]> Shamusworld >> Repos - rmac/blobdiff - macro.c
EOY minor update.
[rmac] / macro.c
diff --git a/macro.c b/macro.c
index 2ebaf52e0455773fd5305009ca029f50e0332399..22bd0d9f7c444d4824228cab8307e82c648c9f85 100644 (file)
--- a/macro.c
+++ b/macro.c
@@ -1,7 +1,7 @@
 //
 // RMAC - Reboot's Macro Assembler for all Atari computers
 // MACRO.C - Macro Definition and Invocation
-// 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
 //
 
 LONG curuniq;                          // Current macro's unique number
 int macnum;                                    // Unique number for macro definition
-TOKEN * argPtrs[128];          // 128 arguments ought to be enough for anyone
-static int argp;
 
 static LONG macuniq;           // Unique-per-macro number
 static SYM * curmac;           // Macro currently being defined
-static VALUE argno;                    // Formal argument count
+static uint32_t argno;                 // Formal argument count
 
-static LONG * firstrpt;                // First .rept line
-static LONG * nextrpt;         // Last .rept line
+static LLIST * firstrpt;       // First .rept line
+static LLIST * nextrpt;                // Last .rept line
 static int rptlevel;           // .rept nesting level
 
+// Function prototypes
+static int KWMatch(char *, char *);
+static int LNCatch(int (*)(), char *);
+
 
 //
 // Initialize macro processor
@@ -38,7 +40,6 @@ void InitMacro(void)
 {
        macuniq = 0;
        macnum = 1;
-       argp = 0;
 }
 
 
@@ -51,11 +52,13 @@ void InitMacro(void)
 int ExitMacro(void)
 {
 WARNING(!!! Bad macro exiting !!!)
-
 /*
 This is a problem. Currently, the argument logic just keeps the current
 arguments and doesn't save anything if a new macro is called in the middle
 of another (nested macros). Need to fix that somehow.
+
+Is this still true, now that we have IMACROs with TOKENSTREAMs in them? Need to
+check it out for sure...!
 */
        // Pop intervening include files and .rept blocks
        while (cur_inobj != NULL && cur_inobj->in_type != SRC_IMACRO)
@@ -72,11 +75,7 @@ of another (nested macros). Need to fix that somehow.
        IMACRO * imacro = cur_inobj->inobj.imacro;
        curuniq = imacro->im_olduniq;
 
-//     /*TOKEN ** p = */argp--;
-//     argp = (TOKEN **)*argp;
-       DEBUG printf("ExitMacro: argp: %d -> ", argp);
-       argp -= imacro->im_nargs;
-       DEBUG printf("%d (nargs = %d)\n", argp, imacro->im_nargs);
+       DEBUG { printf("ExitMacro: nargs = %d\n", imacro->im_nargs); }
 
        return fpop();
 }
@@ -87,12 +86,10 @@ of another (nested macros). Need to fix that somehow.
 //
 int defmac2(char * argname)
 {
-       SYM * arg;
-
        if (lookup(argname, MACARG, (int)curmac->sattr) != NULL)
                return error("multiple formal argument definition");
 
-       arg = NewSymbol(argname, MACARG, (int)curmac->sattr);
+       SYM * arg = NewSymbol(argname, MACARG, (int)curmac->sattr);
        arg->svalue = argno++;
 
        return OK;
@@ -107,69 +104,35 @@ int defmac2(char * argname)
 //
 int defmac1(char * ln, int notEndFlag)
 {
-//     PTR p;
-//     LONG len;
-
        if (list_flag)
        {
-               listeol();                                                      // Flush previous source line
-               lstout('.');                                            // Mark macro definition with period
+               listeol();              // Flush previous source line
+               lstout('.');    // Mark macro definition with period
        }
 
-       // This is just wrong, wrong, wrong. It makes up a weird kind of string with
-       // a pointer on front, and then uses a ** to manage them: This is a recipe
-       // for disaster.
-       // How to manage it then?
-       // Could use a linked list, like Landon uses everywhere else.
-/*
-How it works:
-Allocate a space big enough for the string + NULL + pointer.
-Set the pointer to NULL.
-Copy the string to the space after the pointer.
-If this is the 1st time through, set the SYM * "svalue" to the pointer.
-If this is the 2nd time through, derefence the ** to point to the memory you just allocated.
-Then, set the ** to the location of the memory you allocated for the next pass through.
-
-This is a really low level way to do a linked list, and by bypassing all the safety
-features of the language. Seems like we can do better here.
-*/
        if (notEndFlag)
        {
-#if 0
-               len = strlen(ln) + 1 + sizeof(LONG);
-               p.cp = malloc(len);
-               *p.lp = 0;
-               strcpy(p.cp + sizeof(LONG), ln);
-
-               // Link line of text onto end of list
-               if (curmln == NULL)
-                       curmac->svalue = p.cp;
-               else
-                       *curmln = p.cp;
-
-               curmln = (char **)p.cp;
-               return 1;                                                       // Keep looking
-#else
                if (curmac->lineList == NULL)
                {
-                       curmac->lineList = malloc(sizeof(struct LineList));
+                       curmac->lineList = malloc(sizeof(LLIST));
                        curmac->lineList->next = NULL;
                        curmac->lineList->line = strdup(ln);
+                       curmac->lineList->lineno = curlineno;
                        curmac->last = curmac->lineList;
                }
                else
                {
-                       curmac->last->next = malloc(sizeof(struct LineList));
+                       curmac->last->next = malloc(sizeof(LLIST));
                        curmac->last->next->next = NULL;
                        curmac->last->next->line = strdup(ln);
+                       curmac->lineList->lineno = curlineno;
                        curmac->last = curmac->last->next;
                }
 
-               return 1;                                                       // Keep looking
-#endif
+               return 1;               // Keep looking
        }
 
-       return 0;                                                               // Stop looking at the end
+       return 0;                       // Stop looking; at the end
 }
 
 
@@ -184,7 +147,8 @@ features of the language. Seems like we can do better here.
 //  Helper functions:
 //  -----------------
 //  `defmac1' adds lines of text to the macro definition
-//  `defmac2' processes the formal arguments (and sticks them into the symbol table)
+//  `defmac2' processes the formal arguments (and sticks them into the symbol
+//   table)
 //
 int DefineMacro(void)
 {
@@ -208,14 +172,14 @@ int DefineMacro(void)
        {
                argno = 0;
                symlist(defmac2);
-               at_eol();
+               ErrorIfNotAtEOL();
        }
 
        // Suck in the macro definition; we're looking for an ENDM symbol on a line
        // by itself to terminate the definition.
 //     curmln = NULL;
        curmac->lineList = NULL;
-       lncatch(defmac1, "endm ");
+       LNCatch(defmac1, "endm ");
 
        return 0;
 }
@@ -224,61 +188,67 @@ int DefineMacro(void)
 //
 // Add lines to a .rept definition
 //
-int defr1(char * ln, int kwno)
+int defr1(char * line, int kwno)
 {
-       LONG len;
-       LONG * p;
-
        if (list_flag)
        {
-               listeol();                                                              // Flush previous source line
-               lstout('#');                                                    // Mark this a 'rept' block
+               listeol();                      // Flush previous source line
+               lstout('#');            // Mark this a 'rept' block
        }
 
-       switch (kwno)
+       if (kwno == 0)                  // .endr
        {
-       case 0:                                                                         // .endr
                if (--rptlevel == 0)
-               return(0);
-               goto addln;
-       case 1:                                                                         // .rept
+                       return 0;
+       }
+       else if (kwno == 1)             // .rept
                rptlevel++;
-       default:
-//MORE stupidity here...
-WARNING(!!! Casting (char *) as LONG !!!)
-       addln:
-               // Allocate length of line + 1('\0') + LONG
-               len = strlen(ln) + 1 + sizeof(LONG);
-//             p = (LONG *)amem(len);
-               p = (LONG *)malloc(len);
-               *p = 0;
 
-               strcpy((char *)(p + 1), ln);
+//DEBUG { printf("  defr1: line=\"%s\", kwno=%d, rptlevel=%d\n", line, kwno, rptlevel); }
 
-               if (nextrpt == NULL)
-               {
-                       firstrpt = p;           // First line of rept statement
-               }
-               else
-               {
-                       *nextrpt = (LONG)p;
-               }
-
-               nextrpt = p;
-
-               return rptlevel;
+#if 0
+//MORE stupidity here...
+WARNING("!!! Casting (char *) as LONG !!!")
+       // Allocate length of line + 1('\0') + LONG
+       LONG * p = (LONG *)malloc(strlen(line) + 1 + sizeof(LONG));
+       *p = 0;
+       strcpy((char *)(p + 1), line);
+
+       if (nextrpt == NULL)
+               firstrpt = p;           // First line of rept statement
+       else
+               *nextrpt = (LONG)p;
+
+       nextrpt = p;
+#else
+       if (firstrpt == NULL)
+       {
+               firstrpt = malloc(sizeof(LLIST));
+               firstrpt->next = NULL;
+               firstrpt->line = strdup(line);
+               firstrpt->lineno = curlineno;
+               nextrpt = firstrpt;
        }
+       else
+       {
+               nextrpt->next = malloc(sizeof(LLIST));
+               nextrpt->next->next = NULL;
+               nextrpt->next->line = strdup(line);
+               nextrpt->next->lineno = curlineno;
+               nextrpt = nextrpt->next;
+       }
+#endif
+
+       return rptlevel;
 }
 
 
 //
-// Define a .rept block, this gets hairy because they can be nested
+// Handle a .rept block; this gets hairy because they can be nested
 //
-int defrept(void)
+int HandleRept(void)
 {
-       INOBJ * inobj;
-       IREPT * irept;
-       VALUE eval;
+       uint64_t eval;
 
        // Evaluate repeat expression
        if (abs_expr(&eval) != OK)
@@ -288,16 +258,17 @@ int defrept(void)
        firstrpt = NULL;
        nextrpt = NULL;
        rptlevel = 1;
-       lncatch(defr1, "endr rept ");
+       LNCatch(defr1, "endr rept ");
 
+//DEBUG { printf("HandleRept: firstrpt=$%X\n", firstrpt); }
        // Alloc and init input object
        if (firstrpt)
        {
-               inobj = a_inobj(SRC_IREPT);                             // Create a new REPT input object
-               irept = inobj->inobj.irept;
+               INOBJ * inobj = a_inobj(SRC_IREPT);     // Create a new REPT input object
+               IREPT * irept = inobj->inobj.irept;
                irept->ir_firstln = firstrpt;
                irept->ir_nextln = NULL;
-               irept->ir_count = eval;
+               irept->ir_count = (uint32_t)eval;
        }
 
        return 0;
@@ -305,30 +276,26 @@ int defrept(void)
 
 
 //
-// Hand off lines of text to the function `lnfunc' until a line containing one
-// of the directives in `dirlist' is encountered. Return the number of the
-// keyword encountered (0..n)
+// Hand off lines of text to the function 'lnfunc' until a line containing one
+// of the directives in 'dirlist' is encountered.
 //
-// `dirlist' contains null-seperated terminated keywords.  A final null
-// terminates the list. Directives are compared to the keywords without regard
-// to case.
+// 'dirlist' contains space-separated terminated keywords. A final space
+// terminates the list. Directives are case-insensitively compared to the
+// keywords.
 //
-// If `lnfunc' is NULL, then lines are simply skipped.
-// If `lnfunc' returns an error, processing is stopped.
+// If 'lnfunc' is NULL, then lines are simply skipped.
+// If 'lnfunc' returns an error, processing is stopped.
 //
-// `lnfunc' is called with an argument of -1 for every line but the last one,
+// 'lnfunc' is called with an argument of -1 for every line but the last one,
 // when it is called with an argument of the keyword number that caused the
 // match.
 //
-int lncatch(int (* lnfunc)(), char * dirlist)
+static int LNCatch(int (* lnfunc)(), char * dirlist)
 {
-       char * p;
-       int k;
-
        if (lnfunc != NULL)
-               lnsave++;                                                               // Tell tokenizer to keep lines
+               lnsave++;                       // Tell tokenizer to keep lines
 
-       for(;;)
+       while (1)
        {
                if (TokenizeLine() == TKEOF)
                {
@@ -336,31 +303,36 @@ int lncatch(int (* lnfunc)(), char * dirlist)
                        fatal("cannot continue");
                }
 
+               DEBUG { DumpTokenBuffer(); }
+
                // Test for end condition.  Two cases to handle:
                //            <directive>
                //    symbol: <directive>
-               p = NULL;
-               k = -1;
+               char * p = NULL;
+               int k = -1;
 
                if (*tok == SYMBOL)
                {
+                       // A string followed by a colon or double colon is a symbol and
+                       // *not* a directive, see if we can find the directive after it
                        if ((tok[2] == ':' || tok[2] == DCOLON))
                        {
-                               if (tok[3] == SYMBOL)                   // label: symbol
+                               if (tok[3] == SYMBOL)
                                        p = string[tok[4]];
                        }
                        else
                        {
-                               p = string[tok[1]];                             // Symbol
+                               // Otherwise, just grab the directive
+                               p = string[tok[1]];
                        }
                }
 
                if (p != NULL)
                {
-                       if (*p == '.')                                          // ignore leading '.'s
+                       if (*p == '.')          // Ignore leading periods
                                p++;
 
-                       k = kwmatch(p, dirlist);
+                       k = KWMatch(p, dirlist);
                }
 
                // Hand-off line to function
@@ -369,35 +341,30 @@ int lncatch(int (* lnfunc)(), char * dirlist)
                if (lnfunc != NULL)
                        k = (*lnfunc)(lnbuf, k);
 
-               if (!k)
+               if (k == 0)
                        break;
        }
 
        if (lnfunc != NULL)
-               lnsave--;                                                               // Tell tokenizer to stop keeping lines
+               lnsave--;                               // Tell tokenizer to stop keeping lines
 
        return 0;
 }
 
 
 //
-// See if the string `kw' matches one of the keywords in `kwlist'.  If so,
-// return the number of the keyword matched.  Return -1 if there was no match.
+// See if the string `kw' matches one of the keywords in `kwlist'. If so,
+// return the number of the keyword matched. Return -1 if there was no match.
 // Strings are compared without regard for case.
 //
-int kwmatch(char * kw, char * kwlist)
+static int KWMatch(char * kw, char * kwlist)
 {
-       char * p;
-       char c1;
-       char c2;
-       int k;
-
-       for(k=0; *kwlist; ++k)
+       for(int k=0; *kwlist; k++)
        {
-               for(p=kw;;)
+               for(char * p=kw;;)
                {
-                       c1 = *kwlist++;
-                       c2 = *p++;
+                       char c1 = *kwlist++;
+                       char c2 = *p++;
 
                        if (c2 >= 'A' && c2 <= 'Z')
                                c2 += 32;
@@ -410,7 +377,7 @@ int kwmatch(char * kw, char * kwlist)
                }
 
                // Skip to beginning of next keyword in `kwlist'
-               while (*kwlist && *kwlist != ' ')
+               while (*kwlist && (*kwlist != ' '))
                        ++kwlist;
 
                if (*kwlist== ' ')
@@ -422,180 +389,87 @@ int kwmatch(char * kw, char * kwlist)
 
 
 //
-// Invoke a macro
-// o  parse, count and copy arguments
-// o  push macro's string-stream
+// Invoke a macro by creating a new IMACRO object & chopping up the arguments
 //
 int InvokeMacro(SYM * mac, WORD siz)
 {
-       TOKEN * p = NULL;
-       int dry_run;
-       WORD arg_siz = 0;
-//     TOKEN ** argptr = NULL;
-//Doesn't need to be global! (or does it???--it does)
-//     argp = 0;
-       DEBUG printf("InvokeMacro: argp: %d -> ", argp);
-
-       INOBJ * inobj = a_inobj(SRC_IMACRO);            // Alloc and init IMACRO
+       DEBUG { printf("InvokeMacro: arguments="); DumpTokens(tok); }
+
+       INOBJ * inobj = a_inobj(SRC_IMACRO);    // Alloc and init IMACRO
        IMACRO * imacro = inobj->inobj.imacro;
-       imacro->im_siz = siz;
-       WORD nargs = 0;
-       TOKEN * beg_tok = tok;                                          // 'tok' comes from token.c
-       TOKEN * startOfArg;
-       TOKEN * dest;
-       int stringNum = 0;
-       int argumentNum = 0;
-
-       for(dry_run=1; ; dry_run--)
+       uint16_t nargs = 0;
+
+       // Chop up the arguments, if any (tok comes from token.c, which at this
+       // point points at the macro argument token stream)
+       if (*tok != EOL)
        {
-               for(tok=beg_tok; *tok!=EOL;)
+               // Parse out the arguments and set them up correctly
+               TOKEN * p = imacro->argument[nargs].token;
+               int stringNum = 0;
+
+               while (*tok != EOL)
                {
-                       if (dry_run)
-                               nargs++;
-                       else
+                       if (*tok == ACONST)
                        {
-#if 0
-                               *argptr++ = p;
-#else
-                               argPtrs[argp++] = p;
-                               startOfArg = p;
-#endif
+                               for(int i=0; i<3; i++)
+                                       *p++ = *tok++;
                        }
-
-                       // Keep going while tok isn't pointing at a comma or EOL
-                       while (*tok != ',' && *tok != EOL)
+                       else if (*tok == CONST)         // Constants are 64-bits
                        {
-                               // Skip over backslash character, unless it's followed by an EOL
-                               if (*tok == '\\' && tok[1] != EOL)
-                                       tok++;
-
-                               switch (*tok)
-                               {
-                               case CONST:
-                               case SYMBOL:
-//Shamus: Possible bug. ACONST has 2 tokens after it, not just 1
-                               case ACONST:
-                                       if (dry_run)
-                                       {
-                                               arg_siz += sizeof(TOKEN);
-                                               tok++;
-                                       }
-                                       else
-                                       {
-                                               *p++ = *tok++;
-                                       }
-                               // FALLTHROUGH (picks up the arg after a CONST, SYMBOL or ACONST)
-                               default:
-                                       if (dry_run)
-                                       {
-                                               arg_siz += sizeof(TOKEN);
-                                               tok++;
-                                       }
-                                       else
-                                       {
-                                               *p++ = *tok++;
-                                       }
-
-                                       break;
-                               }
+                               *p++ = *tok++;                  // Token
+                               uint64_t *p64 = (uint64_t *)p;
+                               uint64_t *tok64 = (uint64_t *)tok;
+                               *p64++ = *tok64++;
+                               tok = (TOKEN *)tok64;
+                               p = (uint32_t *)p64;
                        }
-
-                       // We hit the comma or EOL, so count/stuff it
-                       if (dry_run)
-                               arg_siz += sizeof(TOKEN);
-                       else
+                       else if ((*tok == STRING) || (*tok == SYMBOL))
+                       {
+                               *p++ = *tok++;
+                               imacro->argument[nargs].string[stringNum] = strdup(string[*tok++]);
+                               *p++ = stringNum++;
+                       }
+                       else if (*tok == ',')
+                       {
+                               // Comma delimiter was found, so set up for next argument
                                *p++ = EOL;
-
-                       // If we hit the comma instead of an EOL, skip over it
-                       if (*tok == ',')
                                tok++;
-
-                       // Do our QnD token grabbing (this will be redone once we get all
-                       // the data structures fixed as this is a really dirty hack)
-                       if (!dry_run)
-                       {
-                               dest = imacro->argument[argumentNum].token;
                                stringNum = 0;
-
-                               do
-                               {
-                                       // Remap strings to point the IMACRO internal token storage
-                                       if (*startOfArg == SYMBOL || *startOfArg == STRING)
-                                       {
-                                               *dest++ = *startOfArg++;
-                                               imacro->argument[argumentNum].string[stringNum] = strdup(string[*startOfArg++]);
-                                               *dest++ = stringNum++;
-                                       }
-                                       else
-                                               *dest++ = *startOfArg++;
-                               }
-                               while (*startOfArg != EOL);
-
-                               *dest = *startOfArg;            // Copy EOL...
-                               argumentNum++;
+                               nargs++;
+                               p = imacro->argument[nargs].token;
+                       }
+                       else
+                       {
+                               *p++ = *tok++;
                        }
                }
 
-               // Allocate space for argument ptrs and so on and then go back and
-               // construct the arg frame
-               if (dry_run)
-               {
-                       if (nargs != 0)
-                               p = (TOKEN *)malloc(arg_siz);
-//                             p = (TOKEN *)malloc(arg_siz + sizeof(TOKEN));
-
-/*
-Shamus:
-This construct is meant to deal with nested macros, so the simple minded way
-we deal with them now won't work. :-/ Have to think about how to fix.
-What we could do is simply move the argp with each call, and move it back by
-the number of arguments in the macro that's ending. That would solve the
-problem nicely.
-[Which we do now. But that uncovered another problem: the token strings are all
-stale by the time a nested macro gets to the end. But they're supposed to be
-symbols, which means if we put symbol references into the argument token
-streams, we can alleviate this problem.]
-*/
-#if 0
-                       argptr = (TOKEN **)malloc((nargs + 1) * sizeof(LONG));
-                       *argptr++ = (TOKEN *)argp;
-                       argp = argptr;
-#else
-                       // We don't need to do anything here since we already advance argp
-                       // when parsing the arguments.
-//                     argp += nargs;
-#endif
-               }
-               else
-                       break;
+               // Make sure to stuff the final EOL (otherwise, it will be skipped)
+               *p++ = EOL;
+               nargs++;
        }
 
-       DEBUG printf("%d\n", argp);
-
-       // Setup imacro:
-       // o  # arguments;
-       // o  -> macro symbol;
-       // o  -> macro definition string list;
-       // o  save 'curuniq', to be restored when the macro pops;
-       // o  bump `macuniq' counter and set 'curuniq' to it;
+       // Setup IMACRO:
+       //  o  # arguments;
+       //  o  -> macro symbol;
+       //  o  -> macro definition string list;
+       //  o  save 'curuniq', to be restored when the macro pops;
+       //  o  bump `macuniq' counter and set 'curuniq' to it;
        imacro->im_nargs = nargs;
        imacro->im_macro = mac;
-//     imacro->im_nextln = (TOKEN *)mac->svalue;
+       imacro->im_siz = siz;
        imacro->im_nextln = mac->lineList;
        imacro->im_olduniq = curuniq;
        curuniq = macuniq++;
-       imacro->argBase = argp - nargs;                 // Shamus: keep track of argument base
 
        DEBUG
        {
-               printf("nargs=%d\n", nargs);
+               printf("# args = %d\n", nargs);
 
-               for(nargs=0; nargs<imacro->im_nargs; nargs++)
+               for(uint16_t i=0; i<nargs; i++)
                {
-                       printf("arg%d=", nargs);
-//                     dumptok(argp[imacro->im_nargs - nargs - 1]);
-//                     dumptok(argPtrs[imacro->im_nargs - nargs - 1]);
-                       dumptok(argPtrs[(argp - imacro->im_nargs) + nargs]);
+                       printf("arg%d=", i);
+                       DumpTokens(imacro->argument[i].token);
                }
        }