
4208 IncPath *j = ipath;
4209 while (j->next != NULL)
4210 j = j->next;
4211 j->next = i;
4212 } else {
4213 ipath = i;
4217 void pp_pre_include(char *fname)
4219 Token *inc, *space, *name;
4220 Line *l;
4222 name = new_Token(NULL, TOK_INTERNAL_STRING, fname, 0);
4223 space = new_Token(name, TOK_WHITESPACE, NULL, 0);
4224 inc = new_Token(space, TOK_PREPROC_ID, "%include", 0);
4226 l = nasm_malloc(sizeof(Line));
4227 l->next = predef;
4228 l->first = inc;
4229 l->finishes = false;
4230 predef = l;
4233 void pp_pre_define(char *definition)
4235 Token *def, *space;
4236 Line *l;
4237 char *equals;
4239 equals = strchr(definition, '=');
4240 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
4241 def = new_Token(space, TOK_PREPROC_ID, "%define", 0);
4242 if (equals)
4243 *equals = ' ';
4244 space->next = tokenize(definition);
4245 if (equals)
4246 *equals = '=';
4248 l = nasm_malloc(sizeof(Line));
4249 l->next = predef;
4250 l->first = def;
4251 l->finishes = false;
4252 predef = l;
4255 void pp_pre_undefine(char *definition)
4257 Token *def, *space;
4258 Line *l;
4260 space = new_Token(NULL, TOK_WHITESPACE, NULL, 0);
4261 def = new_Token(space, TOK_PREPROC_ID, "%undef", 0);
4262 space->next = tokenize(definition);
4264 l = nasm_malloc(sizeof(Line));
4265 l->next = predef;
4266 l->first = def;
4267 l->finishes = false;
4268 predef = l;
4272 * Added by Keith Kanios:
4274 * This function is used to assist with "runtime" preprocessor
4275 * directives. (e.g. pp_runtime("%define __BITS__ 64");)
4277 * ERRORS ARE IGNORED HERE, SO MAKE COMPLETELY SURE THAT YOU
4278 * PASS A VALID STRING TO THIS FUNCTION!!!!!
4281 void pp_runtime(char *definition)
4283 Token *def;
4285 def = tokenize(definition);
4286 if(do_directive(def) == NO_DIRECTIVE_FOUND)
4287 free_tlist(def);
4291 void pp_extra_stdmac(const char **macros)
4293 extrastdmac = macros;
4296 static void make_tok_num(Token * tok, int64_t val)
4298 char numbuf[20];
4299 snprintf(numbuf, sizeof(numbuf), "%"PRId64"", val);
4300 tok->text = nasm_strdup(numbuf);
4301 tok->type = TOK_NUMBER;
4304 Preproc nasmpp = {
4305 pp_reset,
4306 pp_getline,
4307 pp_cleanup