X-Git-Url: http://shamusworld.gotdns.org/cgi-bin/gitweb.cgi?p=rmac;a=blobdiff_plain;f=token.c;h=55dbc290980722d06ed144755e43bb85587965a5;hp=27eedc3d849b500d57305740b9d8d910cba6f476;hb=c38505ee4b2a0de59926107e52fb8bb84041a0e4;hpb=c9541a9d4f01e190ffefc20ce07896c7f04d9557 diff --git a/token.c b/token.c index 27eedc3..55dbc29 100644 --- a/token.c +++ b/token.c @@ -1,12 +1,14 @@ // // RMAC - Reboot's Macro Assembler for all Atari computers // TOKEN.C - Token Handling -// Copyright (C) 199x Landon Dyer, 2011-2017 Reboot and Friends +// Copyright (C) 199x Landon Dyer, 2011-2020 Reboot and Friends // RMAC derived from MADMAC v1.07 Written by Landon Dyer, 1986 // Source utilised with the kind permission of Landon Dyer // #include "token.h" + +#include #include "direct.h" #include "error.h" #include "macro.h" @@ -35,7 +37,8 @@ WORD cfileno; // Current file number TOKEN * tok; // Ptr to current token TOKEN * etok; // Ptr past last token in tokbuf[] TOKEN tokeol[1] = {EOL}; // Bailout end-of-line token -char * string[TOKBUFSIZE*2]; // Token buffer string pointer storage +char * string[TOKBUFSIZE*2];// Token buffer string pointer storage +int optimizeOff; // Optimization override flag // File record, used to maintain a list of every include file ever visited #define FILEREC struct _filerec @@ -48,12 +51,12 @@ FILEREC FILEREC * filerec; FILEREC * last_fr; -INOBJ * cur_inobj; // Ptr current input obj (IFILE/IMACRO) -static INOBJ * f_inobj; // Ptr list of free INOBJs -static IFILE * f_ifile; // Ptr list of free IFILEs -static IMACRO * f_imacro; // Ptr list of free IMACROs +INOBJ * cur_inobj; // Ptr current input obj (IFILE/IMACRO) +static INOBJ * f_inobj; // Ptr list of free INOBJs +static IFILE * f_ifile; // Ptr list of free IFILEs +static IMACRO * f_imacro; // Ptr list of free IMACROs -static TOKEN tokbuf[TOKBUFSIZE]; // Token buffer (stack-like, all files) +static TOKEN tokbuf[TOKBUFSIZE]; // Token buffer (stack-like, all files) uint8_t chrtab[0x100] = { ILLEG, ILLEG, ILLEG, ILLEG, // NUL SOH STX ETX @@ -205,13 +208,13 @@ void InitTokenizer(void) dotxtab['W'] = DOTW; dotxtab['l'] = DOTL; // .l .L dotxtab['L'] = DOTL; - dotxtab['i'] = DOTI; // .i .I (???) + dotxtab['i'] = DOTI; // .i .I (WTF is this???) dotxtab['I'] = DOTI; - dotxtab['D'] = DOTD; // .d .D (quad word) + dotxtab['D'] = DOTD; // .d .D (double) dotxtab['d'] = DOTD; dotxtab['S'] = DOTS; // .s .S dotxtab['s'] = DOTS; - dotxtab['Q'] = DOTQ; // .q .Q + dotxtab['Q'] = DOTQ; // .q .Q (quad word) dotxtab['q'] = DOTQ; dotxtab['X'] = DOTX; // .x .x dotxtab['x'] = DOTX; @@ -389,7 +392,7 @@ int ExpandMacro(char * src, char * dest, int destsiz) *dst++ = *s++; continue; case '?': // \? set `questmark' flag - ++s; + s++; questmark = 1; break; case '#': // \#, number of arguments @@ -475,24 +478,18 @@ copy_d: // macro invocation) then it is ignored. i = (int)arg->svalue; arg_num: - DEBUG { printf("~argnumber=%d (argBase=%u)\n", i, imacro->argBase); } + DEBUG { printf("~argnumber=%d\n", i); } tk = NULL; if (i < imacro->im_nargs) { -#if 0 -// tk = argp[i]; -// tk = argPtrs[i]; - tk = argPtrs[imacro->argBase + i]; -#else tk = imacro->argument[i].token; symbolString = imacro->argument[i].string; //DEBUG //{ // printf("ExM: Preparing to parse argument #%u...\n", i); -// dumptok(tk); +// DumpTokens(tk); //} -#endif } // \?arg yields: @@ -534,22 +531,12 @@ arg_num: switch ((int)*tk++) { case SYMBOL: -#if 0 -// d = (char *)*tk++; - d = string[*tk++]; -#else - // This fix should be done for strings too d = symbolString[*tk++]; DEBUG { printf("ExM: SYMBOL=\"%s\"", d); } -#endif break; case STRING: -#if 0 -// d = (char *)*tk++; - d = string[*tk++]; -#else d = symbolString[*tk++]; -#endif + if (dst >= edst) goto overflow; @@ -573,7 +560,9 @@ DEBUG { printf("ExM: SYMBOL=\"%s\"", d); } // to choke on legitimate code... Need to investigate this further // before changing anything else here! case CONST: - sprintf(numbuf, "$%lx", (long unsigned int)*tk++); +// sprintf(numbuf, "$%lx", (uint64_t)*tk++); + sprintf(numbuf, "$%" PRIX64, (uint64_t)*tk++); + tk++; d = numbuf; break; case DEQUALS: @@ -615,6 +604,9 @@ DEBUG { printf("ExM: SYMBOL=\"%s\"", d); } case CR_ABSCOUNT: d = "^^abscount"; break; + case CR_FILESIZE: + d = "^^filesize"; + break; case CR_DATE: d = "^^date"; break; @@ -718,6 +710,11 @@ char * GetNextRepeatLine(void) // strp = irept->ir_nextln; } + // Mark the current macro line in the irept object + // This is probably overkill - a global variable + // would suffice here (it only gets used during + // error reporting anyway) + irept->lineno = irept->ir_nextln->lineno; // strcpy(irbuf, (char *)(irept->ir_nextln + 1)); strcpy(irbuf, irept->ir_nextln->line); @@ -830,17 +827,11 @@ DEBUG { printf("[fpop: (post) cfileno=%d ifile->ifno=%d]\n", (int)cfileno, (int) case SRC_IREPT: // Pop and release an IREPT { DEBUG { printf("dealloc IREPT\n"); } -// LONG * p = inobj->inobj.irept->ir_firstln; LLIST * p = inobj->inobj.irept->ir_firstln; // Deallocate repeat lines while (p != NULL) { -// Shamus: ggn confirmed that this will cause a segfault on 64-bit versions of -// RMAC. This is just stupid and wrong anyway, so we need to fix crapola -// like this... -// LONG * p1 = (LONG *)*p; -// p = p1; free(p->line); p = p->next; } @@ -956,11 +947,12 @@ int TokenizeLine(void) { uint8_t * ln = NULL; // Ptr to current position in line uint8_t * p; // Random character ptr - TOKEN * tk; // Token-deposit ptr + PTR tk; // Token-deposit ptr int state = 0; // State for keyword detector int j = 0; // Var for keyword detector uint8_t c; // Random char - VALUE v; // Random value + uint64_t v; // Random value + uint32_t cursize = 0; // Current line's size (.b, .w, .l, .s, .q, .d) uint8_t * nullspot = NULL; // Spot to clobber for SYMBOL termination int stuffnull; // 1:terminate SYMBOL '\0' at *nullspot uint8_t c1; @@ -968,7 +960,7 @@ int TokenizeLine(void) retry: - if (cur_inobj == NULL) // Return EOF if input stack is empty + if (cur_inobj == NULL) // Return EOF if input stack is empty return TKEOF; // Get another line of input from the current input source: a file, a @@ -984,8 +976,8 @@ retry: if ((ln = GetNextLine()) == NULL) { DEBUG { printf("TokenizeLine: Calling fpop() from SRC_IFILE...\n"); } - if (fpop() == 0) // Pop input level - goto retry; // Try for more lines + if (fpop() == 0) // Pop input level + goto retry; // Try for more lines else { ifent->if_prev = (IFENT *)-1; //Signal Assemble() that we have reached EOF with unbalanced if/endifs @@ -993,7 +985,7 @@ DEBUG { printf("TokenizeLine: Calling fpop() from SRC_IFILE...\n"); } } } - curlineno++; // Bump line number + curlineno++; // Bump line number lntag = SPACE; if (as68_flag) @@ -1039,7 +1031,7 @@ DEBUG { printf("TokenizeLine: Calling fpop() from SRC_IFILE...\n"); } case SRC_IREPT: if ((ln = GetNextRepeatLine()) == NULL) { -DEBUG { printf("TokenizeLine: Calling fpop() from SRC_IREPT...\n"); } + DEBUG { printf("TokenizeLine: Calling fpop() from SRC_IREPT...\n"); } fpop(); goto retry; } @@ -1056,7 +1048,7 @@ DEBUG { printf("TokenizeLine: Calling fpop() from SRC_IREPT...\n"); } // General housekeeping tok = tokeol; // Set "tok" to EOL in case of error - tk = etok; // Reset token ptr + tk.u32 = etok; // Reset token ptr stuffnull = 0; // Don't stuff nulls totlines++; // Bump total #lines assembled @@ -1065,6 +1057,16 @@ DEBUG { printf("TokenizeLine: Calling fpop() from SRC_IREPT...\n"); } if (*ln == '*' || *ln == ';' || ((*ln == '/') && (*(ln + 1) == '/'))) goto goteol; + // And here we have a very ugly hack for signalling a single line 'turn off + // optimization'. There's really no nice way to do this, so hack it is! + optimizeOff = 0; // Default is to take optimizations as they come + + if (*ln == '!') + { + optimizeOff = 1; // Signal that we don't want to optimize this line + ln++; // & skip over the darned thing + } + // Main tokenization loop; // o skip whitespace; // o handle end-of-line; @@ -1109,7 +1111,7 @@ DEBUG { printf("TokenizeLine: Calling fpop() from SRC_IREPT...\n"); } // token stream: ln++; stuffnull = 0; - *tk++ = (TOKEN)dotxtab[*ln++]; + *tk.u32++ = (TOKEN)dotxtab[*ln++]; continue; } } @@ -1127,14 +1129,15 @@ DEBUG { printf("TokenizeLine: Calling fpop() from SRC_IREPT...\n"); } *ln++ = EOS; // Terminate symbol stuffnull = 0; // And never try it again - // Character following the `.' must have a DOT attribute, and + // Character following the '.' must have a DOT attribute, and // the chararacter after THAT one must not have a start-symbol // attribute (to prevent symbols that look like, for example, // "zingo.barf", which might be a good idea anyway....) if (((chrtab[*ln] & DOT) == 0) || (dotxtab[*ln] == 0)) return error("[bwsl] must follow '.' in symbol"); - v = (VALUE)dotxtab[*ln++]; + v = (uint32_t)dotxtab[*ln++]; + cursize = (uint32_t)v; if (chrtab[*ln] & CTSYM) return error("misuse of '.'; not allowed in symbols"); @@ -1170,7 +1173,7 @@ DEBUG { printf("TokenizeLine: Calling fpop() from SRC_IREPT...\n"); } } // Make j = -1 if user tries to use a RISC register while in 68K mode - if (!(rgpu || rdsp) && ((TOKEN)j >= KW_R0 && (TOKEN)j <= KW_R31)) + if (!(rgpu || rdsp || dsp56001) && ((TOKEN)j >= KW_R0 && (TOKEN)j <= KW_R31)) { j = -1; } @@ -1191,28 +1194,21 @@ DEBUG { printf("TokenizeLine: Calling fpop() from SRC_IREPT...\n"); } // If not tokenized keyword OR token was not found if ((j < 0) || (state < 0)) { - *tk++ = SYMBOL; -//#warning -//problem here: nullspot is a char * but TOKEN is a uint32_t. On a 64-bit -//system, this will cause all kinds of mischief. -#if 0 - *tk++ = (TOKEN)nullspot; -#else + *tk.u32++ = SYMBOL; string[stringNum] = nullspot; - *tk++ = stringNum; + *tk.u32++ = stringNum; stringNum++; -#endif } else { - *tk++ = (TOKEN)j; + *tk.u32++ = (TOKEN)j; stuffnull = 0; } - if (v) // Record attribute token (if any) - *tk++ = (TOKEN)v; + if (v) // Record attribute token (if any) + *tk.u32++ = (TOKEN)v; - if (stuffnull) // Arrange for string termination on next pass + if (stuffnull) // Arrange for string termination on next pass nullspot = ln; continue; @@ -1221,7 +1217,7 @@ DEBUG { printf("TokenizeLine: Calling fpop() from SRC_IREPT...\n"); } // Handle identity tokens if (c & SELF) { - *tk++ = *ln++; + *tk.u32++ = *ln++; continue; } @@ -1233,27 +1229,27 @@ DEBUG { printf("TokenizeLine: Calling fpop() from SRC_IREPT...\n"); } case '!': // ! or != if (*ln == '=') { - *tk++ = NE; - ++ln; + *tk.u32++ = NE; + ln++; } else - *tk++ = '!'; + *tk.u32++ = '!'; continue; case '\'': // 'string' if (m6502) { // Hardcoded for now, maybe this will change in the future - *tk++ = STRINGA8; + *tk.u32++ = STRINGA8; goto dostring; } // Fall through case '\"': // "string" - *tk++ = STRING; + *tk.u32++ = STRING; dostring: c1 = ln[-1]; string[stringNum] = ln; - *tk++ = stringNum; + *tk.u32++ = stringNum; stringNum++; for(p=ln; *ln!=EOS && *ln!=c1;) @@ -1293,6 +1289,15 @@ dostring: case '\\': c = '\\'; break; + case '{': + // If we're evaluating a macro + // this is valid because it's + // a parameter expansion + case '!': + // If we're evaluating a macro + // this is valid and expands to + // "dot-size" + break; default: warn("bad backslash code in string"); ln--; @@ -1317,109 +1322,85 @@ dostring: while (hextab[*ln] >= 0) v = (v << 4) + (int)hextab[*ln++]; + *tk.u32++ = CONST; + *tk.u64++ = v; + if (*ln == '.') { - if (obj_format == BSD) + if ((*(ln + 1) == 'w') || (*(ln + 1) == 'W')) { - if ((*(ln + 1) & 0xDF) == 'B') - { - v &= 0x000000FF; - ln += 2; - } - else if ((*(ln + 1) & 0xDF) == 'W') - { - v &= 0x0000FFFF; - ln += 2; - } - else if ((*(ln + 1) & 0xDF) == 'L') - { - ln += 2; - } + *tk.u32++ = DOTW; + ln += 2; } - } - - *tk++ = CONST; - *tk++ = v; - - if (obj_format == ALCYON) - { - if (*ln == '.') + else if ((*(ln + 1) == 'l') || (*(ln + 1) == 'L')) { - if ((*(ln + 1) == 'w') || (*(ln + 1) == 'W')) - { - *tk++ = DOTW; - ln += 2; - } - else if ((*(ln + 1) == 'l') || (*(ln + 1) == 'L')) - { - *tk++ = DOTL; - ln += 2; - } + *tk.u32++ = DOTL; + ln += 2; } } } else - *tk++ = '$'; + *tk.u32++ = '$'; continue; case '<': // < or << or <> or <= switch (*ln) { case '<': - *tk++ = SHL; - ++ln; + *tk.u32++ = SHL; + ln++; continue; case '>': - *tk++ = NE; - ++ln; + *tk.u32++ = NE; + ln++; continue; case '=': - *tk++ = LE; - ++ln; + *tk.u32++ = LE; + ln++; continue; default: - *tk++ = '<'; + *tk.u32++ = '<'; continue; } case ':': // : or :: if (*ln == ':') { - *tk++ = DCOLON; - ++ln; + *tk.u32++ = DCOLON; + ln++; } else - *tk++ = ':'; + *tk.u32++ = ':'; continue; case '=': // = or == if (*ln == '=') { - *tk++ = DEQUALS; - ++ln; + *tk.u32++ = DEQUALS; + ln++; } else - *tk++ = '='; + *tk.u32++ = '='; continue; case '>': // > or >> or >= switch (*ln) { case '>': - *tk++ = SHR; + *tk.u32++ = SHR; ln++; continue; case '=': - *tk++ = GE; + *tk.u32++ = GE; ln++; continue; default: - *tk++ = '>'; + *tk.u32++ = '>'; continue; } case '%': // % or binary constant if (*ln < '0' || *ln > '1') { - *tk++ = '%'; + *tk.u32++ = '%'; continue; } @@ -1444,17 +1425,18 @@ dostring: if ((*(ln + 1) == 'l') || (*(ln + 1) == 'L')) { + v &= 0xFFFFFFFF; ln += 2; } } - *tk++ = CONST; - *tk++ = v; + *tk.u32++ = CONST; + *tk.u64++ = v; continue; case '@': // @ or octal constant if (*ln < '0' || *ln > '7') { - *tk++ = '@'; + *tk.u32++ = '@'; continue; } @@ -1465,31 +1447,32 @@ dostring: if (*ln == '.') { - if ((*(ln+1) == 'b') || (*(ln+1) == 'B')) + if ((*(ln + 1) == 'b') || (*(ln + 1) == 'B')) { v &= 0x000000FF; ln += 2; } - if ((*(ln+1) == 'w') || (*(ln+1) == 'W')) + if ((*(ln + 1) == 'w') || (*(ln + 1) == 'W')) { v &= 0x0000FFFF; ln += 2; } - if ((*(ln+1) == 'l') || (*(ln+1) == 'L')) + if ((*(ln + 1) == 'l') || (*(ln + 1) == 'L')) { + v &= 0xFFFFFFFF; ln += 2; } } - *tk++ = CONST; - *tk++ = v; + *tk.u32++ = CONST; + *tk.u64++ = v; continue; case '^': // ^ or ^^ if (*ln != '^') { - *tk++ = '^'; + *tk.u32++ = '^'; continue; } @@ -1535,7 +1518,7 @@ dostring: continue; } - *tk++ = (TOKEN)j; + *tk.u32++ = (TOKEN)j; continue; default: interror(2); // Bad MULTX entry in chrtab @@ -1546,6 +1529,7 @@ dostring: // Handle decimal constant if (c & DIGIT) { + uint8_t * numStart = ln; v = 0; while ((int)chrtab[*ln] & DIGIT) @@ -1558,20 +1542,55 @@ dostring: { v &= 0x000000FF; ln += 2; + *tk.u32++ = CONST; + *tk.u64++ = v; + *tk.u32++ = DOTB; } else if ((*(ln + 1) == 'w') || (*(ln + 1) == 'W')) { v &= 0x0000FFFF; ln += 2; + *tk.u32++ = CONST; + *tk.u64++ = v; + *tk.u32++ = DOTW; } else if ((*(ln + 1) == 'l') || (*(ln + 1) == 'L')) { + v &= 0xFFFFFFFF; ln += 2; + *tk.u32++ = CONST; + *tk.u64++ = v; + *tk.u32++ = DOTL; } + else if ((int)chrtab[*(ln + 1)] & DIGIT) + { + // Hey, more digits after the dot, so we assume it's a + // floating point number of some kind... numEnd will point + // to the first non-float character after it's done + char * numEnd; + errno = 0; + double f = strtod(numStart, &numEnd); + ln = (uint8_t *)numEnd; + + if (errno != 0) + return error("floating point parse error"); + + // N.B.: We use the C compiler's internal double + // representation for all internal float calcs and + // are reasonably sure that the size of said double + // is 8 bytes long (which we check for in fltpoint.c) + *tk.u32++ = FCONST; + *tk.dp = f; + tk.u64++; + continue; + } + } + else + { + *tk.u32++ = CONST; + *tk.u64++ = v; } - *tk++ = CONST; - *tk++ = v; //printf("CONST: %i\n", v); continue; } @@ -1583,12 +1602,12 @@ dostring: // Terminate line of tokens and return "success." goteol: - tok = etok; // Set tok to beginning of line + tok = etok; // Set tok to beginning of line - if (stuffnull) // Terminate last SYMBOL + if (stuffnull) // Terminate last SYMBOL *nullspot = EOS; - *tk++ = EOL; + *tk.u32++ = EOL; return OK; } @@ -1659,22 +1678,116 @@ int d_goto(WORD unused) } +void DumpToken(TOKEN t) +{ + if (t == COLON) + printf("[COLON]"); + else if (t == CONST) + printf("[CONST]"); + else if (t == FCONST) + printf("[FCONST]"); + else if (t == ACONST) + printf("[ACONST]"); + else if (t == STRING) + printf("[STRING]"); + else if (t == SYMBOL) + printf("[SYMBOL]"); + else if (t == EOS) + printf("[EOS]"); + else if (t == TKEOF) + printf("[TKEOF]"); + else if (t == DEQUALS) + printf("[DEQUALS]"); + else if (t == SET) + printf("[SET]"); + else if (t == REG) + printf("[REG]"); + else if (t == DCOLON) + printf("[DCOLON]"); + else if (t == GE) + printf("[GE]"); + else if (t == LE) + printf("[LE]"); + else if (t == NE) + printf("[NE]"); + else if (t == SHR) + printf("[SHR]"); + else if (t == SHL) + printf("[SHL]"); + else if (t == UNMINUS) + printf("[UNMINUS]"); + else if (t == DOTB) + printf("[DOTB]"); + else if (t == DOTW) + printf("[DOTW]"); + else if (t == DOTL) + printf("[DOTL]"); + else if (t == DOTQ) + printf("[DOTQ]"); + else if (t == DOTS) + printf("[DOTS]"); + else if (t == DOTD) + printf("[DOTD]"); + else if (t == DOTI) + printf("[DOTI]"); + else if (t == ENDEXPR) + printf("[ENDEXPR]"); + else if (t == CR_ABSCOUNT) + printf("[CR_ABSCOUNT]"); + else if (t == CR_FILESIZE) + printf("[CR_FILESIZE]"); + else if (t == CR_DEFINED) + printf("[CR_DEFINED]"); + else if (t == CR_REFERENCED) + printf("[CR_REFERENCED]"); + else if (t == CR_STREQ) + printf("[CR_STREQ]"); + else if (t == CR_MACDEF) + printf("[CR_MACDEF]"); + else if (t == CR_TIME) + printf("[CR_TIME]"); + else if (t == CR_DATE) + printf("[CR_DATE]"); + else if (t >= 0x20 && t <= 0x2F) + printf("[%c]", (char)t); + else if (t >= 0x3A && t <= 0x3F) + printf("[%c]", (char)t); + else if (t >= 0x80 && t <= 0x87) + printf("[D%u]", ((uint32_t)t) - 0x80); + else if (t >= 0x88 && t <= 0x8F) + printf("[A%u]", ((uint32_t)t) - 0x88); + else + printf("[%X:%c]", (uint32_t)t, (char)t); +} + + void DumpTokenBuffer(void) { - TOKEN * t; printf("Tokens [%X]: ", sloc); - for(t=tokbuf; *t!=EOL; t++) + for(TOKEN * t=tokbuf; *t!=EOL; t++) { if (*t == COLON) printf("[COLON]"); else if (*t == CONST) { - t++; - printf("[CONST: $%X]", (uint32_t)*t); + PTR tp; + tp.u32 = t + 1; + printf("[CONST: $%lX]", *tp.u64); + t += 2; + } + else if (*t == FCONST) + { + PTR tp; + tp.u32 = t + 1; + printf("[FCONST: $%lX]", *tp.u64); + t += 2; } else if (*t == ACONST) - printf("[ACONST]"); + { + printf("[ACONST: $%X, $%X]", (uint32_t)t[1], (uint32_t)t[2]); + t += 2; + } else if (*t == STRING) { t++; @@ -1715,12 +1828,20 @@ void DumpTokenBuffer(void) printf("[DOTW]"); else if (*t == DOTL) printf("[DOTL]"); + else if (*t == DOTQ) + printf("[DOTQ]"); + else if (*t == DOTS) + printf("[DOTS]"); + else if (*t == DOTD) + printf("[DOTD]"); else if (*t == DOTI) printf("[DOTI]"); else if (*t == ENDEXPR) printf("[ENDEXPR]"); else if (*t == CR_ABSCOUNT) printf("[CR_ABSCOUNT]"); + else if (*t == CR_FILESIZE) + printf("[CR_FILESIZE]"); else if (*t == CR_DEFINED) printf("[CR_DEFINED]"); else if (*t == CR_REFERENCED)