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 setconfig(argv[0]); /* the path to the include files */
308 lcl_ctrlchar = sc_ctrlchar;
309 lcl_packstr = sc_packstr;
310 lcl_needsemicolon = sc_needsemicolon;
311 lcl_tabsize = sc_tabsize;
312 inpf = inpf_org = (FILE *) sc_opensrc(inpfname);
314 error(100, inpfname);
316 outf = (FILE *) sc_openasm(fd_out); /* first write to assembler
317 * file (may be temporary) */
319 error(101, outfname);
320 /* immediately open the binary file, for other programs to check */
321 binf = (FILE *) sc_openbin(binfname);
323 error(101, binfname);
324 setconstants(); /* set predefined constants and tagnames */
325 for (i = 0; i < skipinput; i++) /* skip lines in the input file */
326 if (sc_readsrc(inpf, pline, sLINEMAX))
327 fline++; /* keep line number up to date */
329 sc_status = statFIRST;
330 /* do the first pass through the file */
331 inpfmark = sc_getpossrc(inpf);
332 if (incfname[0] != '\0')
334 if (strcmp(incfname, sDEF_PREFIX) == 0)
336 plungefile(incfname, FALSE, TRUE); /* parse "default.inc" */
340 if (!plungequalifiedfile(incfname)) /* parse "prefix" include
342 error(100, incfname); /* cannot read from ... (fatal error) */
345 preprocess(); /* fetch first line */
346 parse(); /* process all input */
349 sc_status = statWRITE; /* set, to enable warnings */
351 /* ??? for re-parsing the listing file instead of the original source
352 * file (and doing preprocessing twice):
353 * - close input file, close listing file
354 * - re-open listing file for reading (inpf)
355 * - open assembler file (outf)
358 /* reset "defined" flag of all functions and global variables */
359 reduce_referrers(&glbtab);
360 delete_symbols(&glbtab, 0, TRUE, FALSE);
361 #if !defined NO_DEFINE
365 sc_ctrlchar = lcl_ctrlchar;
366 sc_packstr = lcl_packstr;
367 sc_needsemicolon = lcl_needsemicolon;
368 sc_tabsize = lcl_tabsize;
370 /* reset the source file */
373 sc_resetsrc(inpf, inpfmark); /* reset file position */
374 fline = skipinput; /* reset line number */
375 lexinit(); /* clear internal flags of lex() */
376 sc_status = statWRITE; /* allow to write --this variable was reset
377 * by resetglobals() */
379 setfile(inpfname, fnumber);
380 if (incfname[0] != '\0')
382 if (strcmp(incfname, sDEF_PREFIX) == 0)
383 plungefile(incfname, FALSE, TRUE); /* parse "default.inc" (again) */
385 plungequalifiedfile(incfname); /* parse implicit include
388 preprocess(); /* fetch first line */
389 parse(); /* process all input */
390 /* inpf is already closed when readline() attempts to pop of a file */
391 writetrailer(); /* write remaining stuff */
393 entry = testsymbols(&glbtab, 0, TRUE, FALSE); /* test for unused
394 * or undefined functions and variables */
396 error(13); /* no entry point (no public functions) */
399 if (inpf) /* main source file is not closed, do it now */
401 /* write the binary file (the file is already open) */
402 if (errnum == 0 && jmpcode == 0)
404 assert(binf != NULL);
405 sc_resetasm(outf); /* flush and loop back, for reading */
406 assemble(binf, outf); /* assembler file is now input */
412 sc_closebin(binf, errnum != 0);
420 assert(jmpcode != 0 || loctab.next == NULL); /* on normal flow,
422 * should already have been deleted */
423 delete_symbols(&loctab, 0, TRUE, TRUE); /* delete local variables
424 * if not yet done (i.e.
425 * on a fatal error) */
426 delete_symbols(&glbtab, 0, TRUE, TRUE);
427 delete_consttable(&tagname_tab);
428 delete_consttable(&libname_tab);
431 #if !defined NO_DEFINE
436 printf("\n%d Error%s.\n", errnum, (errnum > 1) ? "s" : "");
439 else if (warnnum != 0)
441 printf("\n%d Warning%s.\n", warnnum, (warnnum > 1) ? "s" : "");
452 sc_addconstant(char *name, cell value, int tag)
454 errorset(sFORCESET); /* make sure error engine is silenced */
455 sc_status = statIDLE;
456 add_constant(name, value, sGLOBAL, tag);
461 sc_addtag(char *name)
469 /* no tagname was given, check for one */
470 if (lex(&val, &name) != tLABEL)
473 return 0; /* untagged */
478 ptr = tagname_tab.next;
481 tag = (int)(ptr->value & TAGMASK);
482 if (strcmp(name, ptr->name) == 0)
483 return tag; /* tagname is known, return its sequence number */
484 tag &= (int)~FIXEDTAG;
490 /* tagname currently unknown, add it */
491 tag = last + 1; /* guaranteed not to exist already */
492 if (sc_isupper(*name))
493 tag |= (int)FIXEDTAG;
494 append_constval(&tagname_tab, name, (cell) tag, 0);
501 /* reset the subset of global variables that is modified by the
503 curfunc = NULL; /* pointer to current function */
504 lastst = 0; /* last executed statement type */
505 nestlevel = 0; /* number of active (open) compound statements */
506 rettype = 0; /* the type that a "return" expression should have */
507 litidx = 0; /* index to literal table */
508 stgidx = 0; /* index to the staging buffer */
509 labnum = 0; /* number of (internal) labels */
510 staging = 0; /* true if staging output */
511 declared = 0; /* number of local cells declared */
512 glb_declared = 0; /* number of global cells declared */
513 code_idx = 0; /* number of bytes with generated code */
514 ntv_funcid = 0; /* incremental number of native function */
515 curseg = 0; /* 1 if currently parsing CODE, 2 if parsing DATA */
516 freading = FALSE; /* no input file ready yet */
517 fline = 0; /* the line number in the current file */
518 fnumber = 0; /* the file number in the file table (debugging) */
519 fcurrent = 0; /* current file being processed (debugging) */
520 intest = 0; /* true if inside a test */
521 sideeffect = 0; /* true if an expression causes a side-effect */
522 stmtindent = 0; /* current indent of the statement */
523 indent_nowarn = TRUE; /* do not skip warning "217 loose indentation" */
524 sc_allowtags = TRUE; /* allow/detect tagnames */
525 sc_status = statIDLE;
533 skipinput = 0; /* number of lines to skip from the first
535 sc_ctrlchar = CTRL_CHAR; /* the escape character */
536 litmax = sDEF_LITMAX; /* current size of the literal table */
537 errnum = 0; /* number of errors */
538 warnnum = 0; /* number of warnings */
539 /* sc_debug=sCHKBOUNDS; by default: bounds checking+assertions */
540 sc_debug = 0; /* by default: no debug */
541 charbits = 8; /* a "char" is 8 bits */
542 sc_packstr = FALSE; /* strings are unpacked by default */
543 /* sc_compress=TRUE; compress output bytecodes */
544 sc_compress = FALSE; /* compress output bytecodes */
545 sc_needsemicolon = FALSE; /* semicolon required to terminate
548 sc_stksize = sDEF_AMXSTACK; /* default stack size */
549 sc_tabsize = 8; /* assume a TAB is 8 spaces */
550 sc_rationaltag = 0; /* assume no support for rational numbers */
551 rational_digits = 0; /* number of fractional digits */
553 outfname[0] = '\0'; /* output file name */
554 inpf = NULL; /* file read from */
555 inpfname = NULL; /* pointer to name of the file currently
557 outf = NULL; /* file written to */
558 litq = NULL; /* the literal queue */
559 glbtab.next = NULL; /* clear global variables/constants table */
560 loctab.next = NULL; /* " local " / " " */
561 tagname_tab.next = NULL; /* tagname table */
562 libname_tab.next = NULL; /* library table (#pragma library "..."
565 pline[0] = '\0'; /* the line read from the input file */
566 lptr = NULL; /* points to the current position in "pline" */
567 curlibrary = NULL; /* current library */
568 inpf_org = NULL; /* main source file */
570 wqptr = wq; /* initialize while queue pointer */
575 parseoptions(int argc, char **argv, char *iname, char *oname,
576 char *pname __UNUSED__, char *rname __UNUSED__)
582 /* use embryo include dir always */
583 snprintf(str, sizeof(str), "%s/include/", e_prefix_data_get());
587 for (i = 1; i < argc; i++)
589 if (!strcmp (argv[i], "-i") && (i + 1 < argc) && *argv[i + 1])
591 /* include directory */
593 strncpy(str, argv[i], sizeof(str));
596 if (str[len - 1] != DIRSEP_CHAR)
598 str[len] = DIRSEP_CHAR;
604 else if (!strcmp (argv[i], "-o") && (i + 1 < argc) && *argv[i + 1])
608 strcpy(oname, argv[i]); /* FIXME */
610 else if (!strcmp (argv[i], "-S") && (i + 1 < argc) && *argv[i + 1])
614 stack_size = atoi(argv[i]);
617 sc_stksize = (cell) stack_size;
624 strcpy(iname, argv[i]); /* FIXME */
628 /* only allow one input filename */
635 setopt(int argc, char **argv, char *iname, char *oname,
636 char *pname, char *rname)
642 strcpy(pname, sDEF_PREFIX);
644 parseoptions(argc, argv, iname, oname, pname, rname);
645 if (iname[0] == '\0')
650 setconfig(char *root)
656 path[sizeof(path) - 1] = 0;
658 /* add the default "include" directory */
661 /* path + filename (hopefully) */
662 strncpy(path, root, sizeof(path) - 1);
663 path[sizeof(path) - 1] = 0;
665 /* terminate just behind last \ or : */
666 if ((ptr = strrchr(path, DIRSEP_CHAR))
667 || (ptr = strchr(path, ':')))
669 /* If there was no terminating "\" or ":",
670 * the filename probably does not
671 * contain the path; so we just don't add it
672 * to the list in that case
675 if (strlen(path) < (sizeof(path) - 1 - 7))
677 strcat(path, "include");
680 path[len] = DIRSEP_CHAR;
681 path[len + 1] = '\0';
689 printf("Usage: embryo_cc <filename> [options]\n\n");
690 printf("Options:\n");
693 (" -A<num> alignment in bytes of the data segment and the\
697 (" -a output assembler code (skip code generation\
701 (" -C[+/-] compact encoding for output file (default=%c)\n",
702 sc_compress ? '+' : '-');
703 printf(" -c8 [default] a character is 8-bits\
704 (ASCII/ISO Latin-1)\n");
706 printf(" -c16 a character is 16-bits (Unicode)\n");
707 #if defined dos_setdrive
708 printf(" -Dpath active directory path\n");
711 (" -d0 no symbolic information, no run-time checks\n");
712 printf(" -d1 [default] run-time checks, no symbolic\
715 (" -d2 full debug information and dynamic checking\n");
716 printf(" -d3 full debug information, dynamic checking,\
719 printf(" -i <name> path for include files\n");
721 printf(" -l create list file (preprocess only)\n");
723 printf(" -o <name> set base name of output file\n");
726 (" -P[+/-] strings are \"packed\" by default (default=%c)\n",
727 sc_packstr ? '+' : '-');
728 printf(" -p<name> set name of \"prefix\" file\n");
733 (" -S <num> stack/heap size in cells (default=%d, min=65)\n",
736 printf(" -s<num> skip lines from the input file\n");
738 (" -t<num> TAB indent size (in character positions)\n");
739 printf(" -\\ use '\\' for escape characters\n");
740 printf(" -^ use '^' for escape characters\n");
741 printf(" -;[+/-] require a semicolon to end each statement\
742 (default=%c)\n", sc_needsemicolon ? '+' : '-');
745 (" sym=val define constant \"sym\" with value \"val\"\n");
746 printf(" sym= define constant \"sym\" with value 0\n");
748 longjmp(errbuf, 3); /* user abort */
756 assert(sc_status == statIDLE);
757 append_constval(&tagname_tab, "_", 0, 0); /* "untagged" */
758 append_constval(&tagname_tab, "bool", 1, 0);
760 add_constant("true", 1, sGLOBAL, 1); /* boolean flags */
761 add_constant("false", 0, sGLOBAL, 1);
762 add_constant("EOS", 0, sGLOBAL, 0); /* End Of String, or '\0' */
763 add_constant("cellbits", 32, sGLOBAL, 0);
764 add_constant("cellmax", INT_MAX, sGLOBAL, 0);
765 add_constant("cellmin", INT_MIN, sGLOBAL, 0);
766 add_constant("charbits", charbits, sGLOBAL, 0);
767 add_constant("charmin", 0, sGLOBAL, 0);
768 add_constant("charmax", (charbits == 16) ? 0xffff : 0xff, sGLOBAL, 0);
770 add_constant("__Small", VERSION_INT, sGLOBAL, 0);
773 if ((sc_debug & (sCHKBOUNDS | sSYMBOLIC)) == (sCHKBOUNDS | sSYMBOLIC))
775 else if ((sc_debug & sCHKBOUNDS) == sCHKBOUNDS)
777 add_constant("debug", debug, sGLOBAL, 0);
780 /* parse - process all input text
782 * At this level, only static declarations and function definitions
788 int tok, tag, fconst, fstock, fstatic;
794 /* first try whether a declaration possibly is native or public */
795 tok = lex(&val, &str); /* read in (new) token */
802 fconst = matchtoken(tCONST);
803 declglb(NULL, 0, FALSE, FALSE, FALSE, fconst);
806 /* This can be a static function or a static global variable;
807 * we know which of the two as soon as we have parsed up to the
808 * point where an opening parenthesis of a function would be
809 * expected. To back out after deciding it was a declaration of
810 * a static variable after all, we have to store the symbol name
813 fstock = matchtoken(tSTOCK);
814 fconst = matchtoken(tCONST);
815 tag = sc_addtag(NULL);
816 tok = lex(&val, &str);
817 if (tok == tNATIVE || tok == tPUBLIC)
819 error(42); /* invalid combination of class specifiers */
822 declfuncvar(tok, str, tag, FALSE, TRUE, fstock, fconst);
831 /* This can be a public function or a public variable;
832 * see the comment above (for static functions/variables)
835 fconst = matchtoken(tCONST);
836 tag = sc_addtag(NULL);
837 tok = lex(&val, &str);
838 if (tok == tNATIVE || tok == tSTOCK || tok == tSTATIC)
840 error(42); /* invalid combination of class specifiers */
843 declfuncvar(tok, str, tag, TRUE, FALSE, FALSE, fconst);
846 /* This can be a stock function or a stock *global) variable;
847 * see the comment above (for static functions/variables) for
850 fstatic = matchtoken(tSTATIC);
851 fconst = matchtoken(tCONST);
852 tag = sc_addtag(NULL);
853 tok = lex(&val, &str);
854 if (tok == tNATIVE || tok == tPUBLIC)
856 error(42); /* invalid combination of class specifiers */
859 declfuncvar(tok, str, tag, FALSE, fstatic, TRUE, fconst);
865 if (!newfunc(NULL, -1, FALSE, FALSE, FALSE))
867 error(10); /* illegal function or declaration */
868 lexclr(TRUE); /* drop the rest of the line */
872 funcstub(TRUE); /* create a dummy function */
878 error(54); /* unmatched closing brace */
881 error(55); /* start of function body without function header */
886 error(10); /* illegal function or declaration */
887 lexclr(TRUE); /* drop the rest of the line */
895 * Dump the literal pool (strings etc.)
897 * Global references: litidx (referred to only)
907 /* should be in the data segment */
910 j = 16; /* 16 values per line */
911 while (j && k < litidx)
913 outval(litq[k], FALSE);
917 if (j == 0 || k >= litidx)
918 stgwrite("\n"); /* force a newline after 10 dumps */
919 /* Note: stgwrite() buffers a line until it is complete. It recognizes
920 * the end of line as a sequence of "\n\0", so something like "\n\t"
921 * so should not be passed to stgwrite().
929 * Dump zero's for default initial values
945 stgwrite((i == 0 || count == 0) ? "\n" : " ");
946 if (i == 0 && count > 0)
952 aligndata(int numbytes)
954 if ((((glb_declared + litidx) * sizeof(cell)) % numbytes) != 0)
956 while ((((glb_declared + litidx) * sizeof(cell)) % numbytes) != 0)
963 declfuncvar(int tok, char *symname, int tag, int fpublic, int fstatic,
964 int fstock, int fconst)
966 char name[sNAMEMAX + 1];
968 if (tok != tSYMBOL && tok != tOPERATOR)
971 error(20, symname); /* invalid symbol name */
974 if (tok == tOPERATOR)
977 if (!newfunc(NULL, tag, fpublic, fstatic, fstock))
978 error(10); /* illegal function or declaration */
982 assert(strlen(symname) <= sNAMEMAX);
983 strcpy(name, symname);
984 if (fconst || !newfunc(name, tag, fpublic, fstatic, fstock))
985 declglb(name, tag, fpublic, fstatic, fstock, fconst);
986 /* if not a static function, try a static variable */
990 /* declglb - declare global symbols
992 * Declare a static (global) variable. Global variables are stored in
995 * global references: glb_declared (altered)
998 declglb(char *firstname, int firsttag, int fpublic, int fstatic,
999 int stock, int fconst)
1001 int ident, tag, ispublic;
1002 int idxtag[sDIMEN_MAX];
1003 char name[sNAMEMAX + 1];
1004 cell val, size, cidx;
1006 int dim[sDIMEN_MAX];
1015 filenum = fcurrent; /* save file number at the start of the
1019 size = 1; /* single size (no array) */
1020 numdim = 0; /* no dimensions */
1024 assert(strlen(firstname) <= sNAMEMAX);
1025 strcpy(name, firstname); /* save symbol name */
1031 tag = sc_addtag(NULL);
1032 if (lex(&val, &str) != tSYMBOL) /* read in (new) token */
1033 error(20, str); /* invalid symbol name */
1034 assert(strlen(str) <= sNAMEMAX);
1035 strcpy(name, str); /* save symbol name */
1037 sym = findglb(name);
1039 sym = findconst(name);
1040 if (sym && (sym->usage & uDEFINE) != 0)
1041 error(21, name); /* symbol already defined */
1043 if (name[0] == PUBLIC_CHAR)
1045 ispublic = TRUE; /* implicitly public variable */
1046 if (stock || fstatic)
1047 error(42); /* invalid combination of class specifiers */
1049 while (matchtoken('['))
1052 if (numdim == sDIMEN_MAX)
1054 error(53); /* exceeding maximum number of dimensions */
1057 if (numdim > 0 && dim[numdim - 1] == 0)
1058 error(52); /* only last dimension may be variable length */
1059 size = needsub(&idxtag[numdim]); /* get size; size==0 for
1061 #if INT_MAX < LONG_MAX
1063 error(105); /* overflow, exceeding capacity */
1066 error(56, name); /* arrays cannot be public */
1067 dim[numdim++] = (int)size;
1069 /* if this variable is never used (which can be detected only in
1070 * the second stage), shut off code generation; make an exception
1071 * for public variables
1073 cidx = 0; /* only to avoid a compiler warning */
1074 if (sc_status == statWRITE && sym
1075 && (sym->usage & (uREAD | uWRITTEN | uPUBLIC)) == 0)
1077 sc_status = statSKIP;
1080 glbdecl = glb_declared;
1083 defsymbol(name, ident, sGLOBAL, sizeof(cell) * glb_declared, tag);
1084 begdseg(); /* real (initialized) data in data segment */
1085 assert(litidx == 0); /* literal queue should be empty */
1089 aligndata(sc_dataalign);
1090 dumplits(); /* dump the literal queue */
1091 sc_alignnext = FALSE;
1092 litidx = 0; /* global initial data is dumped, so restart at zero */
1094 initials(ident, tag, &size, dim, numdim); /* stores values in
1095 * the literal queue */
1098 dumplits(); /* dump the literal queue */
1099 dumpzero((int)size - litidx);
1102 { /* define only if not yet defined */
1104 addvariable(name, sizeof(cell) * glb_declared, ident, sGLOBAL,
1105 tag, dim, numdim, idxtag);
1108 { /* if declared but not yet defined, adjust the
1109 * variable's address */
1110 sym->addr = sizeof(cell) * glb_declared;
1111 sym->usage |= uDEFINE;
1114 sym->usage |= uPUBLIC;
1116 sym->usage |= uCONST;
1118 sym->usage |= uSTOCK;
1120 sym->fnumber = filenum;
1121 if (ident == iARRAY)
1122 for (level = 0; level < numdim; level++)
1123 symbolrange(level, dim[level]);
1124 if (sc_status == statSKIP)
1126 sc_status = statWRITE;
1128 assert(glb_declared == glbdecl);
1132 glb_declared += (int)size; /* add total number of cells */
1135 while (matchtoken(',')); /* enddo *//* more? */
1136 needtoken(tTERM); /* if not comma, must be semicolumn */
1139 /* declloc - declare local symbols
1141 * Declare local (automatic) variables. Since these variables are
1142 * relative to the STACK, there is no switch to the DATA segment.
1143 * These variables cannot be initialized either.
1145 * global references: declared (altered)
1146 * funcstatus (referred to only)
1149 declloc(int fstatic)
1152 int idxtag[sDIMEN_MAX];
1153 char name[sNAMEMAX + 1];
1157 value lval = { NULL, 0, 0, 0, 0, NULL };
1159 int dim[sDIMEN_MAX];
1163 fconst = matchtoken(tCONST);
1168 numdim = 0; /* no dimensions */
1169 tag = sc_addtag(NULL);
1170 if (lex(&val, &str) != tSYMBOL) /* read in (new) token */
1171 error(20, str); /* invalid symbol name */
1172 assert(strlen(str) <= sNAMEMAX);
1173 strcpy(name, str); /* save symbol name */
1174 if (name[0] == PUBLIC_CHAR)
1175 error(56, name); /* local variables cannot be public */
1176 /* Note: block locals may be named identical to locals at higher
1177 * compound blocks (as with standard C); so we must check (and add)
1178 * the "nesting level" of local variables to verify the
1179 * multi-definition of symbols.
1181 if ((sym = findloc(name)) && sym->compound == nestlevel)
1182 error(21, name); /* symbol already defined */
1183 /* Although valid, a local variable whose name is equal to that
1184 * of a global variable or to that of a local variable at a lower
1185 * level might indicate a bug.
1187 if (((sym = findloc(name)) && sym->compound != nestlevel)
1189 error(219, name); /* variable shadows another symbol */
1190 while (matchtoken('['))
1193 if (numdim == sDIMEN_MAX)
1195 error(53); /* exceeding maximum number of dimensions */
1198 if (numdim > 0 && dim[numdim - 1] == 0)
1199 error(52); /* only last dimension may be variable length */
1200 size = needsub(&idxtag[numdim]); /* get size; size==0 for "var[]" */
1201 #if INT_MAX < LONG_MAX
1203 error(105); /* overflow, exceeding capacity */
1205 dim[numdim++] = (int)size;
1207 if (ident == iARRAY || fstatic)
1211 aligndata(sc_dataalign);
1212 sc_alignnext = FALSE;
1214 cur_lit = litidx; /* save current index in the literal table */
1215 initials(ident, tag, &size, dim, numdim);
1217 return ident; /* error message already given */
1221 /* reserve memory (on the stack) for the variable */
1224 /* write zeros for uninitialized fields */
1225 while (litidx < cur_lit + size)
1228 addvariable(name, (cur_lit + glb_declared) * sizeof(cell),
1229 ident, sSTATIC, tag, dim, numdim, idxtag);
1230 defsymbol(name, ident, sSTATIC,
1231 (cur_lit + glb_declared) * sizeof(cell), tag);
1235 declared += (int)size; /* variables are put on stack,
1236 * adjust "declared" */
1238 addvariable(name, -declared * sizeof(cell), ident, sLOCAL, tag,
1239 dim, numdim, idxtag);
1240 defsymbol(name, ident, sLOCAL, -declared * sizeof(cell), tag);
1241 modstk(-(int)size * sizeof(cell));
1243 /* now that we have reserved memory for the variable, we can
1244 * proceed to initialize it */
1245 sym->compound = nestlevel; /* for multiple declaration/shadowing */
1247 sym->usage |= uCONST;
1248 if (ident == iARRAY)
1249 for (level = 0; level < numdim; level++)
1250 symbolrange(level, dim[level]);
1252 { /* static variables already initialized */
1253 if (ident == iVARIABLE)
1255 /* simple variable, also supports initialization */
1256 int ctag = tag; /* set to "tag" by default */
1257 int explicit_init = FALSE; /* is the variable explicitly
1259 if (matchtoken('='))
1261 doexpr(FALSE, FALSE, FALSE, FALSE, &ctag, TRUE);
1262 explicit_init = TRUE;
1266 const1(0); /* uninitialized variable, set to zero */
1268 /* now try to save the value (still in PRI) in the variable */
1270 lval.ident = iVARIABLE;
1273 check_userop(NULL, ctag, lval.tag, 2, NULL, &ctag);
1275 endexpr(TRUE); /* full expression ends after the store */
1276 if (!matchtag(tag, ctag, TRUE))
1277 error(213); /* tag mismatch */
1278 /* if the variable was not explicitly initialized, reset the
1279 * "uWRITTEN" flag that store() set */
1281 sym->usage &= ~uWRITTEN;
1286 if (litidx - cur_lit < size)
1287 fillarray(sym, size * sizeof(cell), 0);
1288 if (cur_lit < litidx)
1290 /* check whether the complete array is set to a single value;
1291 * if it is, more compact code can be generated */
1292 cell first = litq[cur_lit];
1295 for (i = cur_lit; i < litidx && litq[i] == first; i++)
1299 /* all values are the same */
1300 fillarray(sym, (litidx - cur_lit) * sizeof(cell),
1302 litidx = cur_lit; /* reset literal table */
1306 /* copy the literals to the array */
1307 const1((cur_lit + glb_declared) * sizeof(cell));
1308 copyarray(sym, (litidx - cur_lit) * sizeof(cell));
1314 while (matchtoken(',')); /* enddo *//* more? */
1315 needtoken(tTERM); /* if not comma, must be semicolumn */
1320 calc_arraysize(int dim[], int numdim, int cur)
1324 return dim[cur] + (dim[cur] * calc_arraysize(dim, numdim, cur + 1));
1329 * Initialize global objects and local arrays.
1330 * size==array cells (count), if 0 on input, the routine counts
1331 * the number of elements
1332 * tag==required tagname id (not the returned tag)
1334 * Global references: litidx (altered)
1337 initials(int ident, int tag, cell * size, int dim[], int numdim)
1340 int curlit = litidx;
1343 if (!matchtoken('='))
1345 if (ident == iARRAY && dim[numdim - 1] == 0)
1347 /* declared as "myvar[];" which is senseless (note: this *does* make
1348 * sense in the case of a iREFARRAY, which is a function parameter)
1350 error(9); /* array has zero length -> invalid size */
1354 /* initialize the indirection tables */
1356 #error Array algorithms for more than 2 dimensions are not implemented
1358 assert(numdim == 2);
1359 *size = calc_arraysize(dim, numdim, 0);
1360 for (d = 0; d < dim[0]; d++)
1361 stowlit((dim[0] + d * (dim[1] - 1)) * sizeof(cell));
1366 if (ident == iVARIABLE)
1370 if (!matchtag(tag, ctag, TRUE))
1371 error(213); /* tag mismatch */
1378 *size = initvector(ident, tag, dim[0], FALSE);
1384 /* The simple algorithm below only works for arrays with one or
1385 * two dimensions. This should be some recursive algorithm.
1387 if (dim[numdim - 1] != 0)
1388 /* set size to (known) full size */
1389 *size = calc_arraysize(dim, numdim, 0);
1390 /* dump indirection tables */
1391 for (d = 0; d < dim[0]; d++)
1393 /* now dump individual vectors */
1396 for (d = 0; d < dim[0]; d++)
1398 litq[curlit + d] = offs * sizeof(cell);
1399 dsize = initvector(ident, tag, dim[1], TRUE);
1403 if (matchtoken('{') || matchtoken(tSTRING))
1404 /* expect a '{' or a string */
1415 *size = litidx - curlit; /* number of elements defined */
1419 * Initialize a single dimensional array
1422 initvector(int ident, int tag, cell size, int fillzero)
1424 cell prev1 = 0, prev2 = 0;
1427 int curlit = litidx;
1429 assert(ident == iARRAY || ident == iREFARRAY);
1430 if (matchtoken('{'))
1434 if (matchtoken('}'))
1435 { /* to allow for trailing ',' after the initialization */
1439 if ((ellips = matchtoken(tELLIPS)) != 0)
1442 prev1 = init(ident, &ctag);
1443 if (!matchtag(tag, ctag, TRUE))
1444 error(213); /* tag mismatch */
1446 while (matchtoken(',')); /* do */
1452 if (!matchtag(tag, ctag, TRUE))
1453 error(213); /* tagname mismatch */
1455 /* fill up the literal queue with a series */
1459 ((litidx - curlit) == 1) ? (cell) 0 : prev1 - prev2;
1460 if (size == 0 || (litidx - curlit) == 0)
1461 error(41); /* invalid ellipsis, array size unknown */
1462 else if ((litidx - curlit) == (int)size)
1463 error(18); /* initialisation data exceeds declared size */
1464 while ((litidx - curlit) < (int)size)
1470 if (fillzero && size > 0)
1472 while ((litidx - curlit) < (int)size)
1477 size = litidx - curlit; /* number of elements defined */
1479 else if (litidx - curlit > (int)size)
1480 { /* e.g. "myvar[3]={1,2,3,4};" */
1481 error(18); /* initialisation data exceeds declared size */
1482 litidx = (int)size + curlit; /* avoid overflow in memory moves */
1489 * Evaluate one initializer.
1492 init(int ident, int *tag)
1496 if (matchtoken(tSTRING))
1498 /* lex() automatically stores strings in the literal table (and
1499 * increases "litidx") */
1500 if (ident == iVARIABLE)
1502 error(6); /* must be assigned to an array */
1503 litidx = 1; /* reset literal queue */
1507 else if (constexpr(&i, tag))
1509 stowlit(i); /* store expression result in literal table */
1516 * Get required array size
1524 if (matchtoken(']')) /* we've already seen "[" */
1525 return 0; /* null size (like "char msg[]") */
1526 constexpr(&val, tag); /* get value (must be constant expression) */
1529 error(9); /* negative array size is invalid; assumed zero */
1533 return val; /* return array size */
1536 /* decl_const - declare a single constant
1540 decl_const(int vclass)
1542 char constname[sNAMEMAX + 1];
1548 tag = sc_addtag(NULL);
1549 if (lex(&val, &str) != tSYMBOL) /* read in (new) token */
1550 error(20, str); /* invalid symbol name */
1551 symbolline = fline; /* save line where symbol was found */
1552 strcpy(constname, str); /* save symbol name */
1554 constexpr(&val, &exprtag); /* get value */
1556 /* add_constant() checks for duplicate definitions */
1557 if (!matchtag(tag, exprtag, FALSE))
1559 /* temporarily reset the line number to where the symbol was
1561 int orgfline = fline;
1564 error(213); /* tagname mismatch */
1567 add_constant(constname, val, vclass, tag);
1570 /* decl_enum - declare enumerated constants
1574 decl_enum(int vclass)
1576 char enumname[sNAMEMAX + 1], constname[sNAMEMAX + 1];
1577 cell val, value, size;
1579 int tok, tag, explicittag;
1580 cell increment, multiplier;
1582 /* get an explicit tag, if any (we need to remember whether an
1583 * explicit tag was passed, even if that explicit tag was "_:", so we
1584 * cannot call sc_addtag() here
1586 if (lex(&val, &str) == tLABEL)
1588 tag = sc_addtag(str);
1595 explicittag = FALSE;
1598 /* get optional enum name (also serves as a tag if no explicit
1600 if (lex(&val, &str) == tSYMBOL)
1601 { /* read in (new) token */
1602 strcpy(enumname, str); /* save enum name (last constant) */
1604 tag = sc_addtag(enumname);
1608 lexpush(); /* analyze again */
1612 /* get increment and multiplier */
1615 if (matchtoken('('))
1617 if (matchtoken(taADD))
1619 constexpr(&increment, NULL);
1621 else if (matchtoken(taMULT))
1623 constexpr(&multiplier, NULL);
1625 else if (matchtoken(taSHL))
1627 constexpr(&val, NULL);
1635 /* go through all constants */
1636 value = 0; /* default starting value */
1639 if (matchtoken('}'))
1640 { /* quick exit if '}' follows ',' */
1644 tok = lex(&val, &str); /* read in (new) token */
1645 if (tok != tSYMBOL && tok != tLABEL)
1646 error(20, str); /* invalid symbol name */
1647 strcpy(constname, str); /* save symbol name */
1648 size = increment; /* default increment of 'val' */
1649 if (tok == tLABEL || matchtoken(':'))
1650 constexpr(&size, NULL); /* get size */
1651 if (matchtoken('='))
1652 constexpr(&value, NULL); /* get value */
1653 /* add_constant() checks whether a variable (global or local) or
1654 * a constant with the same name already exists */
1655 add_constant(constname, value, vclass, tag);
1656 if (multiplier == 1)
1659 value *= size * multiplier;
1661 while (matchtoken(','));
1662 needtoken('}'); /* terminates the constant list */
1663 matchtoken(';'); /* eat an optional ; */
1665 /* set the enum name to the last value plus one */
1666 if (enumname[0] != '\0')
1667 add_constant(enumname, value, vclass, tag);
1671 * Finds a function in the global symbol table or creates a new entry.
1672 * It does some basic processing and error checking.
1675 fetchfunc(char *name, int tag)
1681 if ((sc_debug & sSYMBOLIC) != 0)
1683 offset += opcodes(1) + opargs(3) + nameincells(name);
1684 /* ^^^ The address for the symbol is the code address. But the
1685 * "symbol" instruction itself generates code. Therefore the
1686 * offset is pre-adjusted to the value it will have after the
1687 * symbol instruction.
1690 if ((sym = findglb(name)))
1691 { /* already in symbol table? */
1692 if (sym->ident != iFUNCTN)
1694 error(21, name); /* yes, but not as a function */
1695 return NULL; /* make sure the old symbol is not damaged */
1697 else if ((sym->usage & uDEFINE) != 0)
1699 error(21, name); /* yes, and it's already defined */
1701 else if ((sym->usage & uNATIVE) != 0)
1703 error(21, name); /* yes, and it is an native */
1705 assert(sym->vclass == sGLOBAL);
1706 if ((sym->usage & uDEFINE) == 0)
1708 /* as long as the function stays undefined, update the address
1716 /* don't set the "uDEFINE" flag; it may be a prototype */
1717 sym = addsym(name, offset, iFUNCTN, sGLOBAL, tag, 0);
1718 /* assume no arguments */
1719 sym->dim.arglist = (arginfo *) malloc(1 * sizeof(arginfo));
1720 sym->dim.arglist[0].ident = 0;
1721 /* set library ID to NULL (only for native functions) */
1727 /* This routine adds symbolic information for each argument.
1734 /* At this point, no local variables have been declared. All
1735 * local symbols are function arguments.
1740 assert(sym->ident != iLABEL);
1741 assert(sym->vclass == sLOCAL);
1742 defsymbol(sym->name, sym->ident, sLOCAL, sym->addr, sym->tag);
1743 if (sym->ident == iREFARRAY)
1749 symbolrange(sub->dim.array.level, sub->dim.array.length);
1750 sub = finddepend(sub);
1758 operatorname(char *name)
1764 assert(name != NULL);
1766 /* check the operator */
1767 opertok = lex(&val, &str);
1780 name[0] = (char)opertok;
1803 error(61); /* operator cannot be redefined
1804 * (or bad operator name) */
1812 operatoradjust(int opertok, symbol * sym, char *opername, int resulttag)
1814 int tags[2] = { 0, 0 };
1817 char tmpname[sNAMEMAX + 1];
1823 /* count arguments and save (first two) tags */
1824 while (arg = &sym->dim.arglist[count], arg->ident != 0)
1828 if (arg->numtags > 1)
1829 error(65, count + 1); /* function argument may only have
1831 else if (arg->numtags == 1)
1832 tags[count] = arg->tags[0];
1834 if (opertok == '~' && count == 0)
1836 if (arg->ident != iREFARRAY)
1837 error(73, arg->name); /* must be an array argument */
1841 if (arg->ident != iVARIABLE)
1842 error(66, arg->name); /* must be non-reference argument */
1844 if (arg->hasdefault)
1845 error(59, arg->name); /* arguments of an operator may not
1846 * have a default value */
1850 /* for '!', '++' and '--', count must be 1
1851 * for '-', count may be 1 or 2
1852 * for '=', count must be 1, and the resulttag is also important
1853 * for all other (binary) operators and the special '~'
1854 * operator, count must be 2
1863 error(62); /* number or placement of the operands does
1864 * not fit the operator */
1867 if (count != 1 && count != 2)
1868 error(62); /* number or placement of the operands does
1869 * not fit the operator */
1873 error(62); /* number or placement of the operands does
1874 * not fit the operator */
1878 && ((opertok != '=' && tags[1] == 0) || (opertok == '=' && resulttag == 0)))
1879 error(64); /* cannot change predefined operators */
1881 /* change the operator name */
1882 assert(opername[0] != '\0');
1883 operator_symname(tmpname, opername, tags[0], tags[1], count, resulttag);
1884 if ((oldsym = findglb(tmpname)))
1888 if ((oldsym->usage & uDEFINE) != 0)
1890 char errname[2 * sNAMEMAX + 16];
1892 funcdisplayname(errname, tmpname);
1893 error(21, errname); /* symbol already defined */
1895 sym->usage |= oldsym->usage; /* copy flags from the previous
1897 for (i = 0; i < oldsym->numrefers; i++)
1898 if (oldsym->refer[i])
1899 refer_symbol(sym, oldsym->refer[i]);
1900 delete_symbol(&glbtab, oldsym);
1902 if ((sc_debug & sSYMBOLIC) != 0)
1903 sym->addr += nameincells(tmpname) - nameincells(sym->name);
1904 strcpy(sym->name, tmpname);
1905 sym->hash = namehash(sym->name); /* calculate new hash */
1907 /* operators should return a value, except the '~' operator */
1909 sym->usage |= uRETVALUE;
1915 check_operatortag(int opertok, int resulttag, char *opername)
1917 assert(opername != NULL && opername[0] != '\0');
1927 if (resulttag != sc_addtag("bool"))
1929 error(63, opername, "bool:"); /* operator X requires
1930 * a "bool:" result tag */
1937 error(63, opername, "_:"); /* operator "~" requires
1938 * a "_:" result tag */
1947 tag2str(char *dest, int tag)
1951 sprintf(dest, "0%x", tag);
1952 return sc_isdigit(dest[1]) ? &dest[1] : dest;
1956 operator_symname(char *symname, char *opername, int tag1, int tag2,
1957 int numtags, int resulttag)
1959 char tagstr1[10], tagstr2[10];
1962 assert(numtags >= 1 && numtags <= 2);
1963 opertok = (opername[1] == '\0') ? opername[0] : 0;
1965 sprintf(symname, "%s%s%s", tag2str(tagstr1, resulttag), opername,
1966 tag2str(tagstr2, tag1));
1967 else if (numtags == 1 || opertok == '~')
1968 sprintf(symname, "%s%s", opername, tag2str(tagstr1, tag1));
1970 sprintf(symname, "%s%s%s", tag2str(tagstr1, tag1), opername,
1971 tag2str(tagstr2, tag2));
1976 parse_funcname(char *fname, int *tag1, int *tag2, char *opname)
1981 /* tags are only positive, so if the function name starts with a '-',
1982 * the operator is an unary '-' or '--' operator.
1992 *tag1 = (int)strtol(fname, &ptr, 16);
1993 unary = ptr == fname; /* unary operator if it doesn't start
1994 * with a tag name */
1996 assert(!unary || *tag1 == 0);
1997 assert(*ptr != '\0');
1998 for (name = opname; !sc_isdigit(*ptr);)
2001 *tag2 = (int)strtol(ptr, NULL, 16);
2006 funcdisplayname(char *dest, char *funcname)
2010 constvalue *tagsym[2];
2013 if (sc_isalpha(*funcname) || *funcname == '_' || *funcname == PUBLIC_CHAR
2014 || *funcname == '\0')
2016 if (dest != funcname)
2017 strcpy(dest, funcname);
2021 unary = parse_funcname(funcname, &tags[0], &tags[1], opname);
2022 tagsym[1] = find_constval_byval(&tagname_tab, tags[1]);
2023 assert(tagsym[1] != NULL);
2026 sprintf(dest, "operator%s(%s:)", opname, tagsym[1]->name);
2030 tagsym[0] = find_constval_byval(&tagname_tab, tags[0]);
2031 /* special case: the assignment operator has the return value
2033 if (opname[0] == '=' && opname[1] == '\0')
2034 sprintf(dest, "%s:operator%s(%s:)", tagsym[0]->name, opname,
2037 sprintf(dest, "operator%s(%s:,%s:)", opname, tagsym[0]->name,
2044 funcstub(int native)
2049 char symbolname[sNAMEMAX + 1];
2055 litidx = 0; /* clear the literal pool */
2057 tag = sc_addtag(NULL);
2058 tok = lex(&val, &str);
2061 if (tok == tPUBLIC || tok == tSTOCK || tok == tSTATIC ||
2062 (tok == tSYMBOL && *str == PUBLIC_CHAR))
2063 error(42); /* invalid combination of class specifiers */
2067 if (tok == tPUBLIC || tok == tSTATIC)
2068 tok = lex(&val, &str);
2070 if (tok == tOPERATOR)
2072 opertok = operatorname(symbolname);
2074 return; /* error message already given */
2075 check_operatortag(opertok, tag, symbolname);
2079 if (tok != tSYMBOL && freading)
2081 error(10); /* illegal function or declaration */
2084 strcpy(symbolname, str);
2086 needtoken('('); /* only functions may be native/forward */
2088 sym = fetchfunc(symbolname, tag); /* get a pointer to the
2094 sym->usage = uNATIVE | uRETVALUE | uDEFINE;
2095 sym->x.lib = curlibrary;
2099 /* "declargs()" found the ")" */
2100 if (!operatoradjust(opertok, sym, symbolname, tag))
2101 sym->usage &= ~uDEFINE;
2102 /* for a native operator, also need to specify an "exported"
2103 * function name; for a native function, this is optional
2110 lexpush(); /* push back, for matchtoken() to retrieve again */
2112 if (matchtoken('='))
2114 /* allow number or symbol */
2115 if (matchtoken(tSYMBOL))
2117 tokeninfo(&val, &str);
2118 if (strlen(str) > sEXPMAX)
2120 error(220, str, sEXPMAX);
2121 str[sEXPMAX] = '\0';
2123 insert_alias(sym->name, str);
2127 constexpr(&val, NULL);
2130 * ?? Must mark this address, so that it won't be generated again
2131 * and it won't be written to the output file. At the moment,
2132 * I have assumed that this syntax is only valid if val < 0.
2133 * To properly mix "normal" native functions and indexed native
2134 * functions, one should use negative indices anyway.
2135 * Special code for a negative index in sym->addr exists in
2136 * SC4.C (ffcall()) and in SC6.C (the loops for counting the
2137 * number of native variables and for writing them).
2144 litidx = 0; /* clear the literal pool */
2145 /* clear local variables queue */
2146 delete_symbols(&loctab, 0, TRUE, TRUE);
2149 /* newfunc - begin a function
2151 * This routine is called from "parse" and tries to make a function
2152 * out of the following text
2154 * Global references: funcstatus,lastst,litidx
2157 * declared (altered)
2158 * glb_declared (altered)
2159 * sc_alignnext (altered)
2162 newfunc(char *firstname, int firsttag, int fpublic, int fstatic, int stock)
2165 int argcnt, tok, tag, funcline;
2166 int opertok, opererror;
2167 char symbolname[sNAMEMAX + 1];
2169 cell val, cidx, glbdecl;
2172 litidx = 0; /* clear the literal pool ??? */
2174 lastst = 0; /* no statement yet */
2175 cidx = 0; /* just to avoid compiler warnings */
2177 filenum = fcurrent; /* save file number at start of declaration */
2181 assert(strlen(firstname) <= sNAMEMAX);
2182 strcpy(symbolname, firstname); /* save symbol name */
2187 tag = (firsttag >= 0) ? firsttag : sc_addtag(NULL);
2188 tok = lex(&val, &str);
2190 if (tok == tNATIVE || (tok == tPUBLIC && stock))
2191 error(42); /* invalid combination of class specifiers */
2192 if (tok == tOPERATOR)
2194 opertok = operatorname(symbolname);
2196 return TRUE; /* error message already given */
2197 check_operatortag(opertok, tag, symbolname);
2201 if (tok != tSYMBOL && freading)
2203 error(20, str); /* invalid symbol name */
2206 assert(strlen(str) <= sNAMEMAX);
2207 strcpy(symbolname, str);
2210 /* check whether this is a function or a variable declaration */
2211 if (!matchtoken('('))
2213 /* so it is a function, proceed */
2214 funcline = fline; /* save line at which the function is defined */
2215 if (symbolname[0] == PUBLIC_CHAR)
2217 fpublic = TRUE; /* implicitly public function */
2219 error(42); /* invalid combination of class specifiers */
2221 sym = fetchfunc(symbolname, tag); /* get a pointer to the
2226 sym->usage |= uPUBLIC;
2228 sym->fnumber = filenum;
2229 /* declare all arguments */
2230 argcnt = declargs(sym);
2231 opererror = !operatoradjust(opertok, sym, symbolname, tag);
2232 if (strcmp(symbolname, uMAINFUNC) == 0)
2235 error(5); /* "main()" function may not have any arguments */
2236 sym->usage |= uREAD; /* "main()" is the program's entry point:
2239 /* "declargs()" found the ")"; if a ";" appears after this, it was a
2241 if (matchtoken(';'))
2243 if (!sc_needsemicolon)
2244 error(218); /* old style prototypes used with optional
2246 delete_symbols(&loctab, 0, TRUE, TRUE); /* prototype is done;
2247 * forget everything */
2250 /* so it is not a prototype, proceed */
2251 /* if this is a function that is not referred to (this can only be
2252 * detected in the second stage), shut code generation off */
2253 if (sc_status == statWRITE && (sym->usage & uREAD) == 0)
2255 sc_status = statSKIP;
2257 glbdecl = glb_declared;
2260 sym->usage |= uDEFINE; /* set the definition flag */
2262 sym->usage |= uREAD; /* public functions are always "used" */
2264 sym->usage |= uSTOCK;
2265 if (opertok != 0 && opererror)
2266 sym->usage &= ~uDEFINE;
2267 defsymbol(sym->name, iFUNCTN, sGLOBAL,
2268 code_idx + opcodes(1) + opargs(3) + nameincells(sym->name), tag);
2269 /* ^^^ The address for the symbol is the code address. But the
2270 * "symbol" instruction itself generates code. Therefore the
2271 * offset is pre-adjusted to the value it will have after the
2272 * symbol instruction.
2274 startfunc(sym->name); /* creates stack frame */
2275 if ((sc_debug & sSYMBOLIC) != 0)
2276 setline(funcline, fcurrent);
2279 alignframe(sc_dataalign);
2280 sc_alignnext = FALSE;
2282 declared = 0; /* number of local cells */
2283 rettype = (sym->usage & uRETVALUE); /* set "return type" variable */
2285 define_args(); /* add the symbolic info for the function arguments */
2286 statement(NULL, FALSE);
2287 if ((rettype & uRETVALUE) != 0)
2288 sym->usage |= uRETVALUE;
2291 /* This happens only in a very special (and useless) case, where a
2292 * function has only a single statement in its body (no compound
2293 * block) and that statement declares a new variable
2295 modstk((int)declared * sizeof(cell)); /* remove all local
2299 if ((lastst != tRETURN) && (lastst != tGOTO))
2303 if ((sym->usage & uRETVALUE) != 0)
2305 char symname[2 * sNAMEMAX + 16]; /* allow space for user
2306 * defined operators */
2307 funcdisplayname(symname, sym->name);
2308 error(209, symname); /* function should return a value */
2313 { /* if there are literals defined */
2314 glb_declared += litidx;
2315 begdseg(); /* flip to DATA segment */
2316 dumplits(); /* dump literal strings */
2319 testsymbols(&loctab, 0, TRUE, TRUE); /* test for unused arguments
2321 delete_symbols(&loctab, 0, TRUE, TRUE); /* clear local variables
2323 assert(loctab.next == NULL);
2325 if (sc_status == statSKIP)
2327 sc_status = statWRITE;
2329 glb_declared = glbdecl;
2335 argcompare(arginfo * a1, arginfo * a2)
2339 result = strcmp(a1->name, a2->name) == 0;
2341 result = a1->ident == a2->ident;
2343 result = a1->usage == a2->usage;
2345 result = a1->numtags == a2->numtags;
2350 for (i = 0; i < a1->numtags && result; i++)
2351 result = a1->tags[i] == a2->tags[i];
2354 result = a1->hasdefault == a2->hasdefault;
2357 if (a1->ident == iREFARRAY)
2360 result = a1->defvalue.array.size == a2->defvalue.array.size;
2363 a1->defvalue.array.arraysize == a2->defvalue.array.arraysize;
2364 /* also check the dimensions of both arrays */
2366 result = a1->numdim == a2->numdim;
2367 for (level = 0; result && level < a1->numdim; level++)
2368 result = a1->dim[level] == a2->dim[level];
2369 /* ??? should also check contents of the default array
2370 * (these troubles go away in a 2-pass compiler that forbids
2371 * double declarations, but Small currently does not forbid them)
2378 if ((a1->hasdefault & uSIZEOF) != 0
2379 || (a1->hasdefault & uTAGOF) != 0)
2380 result = a1->hasdefault == a2->hasdefault
2381 && strcmp(a1->defvalue.size.symname,
2382 a2->defvalue.size.symname) == 0
2383 && a1->defvalue.size.level == a2->defvalue.size.level;
2385 result = a1->defvalue.val == a2->defvalue.val;
2389 result = a1->defvalue_tag == a2->defvalue_tag;
2396 * This routine adds an entry in the local symbol table for each
2397 * argument found in the argument list.
2398 * It returns the number of arguments.
2401 declargs(symbol * sym)
2405 int argcnt, oldargcnt, tok, tags[MAXTAGS], numtags;
2407 arginfo arg, *arglist;
2408 char name[sNAMEMAX + 1];
2409 int ident, fpublic, fconst;
2412 /* if the function is already defined earlier, get the number of
2413 * arguments of the existing definition
2416 if ((sym->usage & uPROTOTYPED) != 0)
2417 while (sym->dim.arglist[oldargcnt].ident != 0)
2419 argcnt = 0; /* zero aruments up to now */
2423 fpublic = (sym->usage & uPUBLIC) != 0;
2424 /* the '(' parantheses has already been parsed */
2425 if (!matchtoken(')'))
2428 { /* there are arguments; process them */
2429 /* any legal name increases argument count (and stack offset) */
2430 tok = lex(&val, &ptr);
2437 if (ident != iVARIABLE || numtags > 0)
2438 error(1, "-identifier-", "&");
2442 if (ident != iVARIABLE || numtags > 0)
2443 error(1, "-identifier-", "const");
2448 error(1, "-identifier-", "-tagname-");
2449 tags[0] = sc_addtag(ptr);
2454 error(1, "-identifier-", "-tagname-");
2456 while (numtags < MAXTAGS)
2458 if (!matchtoken('_') && !needtoken(tSYMBOL))
2460 tokeninfo(&val, &ptr);
2461 tags[numtags++] = sc_addtag(ptr);
2462 if (matchtoken('}'))
2467 tok = tLABEL; /* for outer loop:
2468 * flag that we have seen a tagname */
2471 if (argcnt >= sMAXARGS)
2472 error(45); /* too many function arguments */
2473 strcpy(name, ptr); /* save symbol name */
2474 if (name[0] == PUBLIC_CHAR)
2475 error(56, name); /* function arguments cannot be public */
2477 tags[numtags++] = 0; /* default tag */
2479 * base + 0*sizeof(cell) == previous "base"
2480 * base + 1*sizeof(cell) == function return address
2481 * base + 2*sizeof(cell) == number of arguments
2482 * base + 3*sizeof(cell) == first argument of the function
2483 * So the offset of each argument is:
2484 * "(argcnt+3) * sizeof(cell)".
2486 doarg(name, ident, (argcnt + 3) * sizeof(cell), tags, numtags,
2487 fpublic, fconst, &arg);
2488 if (fpublic && arg.hasdefault)
2489 error(59, name); /* arguments of a public function may not
2490 * have a default value */
2491 if ((sym->usage & uPROTOTYPED) == 0)
2493 /* redimension the argument list, add the entry */
2495 (arginfo *) realloc(sym->dim.arglist,
2496 (argcnt + 2) * sizeof(arginfo));
2497 if (!sym->dim.arglist)
2498 error(103); /* insufficient memory */
2499 sym->dim.arglist[argcnt] = arg;
2500 sym->dim.arglist[argcnt + 1].ident = 0; /* keep the list
2505 /* check the argument with the earlier definition */
2506 if (argcnt > oldargcnt
2507 || !argcompare(&sym->dim.arglist[argcnt], &arg))
2508 error(25); /* function definition does not match prototype */
2509 /* may need to free default array argument and the tag list */
2510 if (arg.ident == iREFARRAY && arg.hasdefault)
2511 free(arg.defvalue.array.data);
2512 else if (arg.ident == iVARIABLE
2513 && ((arg.hasdefault & uSIZEOF) != 0
2514 || (arg.hasdefault & uTAGOF) != 0))
2515 free(arg.defvalue.size.symname);
2524 if (ident != iVARIABLE)
2525 error(10); /* illegal function or declaration */
2527 tags[numtags++] = 0; /* default tag */
2528 if ((sym->usage & uPROTOTYPED) == 0)
2530 /* redimension the argument list, add the entry iVARARGS */
2532 (arginfo *) realloc(sym->dim.arglist,
2533 (argcnt + 2) * sizeof(arginfo));
2534 if (!sym->dim.arglist)
2535 error(103); /* insufficient memory */
2536 sym->dim.arglist[argcnt + 1].ident = 0; /* keep the list
2538 sym->dim.arglist[argcnt].ident = iVARARGS;
2539 sym->dim.arglist[argcnt].hasdefault = FALSE;
2540 sym->dim.arglist[argcnt].defvalue.val = 0;
2541 sym->dim.arglist[argcnt].defvalue_tag = 0;
2542 sym->dim.arglist[argcnt].numtags = numtags;
2543 sym->dim.arglist[argcnt].tags =
2544 (int *)malloc(numtags * sizeof tags[0]);
2545 if (!sym->dim.arglist[argcnt].tags)
2546 error(103); /* insufficient memory */
2547 memcpy(sym->dim.arglist[argcnt].tags, tags,
2548 numtags * sizeof tags[0]);
2552 if (argcnt > oldargcnt
2553 || sym->dim.arglist[argcnt].ident != iVARARGS)
2554 error(25); /* function definition does not match prototype */
2559 error(10); /* illegal function or declaration */
2562 while (tok == '&' || tok == tLABEL || tok == tCONST || (tok != tELLIPS && matchtoken(','))); /* more? */
2563 /* if the next token is not ",", it should be ")" */
2566 /* resolve any "sizeof" arguments (now that all arguments are known) */
2567 assert(sym->dim.arglist != NULL);
2568 arglist = sym->dim.arglist;
2569 for (idx = 0; idx < argcnt && arglist[idx].ident != 0; idx++)
2571 if ((arglist[idx].hasdefault & uSIZEOF) != 0
2572 || (arglist[idx].hasdefault & uTAGOF) != 0)
2576 /* Find the argument with the name mentioned after the "sizeof".
2577 * Note that we cannot use findloc here because we need the
2578 * arginfo struct, not the symbol.
2580 ptr = arglist[idx].defvalue.size.symname;
2582 altidx < argcnt && strcmp(ptr, arglist[altidx].name) != 0;
2585 if (altidx >= argcnt)
2587 error(17, ptr); /* undefined symbol */
2591 /* check the level against the number of dimensions */
2592 /* the level must be zero for "tagof" values */
2593 assert(arglist[idx].defvalue.size.level == 0
2594 || (arglist[idx].hasdefault & uSIZEOF) != 0);
2595 if (arglist[idx].defvalue.size.level > 0
2596 && arglist[idx].defvalue.size.level >=
2597 arglist[altidx].numdim)
2598 error(28); /* invalid subscript */
2599 if (arglist[altidx].ident != iREFARRAY)
2601 assert(arglist[altidx].ident == iVARIABLE
2602 || arglist[altidx].ident == iREFERENCE);
2603 error(223, ptr); /* redundant sizeof */
2609 sym->usage |= uPROTOTYPED;
2610 errorset(sRESET); /* reset error flag (clear the "panic mode") */
2614 /* doarg - declare one argument type
2616 * this routine is called from "declargs()" and adds an entry in the
2617 * local symbol table for one argument. "fpublic" indicates whether
2618 * the function for this argument list is public.
2619 * The arguments themselves are never public.
2622 doarg(char *name, int ident, int offset, int tags[], int numtags,
2623 int fpublic, int fconst, arginfo * arg)
2627 int idxtag[sDIMEN_MAX];
2629 strcpy(arg->name, name);
2630 arg->hasdefault = FALSE; /* preset (most common case) */
2631 arg->defvalue.val = 0; /* clear */
2632 arg->defvalue_tag = 0;
2634 if (matchtoken('['))
2636 if (ident == iREFERENCE)
2637 error(67, name); /*illegal declaration ("&name[]" is unsupported) */
2640 if (arg->numdim == sDIMEN_MAX)
2642 error(53); /* exceeding maximum number of dimensions */
2645 /* there is no check for non-zero major dimensions here, only if
2646 * the array parameter has a default value, we enforce that all
2647 * array dimensions, except the last, are non-zero
2649 size = needsub(&idxtag[arg->numdim]); /* may be zero here,
2650 *it is a pointer anyway */
2651 #if INT_MAX < LONG_MAX
2653 error(105); /* overflow, exceeding capacity */
2655 arg->dim[arg->numdim] = (int)size;
2658 while (matchtoken('['));
2659 ident = iREFARRAY; /* "reference to array" (is a pointer) */
2660 if (matchtoken('='))
2664 lexpush(); /* initials() needs the "=" token again */
2665 assert(numtags > 0);
2666 /* for the moment, when a default value is given for the array,
2667 * all dimension sizes, except the last, must be non-zero
2668 * (function initials() requires to know the major dimensions)
2670 for (level = 0; level < arg->numdim - 1; level++)
2671 if (arg->dim[level] == 0)
2672 error(52); /* only last dimension may be variable length */
2673 initials(ident, tags[0], &size, arg->dim, arg->numdim);
2674 assert(size >= litidx);
2675 /* allocate memory to hold the initial values */
2676 arg->defvalue.array.data = (cell *) malloc(litidx * sizeof(cell));
2677 if (arg->defvalue.array.data)
2681 memcpy(arg->defvalue.array.data, litq, litidx * sizeof(cell));
2682 arg->hasdefault = TRUE; /* argument has default value */
2683 arg->defvalue.array.size = litidx;
2684 arg->defvalue.array.addr = -1;
2685 /* calculate size to reserve on the heap */
2686 arg->defvalue.array.arraysize = 1;
2687 for (i = 0; i < arg->numdim; i++)
2688 arg->defvalue.array.arraysize *= arg->dim[i];
2689 if (arg->defvalue.array.arraysize < arg->defvalue.array.size)
2690 arg->defvalue.array.arraysize = arg->defvalue.array.size;
2692 litidx = 0; /* reset */
2697 if (matchtoken('='))
2699 unsigned char size_tag_token;
2701 assert(ident == iVARIABLE || ident == iREFERENCE);
2702 arg->hasdefault = TRUE; /* argument has a default value */
2704 (unsigned char)(matchtoken(tSIZEOF) ? uSIZEOF : 0);
2705 if (size_tag_token == 0)
2707 (unsigned char)(matchtoken(tTAGOF) ? uTAGOF : 0);
2708 if (size_tag_token != 0)
2712 if (ident == iREFERENCE)
2713 error(66, name); /* argument may not be a reference */
2715 while (matchtoken('('))
2717 if (needtoken(tSYMBOL))
2719 /* save the name of the argument whose size id to take */
2723 tokeninfo(&val, &name);
2724 if (!(arg->defvalue.size.symname = strdup(name)))
2725 error(103); /* insufficient memory */
2726 arg->defvalue.size.level = 0;
2727 if (size_tag_token == uSIZEOF)
2729 while (matchtoken('['))
2731 arg->defvalue.size.level += (short)1;
2735 if (ident == iVARIABLE) /* make sure we set this only if
2736 * not a reference */
2737 arg->hasdefault |= size_tag_token; /* uSIZEOF or uTAGOF */
2739 while (paranthese--)
2744 constexpr(&arg->defvalue.val, &arg->defvalue_tag);
2745 assert(numtags > 0);
2746 if (!matchtag(tags[0], arg->defvalue_tag, TRUE))
2747 error(213); /* tagname mismatch */
2751 arg->ident = (char)ident;
2752 arg->usage = (char)(fconst ? uCONST : 0);
2753 arg->numtags = numtags;
2754 arg->tags = (int *)malloc(numtags * sizeof tags[0]);
2756 error(103); /* insufficient memory */
2757 memcpy(arg->tags, tags, numtags * sizeof tags[0]);
2758 argsym = findloc(name);
2761 error(21, name); /* symbol already defined */
2765 if ((argsym = findglb(name)) && argsym->ident != iFUNCTN)
2766 error(219, name); /* variable shadows another symbol */
2767 /* add details of type and address */
2768 assert(numtags > 0);
2769 argsym = addvariable(name, offset, ident, sLOCAL, tags[0],
2770 arg->dim, arg->numdim, idxtag);
2771 argsym->compound = 0;
2772 if (ident == iREFERENCE)
2773 argsym->usage |= uREAD; /* because references are passed back */
2775 argsym->usage |= uREAD; /* arguments of public functions
2776 * are always "used" */
2778 argsym->usage |= uCONST;
2783 count_referrers(symbol * entry)
2788 for (i = 0; i < entry->numrefers; i++)
2789 if (entry->refer[i])
2794 /* Every symbol has a referrer list, that contains the functions that
2795 * use the symbol. Now, if function "apple" is accessed by functions
2796 * "banana" and "citron", but neither function "banana" nor "citron" are
2797 * used by anyone else, then, by inference, function "apple" is not used
2800 reduce_referrers(symbol * root)
2808 for (sym = root->next; sym; sym = sym->next)
2811 continue; /* hierarchical data type */
2812 if (sym->ident == iFUNCTN
2813 && (sym->usage & uNATIVE) == 0
2814 && (sym->usage & uPUBLIC) == 0
2815 && strcmp(sym->name, uMAINFUNC) != 0
2816 && count_referrers(sym) == 0)
2818 sym->usage &= ~(uREAD | uWRITTEN); /* erase usage bits if
2819 * there is no referrer */
2820 /* find all symbols that are referred by this symbol */
2821 for (ref = root->next; ref; ref = ref->next)
2824 continue; /* hierarchical data type */
2825 assert(ref->refer != NULL);
2826 for (i = 0; i < ref->numrefers && ref->refer[i] != sym;
2829 if (i < ref->numrefers)
2831 assert(ref->refer[i] == sym);
2832 ref->refer[i] = NULL;
2837 else if ((sym->ident == iVARIABLE || sym->ident == iARRAY)
2838 && (sym->usage & uPUBLIC) == 0
2839 && !sym->parent && count_referrers(sym) == 0)
2841 sym->usage &= ~(uREAD | uWRITTEN); /* erase usage bits if
2842 * there is no referrer */
2845 /* after removing a symbol, check whether more can be removed */
2847 while (restart > 0);
2850 /* testsymbols - test for unused local or global variables
2852 * "Public" functions are excluded from the check, since these
2853 * may be exported to other object modules.
2854 * Labels are excluded from the check if the argument 'testlabs'
2855 * is 0. Thus, labels are not tested until the end of the function.
2856 * Constants may also be excluded (convenient for global constants).
2858 * When the nesting level drops below "level", the check stops.
2860 * The function returns whether there is an "entry" point for the file.
2861 * This flag will only be 1 when browsing the global symbol table.
2864 testsymbols(symbol * root, int level, int testlabs, int testconst)
2866 char symname[2 * sNAMEMAX + 16];
2869 symbol *sym = root->next;
2871 while (sym && sym->compound >= level)
2878 if ((sym->usage & uDEFINE) == 0)
2879 error(19, sym->name); /* not a label: ... */
2880 else if ((sym->usage & uREAD) == 0)
2881 error(203, sym->name); /* symbol isn't used: ... */
2885 if ((sym->usage & (uDEFINE | uREAD | uNATIVE | uSTOCK)) == uDEFINE)
2887 funcdisplayname(symname, sym->name);
2888 if (symname[0] != '\0')
2889 error(203, symname); /* symbol isn't used ...
2890 * (and not native/stock) */
2892 if ((sym->usage & uPUBLIC) != 0
2893 || strcmp(sym->name, uMAINFUNC) == 0)
2894 entry = TRUE; /* there is an entry point */
2897 if (testconst && (sym->usage & uREAD) == 0)
2898 error(203, sym->name); /* symbol isn't used: ... */
2903 break; /* hierarchical data type */
2904 if ((sym->usage & (uWRITTEN | uREAD | uSTOCK | uPUBLIC)) == 0)
2905 error(203, sym->name); /* symbol isn't used (and not stock
2907 else if ((sym->usage & (uREAD | uSTOCK | uPUBLIC)) == 0)
2908 error(204, sym->name); /* value assigned to symbol is
2910 #if 0 /*// ??? not sure whether it is a good idea to
2911 * force people use "const" */
2912 else if ((sym->usage & (uWRITTEN | uPUBLIC | uCONST)) == 0
2913 && sym->ident == iREFARRAY)
2914 error(214, sym->name); /* make array argument "const" */
2924 calc_array_datasize(symbol * sym, cell * offset)
2928 assert(sym != NULL);
2929 assert(sym->ident == iARRAY || sym->ident == iREFARRAY);
2930 length = sym->dim.array.length;
2931 if (sym->dim.array.level > 0)
2934 calc_array_datasize(finddepend(sym), offset);
2936 *offset = length * (*offset + sizeof(cell));
2938 length *= length * sublength;
2951 destructsymbols(symbol * root, int level)
2954 int savepri = FALSE;
2955 symbol *sym = root->next;
2957 while (sym && sym->compound >= level)
2959 if (sym->ident == iVARIABLE || sym->ident == iARRAY)
2961 char symbolname[16];
2965 /* check that the '~' operator is defined for this tag */
2966 operator_symname(symbolname, "~", sym->tag, 0, 1, 0);
2967 if ((opsym = findglb(symbolname)))
2969 /* save PRI, in case of a return statement */
2972 push1(); /* right-hand operand is in PRI */
2975 /* if the variable is an array, get the number of elements */
2976 if (sym->ident == iARRAY)
2978 elements = calc_array_datasize(sym, &offset);
2979 /* "elements" can be zero when the variable is declared like
2980 * new mytag: myvar[2][] = { {1, 2}, {3, 4} }
2981 * one should declare all dimensions!
2984 error(46, sym->name); /* array size is unknown */
2992 /* call the '~' operator */
2994 addconst(offset); /*add offset to array data to the address */
2996 pushval(2 * sizeof(cell)); /* 2 parameters */
2998 if (sc_status != statSKIP)
2999 markusage(opsym, uREAD); /* do not mark as "used" when this
3000 * call itself is skipped */
3002 opsym->x.lib->value += 1; /* increment "usage count"
3008 /* restore PRI, if it was saved */
3014 insert_constval(constvalue * prev, constvalue * next, char *name,
3015 cell val, short idx)
3019 if (!(cur = (constvalue *)malloc(sizeof(constvalue))))
3020 error(103); /* insufficient memory (fatal error) */
3021 memset(cur, 0, sizeof(constvalue));
3022 strcpy(cur->name, name);
3031 append_constval(constvalue * table, char *name, cell val, short idx)
3033 constvalue *cur, *prev;
3035 /* find the end of the constant table */
3036 for (prev = table, cur = table->next; cur;
3037 prev = cur, cur = cur->next)
3039 return insert_constval(prev, NULL, name, val, idx);
3043 find_constval(constvalue * table, char *name, short idx)
3045 constvalue *ptr = table->next;
3049 if (strcmp(name, ptr->name) == 0 && ptr->index == idx)
3057 find_constval_byval(constvalue * table, cell val)
3059 constvalue *ptr = table->next;
3063 if (ptr->value == val)
3070 #if 0 /* never used */
3072 delete_constval(constvalue * table, char *name)
3074 constvalue *prev = table;
3075 constvalue *cur = prev->next;
3079 if (strcmp(name, cur->name) == 0)
3081 prev->next = cur->next;
3093 delete_consttable(constvalue * table)
3095 constvalue *cur = table->next, *next;
3103 memset(table, 0, sizeof(constvalue));
3108 * Adds a symbol to the #define symbol table.
3111 add_constant(char *name, cell val, int vclass, int tag)
3115 /* Test whether a global or local symbol with the same name exists. Since
3116 * constants are stored in the symbols table, this also finds previously
3117 * defind constants. */
3118 sym = findglb(name);
3120 sym = findloc(name);
3123 /* silently ignore redefinitions of constants with the same value */
3124 if (sym->ident == iCONSTEXPR)
3126 if (sym->addr != val)
3127 error(201, name); /* redefinition of constant (different value) */
3131 error(21, name); /* symbol already defined */
3136 /* constant doesn't exist yet, an entry must be created */
3137 sym = addsym(name, val, iCONSTEXPR, vclass, tag, uDEFINE);
3138 if (sc_status == statIDLE)
3139 sym->usage |= uPREDEF;
3142 /* statement - The Statement Parser
3144 * This routine is called whenever the parser needs to know what
3145 * statement it encounters (i.e. whenever program syntax requires a
3149 statement(int *lastindent, int allow_decl)
3157 error(36); /* empty statement */
3162 tok = lex(&val, &st);
3164 setline(fline, fcurrent);
3165 /* lex() has set stmtindent */
3166 if (lastindent && tok != tLABEL)
3169 if (*lastindent >= 0 && *lastindent != stmtindent &&
3170 !indent_nowarn && sc_tabsize > 0)
3171 error(217); /* loose indentation */
3173 *lastindent = stmtindent;
3174 indent_nowarn = TRUE; /* if warning was blocked, re-enable it */
3189 error(3); /* declaration only valid in a block */
3200 error(3); /* declaration only valid in a block */
3204 if (!matchtoken('}')) /* {} is the empty statement */
3206 /* lastst (for "last statement") does not change */
3209 error(36); /* empty statement */
3233 error(14); /* not in switch */
3273 default: /* non-empty expression */
3274 lexpush(); /* analyze token later */
3275 doexpr(TRUE, TRUE, TRUE, TRUE, NULL, FALSE);
3285 cell save_decl = declared;
3288 nestlevel += 1; /* increase compound statement level */
3289 while (matchtoken('}') == 0)
3290 { /* repeat until compound statement is closed */
3293 needtoken('}'); /* gives error: "expected token }" */
3299 && (lastst == tRETURN || lastst == tBREAK
3300 || lastst == tCONTINUE))
3301 error(225); /* unreachable code */
3302 statement(&indent, TRUE); /* do a statement */
3306 if (lastst != tRETURN)
3307 destructsymbols(&loctab, nestlevel);
3308 if (lastst != tRETURN && lastst != tGOTO)
3309 /* delete local variable space */
3310 modstk((int)(declared - save_decl) * sizeof(cell));
3312 testsymbols(&loctab, nestlevel, FALSE, TRUE); /* look for unused
3314 declared = save_decl;
3315 delete_symbols(&loctab, nestlevel, FALSE, TRUE);
3316 /* erase local symbols, but
3317 * retain block local labels
3318 * (within the function) */
3320 nestlevel -= 1; /* decrease compound statement level */
3325 * Global references: stgidx (referred to only)
3328 doexpr(int comma, int chkeffect, int allowarray, int mark_endexpr,
3329 int *tag, int chkfuncresult)
3331 int constant, idx, ident;
3332 int localstaging = FALSE;
3337 stgset(TRUE); /* start stage-buffering */
3338 localstaging = TRUE;
3339 assert(stgidx == 0);
3342 errorset(sEXPRMARK);
3345 /* on second round through, mark the end of the previous expression */
3349 ident = expression(&constant, &val, tag, chkfuncresult);
3350 if (!allowarray && (ident == iARRAY || ident == iREFARRAY))
3351 error(33, "-unknown-"); /* array must be indexed */
3352 if (chkeffect && !sideeffect)
3353 error(215); /* expression has no effect */
3355 while (comma && matchtoken(',')); /* more? */
3357 endexpr(TRUE); /* optionally, mark the end of the expression */
3358 errorset(sEXPRRELEASE);
3362 stgset(FALSE); /* stop staging */
3369 constexpr(cell * val, int *tag)
3374 stgset(TRUE); /* start stage-buffering */
3375 stgget(&idx, &cidx); /* mark position in code generator */
3376 errorset(sEXPRMARK);
3377 expression(&constant, val, tag, FALSE);
3378 stgdel(idx, cidx); /* scratch generated code */
3379 stgset(FALSE); /* stop stage-buffering */
3381 error(8); /* must be constant expression */
3382 errorset(sEXPRRELEASE);
3388 * In the case a "simple assignment" operator ("=") is used within a
3389 * test, * the warning "possibly unintended assignment" is displayed.
3390 * This routine sets the global variable "intest" to true, it is
3391 * restored upon termination. In the case the assignment was intended,
3392 * use parantheses around the expression to avoid the warning;
3393 * primary() sets "intest" to 0.
3395 * Global references: intest (altered, but restored upon termination)
3398 test(int label, int parens, int invert)
3402 value lval = { NULL, 0, 0, 0, 0, NULL };
3403 int localstaging = FALSE;
3407 stgset(TRUE); /* start staging */
3408 localstaging = TRUE;
3410 stgget(&idx, &cidx); /* should start at zero if started
3416 pushstk((stkitem) intest);
3422 stgget(&idx, &cidx); /* mark position (of last expression) in
3426 tok = matchtoken(',');
3430 while (tok); /* do */
3433 if (lval.ident == iARRAY || lval.ident == iREFARRAY)
3436 (lval.sym->name) ? lval.sym->name : "-unknown-";
3437 error(33, ptr); /* array must be indexed */
3439 if (lval.ident == iCONSTEXPR)
3440 { /* constant expression */
3441 intest = (int)(long)popstk(); /* restore stack */
3444 { /* code always executed */
3445 error(206); /* redundant test: always non-zero */
3449 error(205); /* redundant code: never executed */
3454 stgout(0); /* write "jumplabel" code */
3455 stgset(FALSE); /* stop staging */
3459 if (lval.tag != 0 && lval.tag != sc_addtag("bool"))
3460 if (check_userop(lneg, lval.tag, 0, 1, NULL, &lval.tag))
3461 invert = !invert; /* user-defined ! operator inverted result */
3463 jmp_ne0(label); /* jump to label if true (different from 0) */
3465 jmp_eq0(label); /* jump to label if false (equal to 0) */
3466 endexpr(TRUE); /* end expression (give optimizer a chance) */
3467 intest = (int)(long)popstk(); /* double typecast to avoid warning
3468 * with Microsoft C */
3471 stgout(0); /* output queue from the very beginning (see
3472 * assert() when localstaging is set to TRUE) */
3473 stgset(FALSE); /* stop staging */
3483 ifindent = stmtindent; /* save the indent of the "if" instruction */
3484 flab1 = getlabel(); /* get label number for false branch */
3485 test(flab1, TRUE, FALSE); /*get expression, branch to flab1 if false */
3486 statement(NULL, FALSE); /* if true, do a statement */
3487 if (matchtoken(tELSE) == 0)
3489 setlabel(flab1); /* no, simple if..., print false label */
3493 /* to avoid the "dangling else" error, we want a warning if the "else"
3494 * has a lower indent than the matching "if" */
3496 if (stmtindent < ifindent && sc_tabsize > 0)
3497 error(217); /* loose indentation */
3500 if ((lastst != tRETURN) && (lastst != tGOTO))
3502 setlabel(flab1); /* print false label */
3503 statement(NULL, FALSE); /* do "else" clause */
3504 setlabel(flab2); /* print true label */
3511 int wq[wqSIZE]; /* allocate local queue */
3513 addwhile(wq); /* add entry to queue for "break" */
3514 setlabel(wq[wqLOOP]); /* loop label */
3515 /* The debugger uses the "line" opcode to be able to "break" out of
3516 * a loop. To make sure that each loop has a line opcode, even for the
3517 * tiniest loop, set it below the top of the loop */
3518 setline(fline, fcurrent);
3519 test(wq[wqEXIT], TRUE, FALSE); /* branch to wq[wqEXIT] if false */
3520 statement(NULL, FALSE); /* if so, do a statement */
3521 jumplabel(wq[wqLOOP]); /* and loop to "while" start */
3522 setlabel(wq[wqEXIT]); /* exit label */
3523 delwhile(); /* delete queue entry */
3527 * Note that "continue" will in this case not jump to the top of the
3528 * loop, but to the end: just before the TRUE-or-FALSE testing code.
3533 int wq[wqSIZE], top;
3535 addwhile(wq); /* see "dowhile" for more info */
3536 top = getlabel(); /* make a label first */
3537 setlabel(top); /* loop label */
3538 statement(NULL, FALSE);
3540 setlabel(wq[wqLOOP]); /* "continue" always jumps to WQLOOP. */
3541 setline(fline, fcurrent);
3542 test(wq[wqEXIT], TRUE, FALSE);
3544 setlabel(wq[wqEXIT]);
3552 int wq[wqSIZE], skiplab;
3554 int save_nestlevel, idx;
3557 save_decl = declared;
3558 save_nestlevel = nestlevel;
3561 skiplab = getlabel();
3563 if (matchtoken(';') == 0)
3565 /* new variable declarations are allowed here */
3566 if (matchtoken(tNEW))
3568 /* The variable in expr1 of the for loop is at a
3569 * 'compound statement' level of it own.
3572 declloc(FALSE); /* declare local variable */
3576 doexpr(TRUE, TRUE, TRUE, TRUE, NULL, FALSE); /* expression 1 */
3580 /* Adjust the "declared" field in the "while queue", in case that
3581 * local variables were declared in the first expression of the
3582 * "for" loop. These are deleted in separately, so a "break" or a
3583 * "continue" must ignore these fields.
3586 assert(ptr != NULL);
3587 ptr[wqBRK] = (int)declared;
3588 ptr[wqCONT] = (int)declared;
3589 jumplabel(skiplab); /* skip expression 3 1st time */
3590 setlabel(wq[wqLOOP]); /* "continue" goes to this label: expr3 */
3591 setline(fline, fcurrent);
3592 /* Expressions 2 and 3 are reversed in the generated code:
3593 * expression 3 precedes expression 2.
3594 * When parsing, the code is buffered and marks for
3595 * the start of each expression are insterted in the buffer.
3598 stgset(TRUE); /* start staging */
3599 assert(stgidx == 0);
3601 stgmark(sSTARTREORDER);
3602 stgmark((char)(sEXPRSTART + 0)); /* mark start of 2nd expression
3604 setlabel(skiplab); /*jump to this point after 1st expression */
3605 if (matchtoken(';') == 0)
3607 test(wq[wqEXIT], FALSE, FALSE); /* expression 2
3608 *(jump to wq[wqEXIT] if false) */
3611 stgmark((char)(sEXPRSTART + 1)); /* mark start of 3th expression
3613 if (matchtoken(')') == 0)
3615 doexpr(TRUE, TRUE, TRUE, TRUE, NULL, FALSE); /* expression 3 */
3618 stgmark(sENDREORDER); /* mark end of reversed evaluation */
3620 stgset(FALSE); /* stop staging */
3621 statement(NULL, FALSE);
3622 jumplabel(wq[wqLOOP]);
3623 setlabel(wq[wqEXIT]);
3626 assert(nestlevel >= save_nestlevel);
3627 if (nestlevel > save_nestlevel)
3629 /* Clean up the space and the symbol table for the local
3630 * variable in "expr1".
3632 destructsymbols(&loctab, nestlevel);
3633 modstk((int)(declared - save_decl) * sizeof(cell));
3634 declared = save_decl;
3635 delete_symbols(&loctab, nestlevel, FALSE, TRUE);
3636 nestlevel = save_nestlevel; /* reset 'compound statement'
3641 /* The switch statement is incompatible with its C sibling:
3642 * 1. the cases are not drop through
3643 * 2. only one instruction may appear below each case, use a compound
3644 * instruction to execute multiple instructions
3645 * 3. the "case" keyword accepts a comma separated list of values to
3646 * match, it also accepts a range using the syntax "1 .. 4"
3649 * PRI = expression result
3650 * param = table offset (code segment)
3656 int lbl_table, lbl_exit, lbl_case;
3657 int tok, swdefault, casecount;
3660 constvalue caselist = { NULL, "", 0, 0 }; /*case list starts empty */
3661 constvalue *cse, *csp;
3662 char labelname[sNAMEMAX + 1];
3665 doexpr(TRUE, FALSE, FALSE, TRUE, NULL, FALSE); /* evaluate
3666 * switch expression */
3668 /* generate the code for the switch statement, the label is the
3669 * address of the case table (to be generated later).
3671 lbl_table = getlabel();
3672 lbl_case = 0; /* just to avoid a compiler warning */
3673 ffswitch(lbl_table);
3676 lbl_exit = getlabel(); /*get label number for jumping out of switch */
3681 tok = lex(&val, &str); /* read in (new) token */
3685 if (swdefault != FALSE)
3686 error(15); /* "default" case must be last in switch
3688 lbl_case = getlabel();
3689 sc_allowtags = FALSE; /* do not allow tagnames here */
3694 /* ??? enforce/document that, in a switch, a statement cannot
3695 * start an opening brace (marks the start of a compound
3696 * statement) and search for the right-most colon before that
3698 * Now, by replacing the ':' by a special COLON token, you can
3699 * parse all expressions until that special token.
3702 constexpr(&val, NULL);
3703 /* Search the insertion point (the table is kept in sorted
3704 * order, so that advanced abstract machines can sift the
3705 * case table with a binary search). Check for duplicate
3706 * case values at the same time.
3708 for (csp = &caselist, cse = caselist.next;
3709 cse && cse->value < val;
3710 csp = cse, cse = cse->next)
3712 if (cse && cse->value == val)
3713 error(40, val); /* duplicate "case" label */
3714 /* Since the label is stored as a string in the
3715 * "constvalue", the size of an identifier must
3716 * be at least 8, as there are 8
3717 * hexadecimal digits in a 32-bit number.
3720 #error Length of identifier (sNAMEMAX) too small.
3722 insert_constval(csp, cse, itoh(lbl_case), val, 0);
3723 if (matchtoken(tDBLDOT))
3727 constexpr(&end, NULL);
3729 error(50); /* invalid range */
3730 while (++val <= end)
3733 /* find the new insertion point */
3734 for (csp = &caselist, cse = caselist.next;
3735 cse && cse->value < val;
3736 csp = cse, cse = cse->next)
3738 if (cse && cse->value == val)
3739 error(40, val); /* duplicate "case" label */
3740 insert_constval(csp, cse, itoh(lbl_case), val, 0);
3744 while (matchtoken(','));
3745 needtoken(':'); /* ':' ends the case */
3746 sc_allowtags = TRUE; /* reset */
3748 statement(NULL, FALSE);
3749 jumplabel(lbl_exit);
3752 if (swdefault != FALSE)
3753 error(16); /* multiple defaults in switch */
3754 lbl_case = getlabel();
3758 statement(NULL, FALSE);
3759 /* Jump to lbl_exit, even thouh this is the last clause in the
3760 *switch, because the jump table is generated between the last
3761 * clause of the switch and the exit label.
3763 jumplabel(lbl_exit);
3766 /* nothing, but avoid dropping into "default" */
3770 indent_nowarn = TRUE; /* disable this check */
3771 tok = '}'; /* break out of the loop after an error */
3777 /* verify that the case table is sorted (unfortunately, duplicates can
3778 * occur; there really shouldn't be duplicate cases, but the compiler
3779 * may not crash or drop into an assertion for a user error). */
3780 for (cse = caselist.next; cse && cse->next; cse = cse->next)
3781 ; /* empty. no idea whether this is correct, but we MUST NOT do
3782 * the setlabel(lbl_table) call in the loop body. doing so breaks
3783 * switch statements that only have one case statement following.
3787 /* generate the table here, before lbl_exit (general jump target) */
3788 setlabel(lbl_table);
3790 if (swdefault == FALSE)
3792 /* store lbl_exit as the "none-matched" label in the switch table */
3793 strcpy(labelname, itoh(lbl_exit));
3797 /* lbl_case holds the label of the "default" clause */
3798 strcpy(labelname, itoh(lbl_case));
3800 ffcase(casecount, labelname, TRUE);
3801 /* generate the rest of the table */
3802 for (cse = caselist.next; cse; cse = cse->next)
3803 ffcase(cse->value, cse->name, FALSE);
3806 delete_consttable(&caselist); /* clear list of case labels */
3814 value lval = { NULL, 0, 0, 0, 0, NULL };
3816 if ((sc_debug & sCHKBOUNDS) != 0)
3818 flab1 = getlabel(); /* get label number for "OK" branch */
3819 test(flab1, FALSE, TRUE); /* get expression and branch
3820 * to flab1 if true */
3821 setline(fline, fcurrent); /* make sure we abort on the correct
3823 ffabort(xASSERTION);
3828 stgset(TRUE); /* start staging */
3829 stgget(&idx, &cidx); /* mark position in code generator */
3834 stgdel(idx, cidx); /* just scrap the code */
3836 while (matchtoken(','));
3837 stgset(FALSE); /* stop staging */
3849 if (lex(&val, &st) == tSYMBOL)
3852 jumplabel((int)sym->addr);
3853 sym->usage |= uREAD; /* set "uREAD" bit */
3855 * // ??? if the label is defined (check sym->usage & uDEFINE), check
3856 * // sym->compound (nesting level of the label) against nestlevel;
3857 * // if sym->compound < nestlevel, call the destructor operator
3862 error(20, st); /* illegal symbol name */
3874 tokeninfo(&val, &st); /* retrieve label name again */
3875 if (find_constval(&tagname_tab, st, 0))
3876 error(221, st); /* label name shadows tagname */
3878 setlabel((int)sym->addr);
3879 /* since one can jump around variable declarations or out of compound
3880 * blocks, the stack must be manually adjusted
3882 setstk(-declared * sizeof(cell));
3883 sym->usage |= uDEFINE; /* label is now defined */
3888 * Finds a label from the (local) symbol table or adds one to it.
3889 * Labels are local in scope.
3891 * Note: The "_usage" bit is set to zero. The routines that call
3892 * "fetchlab()" must set this bit accordingly.
3895 fetchlab(char *name)
3899 sym = findloc(name); /* labels are local in scope */
3902 if (sym->ident != iLABEL)
3903 error(19, sym->name); /* not a label: ... */
3907 sym = addsym(name, getlabel(), iLABEL, sLOCAL, 0, 0);
3908 sym->x.declared = (int)declared;
3909 sym->compound = nestlevel;
3916 * Global references: rettype (altered)
3923 if (matchtoken(tTERM) == 0)
3925 if ((rettype & uRETNONE) != 0)
3926 error(208); /* mix "return;" and "return value;" */
3927 doexpr(TRUE, FALSE, FALSE, TRUE, &tag, FALSE);
3929 rettype |= uRETVALUE; /* function returns a value */
3930 /* check tagname with function tagname */
3931 assert(curfunc != NULL);
3932 if (!matchtag(curfunc->tag, tag, TRUE))
3933 error(213); /* tagname mismatch */
3937 /* this return statement contains no expression */
3939 if ((rettype & uRETVALUE) != 0)
3941 char symname[2 * sNAMEMAX + 16]; /* allow space for user
3942 * defined operators */
3943 assert(curfunc != NULL);
3944 funcdisplayname(symname, curfunc->name);
3945 error(209, symname); /* function should return a value */
3947 rettype |= uRETNONE; /* function does not return anything */
3949 destructsymbols(&loctab, 0); /*call destructor for *all* locals */
3950 modstk((int)declared * sizeof(cell)); /* end of function, remove
3951 *all* * local variables*/
3960 ptr = readwhile(); /* readwhile() gives an error if not in loop */
3964 destructsymbols(&loctab, nestlevel);
3965 modstk(((int)declared - ptr[wqBRK]) * sizeof(cell));
3966 jumplabel(ptr[wqEXIT]);
3974 ptr = readwhile(); /* readwhile() gives an error if not in loop */
3978 destructsymbols(&loctab, nestlevel);
3979 modstk(((int)declared - ptr[wqCONT]) * sizeof(cell));
3980 jumplabel(ptr[wqLOOP]);
3986 /* find the tag by value in the table, then set the top bit to mark it
3993 assert((tag & PUBLICTAG) == 0);
3994 for (ptr = tagname_tab.next;
3995 ptr && tag != (int)(ptr->value & TAGMASK); ptr = ptr->next)
3998 ptr->value |= PUBLICTAG;
4007 if (matchtoken(tTERM) == 0)
4009 doexpr(TRUE, FALSE, FALSE, TRUE, &tag, FALSE);
4018 destructsymbols(&loctab, 0); /* call destructor for *all* locals */
4027 if (matchtoken(tTERM) == 0)
4029 doexpr(TRUE, FALSE, FALSE, TRUE, &tag, FALSE);
4046 ptr[wqBRK] = (int)declared; /* stack pointer (for "break") */
4047 ptr[wqCONT] = (int)declared; /* for "continue", possibly adjusted later */
4048 ptr[wqLOOP] = getlabel();
4049 ptr[wqEXIT] = getlabel();
4050 if (wqptr >= (wq + wqTABSZ - wqSIZE))
4051 error(102, "loop table"); /* loop table overflow (too many active loops) */
4054 { /* copy "ptr" to while queue table */
4074 error(24); /* out of context */
4079 return (wqptr - wqSIZE);