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 */
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 /* add the default "include" directory */
659 /* path + filename (hopefully) */
660 strncpy(path, root, sizeof(path) - 1);
661 path[sizeof(path) - 1] = 0;
663 /* terminate just behind last \ or : */
664 if ((ptr = strrchr(path, DIRSEP_CHAR))
665 || (ptr = strchr(path, ':')))
667 /* If there was no terminating "\" or ":",
668 * the filename probably does not
669 * contain the path; so we just don't add it
670 * to the list in that case
673 if (strlen(path) < (sizeof(path) - 1 - 7))
675 strcat(path, "include");
678 path[len] = DIRSEP_CHAR;
679 path[len + 1] = '\0';
687 printf("Usage: embryo_cc <filename> [options]\n\n");
688 printf("Options:\n");
691 (" -A<num> alignment in bytes of the data segment and the\
695 (" -a output assembler code (skip code generation\
699 (" -C[+/-] compact encoding for output file (default=%c)\n",
700 sc_compress ? '+' : '-');
701 printf(" -c8 [default] a character is 8-bits\
702 (ASCII/ISO Latin-1)\n");
704 printf(" -c16 a character is 16-bits (Unicode)\n");
705 #if defined dos_setdrive
706 printf(" -Dpath active directory path\n");
709 (" -d0 no symbolic information, no run-time checks\n");
710 printf(" -d1 [default] run-time checks, no symbolic\
713 (" -d2 full debug information and dynamic checking\n");
714 printf(" -d3 full debug information, dynamic checking,\
717 printf(" -i <name> path for include files\n");
719 printf(" -l create list file (preprocess only)\n");
721 printf(" -o <name> set base name of output file\n");
724 (" -P[+/-] strings are \"packed\" by default (default=%c)\n",
725 sc_packstr ? '+' : '-');
726 printf(" -p<name> set name of \"prefix\" file\n");
731 (" -S <num> stack/heap size in cells (default=%d, min=65)\n",
734 printf(" -s<num> skip lines from the input file\n");
736 (" -t<num> TAB indent size (in character positions)\n");
737 printf(" -\\ use '\\' for escape characters\n");
738 printf(" -^ use '^' for escape characters\n");
739 printf(" -;[+/-] require a semicolon to end each statement\
740 (default=%c)\n", sc_needsemicolon ? '+' : '-');
743 (" sym=val define constant \"sym\" with value \"val\"\n");
744 printf(" sym= define constant \"sym\" with value 0\n");
746 longjmp(errbuf, 3); /* user abort */
754 assert(sc_status == statIDLE);
755 append_constval(&tagname_tab, "_", 0, 0); /* "untagged" */
756 append_constval(&tagname_tab, "bool", 1, 0);
758 add_constant("true", 1, sGLOBAL, 1); /* boolean flags */
759 add_constant("false", 0, sGLOBAL, 1);
760 add_constant("EOS", 0, sGLOBAL, 0); /* End Of String, or '\0' */
761 add_constant("cellbits", 32, sGLOBAL, 0);
762 add_constant("cellmax", INT_MAX, sGLOBAL, 0);
763 add_constant("cellmin", INT_MIN, sGLOBAL, 0);
764 add_constant("charbits", charbits, sGLOBAL, 0);
765 add_constant("charmin", 0, sGLOBAL, 0);
766 add_constant("charmax", (charbits == 16) ? 0xffff : 0xff, sGLOBAL, 0);
768 add_constant("__Small", VERSION_INT, sGLOBAL, 0);
771 if ((sc_debug & (sCHKBOUNDS | sSYMBOLIC)) == (sCHKBOUNDS | sSYMBOLIC))
773 else if ((sc_debug & sCHKBOUNDS) == sCHKBOUNDS)
775 add_constant("debug", debug, sGLOBAL, 0);
778 /* parse - process all input text
780 * At this level, only static declarations and function definitions
786 int tok, tag, fconst, fstock, fstatic;
792 /* first try whether a declaration possibly is native or public */
793 tok = lex(&val, &str); /* read in (new) token */
800 fconst = matchtoken(tCONST);
801 declglb(NULL, 0, FALSE, FALSE, FALSE, fconst);
804 /* This can be a static function or a static global variable;
805 * we know which of the two as soon as we have parsed up to the
806 * point where an opening parenthesis of a function would be
807 * expected. To back out after deciding it was a declaration of
808 * a static variable after all, we have to store the symbol name
811 fstock = matchtoken(tSTOCK);
812 fconst = matchtoken(tCONST);
813 tag = sc_addtag(NULL);
814 tok = lex(&val, &str);
815 if (tok == tNATIVE || tok == tPUBLIC)
817 error(42); /* invalid combination of class specifiers */
820 declfuncvar(tok, str, tag, FALSE, TRUE, fstock, fconst);
829 /* This can be a public function or a public variable;
830 * see the comment above (for static functions/variables)
833 fconst = matchtoken(tCONST);
834 tag = sc_addtag(NULL);
835 tok = lex(&val, &str);
836 if (tok == tNATIVE || tok == tSTOCK || tok == tSTATIC)
838 error(42); /* invalid combination of class specifiers */
841 declfuncvar(tok, str, tag, TRUE, FALSE, FALSE, fconst);
844 /* This can be a stock function or a stock *global) variable;
845 * see the comment above (for static functions/variables) for
848 fstatic = matchtoken(tSTATIC);
849 fconst = matchtoken(tCONST);
850 tag = sc_addtag(NULL);
851 tok = lex(&val, &str);
852 if (tok == tNATIVE || tok == tPUBLIC)
854 error(42); /* invalid combination of class specifiers */
857 declfuncvar(tok, str, tag, FALSE, fstatic, TRUE, fconst);
863 if (!newfunc(NULL, -1, FALSE, FALSE, FALSE))
865 error(10); /* illegal function or declaration */
866 lexclr(TRUE); /* drop the rest of the line */
870 funcstub(TRUE); /* create a dummy function */
876 error(54); /* unmatched closing brace */
879 error(55); /* start of function body without function header */
884 error(10); /* illegal function or declaration */
885 lexclr(TRUE); /* drop the rest of the line */
893 * Dump the literal pool (strings etc.)
895 * Global references: litidx (referred to only)
905 /* should be in the data segment */
908 j = 16; /* 16 values per line */
909 while (j && k < litidx)
911 outval(litq[k], FALSE);
915 if (j == 0 || k >= litidx)
916 stgwrite("\n"); /* force a newline after 10 dumps */
917 /* Note: stgwrite() buffers a line until it is complete. It recognizes
918 * the end of line as a sequence of "\n\0", so something like "\n\t"
919 * so should not be passed to stgwrite().
927 * Dump zero's for default initial values
943 stgwrite((i == 0 || count == 0) ? "\n" : " ");
944 if (i == 0 && count > 0)
950 aligndata(int numbytes)
952 if ((((glb_declared + litidx) * sizeof(cell)) % numbytes) != 0)
954 while ((((glb_declared + litidx) * sizeof(cell)) % numbytes) != 0)
961 declfuncvar(int tok, char *symname, int tag, int fpublic, int fstatic,
962 int fstock, int fconst)
964 char name[sNAMEMAX + 1];
966 if (tok != tSYMBOL && tok != tOPERATOR)
969 error(20, symname); /* invalid symbol name */
972 if (tok == tOPERATOR)
975 if (!newfunc(NULL, tag, fpublic, fstatic, fstock))
976 error(10); /* illegal function or declaration */
980 assert(strlen(symname) <= sNAMEMAX);
981 strcpy(name, symname);
982 if (fconst || !newfunc(name, tag, fpublic, fstatic, fstock))
983 declglb(name, tag, fpublic, fstatic, fstock, fconst);
984 /* if not a static function, try a static variable */
988 /* declglb - declare global symbols
990 * Declare a static (global) variable. Global variables are stored in
993 * global references: glb_declared (altered)
996 declglb(char *firstname, int firsttag, int fpublic, int fstatic,
997 int stock, int fconst)
999 int ident, tag, ispublic;
1000 int idxtag[sDIMEN_MAX];
1001 char name[sNAMEMAX + 1];
1002 cell val, size, cidx;
1004 int dim[sDIMEN_MAX];
1013 filenum = fcurrent; /* save file number at the start of the
1017 size = 1; /* single size (no array) */
1018 numdim = 0; /* no dimensions */
1022 assert(strlen(firstname) <= sNAMEMAX);
1023 strcpy(name, firstname); /* save symbol name */
1029 tag = sc_addtag(NULL);
1030 if (lex(&val, &str) != tSYMBOL) /* read in (new) token */
1031 error(20, str); /* invalid symbol name */
1032 assert(strlen(str) <= sNAMEMAX);
1033 strcpy(name, str); /* save symbol name */
1035 sym = findglb(name);
1037 sym = findconst(name);
1038 if (sym && (sym->usage & uDEFINE) != 0)
1039 error(21, name); /* symbol already defined */
1041 if (name[0] == PUBLIC_CHAR)
1043 ispublic = TRUE; /* implicitly public variable */
1044 if (stock || fstatic)
1045 error(42); /* invalid combination of class specifiers */
1047 while (matchtoken('['))
1050 if (numdim == sDIMEN_MAX)
1052 error(53); /* exceeding maximum number of dimensions */
1055 if (numdim > 0 && dim[numdim - 1] == 0)
1056 error(52); /* only last dimension may be variable length */
1057 size = needsub(&idxtag[numdim]); /* get size; size==0 for
1059 #if INT_MAX < LONG_MAX
1061 error(105); /* overflow, exceeding capacity */
1064 error(56, name); /* arrays cannot be public */
1065 dim[numdim++] = (int)size;
1067 /* if this variable is never used (which can be detected only in
1068 * the second stage), shut off code generation; make an exception
1069 * for public variables
1071 cidx = 0; /* only to avoid a compiler warning */
1072 if (sc_status == statWRITE && sym
1073 && (sym->usage & (uREAD | uWRITTEN | uPUBLIC)) == 0)
1075 sc_status = statSKIP;
1078 glbdecl = glb_declared;
1081 defsymbol(name, ident, sGLOBAL, sizeof(cell) * glb_declared, tag);
1082 begdseg(); /* real (initialized) data in data segment */
1083 assert(litidx == 0); /* literal queue should be empty */
1087 aligndata(sc_dataalign);
1088 dumplits(); /* dump the literal queue */
1089 sc_alignnext = FALSE;
1090 litidx = 0; /* global initial data is dumped, so restart at zero */
1092 initials(ident, tag, &size, dim, numdim); /* stores values in
1093 * the literal queue */
1096 dumplits(); /* dump the literal queue */
1097 dumpzero((int)size - litidx);
1100 { /* define only if not yet defined */
1102 addvariable(name, sizeof(cell) * glb_declared, ident, sGLOBAL,
1103 tag, dim, numdim, idxtag);
1106 { /* if declared but not yet defined, adjust the
1107 * variable's address */
1108 sym->addr = sizeof(cell) * glb_declared;
1109 sym->usage |= uDEFINE;
1112 sym->usage |= uPUBLIC;
1114 sym->usage |= uCONST;
1116 sym->usage |= uSTOCK;
1118 sym->fnumber = filenum;
1119 if (ident == iARRAY)
1120 for (level = 0; level < numdim; level++)
1121 symbolrange(level, dim[level]);
1122 if (sc_status == statSKIP)
1124 sc_status = statWRITE;
1126 assert(glb_declared == glbdecl);
1130 glb_declared += (int)size; /* add total number of cells */
1133 while (matchtoken(',')); /* enddo *//* more? */
1134 needtoken(tTERM); /* if not comma, must be semicolumn */
1137 /* declloc - declare local symbols
1139 * Declare local (automatic) variables. Since these variables are
1140 * relative to the STACK, there is no switch to the DATA segment.
1141 * These variables cannot be initialized either.
1143 * global references: declared (altered)
1144 * funcstatus (referred to only)
1147 declloc(int fstatic)
1150 int idxtag[sDIMEN_MAX];
1151 char name[sNAMEMAX + 1];
1155 value lval = { NULL, 0, 0, 0, 0, NULL };
1157 int dim[sDIMEN_MAX];
1161 fconst = matchtoken(tCONST);
1166 numdim = 0; /* no dimensions */
1167 tag = sc_addtag(NULL);
1168 if (lex(&val, &str) != tSYMBOL) /* read in (new) token */
1169 error(20, str); /* invalid symbol name */
1170 assert(strlen(str) <= sNAMEMAX);
1171 strcpy(name, str); /* save symbol name */
1172 if (name[0] == PUBLIC_CHAR)
1173 error(56, name); /* local variables cannot be public */
1174 /* Note: block locals may be named identical to locals at higher
1175 * compound blocks (as with standard C); so we must check (and add)
1176 * the "nesting level" of local variables to verify the
1177 * multi-definition of symbols.
1179 if ((sym = findloc(name)) && sym->compound == nestlevel)
1180 error(21, name); /* symbol already defined */
1181 /* Although valid, a local variable whose name is equal to that
1182 * of a global variable or to that of a local variable at a lower
1183 * level might indicate a bug.
1185 if (((sym = findloc(name)) && sym->compound != nestlevel)
1187 error(219, name); /* variable shadows another symbol */
1188 while (matchtoken('['))
1191 if (numdim == sDIMEN_MAX)
1193 error(53); /* exceeding maximum number of dimensions */
1196 if (numdim > 0 && dim[numdim - 1] == 0)
1197 error(52); /* only last dimension may be variable length */
1198 size = needsub(&idxtag[numdim]); /* get size; size==0 for "var[]" */
1199 #if INT_MAX < LONG_MAX
1201 error(105); /* overflow, exceeding capacity */
1203 dim[numdim++] = (int)size;
1205 if (ident == iARRAY || fstatic)
1209 aligndata(sc_dataalign);
1210 sc_alignnext = FALSE;
1212 cur_lit = litidx; /* save current index in the literal table */
1213 initials(ident, tag, &size, dim, numdim);
1215 return ident; /* error message already given */
1219 /* reserve memory (on the stack) for the variable */
1222 /* write zeros for uninitialized fields */
1223 while (litidx < cur_lit + size)
1226 addvariable(name, (cur_lit + glb_declared) * sizeof(cell),
1227 ident, sSTATIC, tag, dim, numdim, idxtag);
1228 defsymbol(name, ident, sSTATIC,
1229 (cur_lit + glb_declared) * sizeof(cell), tag);
1233 declared += (int)size; /* variables are put on stack,
1234 * adjust "declared" */
1236 addvariable(name, -declared * sizeof(cell), ident, sLOCAL, tag,
1237 dim, numdim, idxtag);
1238 defsymbol(name, ident, sLOCAL, -declared * sizeof(cell), tag);
1239 modstk(-(int)size * sizeof(cell));
1241 /* now that we have reserved memory for the variable, we can
1242 * proceed to initialize it */
1243 sym->compound = nestlevel; /* for multiple declaration/shadowing */
1245 sym->usage |= uCONST;
1246 if (ident == iARRAY)
1247 for (level = 0; level < numdim; level++)
1248 symbolrange(level, dim[level]);
1250 { /* static variables already initialized */
1251 if (ident == iVARIABLE)
1253 /* simple variable, also supports initialization */
1254 int ctag = tag; /* set to "tag" by default */
1255 int explicit_init = FALSE; /* is the variable explicitly
1257 if (matchtoken('='))
1259 doexpr(FALSE, FALSE, FALSE, FALSE, &ctag, TRUE);
1260 explicit_init = TRUE;
1264 const1(0); /* uninitialized variable, set to zero */
1266 /* now try to save the value (still in PRI) in the variable */
1268 lval.ident = iVARIABLE;
1271 check_userop(NULL, ctag, lval.tag, 2, NULL, &ctag);
1273 endexpr(TRUE); /* full expression ends after the store */
1274 if (!matchtag(tag, ctag, TRUE))
1275 error(213); /* tag mismatch */
1276 /* if the variable was not explicitly initialized, reset the
1277 * "uWRITTEN" flag that store() set */
1279 sym->usage &= ~uWRITTEN;
1284 if (litidx - cur_lit < size)
1285 fillarray(sym, size * sizeof(cell), 0);
1286 if (cur_lit < litidx)
1288 /* check whether the complete array is set to a single value;
1289 * if it is, more compact code can be generated */
1290 cell first = litq[cur_lit];
1293 for (i = cur_lit; i < litidx && litq[i] == first; i++)
1297 /* all values are the same */
1298 fillarray(sym, (litidx - cur_lit) * sizeof(cell),
1300 litidx = cur_lit; /* reset literal table */
1304 /* copy the literals to the array */
1305 const1((cur_lit + glb_declared) * sizeof(cell));
1306 copyarray(sym, (litidx - cur_lit) * sizeof(cell));
1312 while (matchtoken(',')); /* enddo *//* more? */
1313 needtoken(tTERM); /* if not comma, must be semicolumn */
1318 calc_arraysize(int dim[], int numdim, int cur)
1322 return dim[cur] + (dim[cur] * calc_arraysize(dim, numdim, cur + 1));
1327 * Initialize global objects and local arrays.
1328 * size==array cells (count), if 0 on input, the routine counts
1329 * the number of elements
1330 * tag==required tagname id (not the returned tag)
1332 * Global references: litidx (altered)
1335 initials(int ident, int tag, cell * size, int dim[], int numdim)
1338 int curlit = litidx;
1341 if (!matchtoken('='))
1343 if (ident == iARRAY && dim[numdim - 1] == 0)
1345 /* declared as "myvar[];" which is senseless (note: this *does* make
1346 * sense in the case of a iREFARRAY, which is a function parameter)
1348 error(9); /* array has zero length -> invalid size */
1352 /* initialize the indirection tables */
1354 #error Array algorithms for more than 2 dimensions are not implemented
1356 assert(numdim == 2);
1357 *size = calc_arraysize(dim, numdim, 0);
1358 for (d = 0; d < dim[0]; d++)
1359 stowlit((dim[0] + d * (dim[1] - 1)) * sizeof(cell));
1364 if (ident == iVARIABLE)
1368 if (!matchtag(tag, ctag, TRUE))
1369 error(213); /* tag mismatch */
1376 *size = initvector(ident, tag, dim[0], FALSE);
1382 /* The simple algorithm below only works for arrays with one or
1383 * two dimensions. This should be some recursive algorithm.
1385 if (dim[numdim - 1] != 0)
1386 /* set size to (known) full size */
1387 *size = calc_arraysize(dim, numdim, 0);
1388 /* dump indirection tables */
1389 for (d = 0; d < dim[0]; d++)
1391 /* now dump individual vectors */
1394 for (d = 0; d < dim[0]; d++)
1396 litq[curlit + d] = offs * sizeof(cell);
1397 dsize = initvector(ident, tag, dim[1], TRUE);
1401 if (matchtoken('{') || matchtoken(tSTRING))
1402 /* expect a '{' or a string */
1413 *size = litidx - curlit; /* number of elements defined */
1417 * Initialize a single dimensional array
1420 initvector(int ident, int tag, cell size, int fillzero)
1422 cell prev1 = 0, prev2 = 0;
1425 int curlit = litidx;
1427 assert(ident == iARRAY || ident == iREFARRAY);
1428 if (matchtoken('{'))
1432 if (matchtoken('}'))
1433 { /* to allow for trailing ',' after the initialization */
1437 if ((ellips = matchtoken(tELLIPS)) != 0)
1440 prev1 = init(ident, &ctag);
1441 if (!matchtag(tag, ctag, TRUE))
1442 error(213); /* tag mismatch */
1444 while (matchtoken(',')); /* do */
1450 if (!matchtag(tag, ctag, TRUE))
1451 error(213); /* tagname mismatch */
1453 /* fill up the literal queue with a series */
1457 ((litidx - curlit) == 1) ? (cell) 0 : prev1 - prev2;
1458 if (size == 0 || (litidx - curlit) == 0)
1459 error(41); /* invalid ellipsis, array size unknown */
1460 else if ((litidx - curlit) == (int)size)
1461 error(18); /* initialisation data exceeds declared size */
1462 while ((litidx - curlit) < (int)size)
1468 if (fillzero && size > 0)
1470 while ((litidx - curlit) < (int)size)
1475 size = litidx - curlit; /* number of elements defined */
1477 else if (litidx - curlit > (int)size)
1478 { /* e.g. "myvar[3]={1,2,3,4};" */
1479 error(18); /* initialisation data exceeds declared size */
1480 litidx = (int)size + curlit; /* avoid overflow in memory moves */
1487 * Evaluate one initializer.
1490 init(int ident, int *tag)
1494 if (matchtoken(tSTRING))
1496 /* lex() automatically stores strings in the literal table (and
1497 * increases "litidx") */
1498 if (ident == iVARIABLE)
1500 error(6); /* must be assigned to an array */
1501 litidx = 1; /* reset literal queue */
1505 else if (constexpr(&i, tag))
1507 stowlit(i); /* store expression result in literal table */
1514 * Get required array size
1522 if (matchtoken(']')) /* we've already seen "[" */
1523 return 0; /* null size (like "char msg[]") */
1524 constexpr(&val, tag); /* get value (must be constant expression) */
1527 error(9); /* negative array size is invalid; assumed zero */
1531 return val; /* return array size */
1534 /* decl_const - declare a single constant
1538 decl_const(int vclass)
1540 char constname[sNAMEMAX + 1];
1546 tag = sc_addtag(NULL);
1547 if (lex(&val, &str) != tSYMBOL) /* read in (new) token */
1548 error(20, str); /* invalid symbol name */
1549 symbolline = fline; /* save line where symbol was found */
1550 strcpy(constname, str); /* save symbol name */
1552 constexpr(&val, &exprtag); /* get value */
1554 /* add_constant() checks for duplicate definitions */
1555 if (!matchtag(tag, exprtag, FALSE))
1557 /* temporarily reset the line number to where the symbol was
1559 int orgfline = fline;
1562 error(213); /* tagname mismatch */
1565 add_constant(constname, val, vclass, tag);
1568 /* decl_enum - declare enumerated constants
1572 decl_enum(int vclass)
1574 char enumname[sNAMEMAX + 1], constname[sNAMEMAX + 1];
1575 cell val, value, size;
1577 int tok, tag, explicittag;
1578 cell increment, multiplier;
1580 /* get an explicit tag, if any (we need to remember whether an
1581 * explicit tag was passed, even if that explicit tag was "_:", so we
1582 * cannot call sc_addtag() here
1584 if (lex(&val, &str) == tLABEL)
1586 tag = sc_addtag(str);
1593 explicittag = FALSE;
1596 /* get optional enum name (also serves as a tag if no explicit
1598 if (lex(&val, &str) == tSYMBOL)
1599 { /* read in (new) token */
1600 strcpy(enumname, str); /* save enum name (last constant) */
1602 tag = sc_addtag(enumname);
1606 lexpush(); /* analyze again */
1610 /* get increment and multiplier */
1613 if (matchtoken('('))
1615 if (matchtoken(taADD))
1617 constexpr(&increment, NULL);
1619 else if (matchtoken(taMULT))
1621 constexpr(&multiplier, NULL);
1623 else if (matchtoken(taSHL))
1625 constexpr(&val, NULL);
1633 /* go through all constants */
1634 value = 0; /* default starting value */
1637 if (matchtoken('}'))
1638 { /* quick exit if '}' follows ',' */
1642 tok = lex(&val, &str); /* read in (new) token */
1643 if (tok != tSYMBOL && tok != tLABEL)
1644 error(20, str); /* invalid symbol name */
1645 strcpy(constname, str); /* save symbol name */
1646 size = increment; /* default increment of 'val' */
1647 if (tok == tLABEL || matchtoken(':'))
1648 constexpr(&size, NULL); /* get size */
1649 if (matchtoken('='))
1650 constexpr(&value, NULL); /* get value */
1651 /* add_constant() checks whether a variable (global or local) or
1652 * a constant with the same name already exists */
1653 add_constant(constname, value, vclass, tag);
1654 if (multiplier == 1)
1657 value *= size * multiplier;
1659 while (matchtoken(','));
1660 needtoken('}'); /* terminates the constant list */
1661 matchtoken(';'); /* eat an optional ; */
1663 /* set the enum name to the last value plus one */
1664 if (enumname[0] != '\0')
1665 add_constant(enumname, value, vclass, tag);
1669 * Finds a function in the global symbol table or creates a new entry.
1670 * It does some basic processing and error checking.
1673 fetchfunc(char *name, int tag)
1679 if ((sc_debug & sSYMBOLIC) != 0)
1681 offset += opcodes(1) + opargs(3) + nameincells(name);
1682 /* ^^^ The address for the symbol is the code address. But the
1683 * "symbol" instruction itself generates code. Therefore the
1684 * offset is pre-adjusted to the value it will have after the
1685 * symbol instruction.
1688 if ((sym = findglb(name)))
1689 { /* already in symbol table? */
1690 if (sym->ident != iFUNCTN)
1692 error(21, name); /* yes, but not as a function */
1693 return NULL; /* make sure the old symbol is not damaged */
1695 else if ((sym->usage & uDEFINE) != 0)
1697 error(21, name); /* yes, and it's already defined */
1699 else if ((sym->usage & uNATIVE) != 0)
1701 error(21, name); /* yes, and it is an native */
1703 assert(sym->vclass == sGLOBAL);
1704 if ((sym->usage & uDEFINE) == 0)
1706 /* as long as the function stays undefined, update the address
1714 /* don't set the "uDEFINE" flag; it may be a prototype */
1715 sym = addsym(name, offset, iFUNCTN, sGLOBAL, tag, 0);
1716 /* assume no arguments */
1717 sym->dim.arglist = (arginfo *) malloc(1 * sizeof(arginfo));
1718 sym->dim.arglist[0].ident = 0;
1719 /* set library ID to NULL (only for native functions) */
1725 /* This routine adds symbolic information for each argument.
1732 /* At this point, no local variables have been declared. All
1733 * local symbols are function arguments.
1738 assert(sym->ident != iLABEL);
1739 assert(sym->vclass == sLOCAL);
1740 defsymbol(sym->name, sym->ident, sLOCAL, sym->addr, sym->tag);
1741 if (sym->ident == iREFARRAY)
1747 symbolrange(sub->dim.array.level, sub->dim.array.length);
1748 sub = finddepend(sub);
1756 operatorname(char *name)
1762 assert(name != NULL);
1764 /* check the operator */
1765 opertok = lex(&val, &str);
1778 name[0] = (char)opertok;
1801 error(61); /* operator cannot be redefined
1802 * (or bad operator name) */
1810 operatoradjust(int opertok, symbol * sym, char *opername, int resulttag)
1812 int tags[2] = { 0, 0 };
1815 char tmpname[sNAMEMAX + 1];
1821 /* count arguments and save (first two) tags */
1822 while (arg = &sym->dim.arglist[count], arg->ident != 0)
1826 if (arg->numtags > 1)
1827 error(65, count + 1); /* function argument may only have
1829 else if (arg->numtags == 1)
1830 tags[count] = arg->tags[0];
1832 if (opertok == '~' && count == 0)
1834 if (arg->ident != iREFARRAY)
1835 error(73, arg->name); /* must be an array argument */
1839 if (arg->ident != iVARIABLE)
1840 error(66, arg->name); /* must be non-reference argument */
1842 if (arg->hasdefault)
1843 error(59, arg->name); /* arguments of an operator may not
1844 * have a default value */
1848 /* for '!', '++' and '--', count must be 1
1849 * for '-', count may be 1 or 2
1850 * for '=', count must be 1, and the resulttag is also important
1851 * for all other (binary) operators and the special '~'
1852 * operator, count must be 2
1861 error(62); /* number or placement of the operands does
1862 * not fit the operator */
1865 if (count != 1 && count != 2)
1866 error(62); /* number or placement of the operands does
1867 * not fit the operator */
1871 error(62); /* number or placement of the operands does
1872 * not fit the operator */
1876 && ((opertok != '=' && tags[1] == 0) || (opertok == '=' && resulttag == 0)))
1877 error(64); /* cannot change predefined operators */
1879 /* change the operator name */
1880 assert(opername[0] != '\0');
1881 operator_symname(tmpname, opername, tags[0], tags[1], count, resulttag);
1882 if ((oldsym = findglb(tmpname)))
1886 if ((oldsym->usage & uDEFINE) != 0)
1888 char errname[2 * sNAMEMAX + 16];
1890 funcdisplayname(errname, tmpname);
1891 error(21, errname); /* symbol already defined */
1893 sym->usage |= oldsym->usage; /* copy flags from the previous
1895 for (i = 0; i < oldsym->numrefers; i++)
1896 if (oldsym->refer[i])
1897 refer_symbol(sym, oldsym->refer[i]);
1898 delete_symbol(&glbtab, oldsym);
1900 if ((sc_debug & sSYMBOLIC) != 0)
1901 sym->addr += nameincells(tmpname) - nameincells(sym->name);
1902 strcpy(sym->name, tmpname);
1903 sym->hash = namehash(sym->name); /* calculate new hash */
1905 /* operators should return a value, except the '~' operator */
1907 sym->usage |= uRETVALUE;
1913 check_operatortag(int opertok, int resulttag, char *opername)
1915 assert(opername != NULL && opername[0] != '\0');
1925 if (resulttag != sc_addtag("bool"))
1927 error(63, opername, "bool:"); /* operator X requires
1928 * a "bool:" result tag */
1935 error(63, opername, "_:"); /* operator "~" requires
1936 * a "_:" result tag */
1945 tag2str(char *dest, int tag)
1949 sprintf(dest, "0%x", tag);
1950 return isdigit(dest[1]) ? &dest[1] : dest;
1954 operator_symname(char *symname, char *opername, int tag1, int tag2,
1955 int numtags, int resulttag)
1957 char tagstr1[10], tagstr2[10];
1960 assert(numtags >= 1 && numtags <= 2);
1961 opertok = (opername[1] == '\0') ? opername[0] : 0;
1963 sprintf(symname, "%s%s%s", tag2str(tagstr1, resulttag), opername,
1964 tag2str(tagstr2, tag1));
1965 else if (numtags == 1 || opertok == '~')
1966 sprintf(symname, "%s%s", opername, tag2str(tagstr1, tag1));
1968 sprintf(symname, "%s%s%s", tag2str(tagstr1, tag1), opername,
1969 tag2str(tagstr2, tag2));
1974 parse_funcname(char *fname, int *tag1, int *tag2, char *opname)
1979 /* tags are only positive, so if the function name starts with a '-',
1980 * the operator is an unary '-' or '--' operator.
1990 *tag1 = (int)strtol(fname, &ptr, 16);
1991 unary = ptr == fname; /* unary operator if it doesn't start
1992 * with a tag name */
1994 assert(!unary || *tag1 == 0);
1995 assert(*ptr != '\0');
1996 for (name = opname; !isdigit(*ptr);)
1999 *tag2 = (int)strtol(ptr, NULL, 16);
2004 funcdisplayname(char *dest, char *funcname)
2008 constvalue *tagsym[2];
2011 if (isalpha(*funcname) || *funcname == '_' || *funcname == PUBLIC_CHAR
2012 || *funcname == '\0')
2014 if (dest != funcname)
2015 strcpy(dest, funcname);
2019 unary = parse_funcname(funcname, &tags[0], &tags[1], opname);
2020 tagsym[1] = find_constval_byval(&tagname_tab, tags[1]);
2021 assert(tagsym[1] != NULL);
2024 sprintf(dest, "operator%s(%s:)", opname, tagsym[1]->name);
2028 tagsym[0] = find_constval_byval(&tagname_tab, tags[0]);
2029 /* special case: the assignment operator has the return value
2031 if (opname[0] == '=' && opname[1] == '\0')
2032 sprintf(dest, "%s:operator%s(%s:)", tagsym[0]->name, opname,
2035 sprintf(dest, "operator%s(%s:,%s:)", opname, tagsym[0]->name,
2042 funcstub(int native)
2047 char symbolname[sNAMEMAX + 1];
2053 litidx = 0; /* clear the literal pool */
2055 tag = sc_addtag(NULL);
2056 tok = lex(&val, &str);
2059 if (tok == tPUBLIC || tok == tSTOCK || tok == tSTATIC ||
2060 (tok == tSYMBOL && *str == PUBLIC_CHAR))
2061 error(42); /* invalid combination of class specifiers */
2065 if (tok == tPUBLIC || tok == tSTATIC)
2066 tok = lex(&val, &str);
2068 if (tok == tOPERATOR)
2070 opertok = operatorname(symbolname);
2072 return; /* error message already given */
2073 check_operatortag(opertok, tag, symbolname);
2077 if (tok != tSYMBOL && freading)
2079 error(10); /* illegal function or declaration */
2082 strcpy(symbolname, str);
2084 needtoken('('); /* only functions may be native/forward */
2086 sym = fetchfunc(symbolname, tag); /* get a pointer to the
2092 sym->usage = uNATIVE | uRETVALUE | uDEFINE;
2093 sym->x.lib = curlibrary;
2097 /* "declargs()" found the ")" */
2098 if (!operatoradjust(opertok, sym, symbolname, tag))
2099 sym->usage &= ~uDEFINE;
2100 /* for a native operator, also need to specify an "exported"
2101 * function name; for a native function, this is optional
2108 lexpush(); /* push back, for matchtoken() to retrieve again */
2110 if (matchtoken('='))
2112 /* allow number or symbol */
2113 if (matchtoken(tSYMBOL))
2115 tokeninfo(&val, &str);
2116 if (strlen(str) > sEXPMAX)
2118 error(220, str, sEXPMAX);
2119 str[sEXPMAX] = '\0';
2121 insert_alias(sym->name, str);
2125 constexpr(&val, NULL);
2128 * ?? Must mark this address, so that it won't be generated again
2129 * and it won't be written to the output file. At the moment,
2130 * I have assumed that this syntax is only valid if val < 0.
2131 * To properly mix "normal" native functions and indexed native
2132 * functions, one should use negative indices anyway.
2133 * Special code for a negative index in sym->addr exists in
2134 * SC4.C (ffcall()) and in SC6.C (the loops for counting the
2135 * number of native variables and for writing them).
2142 litidx = 0; /* clear the literal pool */
2143 /* clear local variables queue */
2144 delete_symbols(&loctab, 0, TRUE, TRUE);
2147 /* newfunc - begin a function
2149 * This routine is called from "parse" and tries to make a function
2150 * out of the following text
2152 * Global references: funcstatus,lastst,litidx
2155 * declared (altered)
2156 * glb_declared (altered)
2157 * sc_alignnext (altered)
2160 newfunc(char *firstname, int firsttag, int fpublic, int fstatic, int stock)
2163 int argcnt, tok, tag, funcline;
2164 int opertok, opererror;
2165 char symbolname[sNAMEMAX + 1];
2167 cell val, cidx, glbdecl;
2170 litidx = 0; /* clear the literal pool ??? */
2172 lastst = 0; /* no statement yet */
2173 cidx = 0; /* just to avoid compiler warnings */
2175 filenum = fcurrent; /* save file number at start of declaration */
2179 assert(strlen(firstname) <= sNAMEMAX);
2180 strcpy(symbolname, firstname); /* save symbol name */
2185 tag = (firsttag >= 0) ? firsttag : sc_addtag(NULL);
2186 tok = lex(&val, &str);
2188 if (tok == tNATIVE || (tok == tPUBLIC && stock))
2189 error(42); /* invalid combination of class specifiers */
2190 if (tok == tOPERATOR)
2192 opertok = operatorname(symbolname);
2194 return TRUE; /* error message already given */
2195 check_operatortag(opertok, tag, symbolname);
2199 if (tok != tSYMBOL && freading)
2201 error(20, str); /* invalid symbol name */
2204 assert(strlen(str) <= sNAMEMAX);
2205 strcpy(symbolname, str);
2208 /* check whether this is a function or a variable declaration */
2209 if (!matchtoken('('))
2211 /* so it is a function, proceed */
2212 funcline = fline; /* save line at which the function is defined */
2213 if (symbolname[0] == PUBLIC_CHAR)
2215 fpublic = TRUE; /* implicitly public function */
2217 error(42); /* invalid combination of class specifiers */
2219 sym = fetchfunc(symbolname, tag); /* get a pointer to the
2224 sym->usage |= uPUBLIC;
2226 sym->fnumber = filenum;
2227 /* declare all arguments */
2228 argcnt = declargs(sym);
2229 opererror = !operatoradjust(opertok, sym, symbolname, tag);
2230 if (strcmp(symbolname, uMAINFUNC) == 0)
2233 error(5); /* "main()" function may not have any arguments */
2234 sym->usage |= uREAD; /* "main()" is the program's entry point:
2237 /* "declargs()" found the ")"; if a ";" appears after this, it was a
2239 if (matchtoken(';'))
2241 if (!sc_needsemicolon)
2242 error(218); /* old style prototypes used with optional
2244 delete_symbols(&loctab, 0, TRUE, TRUE); /* prototype is done;
2245 * forget everything */
2248 /* so it is not a prototype, proceed */
2249 /* if this is a function that is not referred to (this can only be
2250 * detected in the second stage), shut code generation off */
2251 if (sc_status == statWRITE && (sym->usage & uREAD) == 0)
2253 sc_status = statSKIP;
2255 glbdecl = glb_declared;
2258 sym->usage |= uDEFINE; /* set the definition flag */
2260 sym->usage |= uREAD; /* public functions are always "used" */
2262 sym->usage |= uSTOCK;
2263 if (opertok != 0 && opererror)
2264 sym->usage &= ~uDEFINE;
2265 defsymbol(sym->name, iFUNCTN, sGLOBAL,
2266 code_idx + opcodes(1) + opargs(3) + nameincells(sym->name), tag);
2267 /* ^^^ The address for the symbol is the code address. But the
2268 * "symbol" instruction itself generates code. Therefore the
2269 * offset is pre-adjusted to the value it will have after the
2270 * symbol instruction.
2272 startfunc(sym->name); /* creates stack frame */
2273 if ((sc_debug & sSYMBOLIC) != 0)
2274 setline(funcline, fcurrent);
2277 alignframe(sc_dataalign);
2278 sc_alignnext = FALSE;
2280 declared = 0; /* number of local cells */
2281 rettype = (sym->usage & uRETVALUE); /* set "return type" variable */
2283 define_args(); /* add the symbolic info for the function arguments */
2284 statement(NULL, FALSE);
2285 if ((rettype & uRETVALUE) != 0)
2286 sym->usage |= uRETVALUE;
2289 /* This happens only in a very special (and useless) case, where a
2290 * function has only a single statement in its body (no compound
2291 * block) and that statement declares a new variable
2293 modstk((int)declared * sizeof(cell)); /* remove all local
2297 if ((lastst != tRETURN) && (lastst != tGOTO))
2301 if ((sym->usage & uRETVALUE) != 0)
2303 char symname[2 * sNAMEMAX + 16]; /* allow space for user
2304 * defined operators */
2305 funcdisplayname(symname, sym->name);
2306 error(209, symname); /* function should return a value */
2311 { /* if there are literals defined */
2312 glb_declared += litidx;
2313 begdseg(); /* flip to DATA segment */
2314 dumplits(); /* dump literal strings */
2317 testsymbols(&loctab, 0, TRUE, TRUE); /* test for unused arguments
2319 delete_symbols(&loctab, 0, TRUE, TRUE); /* clear local variables
2321 assert(loctab.next == NULL);
2323 if (sc_status == statSKIP)
2325 sc_status = statWRITE;
2327 glb_declared = glbdecl;
2333 argcompare(arginfo * a1, arginfo * a2)
2337 result = strcmp(a1->name, a2->name) == 0;
2339 result = a1->ident == a2->ident;
2341 result = a1->usage == a2->usage;
2343 result = a1->numtags == a2->numtags;
2348 for (i = 0; i < a1->numtags && result; i++)
2349 result = a1->tags[i] == a2->tags[i];
2352 result = a1->hasdefault == a2->hasdefault;
2355 if (a1->ident == iREFARRAY)
2358 result = a1->defvalue.array.size == a2->defvalue.array.size;
2361 a1->defvalue.array.arraysize == a2->defvalue.array.arraysize;
2362 /* also check the dimensions of both arrays */
2364 result = a1->numdim == a2->numdim;
2365 for (level = 0; result && level < a1->numdim; level++)
2366 result = a1->dim[level] == a2->dim[level];
2367 /* ??? should also check contents of the default array
2368 * (these troubles go away in a 2-pass compiler that forbids
2369 * double declarations, but Small currently does not forbid them)
2376 if ((a1->hasdefault & uSIZEOF) != 0
2377 || (a1->hasdefault & uTAGOF) != 0)
2378 result = a1->hasdefault == a2->hasdefault
2379 && strcmp(a1->defvalue.size.symname,
2380 a2->defvalue.size.symname) == 0
2381 && a1->defvalue.size.level == a2->defvalue.size.level;
2383 result = a1->defvalue.val == a2->defvalue.val;
2387 result = a1->defvalue_tag == a2->defvalue_tag;
2394 * This routine adds an entry in the local symbol table for each
2395 * argument found in the argument list.
2396 * It returns the number of arguments.
2399 declargs(symbol * sym)
2403 int argcnt, oldargcnt, tok, tags[MAXTAGS], numtags;
2405 arginfo arg, *arglist;
2406 char name[sNAMEMAX + 1];
2407 int ident, fpublic, fconst;
2410 /* if the function is already defined earlier, get the number of
2411 * arguments of the existing definition
2414 if ((sym->usage & uPROTOTYPED) != 0)
2415 while (sym->dim.arglist[oldargcnt].ident != 0)
2417 argcnt = 0; /* zero aruments up to now */
2421 fpublic = (sym->usage & uPUBLIC) != 0;
2422 /* the '(' parantheses has already been parsed */
2423 if (!matchtoken(')'))
2426 { /* there are arguments; process them */
2427 /* any legal name increases argument count (and stack offset) */
2428 tok = lex(&val, &ptr);
2435 if (ident != iVARIABLE || numtags > 0)
2436 error(1, "-identifier-", "&");
2440 if (ident != iVARIABLE || numtags > 0)
2441 error(1, "-identifier-", "const");
2446 error(1, "-identifier-", "-tagname-");
2447 tags[0] = sc_addtag(ptr);
2452 error(1, "-identifier-", "-tagname-");
2454 while (numtags < MAXTAGS)
2456 if (!matchtoken('_') && !needtoken(tSYMBOL))
2458 tokeninfo(&val, &ptr);
2459 tags[numtags++] = sc_addtag(ptr);
2460 if (matchtoken('}'))
2465 tok = tLABEL; /* for outer loop:
2466 * flag that we have seen a tagname */
2469 if (argcnt >= sMAXARGS)
2470 error(45); /* too many function arguments */
2471 strcpy(name, ptr); /* save symbol name */
2472 if (name[0] == PUBLIC_CHAR)
2473 error(56, name); /* function arguments cannot be public */
2475 tags[numtags++] = 0; /* default tag */
2477 * base + 0*sizeof(cell) == previous "base"
2478 * base + 1*sizeof(cell) == function return address
2479 * base + 2*sizeof(cell) == number of arguments
2480 * base + 3*sizeof(cell) == first argument of the function
2481 * So the offset of each argument is:
2482 * "(argcnt+3) * sizeof(cell)".
2484 doarg(name, ident, (argcnt + 3) * sizeof(cell), tags, numtags,
2485 fpublic, fconst, &arg);
2486 if (fpublic && arg.hasdefault)
2487 error(59, name); /* arguments of a public function may not
2488 * have a default value */
2489 if ((sym->usage & uPROTOTYPED) == 0)
2491 /* redimension the argument list, add the entry */
2493 (arginfo *) realloc(sym->dim.arglist,
2494 (argcnt + 2) * sizeof(arginfo));
2495 if (!sym->dim.arglist)
2496 error(103); /* insufficient memory */
2497 sym->dim.arglist[argcnt] = arg;
2498 sym->dim.arglist[argcnt + 1].ident = 0; /* keep the list
2503 /* check the argument with the earlier definition */
2504 if (argcnt > oldargcnt
2505 || !argcompare(&sym->dim.arglist[argcnt], &arg))
2506 error(25); /* function definition does not match prototype */
2507 /* may need to free default array argument and the tag list */
2508 if (arg.ident == iREFARRAY && arg.hasdefault)
2509 free(arg.defvalue.array.data);
2510 else if (arg.ident == iVARIABLE
2511 && ((arg.hasdefault & uSIZEOF) != 0
2512 || (arg.hasdefault & uTAGOF) != 0))
2513 free(arg.defvalue.size.symname);
2522 if (ident != iVARIABLE)
2523 error(10); /* illegal function or declaration */
2525 tags[numtags++] = 0; /* default tag */
2526 if ((sym->usage & uPROTOTYPED) == 0)
2528 /* redimension the argument list, add the entry iVARARGS */
2530 (arginfo *) realloc(sym->dim.arglist,
2531 (argcnt + 2) * sizeof(arginfo));
2532 if (!sym->dim.arglist)
2533 error(103); /* insufficient memory */
2534 sym->dim.arglist[argcnt + 1].ident = 0; /* keep the list
2536 sym->dim.arglist[argcnt].ident = iVARARGS;
2537 sym->dim.arglist[argcnt].hasdefault = FALSE;
2538 sym->dim.arglist[argcnt].defvalue.val = 0;
2539 sym->dim.arglist[argcnt].defvalue_tag = 0;
2540 sym->dim.arglist[argcnt].numtags = numtags;
2541 sym->dim.arglist[argcnt].tags =
2542 (int *)malloc(numtags * sizeof tags[0]);
2543 if (!sym->dim.arglist[argcnt].tags)
2544 error(103); /* insufficient memory */
2545 memcpy(sym->dim.arglist[argcnt].tags, tags,
2546 numtags * sizeof tags[0]);
2550 if (argcnt > oldargcnt
2551 || sym->dim.arglist[argcnt].ident != iVARARGS)
2552 error(25); /* function definition does not match prototype */
2557 error(10); /* illegal function or declaration */
2560 while (tok == '&' || tok == tLABEL || tok == tCONST || (tok != tELLIPS && matchtoken(','))); /* more? */
2561 /* if the next token is not ",", it should be ")" */
2564 /* resolve any "sizeof" arguments (now that all arguments are known) */
2565 assert(sym->dim.arglist != NULL);
2566 arglist = sym->dim.arglist;
2567 for (idx = 0; idx < argcnt && arglist[idx].ident != 0; idx++)
2569 if ((arglist[idx].hasdefault & uSIZEOF) != 0
2570 || (arglist[idx].hasdefault & uTAGOF) != 0)
2574 /* Find the argument with the name mentioned after the "sizeof".
2575 * Note that we cannot use findloc here because we need the
2576 * arginfo struct, not the symbol.
2578 ptr = arglist[idx].defvalue.size.symname;
2580 altidx < argcnt && strcmp(ptr, arglist[altidx].name) != 0;
2583 if (altidx >= argcnt)
2585 error(17, ptr); /* undefined symbol */
2589 /* check the level against the number of dimensions */
2590 /* the level must be zero for "tagof" values */
2591 assert(arglist[idx].defvalue.size.level == 0
2592 || (arglist[idx].hasdefault & uSIZEOF) != 0);
2593 if (arglist[idx].defvalue.size.level > 0
2594 && arglist[idx].defvalue.size.level >=
2595 arglist[altidx].numdim)
2596 error(28); /* invalid subscript */
2597 if (arglist[altidx].ident != iREFARRAY)
2599 assert(arglist[altidx].ident == iVARIABLE
2600 || arglist[altidx].ident == iREFERENCE);
2601 error(223, ptr); /* redundant sizeof */
2607 sym->usage |= uPROTOTYPED;
2608 errorset(sRESET); /* reset error flag (clear the "panic mode") */
2612 /* doarg - declare one argument type
2614 * this routine is called from "declargs()" and adds an entry in the
2615 * local symbol table for one argument. "fpublic" indicates whether
2616 * the function for this argument list is public.
2617 * The arguments themselves are never public.
2620 doarg(char *name, int ident, int offset, int tags[], int numtags,
2621 int fpublic, int fconst, arginfo * arg)
2625 int idxtag[sDIMEN_MAX];
2627 strcpy(arg->name, name);
2628 arg->hasdefault = FALSE; /* preset (most common case) */
2629 arg->defvalue.val = 0; /* clear */
2630 arg->defvalue_tag = 0;
2632 if (matchtoken('['))
2634 if (ident == iREFERENCE)
2635 error(67, name); /*illegal declaration ("&name[]" is unsupported) */
2638 if (arg->numdim == sDIMEN_MAX)
2640 error(53); /* exceeding maximum number of dimensions */
2643 /* there is no check for non-zero major dimensions here, only if
2644 * the array parameter has a default value, we enforce that all
2645 * array dimensions, except the last, are non-zero
2647 size = needsub(&idxtag[arg->numdim]); /* may be zero here,
2648 *it is a pointer anyway */
2649 #if INT_MAX < LONG_MAX
2651 error(105); /* overflow, exceeding capacity */
2653 arg->dim[arg->numdim] = (int)size;
2656 while (matchtoken('['));
2657 ident = iREFARRAY; /* "reference to array" (is a pointer) */
2658 if (matchtoken('='))
2662 lexpush(); /* initials() needs the "=" token again */
2663 assert(numtags > 0);
2664 /* for the moment, when a default value is given for the array,
2665 * all dimension sizes, except the last, must be non-zero
2666 * (function initials() requires to know the major dimensions)
2668 for (level = 0; level < arg->numdim - 1; level++)
2669 if (arg->dim[level] == 0)
2670 error(52); /* only last dimension may be variable length */
2671 initials(ident, tags[0], &size, arg->dim, arg->numdim);
2672 assert(size >= litidx);
2673 /* allocate memory to hold the initial values */
2674 arg->defvalue.array.data = (cell *) malloc(litidx * sizeof(cell));
2675 if (arg->defvalue.array.data)
2679 memcpy(arg->defvalue.array.data, litq, litidx * sizeof(cell));
2680 arg->hasdefault = TRUE; /* argument has default value */
2681 arg->defvalue.array.size = litidx;
2682 arg->defvalue.array.addr = -1;
2683 /* calculate size to reserve on the heap */
2684 arg->defvalue.array.arraysize = 1;
2685 for (i = 0; i < arg->numdim; i++)
2686 arg->defvalue.array.arraysize *= arg->dim[i];
2687 if (arg->defvalue.array.arraysize < arg->defvalue.array.size)
2688 arg->defvalue.array.arraysize = arg->defvalue.array.size;
2690 litidx = 0; /* reset */
2695 if (matchtoken('='))
2697 unsigned char size_tag_token;
2699 assert(ident == iVARIABLE || ident == iREFERENCE);
2700 arg->hasdefault = TRUE; /* argument has a default value */
2702 (unsigned char)(matchtoken(tSIZEOF) ? uSIZEOF : 0);
2703 if (size_tag_token == 0)
2705 (unsigned char)(matchtoken(tTAGOF) ? uTAGOF : 0);
2706 if (size_tag_token != 0)
2710 if (ident == iREFERENCE)
2711 error(66, name); /* argument may not be a reference */
2713 while (matchtoken('('))
2715 if (needtoken(tSYMBOL))
2717 /* save the name of the argument whose size id to take */
2721 tokeninfo(&val, &name);
2722 if (!(arg->defvalue.size.symname = strdup(name)))
2723 error(103); /* insufficient memory */
2724 arg->defvalue.size.level = 0;
2725 if (size_tag_token == uSIZEOF)
2727 while (matchtoken('['))
2729 arg->defvalue.size.level += (short)1;
2733 if (ident == iVARIABLE) /* make sure we set this only if
2734 * not a reference */
2735 arg->hasdefault |= size_tag_token; /* uSIZEOF or uTAGOF */
2737 while (paranthese--)
2742 constexpr(&arg->defvalue.val, &arg->defvalue_tag);
2743 assert(numtags > 0);
2744 if (!matchtag(tags[0], arg->defvalue_tag, TRUE))
2745 error(213); /* tagname mismatch */
2749 arg->ident = (char)ident;
2750 arg->usage = (char)(fconst ? uCONST : 0);
2751 arg->numtags = numtags;
2752 arg->tags = (int *)malloc(numtags * sizeof tags[0]);
2754 error(103); /* insufficient memory */
2755 memcpy(arg->tags, tags, numtags * sizeof tags[0]);
2756 argsym = findloc(name);
2759 error(21, name); /* symbol already defined */
2763 if ((argsym = findglb(name)) && argsym->ident != iFUNCTN)
2764 error(219, name); /* variable shadows another symbol */
2765 /* add details of type and address */
2766 assert(numtags > 0);
2767 argsym = addvariable(name, offset, ident, sLOCAL, tags[0],
2768 arg->dim, arg->numdim, idxtag);
2769 argsym->compound = 0;
2770 if (ident == iREFERENCE)
2771 argsym->usage |= uREAD; /* because references are passed back */
2773 argsym->usage |= uREAD; /* arguments of public functions
2774 * are always "used" */
2776 argsym->usage |= uCONST;
2781 count_referrers(symbol * entry)
2786 for (i = 0; i < entry->numrefers; i++)
2787 if (entry->refer[i])
2792 /* Every symbol has a referrer list, that contains the functions that
2793 * use the symbol. Now, if function "apple" is accessed by functions
2794 * "banana" and "citron", but neither function "banana" nor "citron" are
2795 * used by anyone else, then, by inference, function "apple" is not used
2798 reduce_referrers(symbol * root)
2806 for (sym = root->next; sym; sym = sym->next)
2809 continue; /* hierarchical data type */
2810 if (sym->ident == iFUNCTN
2811 && (sym->usage & uNATIVE) == 0
2812 && (sym->usage & uPUBLIC) == 0
2813 && strcmp(sym->name, uMAINFUNC) != 0
2814 && count_referrers(sym) == 0)
2816 sym->usage &= ~(uREAD | uWRITTEN); /* erase usage bits if
2817 * there is no referrer */
2818 /* find all symbols that are referred by this symbol */
2819 for (ref = root->next; ref; ref = ref->next)
2822 continue; /* hierarchical data type */
2823 assert(ref->refer != NULL);
2824 for (i = 0; i < ref->numrefers && ref->refer[i] != sym;
2827 if (i < ref->numrefers)
2829 assert(ref->refer[i] == sym);
2830 ref->refer[i] = NULL;
2835 else if ((sym->ident == iVARIABLE || sym->ident == iARRAY)
2836 && (sym->usage & uPUBLIC) == 0
2837 && !sym->parent && count_referrers(sym) == 0)
2839 sym->usage &= ~(uREAD | uWRITTEN); /* erase usage bits if
2840 * there is no referrer */
2843 /* after removing a symbol, check whether more can be removed */
2845 while (restart > 0);
2848 /* testsymbols - test for unused local or global variables
2850 * "Public" functions are excluded from the check, since these
2851 * may be exported to other object modules.
2852 * Labels are excluded from the check if the argument 'testlabs'
2853 * is 0. Thus, labels are not tested until the end of the function.
2854 * Constants may also be excluded (convenient for global constants).
2856 * When the nesting level drops below "level", the check stops.
2858 * The function returns whether there is an "entry" point for the file.
2859 * This flag will only be 1 when browsing the global symbol table.
2862 testsymbols(symbol * root, int level, int testlabs, int testconst)
2864 char symname[2 * sNAMEMAX + 16];
2867 symbol *sym = root->next;
2869 while (sym && sym->compound >= level)
2876 if ((sym->usage & uDEFINE) == 0)
2877 error(19, sym->name); /* not a label: ... */
2878 else if ((sym->usage & uREAD) == 0)
2879 error(203, sym->name); /* symbol isn't used: ... */
2883 if ((sym->usage & (uDEFINE | uREAD | uNATIVE | uSTOCK)) == uDEFINE)
2885 funcdisplayname(symname, sym->name);
2886 if (symname[0] != '\0')
2887 error(203, symname); /* symbol isn't used ...
2888 * (and not native/stock) */
2890 if ((sym->usage & uPUBLIC) != 0
2891 || strcmp(sym->name, uMAINFUNC) == 0)
2892 entry = TRUE; /* there is an entry point */
2895 if (testconst && (sym->usage & uREAD) == 0)
2896 error(203, sym->name); /* symbol isn't used: ... */
2901 break; /* hierarchical data type */
2902 if ((sym->usage & (uWRITTEN | uREAD | uSTOCK | uPUBLIC)) == 0)
2903 error(203, sym->name); /* symbol isn't used (and not stock
2905 else if ((sym->usage & (uREAD | uSTOCK | uPUBLIC)) == 0)
2906 error(204, sym->name); /* value assigned to symbol is
2908 #if 0 /*// ??? not sure whether it is a good idea to
2909 * force people use "const" */
2910 else if ((sym->usage & (uWRITTEN | uPUBLIC | uCONST)) == 0
2911 && sym->ident == iREFARRAY)
2912 error(214, sym->name); /* make array argument "const" */
2922 calc_array_datasize(symbol * sym, cell * offset)
2926 assert(sym != NULL);
2927 assert(sym->ident == iARRAY || sym->ident == iREFARRAY);
2928 length = sym->dim.array.length;
2929 if (sym->dim.array.level > 0)
2932 calc_array_datasize(finddepend(sym), offset);
2934 *offset = length * (*offset + sizeof(cell));
2936 length *= length * sublength;
2949 destructsymbols(symbol * root, int level)
2952 int savepri = FALSE;
2953 symbol *sym = root->next;
2955 while (sym && sym->compound >= level)
2957 if (sym->ident == iVARIABLE || sym->ident == iARRAY)
2959 char symbolname[16];
2963 /* check that the '~' operator is defined for this tag */
2964 operator_symname(symbolname, "~", sym->tag, 0, 1, 0);
2965 if ((opsym = findglb(symbolname)))
2967 /* save PRI, in case of a return statement */
2970 push1(); /* right-hand operand is in PRI */
2973 /* if the variable is an array, get the number of elements */
2974 if (sym->ident == iARRAY)
2976 elements = calc_array_datasize(sym, &offset);
2977 /* "elements" can be zero when the variable is declared like
2978 * new mytag: myvar[2][] = { {1, 2}, {3, 4} }
2979 * one should declare all dimensions!
2982 error(46, sym->name); /* array size is unknown */
2990 /* call the '~' operator */
2992 addconst(offset); /*add offset to array data to the address */
2994 pushval(2 * sizeof(cell)); /* 2 parameters */
2996 if (sc_status != statSKIP)
2997 markusage(opsym, uREAD); /* do not mark as "used" when this
2998 * call itself is skipped */
3000 opsym->x.lib->value += 1; /* increment "usage count"
3006 /* restore PRI, if it was saved */
3012 insert_constval(constvalue * prev, constvalue * next, char *name,
3013 cell val, short index)
3017 if (!(cur = (constvalue *)malloc(sizeof(constvalue))))
3018 error(103); /* insufficient memory (fatal error) */
3019 memset(cur, 0, sizeof(constvalue));
3020 strcpy(cur->name, name);
3029 append_constval(constvalue * table, char *name, cell val, short index)
3031 constvalue *cur, *prev;
3033 /* find the end of the constant table */
3034 for (prev = table, cur = table->next; cur;
3035 prev = cur, cur = cur->next)
3037 return insert_constval(prev, NULL, name, val, index);
3041 find_constval(constvalue * table, char *name, short index)
3043 constvalue *ptr = table->next;
3047 if (strcmp(name, ptr->name) == 0 && ptr->index == index)
3055 find_constval_byval(constvalue * table, cell val)
3057 constvalue *ptr = table->next;
3061 if (ptr->value == val)
3068 #if 0 /* never used */
3070 delete_constval(constvalue * table, char *name)
3072 constvalue *prev = table;
3073 constvalue *cur = prev->next;
3077 if (strcmp(name, cur->name) == 0)
3079 prev->next = cur->next;
3091 delete_consttable(constvalue * table)
3093 constvalue *cur = table->next, *next;
3101 memset(table, 0, sizeof(constvalue));
3106 * Adds a symbol to the #define symbol table.
3109 add_constant(char *name, cell val, int vclass, int tag)
3113 /* Test whether a global or local symbol with the same name exists. Since
3114 * constants are stored in the symbols table, this also finds previously
3115 * defind constants. */
3116 sym = findglb(name);
3118 sym = findloc(name);
3121 /* silently ignore redefinitions of constants with the same value */
3122 if (sym->ident == iCONSTEXPR)
3124 if (sym->addr != val)
3125 error(201, name); /* redefinition of constant (different value) */
3129 error(21, name); /* symbol already defined */
3134 /* constant doesn't exist yet, an entry must be created */
3135 sym = addsym(name, val, iCONSTEXPR, vclass, tag, uDEFINE);
3136 if (sc_status == statIDLE)
3137 sym->usage |= uPREDEF;
3140 /* statement - The Statement Parser
3142 * This routine is called whenever the parser needs to know what
3143 * statement it encounters (i.e. whenever program syntax requires a
3147 statement(int *lastindent, int allow_decl)
3155 error(36); /* empty statement */
3160 tok = lex(&val, &st);
3162 setline(fline, fcurrent);
3163 /* lex() has set stmtindent */
3164 if (lastindent && tok != tLABEL)
3167 if (*lastindent >= 0 && *lastindent != stmtindent &&
3168 !indent_nowarn && sc_tabsize > 0)
3169 error(217); /* loose indentation */
3171 *lastindent = stmtindent;
3172 indent_nowarn = TRUE; /* if warning was blocked, re-enable it */
3187 error(3); /* declaration only valid in a block */
3198 error(3); /* declaration only valid in a block */
3202 if (!matchtoken('}')) /* {} is the empty statement */
3204 /* lastst (for "last statement") does not change */
3207 error(36); /* empty statement */
3231 error(14); /* not in switch */
3271 default: /* non-empty expression */
3272 lexpush(); /* analyze token later */
3273 doexpr(TRUE, TRUE, TRUE, TRUE, NULL, FALSE);
3283 cell save_decl = declared;
3286 nestlevel += 1; /* increase compound statement level */
3287 while (matchtoken('}') == 0)
3288 { /* repeat until compound statement is closed */
3291 needtoken('}'); /* gives error: "expected token }" */
3297 && (lastst == tRETURN || lastst == tBREAK
3298 || lastst == tCONTINUE))
3299 error(225); /* unreachable code */
3300 statement(&indent, TRUE); /* do a statement */
3304 if (lastst != tRETURN)
3305 destructsymbols(&loctab, nestlevel);
3306 if (lastst != tRETURN && lastst != tGOTO)
3307 /* delete local variable space */
3308 modstk((int)(declared - save_decl) * sizeof(cell));
3310 testsymbols(&loctab, nestlevel, FALSE, TRUE); /* look for unused
3312 declared = save_decl;
3313 delete_symbols(&loctab, nestlevel, FALSE, TRUE);
3314 /* erase local symbols, but
3315 * retain block local labels
3316 * (within the function) */
3318 nestlevel -= 1; /* decrease compound statement level */
3323 * Global references: stgidx (referred to only)
3326 doexpr(int comma, int chkeffect, int allowarray, int mark_endexpr,
3327 int *tag, int chkfuncresult)
3329 int constant, index, ident;
3330 int localstaging = FALSE;
3335 stgset(TRUE); /* start stage-buffering */
3336 localstaging = TRUE;
3337 assert(stgidx == 0);
3340 errorset(sEXPRMARK);
3343 /* on second round through, mark the end of the previous expression */
3344 if (index != stgidx)
3347 ident = expression(&constant, &val, tag, chkfuncresult);
3348 if (!allowarray && (ident == iARRAY || ident == iREFARRAY))
3349 error(33, "-unknown-"); /* array must be indexed */
3350 if (chkeffect && !sideeffect)
3351 error(215); /* expression has no effect */
3353 while (comma && matchtoken(',')); /* more? */
3355 endexpr(TRUE); /* optionally, mark the end of the expression */
3356 errorset(sEXPRRELEASE);
3360 stgset(FALSE); /* stop staging */
3367 constexpr(cell * val, int *tag)
3369 int constant, index;
3372 stgset(TRUE); /* start stage-buffering */
3373 stgget(&index, &cidx); /* mark position in code generator */
3374 errorset(sEXPRMARK);
3375 expression(&constant, val, tag, FALSE);
3376 stgdel(index, cidx); /* scratch generated code */
3377 stgset(FALSE); /* stop stage-buffering */
3379 error(8); /* must be constant expression */
3380 errorset(sEXPRRELEASE);
3386 * In the case a "simple assignment" operator ("=") is used within a
3387 * test, * the warning "possibly unintended assignment" is displayed.
3388 * This routine sets the global variable "intest" to true, it is
3389 * restored upon termination. In the case the assignment was intended,
3390 * use parantheses around the expression to avoid the warning;
3391 * primary() sets "intest" to 0.
3393 * Global references: intest (altered, but restored upon termination)
3396 test(int label, int parens, int invert)
3400 value lval = { NULL, 0, 0, 0, 0, NULL };
3401 int localstaging = FALSE;
3405 stgset(TRUE); /* start staging */
3406 localstaging = TRUE;
3408 stgget(&index, &cidx); /* should start at zero if started
3414 pushstk((stkitem) intest);
3420 stgget(&index, &cidx); /* mark position (of last expression) in
3424 tok = matchtoken(',');
3428 while (tok); /* do */
3431 if (lval.ident == iARRAY || lval.ident == iREFARRAY)
3434 (lval.sym->name) ? lval.sym->name : "-unknown-";
3435 error(33, ptr); /* array must be indexed */
3437 if (lval.ident == iCONSTEXPR)
3438 { /* constant expression */
3439 intest = (int)(long)popstk(); /* restore stack */
3440 stgdel(index, cidx);
3442 { /* code always executed */
3443 error(206); /* redundant test: always non-zero */
3447 error(205); /* redundant code: never executed */
3452 stgout(0); /* write "jumplabel" code */
3453 stgset(FALSE); /* stop staging */
3457 if (lval.tag != 0 && lval.tag != sc_addtag("bool"))
3458 if (check_userop(lneg, lval.tag, 0, 1, NULL, &lval.tag))
3459 invert = !invert; /* user-defined ! operator inverted result */
3461 jmp_ne0(label); /* jump to label if true (different from 0) */
3463 jmp_eq0(label); /* jump to label if false (equal to 0) */
3464 endexpr(TRUE); /* end expression (give optimizer a chance) */
3465 intest = (int)(long)popstk(); /* double typecast to avoid warning
3466 * with Microsoft C */
3469 stgout(0); /* output queue from the very beginning (see
3470 * assert() when localstaging is set to TRUE) */
3471 stgset(FALSE); /* stop staging */
3481 ifindent = stmtindent; /* save the indent of the "if" instruction */
3482 flab1 = getlabel(); /* get label number for false branch */
3483 test(flab1, TRUE, FALSE); /*get expression, branch to flab1 if false */
3484 statement(NULL, FALSE); /* if true, do a statement */
3485 if (matchtoken(tELSE) == 0)
3487 setlabel(flab1); /* no, simple if..., print false label */
3491 /* to avoid the "dangling else" error, we want a warning if the "else"
3492 * has a lower indent than the matching "if" */
3494 if (stmtindent < ifindent && sc_tabsize > 0)
3495 error(217); /* loose indentation */
3498 if ((lastst != tRETURN) && (lastst != tGOTO))
3500 setlabel(flab1); /* print false label */
3501 statement(NULL, FALSE); /* do "else" clause */
3502 setlabel(flab2); /* print true label */
3509 int wq[wqSIZE]; /* allocate local queue */
3511 addwhile(wq); /* add entry to queue for "break" */
3512 setlabel(wq[wqLOOP]); /* loop label */
3513 /* The debugger uses the "line" opcode to be able to "break" out of
3514 * a loop. To make sure that each loop has a line opcode, even for the
3515 * tiniest loop, set it below the top of the loop */
3516 setline(fline, fcurrent);
3517 test(wq[wqEXIT], TRUE, FALSE); /* branch to wq[wqEXIT] if false */
3518 statement(NULL, FALSE); /* if so, do a statement */
3519 jumplabel(wq[wqLOOP]); /* and loop to "while" start */
3520 setlabel(wq[wqEXIT]); /* exit label */
3521 delwhile(); /* delete queue entry */
3525 * Note that "continue" will in this case not jump to the top of the
3526 * loop, but to the end: just before the TRUE-or-FALSE testing code.
3531 int wq[wqSIZE], top;
3533 addwhile(wq); /* see "dowhile" for more info */
3534 top = getlabel(); /* make a label first */
3535 setlabel(top); /* loop label */
3536 statement(NULL, FALSE);
3538 setlabel(wq[wqLOOP]); /* "continue" always jumps to WQLOOP. */
3539 setline(fline, fcurrent);
3540 test(wq[wqEXIT], TRUE, FALSE);
3542 setlabel(wq[wqEXIT]);
3550 int wq[wqSIZE], skiplab;
3552 int save_nestlevel, index;
3555 save_decl = declared;
3556 save_nestlevel = nestlevel;
3559 skiplab = getlabel();
3561 if (matchtoken(';') == 0)
3563 /* new variable declarations are allowed here */
3564 if (matchtoken(tNEW))
3566 /* The variable in expr1 of the for loop is at a
3567 * 'compound statement' level of it own.
3570 declloc(FALSE); /* declare local variable */
3574 doexpr(TRUE, TRUE, TRUE, TRUE, NULL, FALSE); /* expression 1 */
3578 /* Adjust the "declared" field in the "while queue", in case that
3579 * local variables were declared in the first expression of the
3580 * "for" loop. These are deleted in separately, so a "break" or a
3581 * "continue" must ignore these fields.
3584 assert(ptr != NULL);
3585 ptr[wqBRK] = (int)declared;
3586 ptr[wqCONT] = (int)declared;
3587 jumplabel(skiplab); /* skip expression 3 1st time */
3588 setlabel(wq[wqLOOP]); /* "continue" goes to this label: expr3 */
3589 setline(fline, fcurrent);
3590 /* Expressions 2 and 3 are reversed in the generated code:
3591 * expression 3 precedes expression 2.
3592 * When parsing, the code is buffered and marks for
3593 * the start of each expression are insterted in the buffer.
3596 stgset(TRUE); /* start staging */
3597 assert(stgidx == 0);
3599 stgmark(sSTARTREORDER);
3600 stgmark((char)(sEXPRSTART + 0)); /* mark start of 2nd expression
3602 setlabel(skiplab); /*jump to this point after 1st expression */
3603 if (matchtoken(';') == 0)
3605 test(wq[wqEXIT], FALSE, FALSE); /* expression 2
3606 *(jump to wq[wqEXIT] if false) */
3609 stgmark((char)(sEXPRSTART + 1)); /* mark start of 3th expression
3611 if (matchtoken(')') == 0)
3613 doexpr(TRUE, TRUE, TRUE, TRUE, NULL, FALSE); /* expression 3 */
3616 stgmark(sENDREORDER); /* mark end of reversed evaluation */
3618 stgset(FALSE); /* stop staging */
3619 statement(NULL, FALSE);
3620 jumplabel(wq[wqLOOP]);
3621 setlabel(wq[wqEXIT]);
3624 assert(nestlevel >= save_nestlevel);
3625 if (nestlevel > save_nestlevel)
3627 /* Clean up the space and the symbol table for the local
3628 * variable in "expr1".
3630 destructsymbols(&loctab, nestlevel);
3631 modstk((int)(declared - save_decl) * sizeof(cell));
3632 declared = save_decl;
3633 delete_symbols(&loctab, nestlevel, FALSE, TRUE);
3634 nestlevel = save_nestlevel; /* reset 'compound statement'
3639 /* The switch statement is incompatible with its C sibling:
3640 * 1. the cases are not drop through
3641 * 2. only one instruction may appear below each case, use a compound
3642 * instruction to execute multiple instructions
3643 * 3. the "case" keyword accepts a comma separated list of values to
3644 * match, it also accepts a range using the syntax "1 .. 4"
3647 * PRI = expression result
3648 * param = table offset (code segment)
3654 int lbl_table, lbl_exit, lbl_case;
3655 int tok, swdefault, casecount;
3658 constvalue caselist = { NULL, "", 0, 0 }; /*case list starts empty */
3659 constvalue *cse, *csp;
3660 char labelname[sNAMEMAX + 1];
3663 doexpr(TRUE, FALSE, FALSE, TRUE, NULL, FALSE); /* evaluate
3664 * switch expression */
3666 /* generate the code for the switch statement, the label is the
3667 * address of the case table (to be generated later).
3669 lbl_table = getlabel();
3670 lbl_case = 0; /* just to avoid a compiler warning */
3671 ffswitch(lbl_table);
3674 lbl_exit = getlabel(); /*get label number for jumping out of switch */
3679 tok = lex(&val, &str); /* read in (new) token */
3683 if (swdefault != FALSE)
3684 error(15); /* "default" case must be last in switch
3686 lbl_case = getlabel();
3687 sc_allowtags = FALSE; /* do not allow tagnames here */
3692 /* ??? enforce/document that, in a switch, a statement cannot
3693 * start an opening brace (marks the start of a compound
3694 * statement) and search for the right-most colon before that
3696 * Now, by replacing the ':' by a special COLON token, you can
3697 * parse all expressions until that special token.
3700 constexpr(&val, NULL);
3701 /* Search the insertion point (the table is kept in sorted
3702 * order, so that advanced abstract machines can sift the
3703 * case table with a binary search). Check for duplicate
3704 * case values at the same time.
3706 for (csp = &caselist, cse = caselist.next;
3707 cse && cse->value < val;
3708 csp = cse, cse = cse->next)
3710 if (cse && cse->value == val)
3711 error(40, val); /* duplicate "case" label */
3712 /* Since the label is stored as a string in the
3713 * "constvalue", the size of an identifier must
3714 * be at least 8, as there are 8
3715 * hexadecimal digits in a 32-bit number.
3718 #error Length of identifier (sNAMEMAX) too small.
3720 insert_constval(csp, cse, itoh(lbl_case), val, 0);
3721 if (matchtoken(tDBLDOT))
3725 constexpr(&end, NULL);
3727 error(50); /* invalid range */
3728 while (++val <= end)
3731 /* find the new insertion point */
3732 for (csp = &caselist, cse = caselist.next;
3733 cse && cse->value < val;
3734 csp = cse, cse = cse->next)
3736 if (cse && cse->value == val)
3737 error(40, val); /* duplicate "case" label */
3738 insert_constval(csp, cse, itoh(lbl_case), val, 0);
3742 while (matchtoken(','));
3743 needtoken(':'); /* ':' ends the case */
3744 sc_allowtags = TRUE; /* reset */
3746 statement(NULL, FALSE);
3747 jumplabel(lbl_exit);
3750 if (swdefault != FALSE)
3751 error(16); /* multiple defaults in switch */
3752 lbl_case = getlabel();
3756 statement(NULL, FALSE);
3757 /* Jump to lbl_exit, even thouh this is the last clause in the
3758 *switch, because the jump table is generated between the last
3759 * clause of the switch and the exit label.
3761 jumplabel(lbl_exit);
3764 /* nothing, but avoid dropping into "default" */
3768 indent_nowarn = TRUE; /* disable this check */
3769 tok = '}'; /* break out of the loop after an error */
3775 /* verify that the case table is sorted (unfortunately, duplicates can
3776 * occur; there really shouldn't be duplicate cases, but the compiler
3777 * may not crash or drop into an assertion for a user error). */
3778 for (cse = caselist.next; cse && cse->next; cse = cse->next)
3779 ; /* empty. no idea whether this is correct, but we MUST NOT do
3780 * the setlabel(lbl_table) call in the loop body. doing so breaks
3781 * switch statements that only have one case statement following.
3785 /* generate the table here, before lbl_exit (general jump target) */
3786 setlabel(lbl_table);
3788 if (swdefault == FALSE)
3790 /* store lbl_exit as the "none-matched" label in the switch table */
3791 strcpy(labelname, itoh(lbl_exit));
3795 /* lbl_case holds the label of the "default" clause */
3796 strcpy(labelname, itoh(lbl_case));
3798 ffcase(casecount, labelname, TRUE);
3799 /* generate the rest of the table */
3800 for (cse = caselist.next; cse; cse = cse->next)
3801 ffcase(cse->value, cse->name, FALSE);
3804 delete_consttable(&caselist); /* clear list of case labels */
3812 value lval = { NULL, 0, 0, 0, 0, NULL };
3814 if ((sc_debug & sCHKBOUNDS) != 0)
3816 flab1 = getlabel(); /* get label number for "OK" branch */
3817 test(flab1, FALSE, TRUE); /* get expression and branch
3818 * to flab1 if true */
3819 setline(fline, fcurrent); /* make sure we abort on the correct
3821 ffabort(xASSERTION);
3826 stgset(TRUE); /* start staging */
3827 stgget(&index, &cidx); /* mark position in code generator */
3832 stgdel(index, cidx); /* just scrap the code */
3834 while (matchtoken(','));
3835 stgset(FALSE); /* stop staging */
3847 if (lex(&val, &st) == tSYMBOL)
3850 jumplabel((int)sym->addr);
3851 sym->usage |= uREAD; /* set "uREAD" bit */
3853 * // ??? if the label is defined (check sym->usage & uDEFINE), check
3854 * // sym->compound (nesting level of the label) against nestlevel;
3855 * // if sym->compound < nestlevel, call the destructor operator
3860 error(20, st); /* illegal symbol name */
3872 tokeninfo(&val, &st); /* retrieve label name again */
3873 if (find_constval(&tagname_tab, st, 0))
3874 error(221, st); /* label name shadows tagname */
3876 setlabel((int)sym->addr);
3877 /* since one can jump around variable declarations or out of compound
3878 * blocks, the stack must be manually adjusted
3880 setstk(-declared * sizeof(cell));
3881 sym->usage |= uDEFINE; /* label is now defined */
3886 * Finds a label from the (local) symbol table or adds one to it.
3887 * Labels are local in scope.
3889 * Note: The "_usage" bit is set to zero. The routines that call
3890 * "fetchlab()" must set this bit accordingly.
3893 fetchlab(char *name)
3897 sym = findloc(name); /* labels are local in scope */
3900 if (sym->ident != iLABEL)
3901 error(19, sym->name); /* not a label: ... */
3905 sym = addsym(name, getlabel(), iLABEL, sLOCAL, 0, 0);
3906 sym->x.declared = (int)declared;
3907 sym->compound = nestlevel;
3914 * Global references: rettype (altered)
3921 if (matchtoken(tTERM) == 0)
3923 if ((rettype & uRETNONE) != 0)
3924 error(208); /* mix "return;" and "return value;" */
3925 doexpr(TRUE, FALSE, FALSE, TRUE, &tag, FALSE);
3927 rettype |= uRETVALUE; /* function returns a value */
3928 /* check tagname with function tagname */
3929 assert(curfunc != NULL);
3930 if (!matchtag(curfunc->tag, tag, TRUE))
3931 error(213); /* tagname mismatch */
3935 /* this return statement contains no expression */
3937 if ((rettype & uRETVALUE) != 0)
3939 char symname[2 * sNAMEMAX + 16]; /* allow space for user
3940 * defined operators */
3941 assert(curfunc != NULL);
3942 funcdisplayname(symname, curfunc->name);
3943 error(209, symname); /* function should return a value */
3945 rettype |= uRETNONE; /* function does not return anything */
3947 destructsymbols(&loctab, 0); /*call destructor for *all* locals */
3948 modstk((int)declared * sizeof(cell)); /* end of function, remove
3949 *all* * local variables*/
3958 ptr = readwhile(); /* readwhile() gives an error if not in loop */
3962 destructsymbols(&loctab, nestlevel);
3963 modstk(((int)declared - ptr[wqBRK]) * sizeof(cell));
3964 jumplabel(ptr[wqEXIT]);
3972 ptr = readwhile(); /* readwhile() gives an error if not in loop */
3976 destructsymbols(&loctab, nestlevel);
3977 modstk(((int)declared - ptr[wqCONT]) * sizeof(cell));
3978 jumplabel(ptr[wqLOOP]);
3984 /* find the tag by value in the table, then set the top bit to mark it
3991 assert((tag & PUBLICTAG) == 0);
3992 for (ptr = tagname_tab.next;
3993 ptr && tag != (int)(ptr->value & TAGMASK); ptr = ptr->next)
3996 ptr->value |= PUBLICTAG;
4005 if (matchtoken(tTERM) == 0)
4007 doexpr(TRUE, FALSE, FALSE, TRUE, &tag, FALSE);
4016 destructsymbols(&loctab, 0); /* call destructor for *all* locals */
4025 if (matchtoken(tTERM) == 0)
4027 doexpr(TRUE, FALSE, FALSE, TRUE, &tag, FALSE);
4044 ptr[wqBRK] = (int)declared; /* stack pointer (for "break") */
4045 ptr[wqCONT] = (int)declared; /* for "continue", possibly adjusted later */
4046 ptr[wqLOOP] = getlabel();
4047 ptr[wqEXIT] = getlabel();
4048 if (wqptr >= (wq + wqTABSZ - wqSIZE))
4049 error(102, "loop table"); /* loop table overflow (too many active loops) */
4052 { /* copy "ptr" to while queue table */
4072 error(24); /* out of context */
4077 return (wqptr - wqSIZE);