2 // RMAC - Reboot's Macro Assembler for the Atari Jaguar Console System
3 // TOKEN.C - Token Handling
4 // Copyright (C) 199x Landon Dyer, 2011-2017 Reboot and Friends
5 // RMAC derived from MADMAC v1.07 Written by Landon Dyer, 1986
6 // Source utilised with the kind permission of Landon Dyer
17 #define DECL_KW // Declare keyword arrays
18 #define DEF_KW // Declare keyword values
19 #include "kwtab.h" // Incl generated keyword tables & defs
22 int lnsave; // 1; strcpy() text of current line
23 uint16_t curlineno; // Current line number (64K max currently)
24 int totlines; // Total # of lines
25 int mjump_align = 0; // mjump alignment flag
26 char lntag; // Line tag
27 char * curfname; // Current filename
28 char tolowertab[128]; // Uppercase ==> lowercase
29 int8_t hextab[128]; // Table of hex values
30 char dotxtab[128]; // Table for ".b", ".s", etc.
31 char irbuf[LNSIZ]; // Text for .rept block line
32 char lnbuf[LNSIZ]; // Text of current line
33 WORD filecount; // Unique file number counter
34 WORD cfileno; // Current file number
35 TOKEN * tok; // Ptr to current token
36 TOKEN * etok; // Ptr past last token in tokbuf[]
37 TOKEN tokeol[1] = {EOL}; // Bailout end-of-line token
38 char * string[TOKBUFSIZE*2]; // Token buffer string pointer storage
40 // File record, used to maintain a list of every include file ever visited
41 #define FILEREC struct _filerec
51 INOBJ * cur_inobj; // Ptr current input obj (IFILE/IMACRO)
52 static INOBJ * f_inobj; // Ptr list of free INOBJs
53 static IFILE * f_ifile; // Ptr list of free IFILEs
54 static IMACRO * f_imacro; // Ptr list of free IMACROs
56 static TOKEN tokbuf[TOKBUFSIZE]; // Token buffer (stack-like, all files)
59 ILLEG, ILLEG, ILLEG, ILLEG, // NUL SOH STX ETX
60 ILLEG, ILLEG, ILLEG, ILLEG, // EOT ENQ ACK BEL
61 ILLEG, WHITE, ILLEG, ILLEG, // BS HT LF VT
62 WHITE, ILLEG, ILLEG, ILLEG, // FF CR SO SI
64 ILLEG, ILLEG, ILLEG, ILLEG, // DLE DC1 DC2 DC3
65 ILLEG, ILLEG, ILLEG, ILLEG, // DC4 NAK SYN ETB
66 ILLEG, ILLEG, ILLEG, ILLEG, // CAN EM SUB ESC
67 ILLEG, ILLEG, ILLEG, ILLEG, // FS GS RS US
69 WHITE, MULTX, MULTX, SELF, // SP ! " #
70 MULTX+CTSYM, MULTX, SELF, MULTX, // $ % & '
71 SELF, SELF, SELF, SELF, // ( ) * +
72 SELF, SELF, STSYM, SELF, // , - . /
74 DIGIT+HDIGIT+CTSYM, DIGIT+HDIGIT+CTSYM, // 0 1
75 DIGIT+HDIGIT+CTSYM, DIGIT+HDIGIT+CTSYM, // 2 3
76 DIGIT+HDIGIT+CTSYM, DIGIT+HDIGIT+CTSYM, // 4 5
77 DIGIT+HDIGIT+CTSYM, DIGIT+HDIGIT+CTSYM, // 6 7
78 DIGIT+HDIGIT+CTSYM, DIGIT+HDIGIT+CTSYM, // 8 9
80 MULTX, MULTX, MULTX, STSYM+CTSYM, // < = > ?
82 MULTX, STSYM+CTSYM+HDIGIT, // @ A
83 (char)((BYTE)DOT)+STSYM+CTSYM+HDIGIT, STSYM+CTSYM+HDIGIT, // B C
84 STSYM+CTSYM+HDIGIT, STSYM+CTSYM+HDIGIT, // D E
85 STSYM+CTSYM+HDIGIT, STSYM+CTSYM, // F G
86 STSYM+CTSYM, STSYM+CTSYM, STSYM+CTSYM, STSYM+CTSYM, // H I J K
87 (char)((BYTE)DOT)+STSYM+CTSYM, STSYM+CTSYM, STSYM+CTSYM, STSYM+CTSYM, // L M N O
89 STSYM+CTSYM, STSYM+CTSYM, STSYM+CTSYM, (char)((BYTE)DOT)+STSYM+CTSYM, // P Q R S
90 STSYM+CTSYM, STSYM+CTSYM, STSYM+CTSYM, (char)((BYTE)DOT)+STSYM+CTSYM, // T U V W
91 STSYM+CTSYM, STSYM+CTSYM, STSYM+CTSYM, SELF, // X Y Z [
92 SELF, SELF, MULTX, STSYM+CTSYM, // \ ] ^ _
94 ILLEG, STSYM+CTSYM+HDIGIT, // ` a
95 (char)((BYTE)DOT)+STSYM+CTSYM+HDIGIT, STSYM+CTSYM+HDIGIT, // b c
96 STSYM+CTSYM+HDIGIT, STSYM+CTSYM+HDIGIT, // d e
97 STSYM+CTSYM+HDIGIT, STSYM+CTSYM, // f g
98 STSYM+CTSYM, STSYM+CTSYM, STSYM+CTSYM, STSYM+CTSYM, // h i j k
99 (char)((BYTE)DOT)+STSYM+CTSYM, STSYM+CTSYM, STSYM+CTSYM, STSYM+CTSYM, // l m n o
101 STSYM+CTSYM, STSYM+CTSYM, STSYM+CTSYM, (char)((BYTE)DOT)+STSYM+CTSYM, // p q r s
102 STSYM+CTSYM, STSYM+CTSYM, STSYM+CTSYM, (char)((BYTE)DOT)+STSYM+CTSYM, // t u v w
103 STSYM+CTSYM, STSYM+CTSYM, STSYM+CTSYM, SELF, // x y z {
104 SELF, SELF, SELF, ILLEG // | } ~ DEL
107 // Names of registers
108 static char * regname[] = {
109 "d0", "d1", "d2", "d3", "d4", "d5", "d6", "d7",
110 "a0", "a1", "a2", "a3", "a4", "a5", "a6", "a7",
111 "pc", "ssp", "usp", "sr", "ccr"
114 static char * riscregname[] = {
115 "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7",
116 "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15",
117 "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23",
118 "r24", "r25", "r26", "r27", "r28", "r29", "r30", "r31"
123 // Initialize tokenizer
125 void InitTokenizer(void)
128 char * htab = "0123456789abcdefABCDEF"; // Hex character table
130 lnsave = 0; // Don't save lines
131 curfname = ""; // No file, empty filename
132 filecount = (WORD)-1;
133 cfileno = (WORD)-1; // cfileno gets bumped to 0
145 // Initialize hex, "dot" and tolower tables
150 tolowertab[i] = (char)i;
153 for(i=0; htab[i]!=EOS; i++)
154 hextab[htab[i]] = (char)((i < 16) ? i : i - 6);
156 for(i='A'; i<='Z'; i++)
157 tolowertab[i] |= 0x20;
159 // These characters are legal immediately after a period
160 dotxtab['b'] = DOTB; // .b .B .s .S
164 dotxtab['w'] = DOTW; // .w .W
166 dotxtab['l'] = DOTL; // .l .L
168 dotxtab['i'] = DOTI; // .i .I (???)
173 void SetFilenameForErrorReporting(void)
177 // Check for absolute top filename (this should never happen)
180 curfname = "(*top*)";
184 FILEREC * fr = filerec;
186 // Advance to the correct record...
187 while (fr != NULL && fnum != 0)
193 // Check for file # record not found (this should never happen either)
196 curfname = "(*NOT FOUND*)";
200 curfname = fr->frec_name;
205 // Allocate an IFILE or IMACRO
207 INOBJ * a_inobj(int typ)
213 // Allocate and initialize INOBJ first
215 inobj = malloc(sizeof(INOBJ));
219 f_inobj = f_inobj->in_link;
224 case SRC_IFILE: // Alloc and init an IFILE
226 ifile = malloc(sizeof(IFILE));
230 f_ifile = f_ifile->if_link;
233 inobj->inobj.ifile = ifile;
235 case SRC_IMACRO: // Alloc and init an IMACRO
236 if (f_imacro == NULL)
237 imacro = malloc(sizeof(IMACRO));
241 f_imacro = f_imacro->im_link;
244 inobj->inobj.imacro = imacro;
246 case SRC_IREPT: // Alloc and init an IREPT
247 inobj->inobj.irept = malloc(sizeof(IREPT));
248 DEBUG printf("alloc IREPT\n");
252 // Install INOBJ on top of input stack
253 inobj->in_ifent = ifent; // Record .if context on entry
254 inobj->in_type = (WORD)typ;
255 inobj->in_otok = tok;
256 inobj->in_etok = etok;
257 inobj->in_link = cur_inobj;
265 // Perform macro substitution from 'orig' to 'dest'. Return OK or some error.
266 // A macro reference is in one of two forms:
267 // \name <non-name-character>
269 // A doubled backslash (\\) is compressed to a single backslash (\).
270 // Argument definitions have been pre-tokenized, so we have to turn them back
271 // into text. This means that numbers, in particular, become hex, regardless of
272 // their representation when the macro was invoked. This is a hack.
273 // A label may appear at the beginning of the line:
274 // :<name><whitespace>
275 // (the colon must be in the first column). These labels are stripped before
276 // macro expansion takes place.
278 int ExpandMacro(char * src, char * dest, int destsiz)
281 int questmark; // \? for testing argument existence
282 char mname[128]; // Assume max size of a formal arg name
283 char numbuf[20]; // Buffer for text of CONSTs
286 char ** symbolString;
288 DEBUG { printf("ExM: src=\"%s\"\n", src); }
290 IMACRO * imacro = cur_inobj->inobj.imacro;
291 int macnum = (int)(imacro->im_macro->sattr);
293 char * dst = dest; // Next dest slot
294 char * edst = dest + destsiz - 1; // End + 1(?) of dest buffer
296 // Check for (and skip over) any "label" on the line
302 while (*s != EOS && !(chrtab[*s] & WHITE))
306 s++; // Skip first whitespace
309 // Expand the rest of the line
312 // Copy single character
318 // Skip comments in case a loose @ or \ is in there
319 // In that case the tokeniser was trying to expand it.
320 if ((*s == ';') || ((*s == '/') && (*(s + 1) == '/')))
325 // Do macro expansion
333 case '\\': // \\, \ (collapse to single backslash)
339 case '?': // \? <macro> set `questmark' flag
343 case '#': // \#, number of arguments
344 sprintf(numbuf, "%d", (int)imacro->im_nargs);
346 case '!': // \! size suffix supplied on invocation
347 switch ((int)imacro->im_siz)
349 case SIZN: d = ""; break;
350 case SIZB: d = ".b"; break;
351 case SIZW: d = ".w"; break;
352 case SIZL: d = ".l"; break;
356 case '~': // ==> unique label string Mnnnn...
357 sprintf(numbuf, "M%u", curuniq);
373 return error("missing argument name");
376 // \n ==> argument number 'n', 0..9
377 if (chrtab[*s] & DIGIT)
387 // Get argument name: \name, \{name}
397 while (chrtab[*s] & CTSYM);
402 for(++s; *s != EOS && *s != '}';)
406 return error("missing '}'");
413 // Lookup the argument and copy its (string) value into the
414 // destination string
415 DEBUG printf("argument='%s'\n", mname);
417 if ((arg = lookup(mname, MACARG, macnum)) == NULL)
418 return errors("undefined argument: '%s'", mname);
421 // Convert a string of tokens (terminated with EOL) back into
422 // text. If an argument is out of range (not specified in the
423 // macro invocation) then it is ignored.
424 i = (int)arg->svalue;
426 DEBUG printf("~argnumber=%d (argBase=%u)\n", i, imacro->argBase);
429 if (i < imacro->im_nargs)
434 tk = argPtrs[imacro->argBase + i];
436 tk = imacro->argument[i].token;
437 symbolString = imacro->argument[i].string;
440 // printf("ExM: Preparing to parse argument #%u...\n", i);
447 // 0 if the argument is empty or non-existant,
448 // 1 if the argument is not empty
451 if (tk == NULL || *tk == EOL)
457 *dst++ = (char)(questmark + '0');
461 // Argument # is in range, so expand it
466 // Reverse-translation from a token number to a string.
467 // This is a hack. It might be better table-driven.
470 if ((*tk >= KW_D0) && !rdsp && !rgpu)
472 d = regname[(int)*tk++ - KW_D0];
475 else if ((*tk >= KW_R0) && (*tk <= KW_R31))
477 d = riscregname[(int)*tk++ - KW_R0];
486 // d = (char *)*tk++;
489 // This fix should be done for strings too
490 d = symbolString[*tk++];
491 DEBUG printf("ExM: SYMBOL=\"%s\"", d);
496 // d = (char *)*tk++;
499 d = symbolString[*tk++];
520 // Shamus: Changing the format specifier from %lx to %ux caused the assembler
521 // to choke on legitimate code... Need to investigate this further
522 // before changing anything else here!
524 sprintf(numbuf, "$%lx", (long unsigned int)*tk++);
588 *dst++ = (char)*(tk - 1);
593 // If 'd' != NULL, copy string to destination
597 DEBUG printf("d='%s'\n", d);
616 DEBUG { printf("ExM: dst=\"%s\"\n", dest); }
621 DEBUG printf("*** OVERFLOW LINE ***\n%s\n", dest);
622 return fatal("line too long as a result of macro expansion");
627 // Get next line of text from a macro
629 char * GetNextMacroLine(void)
631 IMACRO * imacro = cur_inobj->inobj.imacro;
632 // LONG * strp = imacro->im_nextln;
633 struct LineList * strp = imacro->im_nextln;
635 if (strp == NULL) // End-of-macro
638 imacro->im_nextln = strp->next;
639 // ExpandMacro((char *)(strp + 1), imacro->im_lnbuf, LNSIZ);
640 ExpandMacro(strp->line, imacro->im_lnbuf, LNSIZ);
642 return imacro->im_lnbuf;
647 // Get next line of text from a repeat block
649 char * GetNextRepeatLine(void)
651 IREPT * irept = cur_inobj->inobj.irept;
652 LONG * strp = irept->ir_nextln; // initial null
654 // Do repeat at end of .rept block's string list
657 DEBUG printf("back-to-top-of-repeat-block count=%d\n", (int)irept->ir_count);
658 irept->ir_nextln = irept->ir_firstln; // copy first line
660 if (irept->ir_count-- == 0)
662 DEBUG printf("end-repeat-block\n");
666 strp = irept->ir_nextln;
669 strcpy(irbuf, (char *)(irept->ir_nextln + 1));
670 DEBUG printf("repeat line='%s'\n", irbuf);
671 irept->ir_nextln = (LONG *)*strp;
678 // Include a source file used at the root, and for ".include" files
680 int include(int handle, char * fname)
684 printf("[include: %s, cfileno=%u]\n", fname, cfileno);
686 // Alloc and initialize include-descriptors
687 INOBJ * inobj = a_inobj(SRC_IFILE);
688 IFILE * ifile = inobj->inobj.ifile;
690 ifile->ifhandle = handle; // Setup file handle
691 ifile->ifind = ifile->ifcnt = 0; // Setup buffer indices
692 ifile->ifoldlineno = curlineno; // Save old line number
693 ifile->ifoldfname = curfname; // Save old filename
694 ifile->ifno = cfileno; // Save old file number
696 // NB: This *must* be preincrement, we're adding one to the filecount here!
697 cfileno = ++filecount; // Compute NEW file number
698 curfname = strdup(fname); // Set current filename (alloc storage)
699 curlineno = 0; // Start on line zero
701 // Add another file to the file-record
702 FILEREC * fr = (FILEREC *)malloc(sizeof(FILEREC));
703 fr->frec_next = NULL;
704 fr->frec_name = curfname;
707 filerec = fr; // Add first filerec
709 last_fr->frec_next = fr; // Append to list of filerecs
712 DEBUG printf("[include: curfname: %s, cfileno=%u]\n", curfname, cfileno);
719 // Pop the current input level
726 INOBJ * inobj = cur_inobj;
730 // Pop IFENT levels until we reach the conditional assembly context we
731 // were at when the input object was entered.
732 int numUnmatched = 0;
734 while (ifent != inobj->in_ifent)
736 if (d_endif() != 0) // Something bad happened during endif parsing?
737 return -1; // If yes, bail instead of getting stuck in a loop
742 // Give a warning to the user that we had to wipe their bum for them
743 if (numUnmatched > 0)
744 warni("missing %d .endif(s)", numUnmatched);
746 tok = inobj->in_otok; // Restore tok and otok
747 etok = inobj->in_etok;
749 switch (inobj->in_type)
751 case SRC_IFILE: // Pop and release an IFILE
753 printf("[Leaving: %s]\n", curfname);
755 ifile = inobj->inobj.ifile;
756 ifile->if_link = f_ifile;
758 close(ifile->ifhandle); // Close source file
759 if (debug) printf("[fpop (pre): curfname=%s]\n", curfname);
760 curfname = ifile->ifoldfname; // Set current filename
761 if (debug) printf("[fpop (post): curfname=%s]\n", curfname);
762 if (debug) printf("[fpop: (pre) cfileno=%d ifile->ifno=%d]\n", (int)cfileno, (int)ifile->ifno);
763 curlineno = ifile->ifoldlineno; // Set current line#
764 DEBUG printf("cfileno=%d ifile->ifno=%d\n", (int)cfileno, (int)ifile->ifno);
765 cfileno = ifile->ifno; // Restore current file number
766 if (debug) printf("[fpop: (post) cfileno=%d ifile->ifno=%d]\n", (int)cfileno, (int)ifile->ifno);
768 case SRC_IMACRO: // Pop and release an IMACRO
769 imacro = inobj->inobj.imacro;
770 imacro->im_link = f_imacro;
773 case SRC_IREPT: // Pop and release an IREPT
774 DEBUG printf("dealloc IREPT\n");
775 p = inobj->inobj.irept->ir_firstln;
786 cur_inobj = inobj->in_link;
787 inobj->in_link = f_inobj;
796 // Get line from file into buf, return NULL on EOF or ptr to the start of a
799 char * GetNextLine(void)
803 int readamt = -1; // 0 if last read() yeilded 0 bytes
804 IFILE * fl = cur_inobj->inobj.ifile;
808 // Scan for next end-of-line; handle stupid text formats by treating
809 // \r\n the same as \n. (lone '\r' at end of buffer means we have to
811 d = &fl->ifbuf[fl->ifind];
813 for(p=d, i=0, j=fl->ifcnt; i<j; i++, p++)
815 if (*p == '\r' || *p == '\n')
822 break; // Need to read more, then look for '\n' to eat
823 else if (p[1] == '\n')
827 // Cover up the newline with end-of-string sentinel
836 // Handle hanging lines by ignoring them (Input file is exhausted, no
837 // \r or \n on last line)
838 // Shamus: This is retarded. Never ignore any input!
839 if (!readamt && fl->ifcnt)
846 // Really should check to see if we're at the end of the buffer!
848 fl->ifbuf[fl->ifind + fl->ifcnt] = '\0';
850 return &fl->ifbuf[fl->ifind];
854 // Truncate and return absurdly long lines.
855 if (fl->ifcnt >= QUANTUM)
857 fl->ifbuf[fl->ifind + fl->ifcnt - 1] = '\0';
859 return &fl->ifbuf[fl->ifind];
862 // Relocate what's left of a line to the beginning of the buffer, and
863 // read some more of the file in; return NULL if the buffer's empty and
867 p = &fl->ifbuf[fl->ifind];
868 d = &fl->ifbuf[fl->ifcnt & 1];
870 for(i=0; i<fl->ifcnt; i++)
873 fl->ifind = fl->ifcnt & 1;
876 readamt = read(fl->ifhandle, &fl->ifbuf[fl->ifind + fl->ifcnt], QUANTUM);
881 if ((fl->ifcnt += readamt) == 0)
890 int TokenizeLine(void)
892 char * ln = NULL; // Ptr to current position in line
893 char * p; // Random character ptr
894 TOKEN * tk; // Token-deposit ptr
895 int state = 0; // State for keyword detector
896 int j = 0; // Var for keyword detector
897 char c; // Random char
898 VALUE v; // Random value
899 char * nullspot = NULL; // Spot to clobber for SYMBOL termination
900 int stuffnull; // 1:terminate SYMBOL '\0' at *nullspot
902 int stringNum = 0; // Pointer to string locations in tokenized line
906 if (cur_inobj == NULL) // Return EOF if input stack is empty
909 // Get another line of input from the current input source: a file, a
910 // macro, or a repeat-block
911 switch (cur_inobj->in_type)
915 // o bump source line number;
916 // o tag the listing-line with a space;
917 // o kludge lines generated by Alcyon C.
919 if ((ln = GetNextLine()) == NULL)
921 if (debug) printf("TokenizeLine: Calling fpop() from SRC_IFILE...\n");
922 if (fpop() == 0) // Pop input level
923 goto retry; // Try for more lines
926 ifent->if_prev = (IFENT *) - 1; //Signal Assemble() that we have reached EOF with unbalanced if/endifs
931 curlineno++; // Bump line number
936 // AS68 compatibility, throw away all lines starting with
937 // back-quotes, tildes, or '*'
938 // On other lines, turn the first '*' into a semi-colon.
939 if (*ln == '`' || *ln == '~' || *ln == '*')
943 for(p=ln; *p!=EOS; p++)
956 // o Handle end-of-macro;
957 // o tag the listing-line with an at (@) sign.
959 if ((ln = GetNextMacroLine()) == NULL)
961 if (ExitMacro() == 0) // Exit macro (pop args, do fpop(), etc)
962 goto retry; // Try for more lines...
964 return TKEOF; // Oops, we got a non zero return code, signal EOF
970 // o Handle end-of-repeat-block;
971 // o tag the listing-line with a pound (#) sign.
973 if ((ln = GetNextRepeatLine()) == NULL)
975 if (debug) printf("TokenizeLine: Calling fpop() from SRC_IREPT...\n");
984 // Save text of the line. We only do this during listings and within
985 // macro-type blocks, since it is expensive to unconditionally copy every
990 // General house-keeping
991 tok = tokeol; // Set "tok" to EOL in case of error
992 tk = etok; // Reset token ptr
993 stuffnull = 0; // Don't stuff nulls
994 totlines++; // Bump total #lines assembled
996 // See if the entire line is a comment. This is a win if the programmer
997 // puts in lots of comments
998 if (*ln == '*' || *ln == ';' || ((*ln == '/') && (*(ln + 1) == '/')))
1001 // Main tokenization loop;
1002 // o skip whitespace;
1003 // o handle end-of-line;
1004 // o handle symbols;
1005 // o handle single-character tokens (operators, etc.);
1006 // o handle multiple-character tokens (constants, strings, etc.).
1009 // Skip whitespace, handle EOL
1010 while ((int)chrtab[*ln] & WHITE)
1013 // Handle EOL, comment with ';'
1014 if (*ln == EOS || *ln == ';'|| ((*ln == '/') && (*(ln + 1) == '/')))
1017 // Handle start of symbol. Symbols are null-terminated in place. The
1018 // termination is always one symbol behind, since there may be no place
1019 // for a null in the case that an operator immediately follows the name.
1024 if (stuffnull) // Terminate old symbol from previous pass
1027 v = 0; // Assume no DOT attrib follows symbol
1030 // In some cases, we need to check for a DOTx at the *beginning*
1031 // of a symbol, as the "start" of the line we're currently looking
1032 // at could be somewhere in the middle of that line!
1035 // Make sure that it's *only* a .[bwsl] following, and not the
1036 // start of a local symbol:
1037 if ((chrtab[*(ln + 1)] & DOT)
1038 && (dotxtab[*(ln + 1)] != 0)
1039 && !(chrtab[*(ln + 2)] & CTSYM))
1041 // We found a legitimate DOTx construct, so add it to the
1045 *tk++ = (TOKEN)dotxtab[*ln++];
1050 p = nullspot = ln++; // Nullspot -> start of this symbol
1052 // Find end of symbol (and compute its length)
1053 for(j=1; (int)chrtab[*ln]&CTSYM; j++)
1056 // Handle "DOT" special forms (like ".b") that follow a normal
1057 // symbol or keyword:
1060 *ln++ = EOS; // Terminate symbol
1061 stuffnull = 0; // And never try it again
1063 // Character following the `.' must have a DOT attribute, and
1064 // the chararacter after THAT one must not have a start-symbol
1065 // attribute (to prevent symbols that look like, for example,
1066 // "zingo.barf", which might be a good idea anyway....)
1067 if (((chrtab[*ln] & DOT) == 0) || (dotxtab[*ln] == 0))
1068 return error("[bwsl] must follow '.' in symbol");
1070 v = (VALUE)dotxtab[*ln++];
1072 if (chrtab[*ln] & CTSYM)
1073 return error("misuse of '.', not allowed in symbols");
1076 // If the symbol is small, check to see if it's really the name of
1080 for(state=0; state>=0;)
1082 j = (int)tolowertab[*p++];
1085 if (kwcheck[j] != state)
1091 if (*p == EOS || p == ln)
1105 // Make j = -1 if user tries to use a RISC register while in 68K mode
1106 if (!(rgpu || rdsp) && ((TOKEN)j >= KW_R0 && (TOKEN)j <= KW_R31))
1111 // Make j = -1 if time, date etc with no preceeding ^^
1112 // defined, referenced, streq, macdef, date and time
1115 case 112: // defined
1116 case 113: // referenced
1124 // If not tokenized keyword OR token was not found
1125 if ((j < 0) || (state < 0))
1129 //problem here: nullspot is a char * but TOKEN is a uint32_t. On a 64-bit
1130 //system, this will cause all kinds of mischief.
1132 *tk++ = (TOKEN)nullspot;
1134 string[stringNum] = nullspot;
1145 if (v) // Record attribute token (if any)
1148 if (stuffnull) // Arrange for string termination on next pass
1154 // Handle identity tokens
1161 // Handle multiple-character tokens
1167 case '!': // ! or !=
1177 case '\'': // 'string'
1180 // Hardcoded for now, maybe this will change in the future
1185 case '\"': // "string"
1189 string[stringNum] = ln;
1193 for(p=ln; *ln!=EOS && *ln!=c1;)
1202 return(error("unterminated string"));
1231 warn("bad backslash code in string");
1241 return error("unterminated string");
1245 case '$': // $, hex constant
1246 if (chrtab[*ln] & HDIGIT)
1250 // Parse the hex value
1251 while (hextab[*ln] >= 0)
1252 v = (v << 4) + (int)hextab[*ln++];
1254 // ggn: Okay, some comments here are in order I think....
1255 // The original madmac sources didn't parse the size at
1256 // this point (i.e. .b/.w/.l). It was probably done at
1257 // another point, although it's unclear to me exactly
1258 // where. So why change this? My understanding (at least
1259 // from what SCPCD said on IRC) is that .w addressing
1260 // formats produce wrong code on jaguar (or doesn't execute
1261 // properly? something like that). So the code was changed
1262 // to mask off the upper bits depending on length (note: I
1263 // don't think .b is valid at all! I only know of .w/.l, so
1264 // this should probably be wiped). Then the code that
1265 // parses the constant and checks to see if it's between
1266 // $ffff0000 and $8000 never got triggered, so yay job
1267 // done! ...now say we want to assemble a st .prg. One of
1268 // the most widely spread optimisations is move.X expr.w,Y
1269 // (or vice versa, or both, anyway...) to access hardware
1270 // registers (which are mapped to $fxxxxx). This botchy
1271 // thing would create "hilarious" code while trying to
1272 // access hardware registers. So I made a condition to see
1273 // if st mode or jaguar is active and apply the both or
1274 // not. One last note: this is hardcoded to get optimised
1275 // for now on ST mode, i.e. it can't generate code like
1276 // move.w $00001234,d0 - it'll always get optimised to
1277 // move.w $1234.w,d0. It's probably ok, but maybe a warning
1278 // should be emitted? Or maybe finding a way to make it not
1279 // auto-optimise? I think it's ok for now...
1282 if (obj_format == BSD)
1284 if ((*(ln + 1) & 0xDF) == 'B')
1289 else if ((*(ln + 1) & 0xDF) == 'W')
1294 else if ((*(ln + 1) & 0xDF) == 'L')
1304 if (obj_format == ALCYON)
1308 if ((*(ln + 1) == 'w') || (*(ln + 1) == 'W'))
1313 else if ((*(ln + 1) == 'l') || (*(ln + 1) == 'L'))
1325 case '<': // < or << or <> or <=
1344 case ':': // : or ::
1354 case '=': // = or ==
1364 case '>': // > or >> or >=
1379 case '%': // % or binary constant
1380 if (*ln < '0' || *ln > '1')
1388 while (*ln >= '0' && *ln <= '1')
1389 v = (v << 1) + *ln++ - '0';
1393 if ((*(ln + 1) == 'b') || (*(ln + 1) == 'B'))
1399 if ((*(ln + 1) == 'w') || (*(ln + 1) == 'W'))
1405 if ((*(ln + 1) == 'l') || (*(ln + 1) == 'L'))
1414 case '@': // @ or octal constant
1415 if (*ln < '0' || *ln > '7')
1423 while (*ln >= '0' && *ln <= '7')
1424 v = (v << 3) + *ln++ - '0';
1428 if ((*(ln+1) == 'b') || (*(ln+1) == 'B'))
1434 if ((*(ln+1) == 'w') || (*(ln+1) == 'W'))
1440 if ((*(ln+1) == 'l') || (*(ln+1) == 'L'))
1449 case '^': // ^ or ^^ <operator-name>
1456 if (((int)chrtab[*++ln] & STSYM) == 0)
1458 error("invalid symbol following ^^");
1464 while ((int)chrtab[*ln] & CTSYM)
1467 for(state=0; state>=0;)
1469 // Get char, convert to lowercase
1472 if (j >= 'A' && j <= 'Z')
1477 if (kwcheck[j] != state)
1483 if (*p == EOS || p == ln)
1492 if (j < 0 || state < 0)
1494 error("unknown symbol following ^^");
1501 interror(2); // Bad MULTX entry in chrtab
1506 // Handle decimal constant
1511 while ((int)chrtab[*ln] & DIGIT)
1512 v = (v * 10) + *ln++ - '0';
1514 // See if there's a .[bwl] after the constant & deal with it if so
1517 if ((*(ln + 1) == 'b') || (*(ln + 1) == 'B'))
1522 else if ((*(ln + 1) == 'w') || (*(ln + 1) == 'W'))
1527 else if ((*(ln + 1) == 'l') || (*(ln + 1) == 'L'))
1535 //printf("CONST: %i\n", v);
1539 // Handle illegal character
1540 return error("illegal character");
1543 // Terminate line of tokens and return "success."
1546 tok = etok; // Set tok to beginning of line
1548 if (stuffnull) // Terminate last SYMBOL
1558 // .GOTO <label> goto directive
1560 // The label is searched for starting from the first line of the current,
1561 // enclosing macro definition. If no enclosing macro exists, an error is
1564 // A label is of the form:
1566 // :<name><whitespace>
1568 // The colon must appear in column 1. The label is stripped prior to macro
1569 // expansion, and is NOT subject to macro expansion. The whitespace may also
1572 int d_goto(WORD unused)
1574 // Setup for the search
1576 return error("missing label");
1578 char * sym = string[tok[1]];
1581 if (cur_inobj->in_type != SRC_IMACRO)
1582 return error("goto not in macro");
1584 IMACRO * imacro = cur_inobj->inobj.imacro;
1585 struct LineList * defln = imacro->im_macro->lineList;
1587 // Attempt to find the label, starting with the first line.
1588 for(; defln!=NULL; defln=defln->next)
1590 // Must start with a colon
1591 if (defln->line[0] == ':')
1593 // Compare names (sleazo string compare)
1595 char * s2 = defln->line;
1597 // Either we will match the strings to EOS on both, or we will
1598 // match EOS on string 1 to whitespace on string 2. Otherwise, we
1600 while ((*s1 == *s2) || ((*s1 == EOS) && (chrtab[*s2] & WHITE)))
1602 // If we reached the end of string 1 (sym), we're done.
1603 // Note that we're also checking for the end of string 2 as
1604 // well, since we've established they're equal above.
1607 // Found the label, set new macro next-line and return.
1608 imacro->im_nextln = defln;
1618 return error("goto label not found");
1622 void DumpTokenBuffer(void)
1625 printf("Tokens [%X]: ", sloc);
1627 for(t=tokbuf; *t!=EOL; t++)
1631 else if (*t == CONST)
1634 printf("[CONST: $%X]", (uint32_t)*t);
1636 else if (*t == ACONST)
1638 else if (*t == STRING)
1641 printf("[STRING:\"%s\"]", string[*t]);
1643 else if (*t == SYMBOL)
1646 printf("[SYMBOL:\"%s\"]", string[*t]);
1650 else if (*t == TKEOF)
1652 else if (*t == DEQUALS)
1653 printf("[DEQUALS]");
1658 else if (*t == DCOLON)
1670 else if (*t == UNMINUS)
1671 printf("[UNMINUS]");
1672 else if (*t == DOTB)
1674 else if (*t == DOTW)
1676 else if (*t == DOTL)
1678 else if (*t == DOTI)
1680 else if (*t == ENDEXPR)
1681 printf("[ENDEXPR]");
1682 else if (*t == CR_ABSCOUNT)
1683 printf("[CR_ABSCOUNT]");
1684 else if (*t == CR_DEFINED)
1685 printf("[CR_DEFINED]");
1686 else if (*t == CR_REFERENCED)
1687 printf("[CR_REFERENCED]");
1688 else if (*t == CR_STREQ)
1689 printf("[CR_STREQ]");
1690 else if (*t == CR_MACDEF)
1691 printf("[CR_MACDEF]");
1692 else if (*t == CR_TIME)
1693 printf("[CR_TIME]");
1694 else if (*t == CR_DATE)
1695 printf("[CR_DATE]");
1696 else if (*t >= 0x20 && *t <= 0x2F)
1697 printf("[%c]", (char)*t);
1698 else if (*t >= 0x3A && *t <= 0x3F)
1699 printf("[%c]", (char)*t);
1700 else if (*t >= 0x80 && *t <= 0x87)
1701 printf("[D%u]", ((uint32_t)*t) - 0x80);
1702 else if (*t >= 0x88 && *t <= 0x8F)
1703 printf("[A%u]", ((uint32_t)*t) - 0x88);
1705 printf("[%X:%c]", (uint32_t)*t, (char)*t);