// TOKEN.C - Token Handling
// Copyright (C) 199x Landon Dyer, 2011-2012 Reboot and Friends
// RMAC derived from MADMAC v1.07 Written by Landon Dyer, 1986
-// Source Utilised with the Kind Permission of Landon Dyer
+// Source utilised with the kind permission of Landon Dyer
//
#include "token.h"
#define DEF_KW // Declare keyword values
#include "kwtab.h" // Incl generated keyword tables & defs
+
int lnsave; // 1; strcpy() text of current line
int curlineno; // Current line number
int totlines; // Total # of lines
};
-// Removing this, provided it doesn't cause unwanted side-effects :-P
-#if 0
//
-// Make `fnum' the Current `curfname'
-// NOTE: This is currently only called from error() in error.c
+// Initialize tokenizer
//
-void setfnum(WORD fnum)
+void InitTokenizer(void)
{
-#if 0
- // NOTE: fnum is ZERO based, this can cause problems if you're not careful!
- FILEREC * fr = filerec;
+ int i; // Iterator
+ char * htab = "0123456789abcdefABCDEF"; // Hex character table
- DEBUG printf("[setfnum: fnum=%u]\n", fnum);
+ lnsave = 0; // Don't save lines
+ curfname = ""; // No file, empty filename
+ filecount = (WORD)-1;
+ cfileno = (WORD)-1; // cfileno gets bumped to 0
+ curlineno = 0;
+ totlines = 0;
+ etok = tokbuf;
+ f_inobj = NULL;
+ f_ifile = NULL;
+ f_imacro = NULL;
+ cur_inobj = NULL;
+ filerec = NULL;
+ last_fr = NULL;
+ lntag = SPACE;
- // Advance to the correct record...
- while (fr != NULL && fnum != 0)
+ // Initialize hex, "dot" and tolower tables
+ for(i=0; i<128; i++)
{
- fr = fr->frec_next;
- fnum--;
+ hextab[i] = -1;
+ dotxtab[i] = 0;
+ tolowertab[i] = (char)i;
}
- if (fr == NULL)
- curfname = "(*top*)";
- else
- curfname = fr->frec_name;
-
- DEBUG printf("[setfnum: curfname=%s]\n", curfname);
-#else
- // Check for absolute top filename (this should never happen)
- if (fnum == -1)
- {
- curfname = "(*top*)";
- return;
- }
+ for(i=0; htab[i]!=EOS; i++)
+ hextab[htab[i]] = (char)((i < 16) ? i : i - 6);
- FILEREC * fr = filerec;
+ for(i='A'; i<='Z'; i++)
+ tolowertab[i] |= 0x20;
- // Advance to the correct record...
- while (fr != NULL && fnum != 0)
- {
- fr = fr->frec_next;
- fnum--;
- }
+ // These characters are legal immediately after a period
+ dotxtab['b'] = DOTB; // .b .B .s .S
+ dotxtab['B'] = DOTB;
+ dotxtab['s'] = DOTB;
+ dotxtab['S'] = DOTB;
+ dotxtab['w'] = DOTW; // .w .W
+ dotxtab['W'] = DOTW;
+ dotxtab['l'] = DOTL; // .l .L
+ dotxtab['L'] = DOTL;
+ dotxtab['i'] = DOTI; // .i .I (???)
+ dotxtab['I'] = DOTI;
+}
- // Check for file # record not found (this should never happen either)
- if (fr == NULL)
- {
- curfname = "(*NOT FOUND*)";
- return;
- }
- curfname = fr->frec_name;
-#endif
-}
-#else
void SetFilenameForErrorReporting(void)
{
WORD fnum = cfileno;
curfname = fr->frec_name;
}
-#endif
//
continue;
}
- if (tk != NULL) // arg # is in range, so expand it
+ // Argument # is in range, so expand it
+ if (tk != NULL)
{
while (*tk != EOL)
{
//
-// Get Next Line of Text from a Macro
+// Get next line of text from a macro
//
-char * getmln(void)
+char * GetNextMacroLine(void)
{
- unsigned source_addr;
+// unsigned source_addr;
IMACRO * imacro = cur_inobj->inobj.imacro;
// LONG * strp = imacro->im_nextln;
// ExpandMacro((char *)(strp + 1), imacro->im_lnbuf, LNSIZ);
ExpandMacro(strp->line, imacro->im_lnbuf, LNSIZ);
- if (!strcmp(imacro->im_macro->sname, "mjump") && !mjump_align)
- {
- // if we need to adjust the alignment of the jump source address to
- // meet the rules of gpu main execution we need to skip the first nop
- // of the macro. This is simpler than trying to insert nop's mid macro.
- source_addr = (orgactive ? orgaddr : sloc);
- source_addr += 8;
-
- if (source_addr % 4)
- {
- strp = imacro->im_nextln;
-
- if (strp == NULL)
- return NULL;
-
-// imacro->im_nextln = (LONG *)*strp;
-// ExpandMacro((char *)(strp + 1), imacro->im_lnbuf, LNSIZ);
- imacro->im_nextln = strp->next;
- ExpandMacro(strp->line, imacro->im_lnbuf, LNSIZ);
- }
-
- mjump_align = 1;
- }
-
return imacro->im_lnbuf;
}
//
-// Get Next Line of Text from a Repeat Block
+// Get next line of text from a repeat block
//
-char * getrln(void)
+char * GetNextRepeatLine(void)
{
IREPT * irept = cur_inobj->inobj.irept;
return NULL;
}
- strp = irept->ir_nextln; //strp
+ strp = irept->ir_nextln;
}
strcpy(irbuf, (char *)(irept->ir_nextln + 1));
//
-// Include a Source File used at the Root, and for ".include" Files
+// Include a source file used at the root, and for ".include" files
//
int include(int handle, char * fname)
{
//
-// Initialize Tokenizer
-//
-void init_token(void)
-{
- int i; // Iterator
- char * htab = "0123456789abcdefABCDEF"; // Hex character table
-
- lnsave = 0; // Don't save lines
- curfname = ""; // No file, empty filename
- filecount = (WORD)-1;
- cfileno = (WORD)-1; // cfileno gets bumped to 0
- curlineno = 0;
- totlines = 0;
- etok = tokbuf;
- f_inobj = NULL;
- f_ifile = NULL;
- f_imacro = NULL;
- cur_inobj = NULL;
- filerec = NULL;
- last_fr = NULL;
- lntag = SPACE;
-
- // Initialize hex, "dot" and tolower tables
- for(i=0; i<128; i++)
- {
- hextab[i] = -1;
- dotxtab[i] = 0;
- tolowertab[i] = (char)i;
- }
-
- for(i=0; htab[i]!=EOS; i++)
- hextab[htab[i]] = (char)((i < 16) ? i : i - 6);
-
- for(i='A'; i<='Z'; i++)
- tolowertab[i] |= 0x20;
-
- // These characters are legal immediately after a period
- dotxtab['b'] = DOTB; // .b .B .s .S
- dotxtab['B'] = DOTB;
- dotxtab['s'] = DOTB;
- dotxtab['S'] = DOTB;
- dotxtab['w'] = DOTW; // .w .W
- dotxtab['W'] = DOTW;
- dotxtab['l'] = DOTL; // .l .L
- dotxtab['L'] = DOTL;
- dotxtab['I'] = DOTI; // .l .L
- dotxtab['I'] = DOTI;
-}
-
-
-//
-// Pop the Current Input Level
+// Pop the current input level
//
int fpop(void)
{
// Get line from file into buf, return NULL on EOF or ptr to the start of a
// null-term line
//
-char * getln(void)
+char * GetNextLine(void)
{
int i, j;
char * p, * d;
// Scan for next end-of-line; handle stupid text formats by treating
// \r\n the same as \n. (lone '\r' at end of buffer means we have to
// check for '\n').
- i = 0;
- j = fl->ifcnt;
d = &fl->ifbuf[fl->ifind];
- for(p=d; i<j; i++, p++)
+ for(p=d, i=0, j=fl->ifcnt; i<j; i++, p++)
{
if (*p == '\r' || *p == '\n')
{
if (*p == '\r')
{
if (i >= j)
- {
- break; // Look for '\n' to eat
- }
+ break; // Need to read more, then look for '\n' to eat
else if (p[1] == '\n')
- {
i++;
- }
}
+ // Cover up the newline with end-of-string sentinel
*p = '\0';
fl->ifind += i;
// Handle hanging lines by ignoring them (Input file is exhausted, no
// \r or \n on last line)
+ // Shamus: This is retarded. Never ignore any input!
if (!readamt && fl->ifcnt)
{
+#if 0
fl->ifcnt = 0;
*p = '\0';
return NULL;
+#else
+ // Really should check to see if we're at the end of the buffer!
+ // :-P
+ fl->ifbuf[fl->ifind + fl->ifcnt] = '\0';
+ fl->ifcnt = 0;
+ return &fl->ifbuf[fl->ifind];
+#endif
}
// Truncate and return absurdly long lines.
fl->ifind = fl->ifcnt & 1;
}
- if ((readamt = read(fl->ifhandle, &fl->ifbuf[fl->ifind + fl->ifcnt], QUANTUM)) < 0)
+ readamt = read(fl->ifhandle, &fl->ifbuf[fl->ifind + fl->ifcnt], QUANTUM);
+
+ if (readamt < 0)
return NULL;
if ((fl->ifcnt += readamt) == 0)
//
-// Tokenize a Line
+// Tokenize a line
//
-int tokln(void)
+int TokenizeLine(void)
{
- char * ln = NULL; // Ptr to current position in line
- char * p; // Random character ptr
- TOKEN * tk; // Token-deposit ptr
- int state = 0; // State for keyword detector
- int j = 0; // Var for keyword detector
- char c; // Random char
- VALUE v; // Random value
- char * nullspot = NULL; // Spot to clobber for SYMBOL terminatn
- int stuffnull; // 1:terminate SYMBOL '\0' at *nullspot
+ char * ln = NULL; // Ptr to current position in line
+ char * p; // Random character ptr
+ TOKEN * tk; // Token-deposit ptr
+ int state = 0; // State for keyword detector
+ int j = 0; // Var for keyword detector
+ char c; // Random char
+ VALUE v; // Random value
+ char * nullspot = NULL; // Spot to clobber for SYMBOL termination
+ int stuffnull; // 1:terminate SYMBOL '\0' at *nullspot
char c1;
- int stringNum = 0; // Pointer to string locations in tokenized line
+ int stringNum = 0; // Pointer to string locations in tokenized line
- retry:
+retry:
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 macro, or a repeat-block
+ // Get another line of input from the current input source: a file, a
+ // macro, or a repeat-block
switch (cur_inobj->in_type)
{
// Include-file:
// o tag the listing-line with a space;
// o kludge lines generated by Alcyon C.
case SRC_IFILE:
- if ((ln = getln()) == NULL)
+ if ((ln = GetNextLine()) == NULL)
{
-if (verb_flag) printf("tokln: Calling fpop() from SRC_IFILE...\n");
+if (verb_flag) printf("TokenizeLine: Calling fpop() from SRC_IFILE...\n");
fpop(); // Pop input level
goto retry; // Try for more lines
}
// o Handle end-of-macro;
// o tag the listing-line with an at (@) sign.
case SRC_IMACRO:
- if ((ln = getmln()) == NULL)
+ if ((ln = GetNextMacroLine()) == NULL)
{
ExitMacro(); // Exit macro (pop args, do fpop(), etc)
goto retry; // Try for more lines...
// o Handle end-of-repeat-block;
// o tag the listing-line with a pound (#) sign.
case SRC_IREPT:
- if ((ln = getrln()) == NULL)
+ if ((ln = GetNextRepeatLine()) == NULL)
{
-if (verb_flag) printf("tokln: Calling fpop() from SRC_IREPT...\n");
+if (verb_flag) printf("TokenizeLine: Calling fpop() from SRC_IREPT...\n");
fpop();
goto retry;
}
strcpy(lnbuf, ln);
// General house-keeping
- tok = tokeol; // Set "tok" to EOL in case of error
- tk = etok; // Reset token ptr
- stuffnull = 0; // Don't stuff nulls
- totlines++; // Bump total #lines assembled
+ tok = tokeol; // Set "tok" to EOL in case of error
+ tk = etok; // Reset token ptr
+ stuffnull = 0; // Don't stuff nulls
+ totlines++; // Bump total #lines assembled
// See if the entire line is a comment. This is a win if the programmer
// puts in lots of comments
if (c & STSYM)
{
- if (stuffnull) // Terminate old symbol from previous pass
+ if (stuffnull) // Terminate old symbol from previous pass
*nullspot = EOS;
- v = 0; // Assume no DOT attrib follows symbol
+ v = 0; // Assume no DOT attrib follows symbol
stuffnull = 1;
- p = nullspot = ln++; // Nullspot -> start of this symbol
+ p = nullspot = ln++; // Nullspot -> start of this symbol
// Find end of symbol (and compute its length)
for(j=1; (int)chrtab[*ln]&CTSYM; j++)
// symbol or keyword:
if (*ln == '.')
{
- *ln++ = EOS; // Terminate symbol
- stuffnull = 0; // And never try it again
+ *ln++ = EOS; // Terminate symbol
+ stuffnull = 0; // And never try it again
// Character following the `.' must have a DOT attribute, and
// the chararacter after THAT one must not have a start-symbol
j = -1;
}
- //make j = -1 if time, date etc with no preceeding ^^
- //defined, referenced, streq, macdef, date and time
+ // 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))
+ {
+ j = -1;
+ }
+
+ // Make j = -1 if time, date etc with no preceeding ^^
+ // defined, referenced, streq, macdef, date and time
switch ((TOKEN)j)
{
case 112: // defined
case 120: // time
case 121: // date
j = -1;
- break;
}
+ // 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.
+//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
{
switch (*ln++)
{
- case '!': // ! or !=
+ case '!': // ! or !=
if (*ln == '=')
{
*tk++ = NE;
*tk++ = '!';
continue;
- case '\'': // 'string'
- case '\"': // "string"
+ case '\'': // 'string'
+ case '\"': // "string"
c1 = ln[-1];
*tk++ = STRING;
//#warning
break;
default:
warn("bad backslash code in string");
- --ln;
+ ln--;
break;
}
}
*p++ = EOS;
continue;
- case '$': // $, hex constant
+ case '$': // $, hex constant
if ((int)chrtab[*ln] & HDIGIT)
{
v = 0;
+ // Parse the hex value
while ((int)hextab[*ln] >= 0)
v = (v << 4) + (int)hextab[*ln++];
+ // ggn: Okay, some comments here are in order I think....
+ // The original madmac sources didn't parse the size at
+ // this point (i.e. .b/.w/.l). It was probably done at
+ // another point, although it's unclear to me exactly
+ // where. So why change this? My understanding (at least
+ // from what SCPCD said on IRC) is that .w addressing
+ // formats produce wrong code on jaguar (or doesn't execute
+ // properly? something like that). So the code was changed
+ // to mask off the upper bits depending on length (note: I
+ // don't think .b is valid at all! I only know of .w/.l, so
+ // this should probably be wiped). Then the code that
+ // parses the constant and checks to see if it's between
+ // $ffff0000 and $8000 never got triggered, so yay job
+ // done! ...now say we want to assemble a st .prg. One of
+ // the most widely spread optimisations is move.X expr.w,Y
+ // (or vice versa, or both, anyway...) to access hardware
+ // registers (which are mapped to $fxxxxx). This botchy
+ // thing would create "hilarious" code while trying to
+ // access hardware registers. So I made a condition to see
+ // if st mode or jaguar is active and apply the both or
+ // not. One last note: this is hardcoded to get optimised
+ // for now on ST mode, i.e. it can't generate code like
+ // move.w $00001234,d0 - it'll always get optimised to
+ // move.w $1234.w,d0. It's probably ok, but maybe a warning
+ // should be emitted? Or maybe finding a way to make it not
+ // auto-optimise? I think it's ok for now...
if (*ln == '.')
{
- if ((*(ln+1) == 'b') || (*(ln+1) == 'B'))
+ if (obj_format == ALCYON)
{
- v &= 0x000000FF;
- ln += 2;
- }
-
- if ((*(ln+1) == 'w') || (*(ln+1) == 'W'))
- {
- v &= 0x0000FFFF;
- ln += 2;
+ if ((*(ln + 1) == 'b') || (*(ln + 1) == 'B') || (*(ln + 1) == 'w') || (*(ln + 1) == 'W') || (*(ln + 1) == 'l') || (*(ln + 1) == 'L'))
+ {
+ ln += 2;
+ }
}
-
- if ((*(ln+1) == 'l') || (*(ln+1) == 'L'))
+ else
{
- ln += 2;
+ if ((*(ln + 1) & 0xDF) == 'B')
+ {
+ v &= 0x000000FF;
+ ln += 2;
+ }
+ else if ((*(ln + 1) & 0xDF) == 'W')
+ {
+ v &= 0x0000FFFF;
+ ln += 2;
+ }
+ else if ((*(ln + 1) & 0xDF) == 'L')
+ {
+ ln += 2;
+ }
}
}
*tk++ = '$';
continue;
- case '<': // < or << or <> or <=
+ case '<': // < or << or <> or <=
switch (*ln)
{
case '<':
*tk++ = '<';
continue;
}
- case ':': // : or ::
+ case ':': // : or ::
if (*ln == ':')
{
*tk++ = DCOLON;
*tk++ = ':';
continue;
- case '=': // = or ==
+ case '=': // = or ==
if (*ln == '=')
{
*tk++ = DEQUALS;
*tk++ = '=';
continue;
- case '>': // > or >> or >=
+ case '>': // > or >> or >=
switch (*ln)
{
case '>':
*tk++ = SHR;
- ++ln;
+ ln++;
continue;
case '=':
*tk++ = GE;
- ++ln;
+ ln++;
continue;
default:
*tk++ = '>';
continue;
}
- case '%': // % or binary constant
+ case '%': // % or binary constant
if (*ln < '0' || *ln > '1')
{
*tk++ = '%';
*tk++ = CONST;
*tk++ = v;
continue;
- case '@': // @ or octal constant
+ case '@': // @ or octal constant
if (*ln < '0' || *ln > '7')
{
*tk++ = '@';
*tk++ = CONST;
*tk++ = v;
continue;
- case '^': // ^ or ^^ <operator-name>
+ case '^': // ^ or ^^ <operator-name>
if (*ln != '^')
{
*tk++ = '^';
*tk++ = (TOKEN)j;
continue;
default:
- interror(2); // Bad MULTX entry in chrtab
+ interror(2); // Bad MULTX entry in chrtab
continue;
}
}
while ((int)chrtab[*ln] & DIGIT)
v = (v * 10) + *ln++ - '0';
- // See if there's a .[bwl] after the constant, & deal with it
+ // See if there's a .[bwl] after the constant & deal with it if so
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'))
+ else if ((*(ln + 1) == 'w') || (*(ln + 1) == 'W'))
{
v &= 0x0000FFFF;
ln += 2;
}
-
- if ((*(ln+1) == 'l') || (*(ln+1) == 'L'))
+ else if ((*(ln + 1) == 'l') || (*(ln + 1) == 'L'))
{
ln += 2;
}
*tk++ = CONST;
*tk++ = v;
+//printf("CONST: %i\n", v);
continue;
}
//int d_goto(void)
int d_goto(WORD unused)
{
-// char * sym; // Label to search for
-// LONG * defln; // Macro definition strings
- char * s1; // Temps for string comparison
- char * s2;
-// IMACRO * imacro; // Macro invocation block
+ char * s1, * s2;
// Setup for the search
if (*tok != SYMBOL)
{
// Compare names (sleazo string compare)
// This string compare is not right. Doesn't check for lengths.
-#warning "!!! Bad string comparison !!!"
+ // (actually it does, but in a crappy, unclear way.)
+WARNING(!!!! Bad string comparison !!!)
s1 = sym;
// s2 = (char *)(defln + 1) + 1;
s2 = defln->line;
return error("goto label not found");
}
+
void DumpTokenBuffer(void)
{
TOKEN * t;
else if (*t == ACONST)
printf("[ACONST]");
else if (*t == STRING)
-// printf("[STRING]");
{
t++;
printf("[STRING:\"%s\"]", string[*t]);
printf("[A%u]", ((uint32_t)*t) - 0x88);
else
printf("[%X:%c]", (uint32_t)*t, (char)*t);
-// printf("[%X]", (uint32_t)*t);
}
printf("[EOL]\n");
}
+