2 * Function and variable definition and declaration, statement parser.
4 * Copyright (c) ITB CompuPhase, 1997-2003
6 * This software is provided "as-is", without any express or implied
7 * warranty. In no event will the authors be held liable for any
8 * damages arising from the use of this software. Permission is granted
9 * to anyone to use this software for any purpose, including commercial
10 * applications, and to alter it and redistribute it freely, subject to
11 * the following restrictions:
13 * 1. The origin of this software must not be misrepresented;
14 * you must not claim that you wrote the original software.
15 * If you use this software in a product, an acknowledgment in the
16 * product documentation would be appreciated but is not required.
17 * 2. Altered source versions must be plainly marked as such, and
18 * must not be misrepresented as being the original software.
19 * 3. This notice may not be removed or altered from any source
43 #endif /* HAVE_EVIL */
45 #include "embryo_cc_sc.h"
46 #include "embryo_cc_prefix.h"
48 #define VERSION_STR "2.4"
49 #define VERSION_INT 240
51 static void resetglobals(void);
52 static void initglobals(void);
53 static void setopt(int argc, char **argv,
54 char *iname, char *oname,
55 char *pname, char *rname);
56 static void setconfig(char *root);
57 static void about(void);
58 static void setconstants(void);
59 static void parse(void);
60 static void dumplits(void);
61 static void dumpzero(int count);
62 static void declfuncvar(int tok, char *symname,
64 int fstatic, int fstock, int fconst);
65 static void declglb(char *firstname, int firsttag,
66 int fpublic, int fstatic, int stock, int fconst);
67 static int declloc(int fstatic);
68 static void decl_const(int table);
69 static void decl_enum(int table);
70 static cell needsub(int *tag);
71 static void initials(int ident, int tag,
72 cell * size, int dim[], int numdim);
73 static cell initvector(int ident, int tag, cell size, int fillzero);
74 static cell init(int ident, int *tag);
75 static void funcstub(int native);
76 static int newfunc(char *firstname, int firsttag,
77 int fpublic, int fstatic, int stock);
78 static int declargs(symbol * sym);
79 static void doarg(char *name, int ident, int offset,
80 int tags[], int numtags,
81 int fpublic, int fconst, arginfo * arg);
82 static void reduce_referrers(symbol * root);
83 static int testsymbols(symbol * root, int level,
84 int testlabs, int testconst);
85 static void destructsymbols(symbol * root, int level);
86 static constvalue *find_constval_byval(constvalue * table, cell val);
87 static void statement(int *lastindent, int allow_decl);
88 static void compound(void);
89 static void doexpr(int comma, int chkeffect,
90 int allowarray, int mark_endexpr,
91 int *tag, int chkfuncresult);
92 static void doassert(void);
93 static void doexit(void);
94 static void test(int label, int parens, int invert);
95 static void doif(void);
96 static void dowhile(void);
97 static void dodo(void);
98 static void dofor(void);
99 static void doswitch(void);
100 static void dogoto(void);
101 static void dolabel(void);
102 static symbol *fetchlab(char *name);
103 static void doreturn(void);
104 static void dobreak(void);
105 static void docont(void);
106 static void dosleep(void);
107 static void addwhile(int *ptr);
108 static void delwhile(void);
109 static int *readwhile(void);
111 static int lastst = 0; /* last executed statement type */
112 static int nestlevel = 0; /* number of active (open) compound statements */
113 static int rettype = 0; /* the type that a "return" expression should have */
114 static int skipinput = 0; /* number of lines to skip from the first input file */
115 static int wq[wqTABSZ]; /* "while queue", internal stack for nested loops */
116 static int *wqptr; /* pointer to next entry */
117 static char binfname[PATH_MAX]; /* binary file name */
120 main(int argc, char *argv[], char *env[] __UNUSED__)
122 e_prefix_determine(argv[0]);
123 return sc_compile(argc, argv);
127 sc_error(int number, char *message, char *filename, int firstline,
128 int lastline, va_list argptr)
130 static char *prefix[3] = { "error", "fatal error", "warning" };
136 pre = prefix[number / 100];
138 fprintf(stderr, "%s(%d -- %d) : %s %03d: ", filename, firstline,
139 lastline, pre, number);
141 fprintf(stderr, "%s(%d) : %s %03d: ", filename, lastline, pre,
144 vfprintf(stderr, message, argptr);
150 sc_opensrc(char *filename)
152 return fopen(filename, "rb");
156 sc_closesrc(void *handle)
158 assert(handle != NULL);
159 fclose((FILE *) handle);
163 sc_resetsrc(void *handle, void *position)
165 assert(handle != NULL);
166 fsetpos((FILE *) handle, (fpos_t *) position);
170 sc_readsrc(void *handle, char *target, int maxchars)
172 return fgets(target, maxchars, (FILE *) handle);
176 sc_getpossrc(void *handle)
178 static fpos_t lastpos; /* may need to have a LIFO stack of
181 fgetpos((FILE *) handle, &lastpos);
186 sc_eofsrc(void *handle)
188 return feof((FILE *) handle);
194 return fdopen(fd, "w+");
198 sc_closeasm(void *handle)
201 fclose((FILE *) handle);
205 sc_resetasm(void *handle)
207 fflush((FILE *) handle);
208 fseek((FILE *) handle, 0, SEEK_SET);
212 sc_writeasm(void *handle, char *st)
214 return fputs(st, (FILE *) handle) >= 0;
218 sc_readasm(void *handle, char *target, int maxchars)
220 return fgets(target, maxchars, (FILE *) handle);
224 sc_openbin(char *filename)
226 return fopen(filename, "wb");
230 sc_closebin(void *handle, int deletefile)
232 fclose((FILE *) handle);
238 sc_resetbin(void *handle)
240 fflush((FILE *) handle);
241 fseek((FILE *) handle, 0, SEEK_SET);
245 sc_writebin(void *handle, void *buffer, int size)
247 return (int)fwrite(buffer, 1, size, (FILE *) handle) == size;
251 sc_lengthbin(void *handle)
253 return ftell((FILE *) handle);
256 /* "main" of the compiler
259 sc_compile(int argc, char *argv[])
261 int entry, i, jmpcode, fd_out;
263 char incfname[PATH_MAX];
264 char reportname[PATH_MAX];
268 int lcl_packstr, lcl_needsemicolon, lcl_tabsize;
271 /* set global variables to their initial value */
275 errorset(sEXPRRELEASE);
278 /* make sure that we clean up on a fatal error; do this before the
279 * first call to error(). */
280 if ((jmpcode = setjmp(errbuf)) != 0)
283 /* allocate memory for fixed tables */
284 inpfname = (char *)malloc(PATH_MAX);
285 litq = (cell *) malloc(litmax * sizeof(cell));
287 error(103); /* insufficient memory */
289 error(103); /* insufficient memory */
291 setopt(argc, argv, inpfname, binfname, incfname, reportname);
293 /* open the output file */
296 tmpdir = getenv("TMPDIR");
297 if (!tmpdir) tmpdir = "/tmp";
299 tmpdir = (char *)evil_tmpdir_get();
300 #endif /* ! HAVE_EVIL */
302 snprintf(outfname, PATH_MAX, "%s/embryo_cc.asm-tmp-XXXXXX", tmpdir);
303 fd_out = mkstemp(outfname);
305 error(101, outfname);
307 unlink (outfname); /* kill this file as soon as it's (f)close'd */
309 setconfig(argv[0]); /* the path to the include files */
310 lcl_ctrlchar = sc_ctrlchar;
311 lcl_packstr = sc_packstr;
312 lcl_needsemicolon = sc_needsemicolon;
313 lcl_tabsize = sc_tabsize;
314 inpf = inpf_org = (FILE *) sc_opensrc(inpfname);
316 error(100, inpfname);
318 outf = (FILE *) sc_openasm(fd_out); /* first write to assembler
319 * file (may be temporary) */
321 error(101, outfname);
322 /* immediately open the binary file, for other programs to check */
323 binf = (FILE *) sc_openbin(binfname);
325 error(101, binfname);
326 setconstants(); /* set predefined constants and tagnames */
327 for (i = 0; i < skipinput; i++) /* skip lines in the input file */
328 if (sc_readsrc(inpf, pline, sLINEMAX))
329 fline++; /* keep line number up to date */
331 sc_status = statFIRST;
332 /* do the first pass through the file */
333 inpfmark = sc_getpossrc(inpf);
334 if (incfname[0] != '\0')
336 if (strcmp(incfname, sDEF_PREFIX) == 0)
338 plungefile(incfname, FALSE, TRUE); /* parse "default.inc" */
342 if (!plungequalifiedfile(incfname)) /* parse "prefix" include
344 error(100, incfname); /* cannot read from ... (fatal error) */
347 preprocess(); /* fetch first line */
348 parse(); /* process all input */
351 sc_status = statWRITE; /* set, to enable warnings */
353 /* ??? for re-parsing the listing file instead of the original source
354 * file (and doing preprocessing twice):
355 * - close input file, close listing file
356 * - re-open listing file for reading (inpf)
357 * - open assembler file (outf)
360 /* reset "defined" flag of all functions and global variables */
361 reduce_referrers(&glbtab);
362 delete_symbols(&glbtab, 0, TRUE, FALSE);
363 #if !defined NO_DEFINE
367 sc_ctrlchar = lcl_ctrlchar;
368 sc_packstr = lcl_packstr;
369 sc_needsemicolon = lcl_needsemicolon;
370 sc_tabsize = lcl_tabsize;
372 /* reset the source file */
375 sc_resetsrc(inpf, inpfmark); /* reset file position */
376 fline = skipinput; /* reset line number */
377 lexinit(); /* clear internal flags of lex() */
378 sc_status = statWRITE; /* allow to write --this variable was reset
379 * by resetglobals() */
381 setfile(inpfname, fnumber);
382 if (incfname[0] != '\0')
384 if (strcmp(incfname, sDEF_PREFIX) == 0)
385 plungefile(incfname, FALSE, TRUE); /* parse "default.inc" (again) */
387 plungequalifiedfile(incfname); /* parse implicit include
390 preprocess(); /* fetch first line */
391 parse(); /* process all input */
392 /* inpf is already closed when readline() attempts to pop of a file */
393 writetrailer(); /* write remaining stuff */
395 entry = testsymbols(&glbtab, 0, TRUE, FALSE); /* test for unused
396 * or undefined functions and variables */
398 error(13); /* no entry point (no public functions) */
401 if (inpf) /* main source file is not closed, do it now */
403 /* write the binary file (the file is already open) */
404 if (errnum == 0 && jmpcode == 0)
406 assert(binf != NULL);
407 sc_resetasm(outf); /* flush and loop back, for reading */
408 assemble(binf, outf); /* assembler file is now input */
413 sc_closebin(binf, errnum != 0);
421 assert(jmpcode != 0 || loctab.next == NULL); /* on normal flow,
423 * should already have been deleted */
424 delete_symbols(&loctab, 0, TRUE, TRUE); /* delete local variables
425 * if not yet done (i.e.
426 * on a fatal error) */
427 delete_symbols(&glbtab, 0, TRUE, TRUE);
428 delete_consttable(&tagname_tab);
429 delete_consttable(&libname_tab);
432 #if !defined NO_DEFINE
437 printf("\n%d Error%s.\n", errnum, (errnum > 1) ? "s" : "");
440 else if (warnnum != 0)
442 printf("\n%d Warning%s.\n", warnnum, (warnnum > 1) ? "s" : "");
453 sc_addconstant(char *name, cell value, int tag)
455 errorset(sFORCESET); /* make sure error engine is silenced */
456 sc_status = statIDLE;
457 add_constant(name, value, sGLOBAL, tag);
462 sc_addtag(char *name)
470 /* no tagname was given, check for one */
471 if (lex(&val, &name) != tLABEL)
474 return 0; /* untagged */
479 ptr = tagname_tab.next;
482 tag = (int)(ptr->value & TAGMASK);
483 if (strcmp(name, ptr->name) == 0)
484 return tag; /* tagname is known, return its sequence number */
485 tag &= (int)~FIXEDTAG;
491 /* tagname currently unknown, add it */
492 tag = last + 1; /* guaranteed not to exist already */
494 tag |= (int)FIXEDTAG;
495 append_constval(&tagname_tab, name, (cell) tag, 0);
502 /* reset the subset of global variables that is modified by the
504 curfunc = NULL; /* pointer to current function */
505 lastst = 0; /* last executed statement type */
506 nestlevel = 0; /* number of active (open) compound statements */
507 rettype = 0; /* the type that a "return" expression should have */
508 litidx = 0; /* index to literal table */
509 stgidx = 0; /* index to the staging buffer */
510 labnum = 0; /* number of (internal) labels */
511 staging = 0; /* true if staging output */
512 declared = 0; /* number of local cells declared */
513 glb_declared = 0; /* number of global cells declared */
514 code_idx = 0; /* number of bytes with generated code */
515 ntv_funcid = 0; /* incremental number of native function */
516 curseg = 0; /* 1 if currently parsing CODE, 2 if parsing DATA */
517 freading = FALSE; /* no input file ready yet */
518 fline = 0; /* the line number in the current file */
519 fnumber = 0; /* the file number in the file table (debugging) */
520 fcurrent = 0; /* current file being processed (debugging) */
521 intest = 0; /* true if inside a test */
522 sideeffect = 0; /* true if an expression causes a side-effect */
523 stmtindent = 0; /* current indent of the statement */
524 indent_nowarn = TRUE; /* do not skip warning "217 loose indentation" */
525 sc_allowtags = TRUE; /* allow/detect tagnames */
526 sc_status = statIDLE;
534 skipinput = 0; /* number of lines to skip from the first
536 sc_ctrlchar = CTRL_CHAR; /* the escape character */
537 litmax = sDEF_LITMAX; /* current size of the literal table */
538 errnum = 0; /* number of errors */
539 warnnum = 0; /* number of warnings */
540 /* sc_debug=sCHKBOUNDS; by default: bounds checking+assertions */
541 sc_debug = 0; /* by default: no debug */
542 charbits = 8; /* a "char" is 8 bits */
543 sc_packstr = FALSE; /* strings are unpacked by default */
544 /* sc_compress=TRUE; compress output bytecodes */
545 sc_compress = FALSE; /* compress output bytecodes */
546 sc_needsemicolon = FALSE; /* semicolon required to terminate
549 sc_stksize = sDEF_AMXSTACK; /* default stack size */
550 sc_tabsize = 8; /* assume a TAB is 8 spaces */
551 sc_rationaltag = 0; /* assume no support for rational numbers */
552 rational_digits = 0; /* number of fractional digits */
554 outfname[0] = '\0'; /* output file name */
555 inpf = NULL; /* file read from */
556 inpfname = NULL; /* pointer to name of the file currently
558 outf = NULL; /* file written to */
559 litq = NULL; /* the literal queue */
560 glbtab.next = NULL; /* clear global variables/constants table */
561 loctab.next = NULL; /* " local " / " " */
562 tagname_tab.next = NULL; /* tagname table */
563 libname_tab.next = NULL; /* library table (#pragma library "..."
566 pline[0] = '\0'; /* the line read from the input file */
567 lptr = NULL; /* points to the current position in "pline" */
568 curlibrary = NULL; /* current library */
569 inpf_org = NULL; /* main source file */
571 wqptr = wq; /* initialize while queue pointer */
576 parseoptions(int argc, char **argv, char *iname, char *oname,
577 char *pname __UNUSED__, char *rname __UNUSED__)
583 /* use embryo include dir always */
584 snprintf(str, sizeof(str), "%s/include/", e_prefix_data_get());
588 for (i = 1; i < argc; i++)
590 if (!strcmp (argv[i], "-i") && (i + 1 < argc) && *argv[i + 1])
592 /* include directory */
594 strncpy(str, argv[i], sizeof(str));
597 if (str[len - 1] != DIRSEP_CHAR)
599 str[len] = DIRSEP_CHAR;
605 else if (!strcmp (argv[i], "-o") && (i + 1 < argc) && *argv[i + 1])
609 strcpy(oname, argv[i]); /* FIXME */
611 else if (!strcmp (argv[i], "-S") && (i + 1 < argc) && *argv[i + 1])
615 stack_size = atoi(argv[i]);
618 sc_stksize = (cell) stack_size;
625 strcpy(iname, argv[i]); /* FIXME */
629 /* only allow one input filename */
636 setopt(int argc, char **argv, char *iname, char *oname,
637 char *pname, char *rname)
643 strcpy(pname, sDEF_PREFIX);
645 parseoptions(argc, argv, iname, oname, pname, rname);
646 if (iname[0] == '\0')
651 setconfig(char *root)
657 /* add the default "include" directory */
660 /* path + filename (hopefully) */
661 strncpy(path, root, sizeof(path) - 1);
662 path[sizeof(path) - 1] = 0;
664 /* terminate just behind last \ or : */
665 if ((ptr = strrchr(path, DIRSEP_CHAR))
666 || (ptr = strchr(path, ':')))
668 /* If there was no terminating "\" or ":",
669 * the filename probably does not
670 * contain the path; so we just don't add it
671 * to the list in that case
674 if (strlen(path) < (sizeof(path) - 1 - 7))
676 strcat(path, "include");
679 path[len] = DIRSEP_CHAR;
680 path[len + 1] = '\0';
688 printf("Usage: embryo_cc <filename> [options]\n\n");
689 printf("Options:\n");
692 (" -A<num> alignment in bytes of the data segment and the\
696 (" -a output assembler code (skip code generation\
700 (" -C[+/-] compact encoding for output file (default=%c)\n",
701 sc_compress ? '+' : '-');
702 printf(" -c8 [default] a character is 8-bits\
703 (ASCII/ISO Latin-1)\n");
705 printf(" -c16 a character is 16-bits (Unicode)\n");
706 #if defined dos_setdrive
707 printf(" -Dpath active directory path\n");
710 (" -d0 no symbolic information, no run-time checks\n");
711 printf(" -d1 [default] run-time checks, no symbolic\
714 (" -d2 full debug information and dynamic checking\n");
715 printf(" -d3 full debug information, dynamic checking,\
718 printf(" -i <name> path for include files\n");
720 printf(" -l create list file (preprocess only)\n");
722 printf(" -o <name> set base name of output file\n");
725 (" -P[+/-] strings are \"packed\" by default (default=%c)\n",
726 sc_packstr ? '+' : '-');
727 printf(" -p<name> set name of \"prefix\" file\n");
732 (" -S <num> stack/heap size in cells (default=%d, min=65)\n",
735 printf(" -s<num> skip lines from the input file\n");
737 (" -t<num> TAB indent size (in character positions)\n");
738 printf(" -\\ use '\\' for escape characters\n");
739 printf(" -^ use '^' for escape characters\n");
740 printf(" -;[+/-] require a semicolon to end each statement\
741 (default=%c)\n", sc_needsemicolon ? '+' : '-');
744 (" sym=val define constant \"sym\" with value \"val\"\n");
745 printf(" sym= define constant \"sym\" with value 0\n");
747 longjmp(errbuf, 3); /* user abort */
755 assert(sc_status == statIDLE);
756 append_constval(&tagname_tab, "_", 0, 0); /* "untagged" */
757 append_constval(&tagname_tab, "bool", 1, 0);
759 add_constant("true", 1, sGLOBAL, 1); /* boolean flags */
760 add_constant("false", 0, sGLOBAL, 1);
761 add_constant("EOS", 0, sGLOBAL, 0); /* End Of String, or '\0' */
762 add_constant("cellbits", 32, sGLOBAL, 0);
763 add_constant("cellmax", INT_MAX, sGLOBAL, 0);
764 add_constant("cellmin", INT_MIN, sGLOBAL, 0);
765 add_constant("charbits", charbits, sGLOBAL, 0);
766 add_constant("charmin", 0, sGLOBAL, 0);
767 add_constant("charmax", (charbits == 16) ? 0xffff : 0xff, sGLOBAL, 0);
769 add_constant("__Small", VERSION_INT, sGLOBAL, 0);
772 if ((sc_debug & (sCHKBOUNDS | sSYMBOLIC)) == (sCHKBOUNDS | sSYMBOLIC))
774 else if ((sc_debug & sCHKBOUNDS) == sCHKBOUNDS)
776 add_constant("debug", debug, sGLOBAL, 0);
779 /* parse - process all input text
781 * At this level, only static declarations and function definitions
787 int tok, tag, fconst, fstock, fstatic;
793 /* first try whether a declaration possibly is native or public */
794 tok = lex(&val, &str); /* read in (new) token */
801 fconst = matchtoken(tCONST);
802 declglb(NULL, 0, FALSE, FALSE, FALSE, fconst);
805 /* This can be a static function or a static global variable;
806 * we know which of the two as soon as we have parsed up to the
807 * point where an opening parenthesis of a function would be
808 * expected. To back out after deciding it was a declaration of
809 * a static variable after all, we have to store the symbol name
812 fstock = matchtoken(tSTOCK);
813 fconst = matchtoken(tCONST);
814 tag = sc_addtag(NULL);
815 tok = lex(&val, &str);
816 if (tok == tNATIVE || tok == tPUBLIC)
818 error(42); /* invalid combination of class specifiers */
821 declfuncvar(tok, str, tag, FALSE, TRUE, fstock, fconst);
830 /* This can be a public function or a public variable;
831 * see the comment above (for static functions/variables)
834 fconst = matchtoken(tCONST);
835 tag = sc_addtag(NULL);
836 tok = lex(&val, &str);
837 if (tok == tNATIVE || tok == tSTOCK || tok == tSTATIC)
839 error(42); /* invalid combination of class specifiers */
842 declfuncvar(tok, str, tag, TRUE, FALSE, FALSE, fconst);
845 /* This can be a stock function or a stock *global) variable;
846 * see the comment above (for static functions/variables) for
849 fstatic = matchtoken(tSTATIC);
850 fconst = matchtoken(tCONST);
851 tag = sc_addtag(NULL);
852 tok = lex(&val, &str);
853 if (tok == tNATIVE || tok == tPUBLIC)
855 error(42); /* invalid combination of class specifiers */
858 declfuncvar(tok, str, tag, FALSE, fstatic, TRUE, fconst);
864 if (!newfunc(NULL, -1, FALSE, FALSE, FALSE))
866 error(10); /* illegal function or declaration */
867 lexclr(TRUE); /* drop the rest of the line */
871 funcstub(TRUE); /* create a dummy function */
877 error(54); /* unmatched closing brace */
880 error(55); /* start of function body without function header */
885 error(10); /* illegal function or declaration */
886 lexclr(TRUE); /* drop the rest of the line */
894 * Dump the literal pool (strings etc.)
896 * Global references: litidx (referred to only)
906 /* should be in the data segment */
909 j = 16; /* 16 values per line */
910 while (j && k < litidx)
912 outval(litq[k], FALSE);
916 if (j == 0 || k >= litidx)
917 stgwrite("\n"); /* force a newline after 10 dumps */
918 /* Note: stgwrite() buffers a line until it is complete. It recognizes
919 * the end of line as a sequence of "\n\0", so something like "\n\t"
920 * so should not be passed to stgwrite().
928 * Dump zero's for default initial values
944 stgwrite((i == 0 || count == 0) ? "\n" : " ");
945 if (i == 0 && count > 0)
951 aligndata(int numbytes)
953 if ((((glb_declared + litidx) * sizeof(cell)) % numbytes) != 0)
955 while ((((glb_declared + litidx) * sizeof(cell)) % numbytes) != 0)
962 declfuncvar(int tok, char *symname, int tag, int fpublic, int fstatic,
963 int fstock, int fconst)
965 char name[sNAMEMAX + 1];
967 if (tok != tSYMBOL && tok != tOPERATOR)
970 error(20, symname); /* invalid symbol name */
973 if (tok == tOPERATOR)
976 if (!newfunc(NULL, tag, fpublic, fstatic, fstock))
977 error(10); /* illegal function or declaration */
981 assert(strlen(symname) <= sNAMEMAX);
982 strcpy(name, symname);
983 if (fconst || !newfunc(name, tag, fpublic, fstatic, fstock))
984 declglb(name, tag, fpublic, fstatic, fstock, fconst);
985 /* if not a static function, try a static variable */
989 /* declglb - declare global symbols
991 * Declare a static (global) variable. Global variables are stored in
994 * global references: glb_declared (altered)
997 declglb(char *firstname, int firsttag, int fpublic, int fstatic,
998 int stock, int fconst)
1000 int ident, tag, ispublic;
1001 int idxtag[sDIMEN_MAX];
1002 char name[sNAMEMAX + 1];
1003 cell val, size, cidx;
1005 int dim[sDIMEN_MAX];
1014 filenum = fcurrent; /* save file number at the start of the
1018 size = 1; /* single size (no array) */
1019 numdim = 0; /* no dimensions */
1023 assert(strlen(firstname) <= sNAMEMAX);
1024 strcpy(name, firstname); /* save symbol name */
1030 tag = sc_addtag(NULL);
1031 if (lex(&val, &str) != tSYMBOL) /* read in (new) token */
1032 error(20, str); /* invalid symbol name */
1033 assert(strlen(str) <= sNAMEMAX);
1034 strcpy(name, str); /* save symbol name */
1036 sym = findglb(name);
1038 sym = findconst(name);
1039 if (sym && (sym->usage & uDEFINE) != 0)
1040 error(21, name); /* symbol already defined */
1042 if (name[0] == PUBLIC_CHAR)
1044 ispublic = TRUE; /* implicitly public variable */
1045 if (stock || fstatic)
1046 error(42); /* invalid combination of class specifiers */
1048 while (matchtoken('['))
1051 if (numdim == sDIMEN_MAX)
1053 error(53); /* exceeding maximum number of dimensions */
1056 if (numdim > 0 && dim[numdim - 1] == 0)
1057 error(52); /* only last dimension may be variable length */
1058 size = needsub(&idxtag[numdim]); /* get size; size==0 for
1060 #if INT_MAX < LONG_MAX
1062 error(105); /* overflow, exceeding capacity */
1065 error(56, name); /* arrays cannot be public */
1066 dim[numdim++] = (int)size;
1068 /* if this variable is never used (which can be detected only in
1069 * the second stage), shut off code generation; make an exception
1070 * for public variables
1072 cidx = 0; /* only to avoid a compiler warning */
1073 if (sc_status == statWRITE && sym
1074 && (sym->usage & (uREAD | uWRITTEN | uPUBLIC)) == 0)
1076 sc_status = statSKIP;
1079 glbdecl = glb_declared;
1082 defsymbol(name, ident, sGLOBAL, sizeof(cell) * glb_declared, tag);
1083 begdseg(); /* real (initialized) data in data segment */
1084 assert(litidx == 0); /* literal queue should be empty */
1088 aligndata(sc_dataalign);
1089 dumplits(); /* dump the literal queue */
1090 sc_alignnext = FALSE;
1091 litidx = 0; /* global initial data is dumped, so restart at zero */
1093 initials(ident, tag, &size, dim, numdim); /* stores values in
1094 * the literal queue */
1097 dumplits(); /* dump the literal queue */
1098 dumpzero((int)size - litidx);
1101 { /* define only if not yet defined */
1103 addvariable(name, sizeof(cell) * glb_declared, ident, sGLOBAL,
1104 tag, dim, numdim, idxtag);
1107 { /* if declared but not yet defined, adjust the
1108 * variable's address */
1109 sym->addr = sizeof(cell) * glb_declared;
1110 sym->usage |= uDEFINE;
1113 sym->usage |= uPUBLIC;
1115 sym->usage |= uCONST;
1117 sym->usage |= uSTOCK;
1119 sym->fnumber = filenum;
1120 if (ident == iARRAY)
1121 for (level = 0; level < numdim; level++)
1122 symbolrange(level, dim[level]);
1123 if (sc_status == statSKIP)
1125 sc_status = statWRITE;
1127 assert(glb_declared == glbdecl);
1131 glb_declared += (int)size; /* add total number of cells */
1134 while (matchtoken(',')); /* enddo *//* more? */
1135 needtoken(tTERM); /* if not comma, must be semicolumn */
1138 /* declloc - declare local symbols
1140 * Declare local (automatic) variables. Since these variables are
1141 * relative to the STACK, there is no switch to the DATA segment.
1142 * These variables cannot be initialized either.
1144 * global references: declared (altered)
1145 * funcstatus (referred to only)
1148 declloc(int fstatic)
1151 int idxtag[sDIMEN_MAX];
1152 char name[sNAMEMAX + 1];
1156 value lval = { NULL, 0, 0, 0, 0, NULL };
1158 int dim[sDIMEN_MAX];
1162 fconst = matchtoken(tCONST);
1167 numdim = 0; /* no dimensions */
1168 tag = sc_addtag(NULL);
1169 if (lex(&val, &str) != tSYMBOL) /* read in (new) token */
1170 error(20, str); /* invalid symbol name */
1171 assert(strlen(str) <= sNAMEMAX);
1172 strcpy(name, str); /* save symbol name */
1173 if (name[0] == PUBLIC_CHAR)
1174 error(56, name); /* local variables cannot be public */
1175 /* Note: block locals may be named identical to locals at higher
1176 * compound blocks (as with standard C); so we must check (and add)
1177 * the "nesting level" of local variables to verify the
1178 * multi-definition of symbols.
1180 if ((sym = findloc(name)) && sym->compound == nestlevel)
1181 error(21, name); /* symbol already defined */
1182 /* Although valid, a local variable whose name is equal to that
1183 * of a global variable or to that of a local variable at a lower
1184 * level might indicate a bug.
1186 if (((sym = findloc(name)) && sym->compound != nestlevel)
1188 error(219, name); /* variable shadows another symbol */
1189 while (matchtoken('['))
1192 if (numdim == sDIMEN_MAX)
1194 error(53); /* exceeding maximum number of dimensions */
1197 if (numdim > 0 && dim[numdim - 1] == 0)
1198 error(52); /* only last dimension may be variable length */
1199 size = needsub(&idxtag[numdim]); /* get size; size==0 for "var[]" */
1200 #if INT_MAX < LONG_MAX
1202 error(105); /* overflow, exceeding capacity */
1204 dim[numdim++] = (int)size;
1206 if (ident == iARRAY || fstatic)
1210 aligndata(sc_dataalign);
1211 sc_alignnext = FALSE;
1213 cur_lit = litidx; /* save current index in the literal table */
1214 initials(ident, tag, &size, dim, numdim);
1216 return ident; /* error message already given */
1220 /* reserve memory (on the stack) for the variable */
1223 /* write zeros for uninitialized fields */
1224 while (litidx < cur_lit + size)
1227 addvariable(name, (cur_lit + glb_declared) * sizeof(cell),
1228 ident, sSTATIC, tag, dim, numdim, idxtag);
1229 defsymbol(name, ident, sSTATIC,
1230 (cur_lit + glb_declared) * sizeof(cell), tag);
1234 declared += (int)size; /* variables are put on stack,
1235 * adjust "declared" */
1237 addvariable(name, -declared * sizeof(cell), ident, sLOCAL, tag,
1238 dim, numdim, idxtag);
1239 defsymbol(name, ident, sLOCAL, -declared * sizeof(cell), tag);
1240 modstk(-(int)size * sizeof(cell));
1242 /* now that we have reserved memory for the variable, we can
1243 * proceed to initialize it */
1244 sym->compound = nestlevel; /* for multiple declaration/shadowing */
1246 sym->usage |= uCONST;
1247 if (ident == iARRAY)
1248 for (level = 0; level < numdim; level++)
1249 symbolrange(level, dim[level]);
1251 { /* static variables already initialized */
1252 if (ident == iVARIABLE)
1254 /* simple variable, also supports initialization */
1255 int ctag = tag; /* set to "tag" by default */
1256 int explicit_init = FALSE; /* is the variable explicitly
1258 if (matchtoken('='))
1260 doexpr(FALSE, FALSE, FALSE, FALSE, &ctag, TRUE);
1261 explicit_init = TRUE;
1265 const1(0); /* uninitialized variable, set to zero */
1267 /* now try to save the value (still in PRI) in the variable */
1269 lval.ident = iVARIABLE;
1272 check_userop(NULL, ctag, lval.tag, 2, NULL, &ctag);
1274 endexpr(TRUE); /* full expression ends after the store */
1275 if (!matchtag(tag, ctag, TRUE))
1276 error(213); /* tag mismatch */
1277 /* if the variable was not explicitly initialized, reset the
1278 * "uWRITTEN" flag that store() set */
1280 sym->usage &= ~uWRITTEN;
1285 if (litidx - cur_lit < size)
1286 fillarray(sym, size * sizeof(cell), 0);
1287 if (cur_lit < litidx)
1289 /* check whether the complete array is set to a single value;
1290 * if it is, more compact code can be generated */
1291 cell first = litq[cur_lit];
1294 for (i = cur_lit; i < litidx && litq[i] == first; i++)
1298 /* all values are the same */
1299 fillarray(sym, (litidx - cur_lit) * sizeof(cell),
1301 litidx = cur_lit; /* reset literal table */
1305 /* copy the literals to the array */
1306 const1((cur_lit + glb_declared) * sizeof(cell));
1307 copyarray(sym, (litidx - cur_lit) * sizeof(cell));
1313 while (matchtoken(',')); /* enddo *//* more? */
1314 needtoken(tTERM); /* if not comma, must be semicolumn */
1319 calc_arraysize(int dim[], int numdim, int cur)
1323 return dim[cur] + (dim[cur] * calc_arraysize(dim, numdim, cur + 1));
1328 * Initialize global objects and local arrays.
1329 * size==array cells (count), if 0 on input, the routine counts
1330 * the number of elements
1331 * tag==required tagname id (not the returned tag)
1333 * Global references: litidx (altered)
1336 initials(int ident, int tag, cell * size, int dim[], int numdim)
1339 int curlit = litidx;
1342 if (!matchtoken('='))
1344 if (ident == iARRAY && dim[numdim - 1] == 0)
1346 /* declared as "myvar[];" which is senseless (note: this *does* make
1347 * sense in the case of a iREFARRAY, which is a function parameter)
1349 error(9); /* array has zero length -> invalid size */
1353 /* initialize the indirection tables */
1355 #error Array algorithms for more than 2 dimensions are not implemented
1357 assert(numdim == 2);
1358 *size = calc_arraysize(dim, numdim, 0);
1359 for (d = 0; d < dim[0]; d++)
1360 stowlit((dim[0] + d * (dim[1] - 1)) * sizeof(cell));
1365 if (ident == iVARIABLE)
1369 if (!matchtag(tag, ctag, TRUE))
1370 error(213); /* tag mismatch */
1377 *size = initvector(ident, tag, dim[0], FALSE);
1383 /* The simple algorithm below only works for arrays with one or
1384 * two dimensions. This should be some recursive algorithm.
1386 if (dim[numdim - 1] != 0)
1387 /* set size to (known) full size */
1388 *size = calc_arraysize(dim, numdim, 0);
1389 /* dump indirection tables */
1390 for (d = 0; d < dim[0]; d++)
1392 /* now dump individual vectors */
1395 for (d = 0; d < dim[0]; d++)
1397 litq[curlit + d] = offs * sizeof(cell);
1398 dsize = initvector(ident, tag, dim[1], TRUE);
1402 if (matchtoken('{') || matchtoken(tSTRING))
1403 /* expect a '{' or a string */
1414 *size = litidx - curlit; /* number of elements defined */
1418 * Initialize a single dimensional array
1421 initvector(int ident, int tag, cell size, int fillzero)
1423 cell prev1 = 0, prev2 = 0;
1426 int curlit = litidx;
1428 assert(ident == iARRAY || ident == iREFARRAY);
1429 if (matchtoken('{'))
1433 if (matchtoken('}'))
1434 { /* to allow for trailing ',' after the initialization */
1438 if ((ellips = matchtoken(tELLIPS)) != 0)
1441 prev1 = init(ident, &ctag);
1442 if (!matchtag(tag, ctag, TRUE))
1443 error(213); /* tag mismatch */
1445 while (matchtoken(',')); /* do */
1451 if (!matchtag(tag, ctag, TRUE))
1452 error(213); /* tagname mismatch */
1454 /* fill up the literal queue with a series */
1458 ((litidx - curlit) == 1) ? (cell) 0 : prev1 - prev2;
1459 if (size == 0 || (litidx - curlit) == 0)
1460 error(41); /* invalid ellipsis, array size unknown */
1461 else if ((litidx - curlit) == (int)size)
1462 error(18); /* initialisation data exceeds declared size */
1463 while ((litidx - curlit) < (int)size)
1469 if (fillzero && size > 0)
1471 while ((litidx - curlit) < (int)size)
1476 size = litidx - curlit; /* number of elements defined */
1478 else if (litidx - curlit > (int)size)
1479 { /* e.g. "myvar[3]={1,2,3,4};" */
1480 error(18); /* initialisation data exceeds declared size */
1481 litidx = (int)size + curlit; /* avoid overflow in memory moves */
1488 * Evaluate one initializer.
1491 init(int ident, int *tag)
1495 if (matchtoken(tSTRING))
1497 /* lex() automatically stores strings in the literal table (and
1498 * increases "litidx") */
1499 if (ident == iVARIABLE)
1501 error(6); /* must be assigned to an array */
1502 litidx = 1; /* reset literal queue */
1506 else if (constexpr(&i, tag))
1508 stowlit(i); /* store expression result in literal table */
1515 * Get required array size
1523 if (matchtoken(']')) /* we've already seen "[" */
1524 return 0; /* null size (like "char msg[]") */
1525 constexpr(&val, tag); /* get value (must be constant expression) */
1528 error(9); /* negative array size is invalid; assumed zero */
1532 return val; /* return array size */
1535 /* decl_const - declare a single constant
1539 decl_const(int vclass)
1541 char constname[sNAMEMAX + 1];
1547 tag = sc_addtag(NULL);
1548 if (lex(&val, &str) != tSYMBOL) /* read in (new) token */
1549 error(20, str); /* invalid symbol name */
1550 symbolline = fline; /* save line where symbol was found */
1551 strcpy(constname, str); /* save symbol name */
1553 constexpr(&val, &exprtag); /* get value */
1555 /* add_constant() checks for duplicate definitions */
1556 if (!matchtag(tag, exprtag, FALSE))
1558 /* temporarily reset the line number to where the symbol was
1560 int orgfline = fline;
1563 error(213); /* tagname mismatch */
1566 add_constant(constname, val, vclass, tag);
1569 /* decl_enum - declare enumerated constants
1573 decl_enum(int vclass)
1575 char enumname[sNAMEMAX + 1], constname[sNAMEMAX + 1];
1576 cell val, value, size;
1578 int tok, tag, explicittag;
1579 cell increment, multiplier;
1581 /* get an explicit tag, if any (we need to remember whether an
1582 * explicit tag was passed, even if that explicit tag was "_:", so we
1583 * cannot call sc_addtag() here
1585 if (lex(&val, &str) == tLABEL)
1587 tag = sc_addtag(str);
1594 explicittag = FALSE;
1597 /* get optional enum name (also serves as a tag if no explicit
1599 if (lex(&val, &str) == tSYMBOL)
1600 { /* read in (new) token */
1601 strcpy(enumname, str); /* save enum name (last constant) */
1603 tag = sc_addtag(enumname);
1607 lexpush(); /* analyze again */
1611 /* get increment and multiplier */
1614 if (matchtoken('('))
1616 if (matchtoken(taADD))
1618 constexpr(&increment, NULL);
1620 else if (matchtoken(taMULT))
1622 constexpr(&multiplier, NULL);
1624 else if (matchtoken(taSHL))
1626 constexpr(&val, NULL);
1634 /* go through all constants */
1635 value = 0; /* default starting value */
1638 if (matchtoken('}'))
1639 { /* quick exit if '}' follows ',' */
1643 tok = lex(&val, &str); /* read in (new) token */
1644 if (tok != tSYMBOL && tok != tLABEL)
1645 error(20, str); /* invalid symbol name */
1646 strcpy(constname, str); /* save symbol name */
1647 size = increment; /* default increment of 'val' */
1648 if (tok == tLABEL || matchtoken(':'))
1649 constexpr(&size, NULL); /* get size */
1650 if (matchtoken('='))
1651 constexpr(&value, NULL); /* get value */
1652 /* add_constant() checks whether a variable (global or local) or
1653 * a constant with the same name already exists */
1654 add_constant(constname, value, vclass, tag);
1655 if (multiplier == 1)
1658 value *= size * multiplier;
1660 while (matchtoken(','));
1661 needtoken('}'); /* terminates the constant list */
1662 matchtoken(';'); /* eat an optional ; */
1664 /* set the enum name to the last value plus one */
1665 if (enumname[0] != '\0')
1666 add_constant(enumname, value, vclass, tag);
1670 * Finds a function in the global symbol table or creates a new entry.
1671 * It does some basic processing and error checking.
1674 fetchfunc(char *name, int tag)
1680 if ((sc_debug & sSYMBOLIC) != 0)
1682 offset += opcodes(1) + opargs(3) + nameincells(name);
1683 /* ^^^ The address for the symbol is the code address. But the
1684 * "symbol" instruction itself generates code. Therefore the
1685 * offset is pre-adjusted to the value it will have after the
1686 * symbol instruction.
1689 if ((sym = findglb(name)))
1690 { /* already in symbol table? */
1691 if (sym->ident != iFUNCTN)
1693 error(21, name); /* yes, but not as a function */
1694 return NULL; /* make sure the old symbol is not damaged */
1696 else if ((sym->usage & uDEFINE) != 0)
1698 error(21, name); /* yes, and it's already defined */
1700 else if ((sym->usage & uNATIVE) != 0)
1702 error(21, name); /* yes, and it is an native */
1704 assert(sym->vclass == sGLOBAL);
1705 if ((sym->usage & uDEFINE) == 0)
1707 /* as long as the function stays undefined, update the address
1715 /* don't set the "uDEFINE" flag; it may be a prototype */
1716 sym = addsym(name, offset, iFUNCTN, sGLOBAL, tag, 0);
1717 /* assume no arguments */
1718 sym->dim.arglist = (arginfo *) malloc(1 * sizeof(arginfo));
1719 sym->dim.arglist[0].ident = 0;
1720 /* set library ID to NULL (only for native functions) */
1726 /* This routine adds symbolic information for each argument.
1733 /* At this point, no local variables have been declared. All
1734 * local symbols are function arguments.
1739 assert(sym->ident != iLABEL);
1740 assert(sym->vclass == sLOCAL);
1741 defsymbol(sym->name, sym->ident, sLOCAL, sym->addr, sym->tag);
1742 if (sym->ident == iREFARRAY)
1748 symbolrange(sub->dim.array.level, sub->dim.array.length);
1749 sub = finddepend(sub);
1757 operatorname(char *name)
1763 assert(name != NULL);
1765 /* check the operator */
1766 opertok = lex(&val, &str);
1779 name[0] = (char)opertok;
1802 error(61); /* operator cannot be redefined
1803 * (or bad operator name) */
1811 operatoradjust(int opertok, symbol * sym, char *opername, int resulttag)
1813 int tags[2] = { 0, 0 };
1816 char tmpname[sNAMEMAX + 1];
1822 /* count arguments and save (first two) tags */
1823 while (arg = &sym->dim.arglist[count], arg->ident != 0)
1827 if (arg->numtags > 1)
1828 error(65, count + 1); /* function argument may only have
1830 else if (arg->numtags == 1)
1831 tags[count] = arg->tags[0];
1833 if (opertok == '~' && count == 0)
1835 if (arg->ident != iREFARRAY)
1836 error(73, arg->name); /* must be an array argument */
1840 if (arg->ident != iVARIABLE)
1841 error(66, arg->name); /* must be non-reference argument */
1843 if (arg->hasdefault)
1844 error(59, arg->name); /* arguments of an operator may not
1845 * have a default value */
1849 /* for '!', '++' and '--', count must be 1
1850 * for '-', count may be 1 or 2
1851 * for '=', count must be 1, and the resulttag is also important
1852 * for all other (binary) operators and the special '~'
1853 * operator, count must be 2
1862 error(62); /* number or placement of the operands does
1863 * not fit the operator */
1866 if (count != 1 && count != 2)
1867 error(62); /* number or placement of the operands does
1868 * not fit the operator */
1872 error(62); /* number or placement of the operands does
1873 * not fit the operator */
1877 && ((opertok != '=' && tags[1] == 0) || (opertok == '=' && resulttag == 0)))
1878 error(64); /* cannot change predefined operators */
1880 /* change the operator name */
1881 assert(opername[0] != '\0');
1882 operator_symname(tmpname, opername, tags[0], tags[1], count, resulttag);
1883 if ((oldsym = findglb(tmpname)))
1887 if ((oldsym->usage & uDEFINE) != 0)
1889 char errname[2 * sNAMEMAX + 16];
1891 funcdisplayname(errname, tmpname);
1892 error(21, errname); /* symbol already defined */
1894 sym->usage |= oldsym->usage; /* copy flags from the previous
1896 for (i = 0; i < oldsym->numrefers; i++)
1897 if (oldsym->refer[i])
1898 refer_symbol(sym, oldsym->refer[i]);
1899 delete_symbol(&glbtab, oldsym);
1901 if ((sc_debug & sSYMBOLIC) != 0)
1902 sym->addr += nameincells(tmpname) - nameincells(sym->name);
1903 strcpy(sym->name, tmpname);
1904 sym->hash = namehash(sym->name); /* calculate new hash */
1906 /* operators should return a value, except the '~' operator */
1908 sym->usage |= uRETVALUE;
1914 check_operatortag(int opertok, int resulttag, char *opername)
1916 assert(opername != NULL && opername[0] != '\0');
1926 if (resulttag != sc_addtag("bool"))
1928 error(63, opername, "bool:"); /* operator X requires
1929 * a "bool:" result tag */
1936 error(63, opername, "_:"); /* operator "~" requires
1937 * a "_:" result tag */
1946 tag2str(char *dest, int tag)
1950 sprintf(dest, "0%x", tag);
1951 return isdigit(dest[1]) ? &dest[1] : dest;
1955 operator_symname(char *symname, char *opername, int tag1, int tag2,
1956 int numtags, int resulttag)
1958 char tagstr1[10], tagstr2[10];
1961 assert(numtags >= 1 && numtags <= 2);
1962 opertok = (opername[1] == '\0') ? opername[0] : 0;
1964 sprintf(symname, "%s%s%s", tag2str(tagstr1, resulttag), opername,
1965 tag2str(tagstr2, tag1));
1966 else if (numtags == 1 || opertok == '~')
1967 sprintf(symname, "%s%s", opername, tag2str(tagstr1, tag1));
1969 sprintf(symname, "%s%s%s", tag2str(tagstr1, tag1), opername,
1970 tag2str(tagstr2, tag2));
1975 parse_funcname(char *fname, int *tag1, int *tag2, char *opname)
1980 /* tags are only positive, so if the function name starts with a '-',
1981 * the operator is an unary '-' or '--' operator.
1991 *tag1 = (int)strtol(fname, &ptr, 16);
1992 unary = ptr == fname; /* unary operator if it doesn't start
1993 * with a tag name */
1995 assert(!unary || *tag1 == 0);
1996 assert(*ptr != '\0');
1997 for (name = opname; !isdigit(*ptr);)
2000 *tag2 = (int)strtol(ptr, NULL, 16);
2005 funcdisplayname(char *dest, char *funcname)
2009 constvalue *tagsym[2];
2012 if (isalpha(*funcname) || *funcname == '_' || *funcname == PUBLIC_CHAR
2013 || *funcname == '\0')
2015 if (dest != funcname)
2016 strcpy(dest, funcname);
2020 unary = parse_funcname(funcname, &tags[0], &tags[1], opname);
2021 tagsym[1] = find_constval_byval(&tagname_tab, tags[1]);
2022 assert(tagsym[1] != NULL);
2025 sprintf(dest, "operator%s(%s:)", opname, tagsym[1]->name);
2029 tagsym[0] = find_constval_byval(&tagname_tab, tags[0]);
2030 /* special case: the assignment operator has the return value
2032 if (opname[0] == '=' && opname[1] == '\0')
2033 sprintf(dest, "%s:operator%s(%s:)", tagsym[0]->name, opname,
2036 sprintf(dest, "operator%s(%s:,%s:)", opname, tagsym[0]->name,
2043 funcstub(int native)
2048 char symbolname[sNAMEMAX + 1];
2054 litidx = 0; /* clear the literal pool */
2056 tag = sc_addtag(NULL);
2057 tok = lex(&val, &str);
2060 if (tok == tPUBLIC || tok == tSTOCK || tok == tSTATIC ||
2061 (tok == tSYMBOL && *str == PUBLIC_CHAR))
2062 error(42); /* invalid combination of class specifiers */
2066 if (tok == tPUBLIC || tok == tSTATIC)
2067 tok = lex(&val, &str);
2069 if (tok == tOPERATOR)
2071 opertok = operatorname(symbolname);
2073 return; /* error message already given */
2074 check_operatortag(opertok, tag, symbolname);
2078 if (tok != tSYMBOL && freading)
2080 error(10); /* illegal function or declaration */
2083 strcpy(symbolname, str);
2085 needtoken('('); /* only functions may be native/forward */
2087 sym = fetchfunc(symbolname, tag); /* get a pointer to the
2093 sym->usage = uNATIVE | uRETVALUE | uDEFINE;
2094 sym->x.lib = curlibrary;
2098 /* "declargs()" found the ")" */
2099 if (!operatoradjust(opertok, sym, symbolname, tag))
2100 sym->usage &= ~uDEFINE;
2101 /* for a native operator, also need to specify an "exported"
2102 * function name; for a native function, this is optional
2109 lexpush(); /* push back, for matchtoken() to retrieve again */
2111 if (matchtoken('='))
2113 /* allow number or symbol */
2114 if (matchtoken(tSYMBOL))
2116 tokeninfo(&val, &str);
2117 if (strlen(str) > sEXPMAX)
2119 error(220, str, sEXPMAX);
2120 str[sEXPMAX] = '\0';
2122 insert_alias(sym->name, str);
2126 constexpr(&val, NULL);
2129 * ?? Must mark this address, so that it won't be generated again
2130 * and it won't be written to the output file. At the moment,
2131 * I have assumed that this syntax is only valid if val < 0.
2132 * To properly mix "normal" native functions and indexed native
2133 * functions, one should use negative indices anyway.
2134 * Special code for a negative index in sym->addr exists in
2135 * SC4.C (ffcall()) and in SC6.C (the loops for counting the
2136 * number of native variables and for writing them).
2143 litidx = 0; /* clear the literal pool */
2144 /* clear local variables queue */
2145 delete_symbols(&loctab, 0, TRUE, TRUE);
2148 /* newfunc - begin a function
2150 * This routine is called from "parse" and tries to make a function
2151 * out of the following text
2153 * Global references: funcstatus,lastst,litidx
2156 * declared (altered)
2157 * glb_declared (altered)
2158 * sc_alignnext (altered)
2161 newfunc(char *firstname, int firsttag, int fpublic, int fstatic, int stock)
2164 int argcnt, tok, tag, funcline;
2165 int opertok, opererror;
2166 char symbolname[sNAMEMAX + 1];
2168 cell val, cidx, glbdecl;
2171 litidx = 0; /* clear the literal pool ??? */
2173 lastst = 0; /* no statement yet */
2174 cidx = 0; /* just to avoid compiler warnings */
2176 filenum = fcurrent; /* save file number at start of declaration */
2180 assert(strlen(firstname) <= sNAMEMAX);
2181 strcpy(symbolname, firstname); /* save symbol name */
2186 tag = (firsttag >= 0) ? firsttag : sc_addtag(NULL);
2187 tok = lex(&val, &str);
2189 if (tok == tNATIVE || (tok == tPUBLIC && stock))
2190 error(42); /* invalid combination of class specifiers */
2191 if (tok == tOPERATOR)
2193 opertok = operatorname(symbolname);
2195 return TRUE; /* error message already given */
2196 check_operatortag(opertok, tag, symbolname);
2200 if (tok != tSYMBOL && freading)
2202 error(20, str); /* invalid symbol name */
2205 assert(strlen(str) <= sNAMEMAX);
2206 strcpy(symbolname, str);
2209 /* check whether this is a function or a variable declaration */
2210 if (!matchtoken('('))
2212 /* so it is a function, proceed */
2213 funcline = fline; /* save line at which the function is defined */
2214 if (symbolname[0] == PUBLIC_CHAR)
2216 fpublic = TRUE; /* implicitly public function */
2218 error(42); /* invalid combination of class specifiers */
2220 sym = fetchfunc(symbolname, tag); /* get a pointer to the
2225 sym->usage |= uPUBLIC;
2227 sym->fnumber = filenum;
2228 /* declare all arguments */
2229 argcnt = declargs(sym);
2230 opererror = !operatoradjust(opertok, sym, symbolname, tag);
2231 if (strcmp(symbolname, uMAINFUNC) == 0)
2234 error(5); /* "main()" function may not have any arguments */
2235 sym->usage |= uREAD; /* "main()" is the program's entry point:
2238 /* "declargs()" found the ")"; if a ";" appears after this, it was a
2240 if (matchtoken(';'))
2242 if (!sc_needsemicolon)
2243 error(218); /* old style prototypes used with optional
2245 delete_symbols(&loctab, 0, TRUE, TRUE); /* prototype is done;
2246 * forget everything */
2249 /* so it is not a prototype, proceed */
2250 /* if this is a function that is not referred to (this can only be
2251 * detected in the second stage), shut code generation off */
2252 if (sc_status == statWRITE && (sym->usage & uREAD) == 0)
2254 sc_status = statSKIP;
2256 glbdecl = glb_declared;
2259 sym->usage |= uDEFINE; /* set the definition flag */
2261 sym->usage |= uREAD; /* public functions are always "used" */
2263 sym->usage |= uSTOCK;
2264 if (opertok != 0 && opererror)
2265 sym->usage &= ~uDEFINE;
2266 defsymbol(sym->name, iFUNCTN, sGLOBAL,
2267 code_idx + opcodes(1) + opargs(3) + nameincells(sym->name), tag);
2268 /* ^^^ The address for the symbol is the code address. But the
2269 * "symbol" instruction itself generates code. Therefore the
2270 * offset is pre-adjusted to the value it will have after the
2271 * symbol instruction.
2273 startfunc(sym->name); /* creates stack frame */
2274 if ((sc_debug & sSYMBOLIC) != 0)
2275 setline(funcline, fcurrent);
2278 alignframe(sc_dataalign);
2279 sc_alignnext = FALSE;
2281 declared = 0; /* number of local cells */
2282 rettype = (sym->usage & uRETVALUE); /* set "return type" variable */
2284 define_args(); /* add the symbolic info for the function arguments */
2285 statement(NULL, FALSE);
2286 if ((rettype & uRETVALUE) != 0)
2287 sym->usage |= uRETVALUE;
2290 /* This happens only in a very special (and useless) case, where a
2291 * function has only a single statement in its body (no compound
2292 * block) and that statement declares a new variable
2294 modstk((int)declared * sizeof(cell)); /* remove all local
2298 if ((lastst != tRETURN) && (lastst != tGOTO))
2302 if ((sym->usage & uRETVALUE) != 0)
2304 char symname[2 * sNAMEMAX + 16]; /* allow space for user
2305 * defined operators */
2306 funcdisplayname(symname, sym->name);
2307 error(209, symname); /* function should return a value */
2312 { /* if there are literals defined */
2313 glb_declared += litidx;
2314 begdseg(); /* flip to DATA segment */
2315 dumplits(); /* dump literal strings */
2318 testsymbols(&loctab, 0, TRUE, TRUE); /* test for unused arguments
2320 delete_symbols(&loctab, 0, TRUE, TRUE); /* clear local variables
2322 assert(loctab.next == NULL);
2324 if (sc_status == statSKIP)
2326 sc_status = statWRITE;
2328 glb_declared = glbdecl;
2334 argcompare(arginfo * a1, arginfo * a2)
2338 result = strcmp(a1->name, a2->name) == 0;
2340 result = a1->ident == a2->ident;
2342 result = a1->usage == a2->usage;
2344 result = a1->numtags == a2->numtags;
2349 for (i = 0; i < a1->numtags && result; i++)
2350 result = a1->tags[i] == a2->tags[i];
2353 result = a1->hasdefault == a2->hasdefault;
2356 if (a1->ident == iREFARRAY)
2359 result = a1->defvalue.array.size == a2->defvalue.array.size;
2362 a1->defvalue.array.arraysize == a2->defvalue.array.arraysize;
2363 /* also check the dimensions of both arrays */
2365 result = a1->numdim == a2->numdim;
2366 for (level = 0; result && level < a1->numdim; level++)
2367 result = a1->dim[level] == a2->dim[level];
2368 /* ??? should also check contents of the default array
2369 * (these troubles go away in a 2-pass compiler that forbids
2370 * double declarations, but Small currently does not forbid them)
2377 if ((a1->hasdefault & uSIZEOF) != 0
2378 || (a1->hasdefault & uTAGOF) != 0)
2379 result = a1->hasdefault == a2->hasdefault
2380 && strcmp(a1->defvalue.size.symname,
2381 a2->defvalue.size.symname) == 0
2382 && a1->defvalue.size.level == a2->defvalue.size.level;
2384 result = a1->defvalue.val == a2->defvalue.val;
2388 result = a1->defvalue_tag == a2->defvalue_tag;
2395 * This routine adds an entry in the local symbol table for each
2396 * argument found in the argument list.
2397 * It returns the number of arguments.
2400 declargs(symbol * sym)
2404 int argcnt, oldargcnt, tok, tags[MAXTAGS], numtags;
2406 arginfo arg, *arglist;
2407 char name[sNAMEMAX + 1];
2408 int ident, fpublic, fconst;
2411 /* if the function is already defined earlier, get the number of
2412 * arguments of the existing definition
2415 if ((sym->usage & uPROTOTYPED) != 0)
2416 while (sym->dim.arglist[oldargcnt].ident != 0)
2418 argcnt = 0; /* zero aruments up to now */
2422 fpublic = (sym->usage & uPUBLIC) != 0;
2423 /* the '(' parantheses has already been parsed */
2424 if (!matchtoken(')'))
2427 { /* there are arguments; process them */
2428 /* any legal name increases argument count (and stack offset) */
2429 tok = lex(&val, &ptr);
2436 if (ident != iVARIABLE || numtags > 0)
2437 error(1, "-identifier-", "&");
2441 if (ident != iVARIABLE || numtags > 0)
2442 error(1, "-identifier-", "const");
2447 error(1, "-identifier-", "-tagname-");
2448 tags[0] = sc_addtag(ptr);
2453 error(1, "-identifier-", "-tagname-");
2455 while (numtags < MAXTAGS)
2457 if (!matchtoken('_') && !needtoken(tSYMBOL))
2459 tokeninfo(&val, &ptr);
2460 tags[numtags++] = sc_addtag(ptr);
2461 if (matchtoken('}'))
2466 tok = tLABEL; /* for outer loop:
2467 * flag that we have seen a tagname */
2470 if (argcnt >= sMAXARGS)
2471 error(45); /* too many function arguments */
2472 strcpy(name, ptr); /* save symbol name */
2473 if (name[0] == PUBLIC_CHAR)
2474 error(56, name); /* function arguments cannot be public */
2476 tags[numtags++] = 0; /* default tag */
2478 * base + 0*sizeof(cell) == previous "base"
2479 * base + 1*sizeof(cell) == function return address
2480 * base + 2*sizeof(cell) == number of arguments
2481 * base + 3*sizeof(cell) == first argument of the function
2482 * So the offset of each argument is:
2483 * "(argcnt+3) * sizeof(cell)".
2485 doarg(name, ident, (argcnt + 3) * sizeof(cell), tags, numtags,
2486 fpublic, fconst, &arg);
2487 if (fpublic && arg.hasdefault)
2488 error(59, name); /* arguments of a public function may not
2489 * have a default value */
2490 if ((sym->usage & uPROTOTYPED) == 0)
2492 /* redimension the argument list, add the entry */
2494 (arginfo *) realloc(sym->dim.arglist,
2495 (argcnt + 2) * sizeof(arginfo));
2496 if (!sym->dim.arglist)
2497 error(103); /* insufficient memory */
2498 sym->dim.arglist[argcnt] = arg;
2499 sym->dim.arglist[argcnt + 1].ident = 0; /* keep the list
2504 /* check the argument with the earlier definition */
2505 if (argcnt > oldargcnt
2506 || !argcompare(&sym->dim.arglist[argcnt], &arg))
2507 error(25); /* function definition does not match prototype */
2508 /* may need to free default array argument and the tag list */
2509 if (arg.ident == iREFARRAY && arg.hasdefault)
2510 free(arg.defvalue.array.data);
2511 else if (arg.ident == iVARIABLE
2512 && ((arg.hasdefault & uSIZEOF) != 0
2513 || (arg.hasdefault & uTAGOF) != 0))
2514 free(arg.defvalue.size.symname);
2523 if (ident != iVARIABLE)
2524 error(10); /* illegal function or declaration */
2526 tags[numtags++] = 0; /* default tag */
2527 if ((sym->usage & uPROTOTYPED) == 0)
2529 /* redimension the argument list, add the entry iVARARGS */
2531 (arginfo *) realloc(sym->dim.arglist,
2532 (argcnt + 2) * sizeof(arginfo));
2533 if (!sym->dim.arglist)
2534 error(103); /* insufficient memory */
2535 sym->dim.arglist[argcnt + 1].ident = 0; /* keep the list
2537 sym->dim.arglist[argcnt].ident = iVARARGS;
2538 sym->dim.arglist[argcnt].hasdefault = FALSE;
2539 sym->dim.arglist[argcnt].defvalue.val = 0;
2540 sym->dim.arglist[argcnt].defvalue_tag = 0;
2541 sym->dim.arglist[argcnt].numtags = numtags;
2542 sym->dim.arglist[argcnt].tags =
2543 (int *)malloc(numtags * sizeof tags[0]);
2544 if (!sym->dim.arglist[argcnt].tags)
2545 error(103); /* insufficient memory */
2546 memcpy(sym->dim.arglist[argcnt].tags, tags,
2547 numtags * sizeof tags[0]);
2551 if (argcnt > oldargcnt
2552 || sym->dim.arglist[argcnt].ident != iVARARGS)
2553 error(25); /* function definition does not match prototype */
2558 error(10); /* illegal function or declaration */
2561 while (tok == '&' || tok == tLABEL || tok == tCONST || (tok != tELLIPS && matchtoken(','))); /* more? */
2562 /* if the next token is not ",", it should be ")" */
2565 /* resolve any "sizeof" arguments (now that all arguments are known) */
2566 assert(sym->dim.arglist != NULL);
2567 arglist = sym->dim.arglist;
2568 for (idx = 0; idx < argcnt && arglist[idx].ident != 0; idx++)
2570 if ((arglist[idx].hasdefault & uSIZEOF) != 0
2571 || (arglist[idx].hasdefault & uTAGOF) != 0)
2575 /* Find the argument with the name mentioned after the "sizeof".
2576 * Note that we cannot use findloc here because we need the
2577 * arginfo struct, not the symbol.
2579 ptr = arglist[idx].defvalue.size.symname;
2581 altidx < argcnt && strcmp(ptr, arglist[altidx].name) != 0;
2584 if (altidx >= argcnt)
2586 error(17, ptr); /* undefined symbol */
2590 /* check the level against the number of dimensions */
2591 /* the level must be zero for "tagof" values */
2592 assert(arglist[idx].defvalue.size.level == 0
2593 || (arglist[idx].hasdefault & uSIZEOF) != 0);
2594 if (arglist[idx].defvalue.size.level > 0
2595 && arglist[idx].defvalue.size.level >=
2596 arglist[altidx].numdim)
2597 error(28); /* invalid subscript */
2598 if (arglist[altidx].ident != iREFARRAY)
2600 assert(arglist[altidx].ident == iVARIABLE
2601 || arglist[altidx].ident == iREFERENCE);
2602 error(223, ptr); /* redundant sizeof */
2608 sym->usage |= uPROTOTYPED;
2609 errorset(sRESET); /* reset error flag (clear the "panic mode") */
2613 /* doarg - declare one argument type
2615 * this routine is called from "declargs()" and adds an entry in the
2616 * local symbol table for one argument. "fpublic" indicates whether
2617 * the function for this argument list is public.
2618 * The arguments themselves are never public.
2621 doarg(char *name, int ident, int offset, int tags[], int numtags,
2622 int fpublic, int fconst, arginfo * arg)
2626 int idxtag[sDIMEN_MAX];
2628 strcpy(arg->name, name);
2629 arg->hasdefault = FALSE; /* preset (most common case) */
2630 arg->defvalue.val = 0; /* clear */
2631 arg->defvalue_tag = 0;
2633 if (matchtoken('['))
2635 if (ident == iREFERENCE)
2636 error(67, name); /*illegal declaration ("&name[]" is unsupported) */
2639 if (arg->numdim == sDIMEN_MAX)
2641 error(53); /* exceeding maximum number of dimensions */
2644 /* there is no check for non-zero major dimensions here, only if
2645 * the array parameter has a default value, we enforce that all
2646 * array dimensions, except the last, are non-zero
2648 size = needsub(&idxtag[arg->numdim]); /* may be zero here,
2649 *it is a pointer anyway */
2650 #if INT_MAX < LONG_MAX
2652 error(105); /* overflow, exceeding capacity */
2654 arg->dim[arg->numdim] = (int)size;
2657 while (matchtoken('['));
2658 ident = iREFARRAY; /* "reference to array" (is a pointer) */
2659 if (matchtoken('='))
2663 lexpush(); /* initials() needs the "=" token again */
2664 assert(numtags > 0);
2665 /* for the moment, when a default value is given for the array,
2666 * all dimension sizes, except the last, must be non-zero
2667 * (function initials() requires to know the major dimensions)
2669 for (level = 0; level < arg->numdim - 1; level++)
2670 if (arg->dim[level] == 0)
2671 error(52); /* only last dimension may be variable length */
2672 initials(ident, tags[0], &size, arg->dim, arg->numdim);
2673 assert(size >= litidx);
2674 /* allocate memory to hold the initial values */
2675 arg->defvalue.array.data = (cell *) malloc(litidx * sizeof(cell));
2676 if (arg->defvalue.array.data)
2680 memcpy(arg->defvalue.array.data, litq, litidx * sizeof(cell));
2681 arg->hasdefault = TRUE; /* argument has default value */
2682 arg->defvalue.array.size = litidx;
2683 arg->defvalue.array.addr = -1;
2684 /* calculate size to reserve on the heap */
2685 arg->defvalue.array.arraysize = 1;
2686 for (i = 0; i < arg->numdim; i++)
2687 arg->defvalue.array.arraysize *= arg->dim[i];
2688 if (arg->defvalue.array.arraysize < arg->defvalue.array.size)
2689 arg->defvalue.array.arraysize = arg->defvalue.array.size;
2691 litidx = 0; /* reset */
2696 if (matchtoken('='))
2698 unsigned char size_tag_token;
2700 assert(ident == iVARIABLE || ident == iREFERENCE);
2701 arg->hasdefault = TRUE; /* argument has a default value */
2703 (unsigned char)(matchtoken(tSIZEOF) ? uSIZEOF : 0);
2704 if (size_tag_token == 0)
2706 (unsigned char)(matchtoken(tTAGOF) ? uTAGOF : 0);
2707 if (size_tag_token != 0)
2711 if (ident == iREFERENCE)
2712 error(66, name); /* argument may not be a reference */
2714 while (matchtoken('('))
2716 if (needtoken(tSYMBOL))
2718 /* save the name of the argument whose size id to take */
2722 tokeninfo(&val, &name);
2723 if (!(arg->defvalue.size.symname = strdup(name)))
2724 error(103); /* insufficient memory */
2725 arg->defvalue.size.level = 0;
2726 if (size_tag_token == uSIZEOF)
2728 while (matchtoken('['))
2730 arg->defvalue.size.level += (short)1;
2734 if (ident == iVARIABLE) /* make sure we set this only if
2735 * not a reference */
2736 arg->hasdefault |= size_tag_token; /* uSIZEOF or uTAGOF */
2738 while (paranthese--)
2743 constexpr(&arg->defvalue.val, &arg->defvalue_tag);
2744 assert(numtags > 0);
2745 if (!matchtag(tags[0], arg->defvalue_tag, TRUE))
2746 error(213); /* tagname mismatch */
2750 arg->ident = (char)ident;
2751 arg->usage = (char)(fconst ? uCONST : 0);
2752 arg->numtags = numtags;
2753 arg->tags = (int *)malloc(numtags * sizeof tags[0]);
2755 error(103); /* insufficient memory */
2756 memcpy(arg->tags, tags, numtags * sizeof tags[0]);
2757 argsym = findloc(name);
2760 error(21, name); /* symbol already defined */
2764 if ((argsym = findglb(name)) && argsym->ident != iFUNCTN)
2765 error(219, name); /* variable shadows another symbol */
2766 /* add details of type and address */
2767 assert(numtags > 0);
2768 argsym = addvariable(name, offset, ident, sLOCAL, tags[0],
2769 arg->dim, arg->numdim, idxtag);
2770 argsym->compound = 0;
2771 if (ident == iREFERENCE)
2772 argsym->usage |= uREAD; /* because references are passed back */
2774 argsym->usage |= uREAD; /* arguments of public functions
2775 * are always "used" */
2777 argsym->usage |= uCONST;
2782 count_referrers(symbol * entry)
2787 for (i = 0; i < entry->numrefers; i++)
2788 if (entry->refer[i])
2793 /* Every symbol has a referrer list, that contains the functions that
2794 * use the symbol. Now, if function "apple" is accessed by functions
2795 * "banana" and "citron", but neither function "banana" nor "citron" are
2796 * used by anyone else, then, by inference, function "apple" is not used
2799 reduce_referrers(symbol * root)
2807 for (sym = root->next; sym; sym = sym->next)
2810 continue; /* hierarchical data type */
2811 if (sym->ident == iFUNCTN
2812 && (sym->usage & uNATIVE) == 0
2813 && (sym->usage & uPUBLIC) == 0
2814 && strcmp(sym->name, uMAINFUNC) != 0
2815 && count_referrers(sym) == 0)
2817 sym->usage &= ~(uREAD | uWRITTEN); /* erase usage bits if
2818 * there is no referrer */
2819 /* find all symbols that are referred by this symbol */
2820 for (ref = root->next; ref; ref = ref->next)
2823 continue; /* hierarchical data type */
2824 assert(ref->refer != NULL);
2825 for (i = 0; i < ref->numrefers && ref->refer[i] != sym;
2828 if (i < ref->numrefers)
2830 assert(ref->refer[i] == sym);
2831 ref->refer[i] = NULL;
2836 else if ((sym->ident == iVARIABLE || sym->ident == iARRAY)
2837 && (sym->usage & uPUBLIC) == 0
2838 && !sym->parent && count_referrers(sym) == 0)
2840 sym->usage &= ~(uREAD | uWRITTEN); /* erase usage bits if
2841 * there is no referrer */
2844 /* after removing a symbol, check whether more can be removed */
2846 while (restart > 0);
2849 /* testsymbols - test for unused local or global variables
2851 * "Public" functions are excluded from the check, since these
2852 * may be exported to other object modules.
2853 * Labels are excluded from the check if the argument 'testlabs'
2854 * is 0. Thus, labels are not tested until the end of the function.
2855 * Constants may also be excluded (convenient for global constants).
2857 * When the nesting level drops below "level", the check stops.
2859 * The function returns whether there is an "entry" point for the file.
2860 * This flag will only be 1 when browsing the global symbol table.
2863 testsymbols(symbol * root, int level, int testlabs, int testconst)
2865 char symname[2 * sNAMEMAX + 16];
2868 symbol *sym = root->next;
2870 while (sym && sym->compound >= level)
2877 if ((sym->usage & uDEFINE) == 0)
2878 error(19, sym->name); /* not a label: ... */
2879 else if ((sym->usage & uREAD) == 0)
2880 error(203, sym->name); /* symbol isn't used: ... */
2884 if ((sym->usage & (uDEFINE | uREAD | uNATIVE | uSTOCK)) == uDEFINE)
2886 funcdisplayname(symname, sym->name);
2887 if (symname[0] != '\0')
2888 error(203, symname); /* symbol isn't used ...
2889 * (and not native/stock) */
2891 if ((sym->usage & uPUBLIC) != 0
2892 || strcmp(sym->name, uMAINFUNC) == 0)
2893 entry = TRUE; /* there is an entry point */
2896 if (testconst && (sym->usage & uREAD) == 0)
2897 error(203, sym->name); /* symbol isn't used: ... */
2902 break; /* hierarchical data type */
2903 if ((sym->usage & (uWRITTEN | uREAD | uSTOCK | uPUBLIC)) == 0)
2904 error(203, sym->name); /* symbol isn't used (and not stock
2906 else if ((sym->usage & (uREAD | uSTOCK | uPUBLIC)) == 0)
2907 error(204, sym->name); /* value assigned to symbol is
2909 #if 0 /*// ??? not sure whether it is a good idea to
2910 * force people use "const" */
2911 else if ((sym->usage & (uWRITTEN | uPUBLIC | uCONST)) == 0
2912 && sym->ident == iREFARRAY)
2913 error(214, sym->name); /* make array argument "const" */
2923 calc_array_datasize(symbol * sym, cell * offset)
2927 assert(sym != NULL);
2928 assert(sym->ident == iARRAY || sym->ident == iREFARRAY);
2929 length = sym->dim.array.length;
2930 if (sym->dim.array.level > 0)
2933 calc_array_datasize(finddepend(sym), offset);
2935 *offset = length * (*offset + sizeof(cell));
2937 length *= length * sublength;
2950 destructsymbols(symbol * root, int level)
2953 int savepri = FALSE;
2954 symbol *sym = root->next;
2956 while (sym && sym->compound >= level)
2958 if (sym->ident == iVARIABLE || sym->ident == iARRAY)
2960 char symbolname[16];
2964 /* check that the '~' operator is defined for this tag */
2965 operator_symname(symbolname, "~", sym->tag, 0, 1, 0);
2966 if ((opsym = findglb(symbolname)))
2968 /* save PRI, in case of a return statement */
2971 push1(); /* right-hand operand is in PRI */
2974 /* if the variable is an array, get the number of elements */
2975 if (sym->ident == iARRAY)
2977 elements = calc_array_datasize(sym, &offset);
2978 /* "elements" can be zero when the variable is declared like
2979 * new mytag: myvar[2][] = { {1, 2}, {3, 4} }
2980 * one should declare all dimensions!
2983 error(46, sym->name); /* array size is unknown */
2991 /* call the '~' operator */
2993 addconst(offset); /*add offset to array data to the address */
2995 pushval(2 * sizeof(cell)); /* 2 parameters */
2997 if (sc_status != statSKIP)
2998 markusage(opsym, uREAD); /* do not mark as "used" when this
2999 * call itself is skipped */
3001 opsym->x.lib->value += 1; /* increment "usage count"
3007 /* restore PRI, if it was saved */
3013 insert_constval(constvalue * prev, constvalue * next, char *name,
3014 cell val, short index)
3018 if (!(cur = (constvalue *)malloc(sizeof(constvalue))))
3019 error(103); /* insufficient memory (fatal error) */
3020 memset(cur, 0, sizeof(constvalue));
3021 strcpy(cur->name, name);
3030 append_constval(constvalue * table, char *name, cell val, short index)
3032 constvalue *cur, *prev;
3034 /* find the end of the constant table */
3035 for (prev = table, cur = table->next; cur;
3036 prev = cur, cur = cur->next)
3038 return insert_constval(prev, NULL, name, val, index);
3042 find_constval(constvalue * table, char *name, short index)
3044 constvalue *ptr = table->next;
3048 if (strcmp(name, ptr->name) == 0 && ptr->index == index)
3056 find_constval_byval(constvalue * table, cell val)
3058 constvalue *ptr = table->next;
3062 if (ptr->value == val)
3069 #if 0 /* never used */
3071 delete_constval(constvalue * table, char *name)
3073 constvalue *prev = table;
3074 constvalue *cur = prev->next;
3078 if (strcmp(name, cur->name) == 0)
3080 prev->next = cur->next;
3092 delete_consttable(constvalue * table)
3094 constvalue *cur = table->next, *next;
3102 memset(table, 0, sizeof(constvalue));
3107 * Adds a symbol to the #define symbol table.
3110 add_constant(char *name, cell val, int vclass, int tag)
3114 /* Test whether a global or local symbol with the same name exists. Since
3115 * constants are stored in the symbols table, this also finds previously
3116 * defind constants. */
3117 sym = findglb(name);
3119 sym = findloc(name);
3122 /* silently ignore redefinitions of constants with the same value */
3123 if (sym->ident == iCONSTEXPR)
3125 if (sym->addr != val)
3126 error(201, name); /* redefinition of constant (different value) */
3130 error(21, name); /* symbol already defined */
3135 /* constant doesn't exist yet, an entry must be created */
3136 sym = addsym(name, val, iCONSTEXPR, vclass, tag, uDEFINE);
3137 if (sc_status == statIDLE)
3138 sym->usage |= uPREDEF;
3141 /* statement - The Statement Parser
3143 * This routine is called whenever the parser needs to know what
3144 * statement it encounters (i.e. whenever program syntax requires a
3148 statement(int *lastindent, int allow_decl)
3156 error(36); /* empty statement */
3161 tok = lex(&val, &st);
3163 setline(fline, fcurrent);
3164 /* lex() has set stmtindent */
3165 if (lastindent && tok != tLABEL)
3168 if (*lastindent >= 0 && *lastindent != stmtindent &&
3169 !indent_nowarn && sc_tabsize > 0)
3170 error(217); /* loose indentation */
3172 *lastindent = stmtindent;
3173 indent_nowarn = TRUE; /* if warning was blocked, re-enable it */
3188 error(3); /* declaration only valid in a block */
3199 error(3); /* declaration only valid in a block */
3203 if (!matchtoken('}')) /* {} is the empty statement */
3205 /* lastst (for "last statement") does not change */
3208 error(36); /* empty statement */
3232 error(14); /* not in switch */
3272 default: /* non-empty expression */
3273 lexpush(); /* analyze token later */
3274 doexpr(TRUE, TRUE, TRUE, TRUE, NULL, FALSE);
3284 cell save_decl = declared;
3287 nestlevel += 1; /* increase compound statement level */
3288 while (matchtoken('}') == 0)
3289 { /* repeat until compound statement is closed */
3292 needtoken('}'); /* gives error: "expected token }" */
3298 && (lastst == tRETURN || lastst == tBREAK
3299 || lastst == tCONTINUE))
3300 error(225); /* unreachable code */
3301 statement(&indent, TRUE); /* do a statement */
3305 if (lastst != tRETURN)
3306 destructsymbols(&loctab, nestlevel);
3307 if (lastst != tRETURN && lastst != tGOTO)
3308 /* delete local variable space */
3309 modstk((int)(declared - save_decl) * sizeof(cell));
3311 testsymbols(&loctab, nestlevel, FALSE, TRUE); /* look for unused
3313 declared = save_decl;
3314 delete_symbols(&loctab, nestlevel, FALSE, TRUE);
3315 /* erase local symbols, but
3316 * retain block local labels
3317 * (within the function) */
3319 nestlevel -= 1; /* decrease compound statement level */
3324 * Global references: stgidx (referred to only)
3327 doexpr(int comma, int chkeffect, int allowarray, int mark_endexpr,
3328 int *tag, int chkfuncresult)
3330 int constant, index, ident;
3331 int localstaging = FALSE;
3336 stgset(TRUE); /* start stage-buffering */
3337 localstaging = TRUE;
3338 assert(stgidx == 0);
3341 errorset(sEXPRMARK);
3344 /* on second round through, mark the end of the previous expression */
3345 if (index != stgidx)
3348 ident = expression(&constant, &val, tag, chkfuncresult);
3349 if (!allowarray && (ident == iARRAY || ident == iREFARRAY))
3350 error(33, "-unknown-"); /* array must be indexed */
3351 if (chkeffect && !sideeffect)
3352 error(215); /* expression has no effect */
3354 while (comma && matchtoken(',')); /* more? */
3356 endexpr(TRUE); /* optionally, mark the end of the expression */
3357 errorset(sEXPRRELEASE);
3361 stgset(FALSE); /* stop staging */
3368 constexpr(cell * val, int *tag)
3370 int constant, index;
3373 stgset(TRUE); /* start stage-buffering */
3374 stgget(&index, &cidx); /* mark position in code generator */
3375 errorset(sEXPRMARK);
3376 expression(&constant, val, tag, FALSE);
3377 stgdel(index, cidx); /* scratch generated code */
3378 stgset(FALSE); /* stop stage-buffering */
3380 error(8); /* must be constant expression */
3381 errorset(sEXPRRELEASE);
3387 * In the case a "simple assignment" operator ("=") is used within a
3388 * test, * the warning "possibly unintended assignment" is displayed.
3389 * This routine sets the global variable "intest" to true, it is
3390 * restored upon termination. In the case the assignment was intended,
3391 * use parantheses around the expression to avoid the warning;
3392 * primary() sets "intest" to 0.
3394 * Global references: intest (altered, but restored upon termination)
3397 test(int label, int parens, int invert)
3401 value lval = { NULL, 0, 0, 0, 0, NULL };
3402 int localstaging = FALSE;
3406 stgset(TRUE); /* start staging */
3407 localstaging = TRUE;
3409 stgget(&index, &cidx); /* should start at zero if started
3415 pushstk((stkitem) intest);
3421 stgget(&index, &cidx); /* mark position (of last expression) in
3425 tok = matchtoken(',');
3429 while (tok); /* do */
3432 if (lval.ident == iARRAY || lval.ident == iREFARRAY)
3435 (lval.sym->name) ? lval.sym->name : "-unknown-";
3436 error(33, ptr); /* array must be indexed */
3438 if (lval.ident == iCONSTEXPR)
3439 { /* constant expression */
3440 intest = (int)(long)popstk(); /* restore stack */
3441 stgdel(index, cidx);
3443 { /* code always executed */
3444 error(206); /* redundant test: always non-zero */
3448 error(205); /* redundant code: never executed */
3453 stgout(0); /* write "jumplabel" code */
3454 stgset(FALSE); /* stop staging */
3458 if (lval.tag != 0 && lval.tag != sc_addtag("bool"))
3459 if (check_userop(lneg, lval.tag, 0, 1, NULL, &lval.tag))
3460 invert = !invert; /* user-defined ! operator inverted result */
3462 jmp_ne0(label); /* jump to label if true (different from 0) */
3464 jmp_eq0(label); /* jump to label if false (equal to 0) */
3465 endexpr(TRUE); /* end expression (give optimizer a chance) */
3466 intest = (int)(long)popstk(); /* double typecast to avoid warning
3467 * with Microsoft C */
3470 stgout(0); /* output queue from the very beginning (see
3471 * assert() when localstaging is set to TRUE) */
3472 stgset(FALSE); /* stop staging */
3482 ifindent = stmtindent; /* save the indent of the "if" instruction */
3483 flab1 = getlabel(); /* get label number for false branch */
3484 test(flab1, TRUE, FALSE); /*get expression, branch to flab1 if false */
3485 statement(NULL, FALSE); /* if true, do a statement */
3486 if (matchtoken(tELSE) == 0)
3488 setlabel(flab1); /* no, simple if..., print false label */
3492 /* to avoid the "dangling else" error, we want a warning if the "else"
3493 * has a lower indent than the matching "if" */
3495 if (stmtindent < ifindent && sc_tabsize > 0)
3496 error(217); /* loose indentation */
3499 if ((lastst != tRETURN) && (lastst != tGOTO))
3501 setlabel(flab1); /* print false label */
3502 statement(NULL, FALSE); /* do "else" clause */
3503 setlabel(flab2); /* print true label */
3510 int wq[wqSIZE]; /* allocate local queue */
3512 addwhile(wq); /* add entry to queue for "break" */
3513 setlabel(wq[wqLOOP]); /* loop label */
3514 /* The debugger uses the "line" opcode to be able to "break" out of
3515 * a loop. To make sure that each loop has a line opcode, even for the
3516 * tiniest loop, set it below the top of the loop */
3517 setline(fline, fcurrent);
3518 test(wq[wqEXIT], TRUE, FALSE); /* branch to wq[wqEXIT] if false */
3519 statement(NULL, FALSE); /* if so, do a statement */
3520 jumplabel(wq[wqLOOP]); /* and loop to "while" start */
3521 setlabel(wq[wqEXIT]); /* exit label */
3522 delwhile(); /* delete queue entry */
3526 * Note that "continue" will in this case not jump to the top of the
3527 * loop, but to the end: just before the TRUE-or-FALSE testing code.
3532 int wq[wqSIZE], top;
3534 addwhile(wq); /* see "dowhile" for more info */
3535 top = getlabel(); /* make a label first */
3536 setlabel(top); /* loop label */
3537 statement(NULL, FALSE);
3539 setlabel(wq[wqLOOP]); /* "continue" always jumps to WQLOOP. */
3540 setline(fline, fcurrent);
3541 test(wq[wqEXIT], TRUE, FALSE);
3543 setlabel(wq[wqEXIT]);
3551 int wq[wqSIZE], skiplab;
3553 int save_nestlevel, index;
3556 save_decl = declared;
3557 save_nestlevel = nestlevel;
3560 skiplab = getlabel();
3562 if (matchtoken(';') == 0)
3564 /* new variable declarations are allowed here */
3565 if (matchtoken(tNEW))
3567 /* The variable in expr1 of the for loop is at a
3568 * 'compound statement' level of it own.
3571 declloc(FALSE); /* declare local variable */
3575 doexpr(TRUE, TRUE, TRUE, TRUE, NULL, FALSE); /* expression 1 */
3579 /* Adjust the "declared" field in the "while queue", in case that
3580 * local variables were declared in the first expression of the
3581 * "for" loop. These are deleted in separately, so a "break" or a
3582 * "continue" must ignore these fields.
3585 assert(ptr != NULL);
3586 ptr[wqBRK] = (int)declared;
3587 ptr[wqCONT] = (int)declared;
3588 jumplabel(skiplab); /* skip expression 3 1st time */
3589 setlabel(wq[wqLOOP]); /* "continue" goes to this label: expr3 */
3590 setline(fline, fcurrent);
3591 /* Expressions 2 and 3 are reversed in the generated code:
3592 * expression 3 precedes expression 2.
3593 * When parsing, the code is buffered and marks for
3594 * the start of each expression are insterted in the buffer.
3597 stgset(TRUE); /* start staging */
3598 assert(stgidx == 0);
3600 stgmark(sSTARTREORDER);
3601 stgmark((char)(sEXPRSTART + 0)); /* mark start of 2nd expression
3603 setlabel(skiplab); /*jump to this point after 1st expression */
3604 if (matchtoken(';') == 0)
3606 test(wq[wqEXIT], FALSE, FALSE); /* expression 2
3607 *(jump to wq[wqEXIT] if false) */
3610 stgmark((char)(sEXPRSTART + 1)); /* mark start of 3th expression
3612 if (matchtoken(')') == 0)
3614 doexpr(TRUE, TRUE, TRUE, TRUE, NULL, FALSE); /* expression 3 */
3617 stgmark(sENDREORDER); /* mark end of reversed evaluation */
3619 stgset(FALSE); /* stop staging */
3620 statement(NULL, FALSE);
3621 jumplabel(wq[wqLOOP]);
3622 setlabel(wq[wqEXIT]);
3625 assert(nestlevel >= save_nestlevel);
3626 if (nestlevel > save_nestlevel)
3628 /* Clean up the space and the symbol table for the local
3629 * variable in "expr1".
3631 destructsymbols(&loctab, nestlevel);
3632 modstk((int)(declared - save_decl) * sizeof(cell));
3633 declared = save_decl;
3634 delete_symbols(&loctab, nestlevel, FALSE, TRUE);
3635 nestlevel = save_nestlevel; /* reset 'compound statement'
3640 /* The switch statement is incompatible with its C sibling:
3641 * 1. the cases are not drop through
3642 * 2. only one instruction may appear below each case, use a compound
3643 * instruction to execute multiple instructions
3644 * 3. the "case" keyword accepts a comma separated list of values to
3645 * match, it also accepts a range using the syntax "1 .. 4"
3648 * PRI = expression result
3649 * param = table offset (code segment)
3655 int lbl_table, lbl_exit, lbl_case;
3656 int tok, swdefault, casecount;
3659 constvalue caselist = { NULL, "", 0, 0 }; /*case list starts empty */
3660 constvalue *cse, *csp;
3661 char labelname[sNAMEMAX + 1];
3664 doexpr(TRUE, FALSE, FALSE, TRUE, NULL, FALSE); /* evaluate
3665 * switch expression */
3667 /* generate the code for the switch statement, the label is the
3668 * address of the case table (to be generated later).
3670 lbl_table = getlabel();
3671 lbl_case = 0; /* just to avoid a compiler warning */
3672 ffswitch(lbl_table);
3675 lbl_exit = getlabel(); /*get label number for jumping out of switch */
3680 tok = lex(&val, &str); /* read in (new) token */
3684 if (swdefault != FALSE)
3685 error(15); /* "default" case must be last in switch
3687 lbl_case = getlabel();
3688 sc_allowtags = FALSE; /* do not allow tagnames here */
3693 /* ??? enforce/document that, in a switch, a statement cannot
3694 * start an opening brace (marks the start of a compound
3695 * statement) and search for the right-most colon before that
3697 * Now, by replacing the ':' by a special COLON token, you can
3698 * parse all expressions until that special token.
3701 constexpr(&val, NULL);
3702 /* Search the insertion point (the table is kept in sorted
3703 * order, so that advanced abstract machines can sift the
3704 * case table with a binary search). Check for duplicate
3705 * case values at the same time.
3707 for (csp = &caselist, cse = caselist.next;
3708 cse && cse->value < val;
3709 csp = cse, cse = cse->next)
3711 if (cse && cse->value == val)
3712 error(40, val); /* duplicate "case" label */
3713 /* Since the label is stored as a string in the
3714 * "constvalue", the size of an identifier must
3715 * be at least 8, as there are 8
3716 * hexadecimal digits in a 32-bit number.
3719 #error Length of identifier (sNAMEMAX) too small.
3721 insert_constval(csp, cse, itoh(lbl_case), val, 0);
3722 if (matchtoken(tDBLDOT))
3726 constexpr(&end, NULL);
3728 error(50); /* invalid range */
3729 while (++val <= end)
3732 /* find the new insertion point */
3733 for (csp = &caselist, cse = caselist.next;
3734 cse && cse->value < val;
3735 csp = cse, cse = cse->next)
3737 if (cse && cse->value == val)
3738 error(40, val); /* duplicate "case" label */
3739 insert_constval(csp, cse, itoh(lbl_case), val, 0);
3743 while (matchtoken(','));
3744 needtoken(':'); /* ':' ends the case */
3745 sc_allowtags = TRUE; /* reset */
3747 statement(NULL, FALSE);
3748 jumplabel(lbl_exit);
3751 if (swdefault != FALSE)
3752 error(16); /* multiple defaults in switch */
3753 lbl_case = getlabel();
3757 statement(NULL, FALSE);
3758 /* Jump to lbl_exit, even thouh this is the last clause in the
3759 *switch, because the jump table is generated between the last
3760 * clause of the switch and the exit label.
3762 jumplabel(lbl_exit);
3765 /* nothing, but avoid dropping into "default" */
3769 indent_nowarn = TRUE; /* disable this check */
3770 tok = '}'; /* break out of the loop after an error */
3776 /* verify that the case table is sorted (unfortunately, duplicates can
3777 * occur; there really shouldn't be duplicate cases, but the compiler
3778 * may not crash or drop into an assertion for a user error). */
3779 for (cse = caselist.next; cse && cse->next; cse = cse->next)
3780 ; /* empty. no idea whether this is correct, but we MUST NOT do
3781 * the setlabel(lbl_table) call in the loop body. doing so breaks
3782 * switch statements that only have one case statement following.
3786 /* generate the table here, before lbl_exit (general jump target) */
3787 setlabel(lbl_table);
3789 if (swdefault == FALSE)
3791 /* store lbl_exit as the "none-matched" label in the switch table */
3792 strcpy(labelname, itoh(lbl_exit));
3796 /* lbl_case holds the label of the "default" clause */
3797 strcpy(labelname, itoh(lbl_case));
3799 ffcase(casecount, labelname, TRUE);
3800 /* generate the rest of the table */
3801 for (cse = caselist.next; cse; cse = cse->next)
3802 ffcase(cse->value, cse->name, FALSE);
3805 delete_consttable(&caselist); /* clear list of case labels */
3813 value lval = { NULL, 0, 0, 0, 0, NULL };
3815 if ((sc_debug & sCHKBOUNDS) != 0)
3817 flab1 = getlabel(); /* get label number for "OK" branch */
3818 test(flab1, FALSE, TRUE); /* get expression and branch
3819 * to flab1 if true */
3820 setline(fline, fcurrent); /* make sure we abort on the correct
3822 ffabort(xASSERTION);
3827 stgset(TRUE); /* start staging */
3828 stgget(&index, &cidx); /* mark position in code generator */
3833 stgdel(index, cidx); /* just scrap the code */
3835 while (matchtoken(','));
3836 stgset(FALSE); /* stop staging */
3848 if (lex(&val, &st) == tSYMBOL)
3851 jumplabel((int)sym->addr);
3852 sym->usage |= uREAD; /* set "uREAD" bit */
3854 * // ??? if the label is defined (check sym->usage & uDEFINE), check
3855 * // sym->compound (nesting level of the label) against nestlevel;
3856 * // if sym->compound < nestlevel, call the destructor operator
3861 error(20, st); /* illegal symbol name */
3873 tokeninfo(&val, &st); /* retrieve label name again */
3874 if (find_constval(&tagname_tab, st, 0))
3875 error(221, st); /* label name shadows tagname */
3877 setlabel((int)sym->addr);
3878 /* since one can jump around variable declarations or out of compound
3879 * blocks, the stack must be manually adjusted
3881 setstk(-declared * sizeof(cell));
3882 sym->usage |= uDEFINE; /* label is now defined */
3887 * Finds a label from the (local) symbol table or adds one to it.
3888 * Labels are local in scope.
3890 * Note: The "_usage" bit is set to zero. The routines that call
3891 * "fetchlab()" must set this bit accordingly.
3894 fetchlab(char *name)
3898 sym = findloc(name); /* labels are local in scope */
3901 if (sym->ident != iLABEL)
3902 error(19, sym->name); /* not a label: ... */
3906 sym = addsym(name, getlabel(), iLABEL, sLOCAL, 0, 0);
3907 sym->x.declared = (int)declared;
3908 sym->compound = nestlevel;
3915 * Global references: rettype (altered)
3922 if (matchtoken(tTERM) == 0)
3924 if ((rettype & uRETNONE) != 0)
3925 error(208); /* mix "return;" and "return value;" */
3926 doexpr(TRUE, FALSE, FALSE, TRUE, &tag, FALSE);
3928 rettype |= uRETVALUE; /* function returns a value */
3929 /* check tagname with function tagname */
3930 assert(curfunc != NULL);
3931 if (!matchtag(curfunc->tag, tag, TRUE))
3932 error(213); /* tagname mismatch */
3936 /* this return statement contains no expression */
3938 if ((rettype & uRETVALUE) != 0)
3940 char symname[2 * sNAMEMAX + 16]; /* allow space for user
3941 * defined operators */
3942 assert(curfunc != NULL);
3943 funcdisplayname(symname, curfunc->name);
3944 error(209, symname); /* function should return a value */
3946 rettype |= uRETNONE; /* function does not return anything */
3948 destructsymbols(&loctab, 0); /*call destructor for *all* locals */
3949 modstk((int)declared * sizeof(cell)); /* end of function, remove
3950 *all* * local variables*/
3959 ptr = readwhile(); /* readwhile() gives an error if not in loop */
3963 destructsymbols(&loctab, nestlevel);
3964 modstk(((int)declared - ptr[wqBRK]) * sizeof(cell));
3965 jumplabel(ptr[wqEXIT]);
3973 ptr = readwhile(); /* readwhile() gives an error if not in loop */
3977 destructsymbols(&loctab, nestlevel);
3978 modstk(((int)declared - ptr[wqCONT]) * sizeof(cell));
3979 jumplabel(ptr[wqLOOP]);
3985 /* find the tag by value in the table, then set the top bit to mark it
3992 assert((tag & PUBLICTAG) == 0);
3993 for (ptr = tagname_tab.next;
3994 ptr && tag != (int)(ptr->value & TAGMASK); ptr = ptr->next)
3997 ptr->value |= PUBLICTAG;
4006 if (matchtoken(tTERM) == 0)
4008 doexpr(TRUE, FALSE, FALSE, TRUE, &tag, FALSE);
4017 destructsymbols(&loctab, 0); /* call destructor for *all* locals */
4026 if (matchtoken(tTERM) == 0)
4028 doexpr(TRUE, FALSE, FALSE, TRUE, &tag, FALSE);
4045 ptr[wqBRK] = (int)declared; /* stack pointer (for "break") */
4046 ptr[wqCONT] = (int)declared; /* for "continue", possibly adjusted later */
4047 ptr[wqLOOP] = getlabel();
4048 ptr[wqEXIT] = getlabel();
4049 if (wqptr >= (wq + wqTABSZ - wqSIZE))
4050 error(102, "loop table"); /* loop table overflow (too many active loops) */
4053 { /* copy "ptr" to while queue table */
4073 error(24); /* out of context */
4078 return (wqptr - wqSIZE);