--- gforth/prim	2006/10/13 17:15:27	1.196
+++ gforth/prim	2011/10/18 13:13:53	1.258
@@ -1,12 +1,12 @@
 \ Gforth primitives
 
-\ Copyright (C) 1995,1996,1997,1998,2000,2003,2004,2005 Free Software Foundation, Inc.
+\ Copyright (C) 1995,1996,1997,1998,2000,2003,2004,2005,2006,2007,2008,2009,2010 Free Software Foundation, Inc.
 
 \ This file is part of Gforth.
 
 \ Gforth is free software; you can redistribute it and/or
 \ modify it under the terms of the GNU General Public License
-\ as published by the Free Software Foundation; either version 2
+\ as published by the Free Software Foundation, either version 3
 \ of the License, or (at your option) any later version.
 
 \ This program is distributed in the hope that it will be useful,
@@ -15,8 +15,7 @@
 \ GNU General Public License for more details.
 
 \ You should have received a copy of the GNU General Public License
-\ along with this program; if not, write to the Free Software
-\ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111, USA.
+\ along with this program. If not, see http://www.gnu.org/licenses/.
 
 
 \ WARNING: This file is processed by m4. Make sure your identifiers
@@ -109,9 +108,9 @@
 \E store-optimization on
 \E ' noop tail-nextp2 ! \ now INST_TAIL just stores, but does not jump
 \E
-\E include-skipped-insts on \ static superinsts include cells for components
-\E                          \ useful for dynamic programming and
-\E                          \ superinsts across entry points
+\E `include-skipped-insts' on \ static superinsts include cells for components
+\E                            \ useful for dynamic programming and
+\E                            \ superinsts across entry points
 
 \ 
 \ 
@@ -203,22 +202,55 @@ INST_TAIL;
 goto *next_code;
 #endif /* defined(NO_IP) */
 
+(dovalue) ( -- w )	gforth-internal	paren_doval
+""run-time routine for constants""
+w = *(Cell *)PFA(CFA);
+#ifdef NO_IP
+INST_TAIL;
+goto *next_code;
+#endif /* defined(NO_IP) */
+
 (dodoes) ( -- a_body R:a_retaddr )	gforth-internal	paren_dodoes
 ""run-time routine for @code{does>}-defined words""
 #ifdef NO_IP
 a_retaddr = next_code;
 a_body = PFA(CFA);
 INST_TAIL;
+#ifdef DEBUG
+fprintf(stderr, "dodoes to %x, push %x\n", a_retaddr, a_body);
+#endif
 goto **(Label *)DOES_CODE1(CFA);
 #else /* !defined(NO_IP) */
 a_retaddr = (Cell *)IP;
 a_body = PFA(CFA);
+#ifdef DEBUG
+fprintf(stderr, "dodoes to %x, push %x\n", a_retaddr, a_body);
+#endif
 SET_IP(DOES_CODE1(CFA));
 #endif /* !defined(NO_IP) */
 
-(does-handler) ( -- )	gforth-internal	paren_does_handler
-""just a slot to have an encoding for the DOESJUMP, 
-which is no longer used anyway (!! eliminate this)""
+(doabicode) ( ... -- ...)	gforth-internal	paren_doabicode
+""run-time routine for @code{ABI-code} definitions""
+abifunc *f = (abifunc *)PFA(CFA);
+Float *fp_mem = fp;
+sp = (*f)(sp, &fp_mem);
+fp = fp_mem;
+#ifdef NO_IP
+INST_TAIL;
+goto *next_code;
+#endif /* defined(NO_IP) */
+
+(do;abicode) ( ... -- ... ) gforth-internal paren_do_semicolon_abi_code
+""run-time routine for @code{;abi-code}-defined words""
+Float *fp_mem = fp;
+Address body = (Address)PFA(CFA);
+semiabifunc *f = (semiabifunc *)DOES_CODE1(CFA);
+sp = (*f)(sp, &fp_mem, body);
+fp = fp_mem;
+#ifdef NO_IP
+INST_TAIL;
+goto *next_code;
+#endif /* defined(NO_IP) */
 
 \F [endif]
 
@@ -248,6 +280,9 @@ SET_IP((Xt *)a_callee);
 
 execute	( xt -- )		core
 ""Perform the semantics represented by the execution token, @i{xt}.""
+#ifdef DEBUG
+fprintf(stderr, "execute %08x\n", xt);
+#endif
 #ifndef NO_IP
 ip=IP;
 #endif
@@ -329,8 +364,6 @@ SET_IP((Xt *)a_target);
 
 \ condbranch(forthname,stackeffect,restline,code1,code2,forthcode)
 \ this is non-syntactical: code must open a brace that is closed by the macro
-\ condbranch(forthname,stackeffect,restline,code1,code2,forthcode)
-\ this is non-syntactical: code must open a brace that is closed by the macro
 define(condbranch,
 $1 ( `#'a_target $2 ) $3
 $4	#ifdef NO_IP
@@ -340,8 +373,10 @@ $5	#ifdef NO_IP
 JUMP(a_target);
 #else
 SET_IP((Xt *)a_target);
+ifelse(condbranch_opt,`1',`INST_TAIL; NEXT_P2;',`/* condbranch_opt=0 */')
 #endif
 }
+ifelse(condbranch_opt,`1',`SUPER_CONTINUE;',`/* condbranch_opt=0 */')
 $6
 
 \+glocals
@@ -355,43 +390,10 @@ $5	lp += nlocals;
 JUMP(a_target);
 #else
 SET_IP((Xt *)a_target);
+ifelse(condbranch_opt,`1',`INST_TAIL; NEXT_P2;',`/* condbranch_opt=0 */')
 #endif
 }
-
-\+
-)
-
-\ version that generates two jumps (not good for PR 15242 workaround)
-define(condbranch_twojump,
-$1 ( `#'a_target $2 ) $3
-$4	#ifdef NO_IP
-INST_TAIL;
-#endif
-$5	#ifdef NO_IP
-JUMP(a_target);
-#else
-SET_IP((Xt *)a_target);
-INST_TAIL; NEXT_P2;
-#endif
-}
-SUPER_CONTINUE;
-$6
-
-\+glocals
-
-$1-lp+!`#' ( `#'a_target `#'nlocals $2 ) $3_lp_plus_store_number
-$4	#ifdef NO_IP
-INST_TAIL;
-#endif
-$5	lp += nlocals;
-#ifdef NO_IP
-JUMP(a_target);
-#else
-SET_IP((Xt *)a_target);
-INST_TAIL; NEXT_P2;
-#endif
-}
-SUPER_CONTINUE;
+ifelse(condbranch_opt,`1',`SUPER_CONTINUE;',`/* condbranch_opt=0 */')
 
 \+
 )
@@ -610,6 +612,24 @@ SET_IP((Xt *)a_target);
      cell+
  THEN  >r ;
 
+(try1)	( ... a_oldhandler a_recovery -- R:a_recovery R:a_sp R:f_fp R:c_lp R:a_oldhandler a_newhandler ) gforth paren_try1
+a_sp = sp-1;
+f_fp = fp;
+c_lp = lp;
+a_newhandler = rp-5;
+
+(throw1) ( ... wball a_handler -- ... wball ) gforth paren_throw1
+rp = a_handler;
+lp = (Address)rp[1];
+fp = (Float *)rp[2];
+sp = (Cell *)rp[3];
+#ifndef NO_IP
+ip=IP;
+#endif
+SUPER_END;
+VM_JUMP(EXEC1(*(Xt *)rp[4]));
+  
+
 \+
 
 \ don't make any assumptions where the return stack is!!
@@ -626,13 +646,13 @@ i'	( R:w R:w2 -- R:w R:w2 w )		gforth		i
   r> r> r> dup itmp ! >r >r >r itmp @ ;
 variable itmp
 
-j	( R:n R:d1 -- n R:n R:d1 )		core
+j	( R:w R:w1 R:w2 -- w R:w R:w1 R:w2 )	core
 :
 \ rp@ cell+ cell+ cell+ @ ;
   r> r> r> r> dup itmp ! >r >r >r >r itmp @ ;
 [IFUNDEF] itmp variable itmp [THEN]
 
-k	( R:n R:d1 R:d2 -- n R:n R:d1 R:d2 )		gforth
+k	( R:w R:w1 R:w2 R:w3 R:w4 -- w R:w R:w1 R:w2 R:w3 R:w4 )	gforth
 :
 \ rp@ [ 5 cells ] Literal + @ ;
   r> r> r> r> r> r> dup itmp ! >r >r >r >r >r >r itmp @ ;
@@ -719,7 +739,7 @@ c2 = toupper(c1);
 :
  dup [char] a - [ char z char a - 1 + ] Literal u<  bl and - ;
 
-capscompare	( c_addr1 u1 c_addr2 u2 -- n )	string
+capscompare	( c_addr1 u1 c_addr2 u2 -- n )	gforth
 ""Compare two strings lexicographically. If they are equal, @i{n} is 0; if
 the first string is smaller, @i{n} is -1; if the first string is larger, @i{n}
 is 1. Currently this is based on the machine's character
@@ -748,6 +768,9 @@ n = n1+n2;
 \ lit+ / lit_plus = lit +
 
 lit+	( n1 #n2 -- n )		new	lit_plus
+#ifdef DEBUG
+fprintf(stderr, "lit+ %08x\n", n2);
+#endif
 n=n1+n2;
 
 \ PFE-0.9.14 has it differently, but the next release will have it as follows
@@ -809,12 +832,21 @@ n = n1*n2;
 
 /	( n1 n2 -- n )		core	slash
 n = n1/n2;
-if(FLOORED_DIV && ((n1^n2) < 0) && (n1%n2 != 0)) n--;
+if (CHECK_DIVISION_SW && n2 == 0)
+  throw(BALL_DIVZERO);
+if (CHECK_DIVISION_SW && n2 == -1 && n1 == CELL_MIN)
+  throw(BALL_RESULTRANGE);
+if (FLOORED_DIV && ((n1^n2) < 0) && (n1%n2 != 0))
+  n--;
 :
  /mod nip ;
 
 mod	( n1 n2 -- n )		core
 n = n1%n2;
+if (CHECK_DIVISION_SW && n2 == 0)
+  throw(BALL_DIVZERO);
+if (CHECK_DIVISION_SW && n2 == -1 && n1 == CELL_MIN)
+  throw(BALL_RESULTRANGE);
 if(FLOORED_DIV && ((n1^n2) < 0) && n!=0) n += n2;
 :
  /mod drop ;
@@ -822,6 +854,10 @@ if(FLOORED_DIV && ((n1^n2) < 0) && n!=0)
 /mod	( n1 n2 -- n3 n4 )		core		slash_mod
 n4 = n1/n2;
 n3 = n1%n2; /* !! is this correct? look into C standard! */
+if (CHECK_DIVISION_SW && n2 == 0)
+  throw(BALL_DIVZERO);
+if (CHECK_DIVISION_SW && n2 == -1 && n1 == CELL_MIN)
+  throw(BALL_RESULTRANGE);
 if (FLOORED_DIV && ((n1^n2) < 0) && n3!=0) {
   n4--;
   n3+=n2;
@@ -836,18 +872,18 @@ DCell d = mmul(n1,n2);
 #else
 DCell d = (DCell)n1 * (DCell)n2;
 #endif
-#ifdef BUGGY_LL_DIV
-DCell r = fmdiv(d,n3);
-n4=DHI(r);
-n5=DLO(r);
-#else
-/* assumes that the processor uses either floored or symmetric division */
-n5 = d/n3;
-n4 = d%n3;
+#ifdef ASM_SM_SLASH_REM
+ASM_SM_SLASH_REM(DLO(d), DHI(d), n3, n4, n5);
 if (FLOORED_DIV && ((DHI(d)^n3)<0) && n4!=0) {
+  if (CHECK_DIVISION && n5 == CELL_MIN)
+    throw(BALL_RESULTRANGE);
   n5--;
   n4+=n3;
 }
+#else
+DCell r = FLOORED_DIV ? fmdiv(d,n3) : smdiv(d,n3);
+n4=DHI(r);
+n5=DLO(r);
 #endif
 :
  >r m* r> fm/mod ;
@@ -859,13 +895,17 @@ DCell d = mmul(n1,n2);
 #else
 DCell d = (DCell)n1 * (DCell)n2;
 #endif
-#ifdef BUGGY_LL_DIV
-DCell r = fmdiv(d,n3);
-n4=DLO(r);
+#ifdef ASM_SM_SLASH_REM
+Cell remainder;
+ASM_SM_SLASH_REM(DLO(d), DHI(d), n3, remainder, n4);
+if (FLOORED_DIV && ((DHI(d)^n3)<0) && remainder!=0) {
+  if (CHECK_DIVISION && n4 == CELL_MIN)
+    throw(BALL_RESULTRANGE);
+  n4--;
+}
 #else
-/* assumes that the processor uses either floored or symmetric division */
-n4 = d/n3;
-if (FLOORED_DIV && ((DHI(d)^n3)<0) && (d%n3)!=0) n4--;
+DCell r = FLOORED_DIV ? fmdiv(d,n3) : smdiv(d,n3);
+n4=DLO(r);
 #endif
 :
  */mod nip ;
@@ -889,10 +929,11 @@ n2 = n1>>1;
 
 fm/mod	( d1 n1 -- n2 n3 )		core		f_m_slash_mod
 ""Floored division: @i{d1} = @i{n3}*@i{n1}+@i{n2}, @i{n1}>@i{n2}>=0 or 0>=@i{n2}>@i{n1}.""
-#ifdef BUGGY_LL_DIV
 #ifdef ASM_SM_SLASH_REM
-ASM_SM_SLASH_REM(d1.lo, d1.hi, n1, n2, n3);
+ASM_SM_SLASH_REM(DLO(d1), DHI(d1), n1, n2, n3);
 if (((DHI(d1)^n1)<0) && n2!=0) {
+  if (CHECK_DIVISION && n3 == CELL_MIN)
+    throw(BALL_RESULTRANGE);
   n3--;
   n2+=n1;
 }
@@ -901,24 +942,6 @@ DCell r = fmdiv(d1,n1);
 n2=DHI(r);
 n3=DLO(r);
 #endif /* !defined(ASM_SM_SLASH_REM) */
-#else
-#ifdef ASM_SM_SLASH_REM4
-ASM_SM_SLASH_REM4(d1, n1, n2, n3);
-if (((DHI(d1)^n1)<0) && n2!=0) {
-  n3--;
-  n2+=n1;
-}
-#else /* !defined(ASM_SM_SLASH_REM4) */
-/* assumes that the processor uses either floored or symmetric division */
-n3 = d1/n1;
-n2 = d1%n1;
-/* note that this 1%-3>0 is optimized by the compiler */
-if (1%-3>0 && ((DHI(d1)^n1)<0) && n2!=0) {
-  n3--;
-  n2+=n1;
-}
-#endif /* !defined(ASM_SM_SLASH_REM4) */
-#endif
 :
  dup >r dup 0< IF  negate >r dnegate r>  THEN
  over       0< IF  tuck + swap  THEN
@@ -927,28 +950,13 @@ if (1%-3>0 && ((DHI(d1)^n1)<0) && n2!=0)
 
 sm/rem	( d1 n1 -- n2 n3 )		core		s_m_slash_rem
 ""Symmetric division: @i{d1} = @i{n3}*@i{n1}+@i{n2}, sign(@i{n2})=sign(@i{d1}) or 0.""
-#ifdef BUGGY_LL_DIV
 #ifdef ASM_SM_SLASH_REM
-ASM_SM_SLASH_REM(d1.lo, d1.hi, n1, n2, n3);
+ASM_SM_SLASH_REM(DLO(d1), DHI(d1), n1, n2, n3);
 #else /* !defined(ASM_SM_SLASH_REM) */
 DCell r = smdiv(d1,n1);
 n2=DHI(r);
 n3=DLO(r);
 #endif /* !defined(ASM_SM_SLASH_REM) */
-#else
-#ifdef ASM_SM_SLASH_REM4
-ASM_SM_SLASH_REM4(d1, n1, n2, n3);
-#else /* !defined(ASM_SM_SLASH_REM4) */
-/* assumes that the processor uses either floored or symmetric division */
-n3 = d1/n1;
-n2 = d1%n1;
-/* note that this 1%-3<0 is optimized by the compiler */
-if (1%-3<0 && ((DHI(d1)^n1)<0) && n2!=0) {
-  n3++;
-  n2-=n1;
-}
-#endif /* !defined(ASM_SM_SLASH_REM4) */
-#endif
 :
  over >r dup >r abs -rot
  dabs rot um/mod
@@ -984,22 +992,13 @@ ud = (UDCell)u1 * (UDCell)u2;
 
 um/mod	( ud u1 -- u2 u3 )		core	u_m_slash_mod
 ""ud=u3*u1+u2, u1>u2>=0""
-#ifdef BUGGY_LL_DIV
 #ifdef ASM_UM_SLASH_MOD
-ASM_UM_SLASH_MOD(ud.lo, ud.hi, u1, u2, u3);
+ASM_UM_SLASH_MOD(DLO(ud), DHI(ud), u1, u2, u3);
 #else /* !defined(ASM_UM_SLASH_MOD) */
 UDCell r = umdiv(ud,u1);
 u2=DHI(r);
 u3=DLO(r);
 #endif /* !defined(ASM_UM_SLASH_MOD) */
-#else
-#ifdef ASM_UM_SLASH_MOD4
-ASM_UM_SLASH_MOD4(ud, u1, u2, u3);
-#else /* !defined(ASM_UM_SLASH_MOD4) */
-u3 = ud/u1;
-u2 = ud%u1;
-#endif /* !defined(ASM_UM_SLASH_MOD4) */
-#endif
 :
    0 swap [ 8 cells 1 + ] literal 0
    ?DO /modstep
@@ -1052,12 +1051,7 @@ d2 = -d1;
 
 d2*	( d1 -- d2 )		double		d_two_star
 ""Shift left by 1; also works on unsigned numbers""
-#ifdef BUGGY_LL_SHIFT
-DLO_IS(d2, DLO(d1)<<1);
-DHI_IS(d2, (DHI(d1)<<1) | (DLO(d1)>>(CELL_BITS-1)));
-#else
-d2 = 2*d1;
-#endif
+d2 = DLSHIFT(d1,1);
 :
  2dup d+ ;
 
@@ -1516,6 +1510,11 @@ for (; f83name1 != NULL; f83name1 = (str
       memcasecmp(c_addr, f83name1->name, u)== 0 /* or inline? */)
     break;
 f83name2=f83name1;
+#ifdef DEBUG
+fprintf(stderr, "F83find ");
+fwrite(c_addr, u, 1, stderr);
+fprintf(stderr, " found %08x\n", f83name2); 
+#endif
 :
     BEGIN  dup WHILE  (find-samelen)  dup  WHILE
 	>r 2dup r@ cell+ char+ capscomp  0=
@@ -1671,8 +1670,6 @@ f = key_query((FILE*)wfileid);
 f = key_query(stdin);
 #endif
 
-\+os
-
 stdin	( -- wfileid )	gforth
 ""The standard input file of the Gforth process.""
 wfileid = (Cell)stdin;
@@ -1685,9 +1682,16 @@ stderr	( -- wfileid )	gforth
 ""The standard error output file of the Gforth process.""
 wfileid = (Cell)stderr;
 
+\+os
+
 form	( -- urows ucols )	gforth
-""The number of lines and columns in the terminal. These numbers may change
-with the window size.""
+""The number of lines and columns in the terminal. These numbers may
+change with the window size.  Note that it depends on the OS whether
+this reflects the actual size and changes with the window size
+(currently only on Unix-like OSs).  On other OSs you just get a
+default, and can tell Gforth the terminal size by setting the
+environment variables @code{COLUMNS} and @code{LINES} before starting
+Gforth.""
 /* we could block SIGWINCH here to get a consistent size, but I don't
  think this is necessary or always beneficial */
 urows=rows;
@@ -1695,7 +1699,11 @@ ucols=cols;
 
 wcwidth	( u -- n )	gforth
 ""The number of fixed-width characters per unicode character u""
+#ifdef HAVE_WCWIDTH
 n = wcwidth(u);
+#else
+n = 1;
+#endif
 
 flush-icache	( c_addr u -- )	gforth	flush_icache
 ""Make sure that the instruction cache of the processor (if there is
@@ -1707,7 +1715,7 @@ supported on your machine (take a look a
 your machine has a separate instruction cache. In such cases,
 @code{flush-icache} does nothing instead of flushing the instruction
 cache.""
-FLUSH_ICACHE(c_addr,u);
+FLUSH_ICACHE((caddr_t)c_addr,u);
 
 (bye)	( n -- )	gforth	paren_bye
 SUPER_END;
@@ -1727,6 +1735,7 @@ c_addr2 = (Char *)getenv(cstr(c_addr1,u1
 u2 = (c_addr2 == NULL ? 0 : strlen((char *)c_addr2));
 
 open-pipe	( c_addr u wfam -- wfileid wior )	gforth	open_pipe
+fflush(stdout);
 wfileid=(Cell)popen(cstr(c_addr,u,1),pfileattr[wfam]); /* ~ expansion of 1st arg? */
 wior = IOR(wfileid==0); /* !! the man page says that errno is not set reliably */
 
@@ -1758,12 +1767,9 @@ nhour =ltime->tm_hour;
 nmin  =ltime->tm_min;
 nsec  =ltime->tm_sec;
 
-ms	( n -- )	facility-ext
+ms	( u -- )	facility-ext
 ""Wait at least @i{n} milli-second.""
-struct timeval timeout;
-timeout.tv_sec=n/1000;
-timeout.tv_usec=1000*(n%1000);
-(void)select(0,0,0,0,&timeout);
+gforth_ms(u);
 
 allocate	( u -- a_addr wior )	memory
 ""Allocate @i{u} address units of contiguous data space. The initial
@@ -1798,6 +1804,8 @@ if (a_addr1==NULL)
 else
   a_addr2 = (Cell *)realloc(a_addr1, u);
 wior = IOR(a_addr2==NULL);	/* !! Define a return code */
+if (a_addr2==NULL)
+  a_addr2 = a_addr1;
 
 strerror	( n -- c_addr u )	gforth
 c_addr = (Char *)strerror(n);
@@ -1810,14 +1818,24 @@ u = strlen((char *)c_addr);
 call-c	( ... w -- ... )	gforth	call_c
 ""Call the C function pointed to by @i{w}. The C function has to
 access the stack itself. The stack pointers are exported in the global
-variables @code{SP} and @code{FP}.""
+variables @code{gforth_SP} and @code{gforth_FP}.""
 /* This is a first attempt at support for calls to C. This may change in
    the future */
+IF_fpTOS(fp[0]=fpTOS);
 gforth_FP=fp;
 gforth_SP=sp;
+gforth_RP=rp;
+gforth_LP=lp;
+#ifdef HAS_LINKBACK
 ((void (*)())w)();
+#else
+((void (*)(void *))w)(gforth_pointers);
+#endif
 sp=gforth_SP;
 fp=gforth_FP;
+rp=gforth_RP;
+lp=gforth_LP;
+IF_fpTOS(fpTOS=fp[0]);
 
 \+
 \+file
@@ -1826,19 +1844,10 @@ close-file	( wfileid -- wior )		file	clo
 wior = IOR(fclose((FILE *)wfileid)==EOF);
 
 open-file	( c_addr u wfam -- wfileid wior )	file	open_file
-wfileid = (Cell)fopen(tilde_cstr(c_addr, u, 1), fileattr[wfam]);
-wior =  IOR(wfileid == 0);
+wfileid = opencreate_file(tilde_cstr(c_addr,u,1), wfam, 0, &wior);
 
 create-file	( c_addr u wfam -- wfileid wior )	file	create_file
-Cell	fd;
-fd = open(tilde_cstr(c_addr, u, 1), O_CREAT|O_TRUNC|ufileattr[wfam], 0666);
-if (fd != -1) {
-  wfileid = (Cell)fdopen(fd, fileattr[wfam]);
-  wior = IOR(wfileid == 0);
-} else {
-  wfileid = 0;
-  wior = IOR(1);
-}
+wfileid = opencreate_file(tilde_cstr(c_addr,u,1), wfam, O_CREAT|O_TRUNC, &wior);
 
 delete-file	( c_addr u -- wior )		file	delete_file
 wior = IOR(unlink(tilde_cstr(c_addr, u, 1))==-1);
@@ -1866,13 +1875,15 @@ wior = IOR(ftruncate(fileno((FILE *)wfil
 read-file	( c_addr u1 wfileid -- u2 wior )	file	read_file
 /* !! fread does not guarantee enough */
 u2 = fread(c_addr, sizeof(Char), u1, (FILE *)wfileid);
+if (u2>0)
+   gf_regetc((FILE *)wfileid);
 wior = FILEIO(u2<u1 && ferror((FILE *)wfileid));
 /* !! is the value of ferror errno-compatible? */
 if (wior)
   clearerr((FILE *)wfileid);
 
 (read-line)	( c_addr u1 wfileid -- u2 flag u3 wior ) file	paren_read_line
-struct Cellquad r = read_line(c_addr, u1, wfileid);
+struct Cellquad r = read_line(c_addr, u1, (FILE *)wfileid);
 u2   = r.n1;
 flag = r.n2;
 u3   = r.n3;
@@ -1964,7 +1975,7 @@ Return an error if this is not possible"
 wior = IOR(chdir(tilde_cstr(c_addr, u, 1)));
 
 get-dir	( c_addr1 u1 -- c_addr2 u2 )	gforth get_dir
-""Store the current directory in the buffer specified by @{c-addr1, u1}.
+""Store the current directory in the buffer specified by @i{c-addr1, u1}.
 If the buffer size is not sufficient, return 0 0""
 c_addr2 = (Char *)getcwd((char *)c_addr1, u1);
 if(c_addr2 != NULL) {
@@ -1973,6 +1984,10 @@ if(c_addr2 != NULL) {
   u2 = 0;
 }
 
+=mkdir ( c_addr u wmode -- wior )        gforth equals_mkdir
+""Create directory @i{c-addr u} with mode @i{wmode}.""
+wior = IOR(mkdir(tilde_cstr(c_addr,u,1),wmode));
+
 \+
 
 newline	( -- c_addr u )	gforth
@@ -2153,7 +2168,7 @@ floor	( r1 -- r2 )	float
 /* !! unclear wording */
 r2 = floor(r1);
 
-fround	( r1 -- r2 )	gforth	f_round
+fround	( r1 -- r2 )	float	f_round
 ""Round to the nearest integral value.""
 r2 = rint(r1);
 
@@ -2176,6 +2191,7 @@ int flag;
 int decpt;
 sig=ecvt(r, u, &decpt, &flag);
 n=(r==0. ? 1 : decpt);
+flag=signbit(r); /* not all ecvt()s do this as desired */
 f1=FLAG(flag!=0);
 f2=FLAG(isdigit((unsigned)(sig[0]))!=0);
 siglen=strlen((char *)sig);
@@ -2438,22 +2454,12 @@ r = fp[u];
 \g syslib
 
 open-lib	( c_addr1 u1 -- u2 )	gforth	open_lib
-#if defined(HAVE_LIBDL) || defined(HAVE_DLOPEN)
-#ifndef RTLD_GLOBAL
-#define RTLD_GLOBAL 0
-#endif
-u2=(UCell) dlopen(cstr(c_addr1, u1, 1), RTLD_GLOBAL | RTLD_LAZY);
-#else
-#  ifdef _WIN32
-u2 = (Cell) GetModuleHandle(cstr(c_addr1, u1, 1));
-#  else
-#warning Define open-lib!
-u2 = 0;
-#  endif
-#endif
+u2 = gforth_dlopen(c_addr1, u1);
 
 lib-sym	( c_addr1 u1 u2 -- u3 )	gforth	lib_sym
-#if defined(HAVE_LIBDL) || defined(HAVE_DLOPEN)
+#ifdef HAVE_LIBLTDL
+u3 = (UCell) lt_dlsym((lt_dlhandle)u2, cstr(c_addr1, u1, 1));
+#elif defined(HAVE_LIBDL) || defined(HAVE_DLOPEN)
 u3 = (UCell) dlsym((void*)u2,cstr(c_addr1, u1, 1));
 #else
 #  ifdef _WIN32
@@ -2493,337 +2499,119 @@ l! ( w c_addr -- )	gforth l_store
 ""Store the bottom 32 bits of @i{w} at @i{c_addr}.""
 *(Tetrabyte*)(c_addr) = w;
 
-\+FFCALL
-
-av-start-void	( c_addr -- )	gforth  av_start_void
-av_start_void(alist, c_addr);
-
-av-start-int	( c_addr -- )	gforth  av_start_int
-av_start_int(alist, c_addr, &irv);
-
-av-start-float	( c_addr -- )	gforth  av_start_float
-av_start_float(alist, c_addr, &frv);
-
-av-start-double	( c_addr -- )	gforth  av_start_double
-av_start_double(alist, c_addr, &drv);
-
-av-start-longlong	( c_addr -- )	gforth  av_start_longlong
-av_start_longlong(alist, c_addr, &llrv);
-
-av-start-ptr	( c_addr -- )	gforth  av_start_ptr
-av_start_ptr(alist, c_addr, void*, &prv);
-
-av-int  ( w -- )  gforth  av_int
-av_int(alist, w);
-
-av-float	( r -- )	gforth  av_float
-av_float(alist, r);
-
-av-double	( r -- )	gforth  av_double
-av_double(alist, r);
-
-av-longlong	( d -- )	gforth  av_longlong
-#ifdef BUGGY_LL_SIZE
-av_longlong(alist, DLO(d));
-#else
-av_longlong(alist, d);
-#endif
-
-av-ptr	( c_addr -- )	gforth  av_ptr
-av_ptr(alist, void*, c_addr);
-
-av-int-r  ( R:w -- )  gforth  av_int_r
-av_int(alist, w);
-
-av-float-r	( -- )	gforth  av_float_r
-float r = *(Float*)lp;
-lp += sizeof(Float);
-av_float(alist, r);
-
-av-double-r	( -- )	gforth  av_double_r
-double r = *(Float*)lp;
-lp += sizeof(Float);
-av_double(alist, r);
-
-av-longlong-r	( R:d -- )	gforth  av_longlong_r
-#ifdef BUGGY_LL_SIZE
-av_longlong(alist, DLO(d));
-#else
-av_longlong(alist, d);
-#endif
-
-av-ptr-r	( R:c_addr -- )	gforth  av_ptr_r
-av_ptr(alist, void*, c_addr);
-
-av-call-void	( ... -- ... )	gforth  av_call_void
-SAVE_REGS
-av_call(alist);
-REST_REGS
-
-av-call-int	( ... -- ... w )	gforth  av_call_int
-SAVE_REGS
-av_call(alist);
-REST_REGS
-w = irv;
-
-av-call-float	( ... -- ... r )	gforth  av_call_float
-SAVE_REGS
-av_call(alist);
-REST_REGS
-r = frv;
-
-av-call-double	( ... -- ... r )	gforth  av_call_double
-SAVE_REGS
-av_call(alist);
-REST_REGS
-r = drv;
-
-av-call-longlong	( ... -- ... d )	gforth  av_call_longlong
-SAVE_REGS
-av_call(alist);
-REST_REGS
-#ifdef BUGGY_LONG_LONG
-DLO_IS(d, llrv);
-DHI_IS(d, 0);
-#else
-d = llrv;
-#endif
-
-av-call-ptr	( ... -- ... c_addr )	gforth  av_call_ptr
-SAVE_REGS
-av_call(alist);
-REST_REGS
-c_addr = prv;
-
-alloc-callback	( a_ip -- c_addr )	gforth	alloc_callback
-c_addr = (char *)alloc_callback(gforth_callback, (Xt *)a_ip);
-
-va-start-void	( -- )	gforth	va_start_void
-va_start_void(gforth_clist);
-
-va-start-int	( -- )	gforth	va_start_int
-va_start_int(gforth_clist);
-
-va-start-longlong	( -- )	gforth	va_start_longlong
-va_start_longlong(gforth_clist);
-
-va-start-ptr	( -- )	gforth	va_start_ptr
-va_start_ptr(gforth_clist, (char *));
-
-va-start-float	( -- )	gforth	va_start_float
-va_start_float(gforth_clist);
-
-va-start-double	( -- )	gforth	va_start_double
-va_start_double(gforth_clist);
-
-va-arg-int	( -- w )	gforth	va_arg_int
-w = va_arg_int(gforth_clist);
-
-va-arg-longlong	( -- d )	gforth	va_arg_longlong
-#ifdef BUGGY_LONG_LONG
-DLO_IS(d, va_arg_longlong(gforth_clist));
-DHI_IS(d, 0);
-#else
-d = va_arg_longlong(gforth_clist);
-#endif
-
-va-arg-ptr	( -- c_addr )	gforth	va_arg_ptr
-c_addr = (char *)va_arg_ptr(gforth_clist,char*);
-
-va-arg-float	( -- r )	gforth	va_arg_float
-r = va_arg_float(gforth_clist);
-
-va-arg-double	( -- r )	gforth	va_arg_double
-r = va_arg_double(gforth_clist);
-
-va-return-void ( -- )	gforth va_return_void
-va_return_void(gforth_clist);
-return 0;
-
-va-return-int ( w -- )	gforth va_return_int
-va_return_int(gforth_clist, w);
-return 0;
-
-va-return-ptr ( c_addr -- )	gforth va_return_ptr
-va_return_ptr(gforth_clist, void *, c_addr);
-return 0;
-
-va-return-longlong ( d -- )	gforth va_return_longlong
-#ifdef BUGGY_LONG_LONG
-va_return_longlong(gforth_clist, d.lo);
-#else
-va_return_longlong(gforth_clist, d);
-#endif
-return 0;
-
-va-return-float ( r -- )	gforth va_return_float
-va_return_float(gforth_clist, r);
-return 0;
-
-va-return-double ( r -- )	gforth va_return_double
-va_return_double(gforth_clist, r);
-return 0;
-
-\+
-
-\+LIBFFI
-
-ffi-type ( n -- a_type )	gforth ffi_type
-static void* ffi_types[] =
-    { &ffi_type_void,
-      &ffi_type_uint8, &ffi_type_sint8,
-      &ffi_type_uint16, &ffi_type_sint16,
-      &ffi_type_uint32, &ffi_type_sint32,
-      &ffi_type_uint64, &ffi_type_sint64,
-      &ffi_type_float, &ffi_type_double, &ffi_type_longdouble,
-      &ffi_type_pointer };
-a_type = ffi_types[n];
-
-ffi-size ( n1 -- n2 )	gforth ffi_size
-static int ffi_sizes[] =
-    { sizeof(ffi_cif), sizeof(ffi_closure) };
-n2 = ffi_sizes[n1];
-
-ffi-prep-cif ( a_atypes n a_rtype a_cif -- w )	gforth ffi_prep_cif
-w = ffi_prep_cif((ffi_cif *)a_cif, FFI_DEFAULT_ABI, n,
-	 (ffi_type *)a_rtype, (ffi_type **)a_atypes);
-
-ffi-call ( a_avalues a_rvalue a_ip a_cif -- )	gforth ffi_call
-SAVE_REGS
-ffi_call((ffi_cif *)a_cif, (void(*)())a_ip, (void *)a_rvalue, (void **)a_avalues);
-REST_REGS
-
-ffi-prep-closure ( a_ip a_cif a_closure -- w )	gforth ffi_prep_closure
-w = ffi_prep_closure((ffi_closure *)a_closure, (ffi_cif *)a_cif, gforth_callback, (void *)a_ip);
-
-ffi-2@ ( a_addr -- d )	gforth ffi_2fetch
-#ifdef BUGGY_LONG_LONG
-DLO_IS(d, *(Cell*)(*a_addr));
-DHI_IS(d, 0);
-#else
-d = *(DCell*)(a_addr);
-#endif
-
-ffi-2! ( d a_addr -- )	gforth ffi_2store
-#ifdef BUGGY_LONG_LONG
-*(Cell*)(a_addr) = DLO(d);
-#else
-*(DCell*)(a_addr) = d;
-#endif
-
-ffi-arg-int ( -- w )	gforth ffi_arg_int
-w = *(int *)(*gforth_clist++);
-
-ffi-arg-long ( -- w )	gforth ffi_arg_long
-w = *(long *)(*gforth_clist++);
-
-ffi-arg-longlong ( -- d )	gforth ffi_arg_longlong
-#ifdef BUGGY_LONG_LONG
-DLO_IS(d, *(Cell*)(*gforth_clist++));
-DHI_IS(d, -(*(Cell*)(*gforth_clist++)<0));
-#else
-d = *(DCell*)(*gforth_clist++);
-#endif
-
-ffi-arg-dlong ( -- d )	gforth ffi_arg_dlong
-#ifdef BUGGY_LONG_LONG
-DLO_IS(d, *(Cell*)(*gforth_clist++));
-DHI_IS(d, -(*(Cell*)(*gforth_clist++)<0));
-#else
-d = *(Cell*)(*gforth_clist++);
-#endif
-
-ffi-arg-ptr ( -- c_addr )	gforth ffi_arg_ptr
-c_addr = *(Char **)(*gforth_clist++);
-
-ffi-arg-float ( -- r )	gforth ffi_arg_float
-r = *(float*)(*gforth_clist++);
-
-ffi-arg-double ( -- r )	gforth ffi_arg_double
-r = *(double*)(*gforth_clist++);
-
-ffi-ret-void ( -- )	gforth ffi_ret_void
-return 0;
-
-ffi-ret-int ( w -- )	gforth ffi_ret_int
-*(int*)(gforth_ritem) = w;
-return 0;
-
-ffi-ret-longlong ( d -- )	gforth ffi_ret_longlong
-#ifdef BUGGY_LONG_LONG
-*(Cell*)(gforth_ritem) = DLO(d);
-#else
-*(DCell*)(gforth_ritem) = d;
-#endif
-return 0;
-
-ffi-ret-dlong ( d -- )	gforth ffi_ret_dlong
-#ifdef BUGGY_LONG_LONG
-*(Cell*)(gforth_ritem) = DLO(d);
-#else
-*(Cell*)(gforth_ritem) = d;
-#endif
-return 0;
-
-ffi-ret-long ( n -- )	gforth ffi_ret_long
-*(Cell*)(gforth_ritem) = n;
-return 0;
-
-ffi-ret-ptr ( c_addr -- )	gforth ffi_ret_ptr
-*(Char **)(gforth_ritem) = c_addr;
-return 0;
-
-ffi-ret-float ( r -- )	gforth ffi_ret_float
-*(float*)(gforth_ritem) = r;
-return 0;
-
-ffi-ret-double ( r -- )	gforth ffi_ret_double
-*(double*)(gforth_ritem) = r;
-return 0;
-
-\+
-
-\+OLDCALL
-
-define(`uploop',
-       `pushdef(`$1', `$2')_uploop(`$1', `$2', `$3', `$4', `$5')`'popdef(`$1')')
-define(`_uploop',
-       `ifelse($1, `$3', `$5',
-	       `$4`'define(`$1', incr($1))_uploop(`$1', `$2', `$3', `$4', `$5')')')
-\ argflist(argnum): Forth argument list
-define(argflist,
-       `ifelse($1, 0, `',
-               `uploop(`_i', 1, $1, `format(`u%d ', _i)', `format(`u%d ', _i)')')')
-\ argdlist(argnum): declare C's arguments
-define(argdlist,
-       `ifelse($1, 0, `',
-               `uploop(`_i', 1, $1, `Cell, ', `Cell')')')
-\ argclist(argnum): pass C's arguments
-define(argclist,
-       `ifelse($1, 0, `',
-               `uploop(`_i', 1, $1, `format(`u%d, ', _i)', `format(`u%d', _i)')')')
-\ icall(argnum)
-define(icall,
-`icall$1	( argflist($1)u -- uret )	gforth
-uret = (SYSCALL(Cell(*)(argdlist($1)))u)(argclist($1));
-
-')
-define(fcall,
-`fcall$1	( argflist($1)u -- rret )	gforth
-rret = (SYSCALL(Float(*)(argdlist($1)))u)(argclist($1));
-
-')
-
-\ close ' to keep fontify happy
-
-uploop(i, 0, 7, `icall(i)')
-icall(20)
-uploop(i, 0, 7, `fcall(i)')
-fcall(20)
+lib-error ( -- c_addr u )       gforth  lib_error
+""Error message for last failed @code{open-lib} or @code{lib-sym}.""
+#ifdef HAVE_LIBLTDL
+c_addr = (Char *)lt_dlerror();
+u = (c_addr == NULL) ? 0 : strlen((char *)c_addr);
+#elif defined(HAVE_LIBDL) || defined(HAVE_DLOPEN)
+c_addr = dlerror();
+u = strlen(c_addr);
+#else
+c_addr = "libltdl is not configured";
+u = strlen(c_addr);
+#endif
+
+be-w! ( w c_addr -- )	gforth w_store_be
+""Store the bottom 16 bits of @i{w} at @i{c_addr} in big endian format.""
+c_addr[0] = w >> 8;
+c_addr[1] = w;
+
+be-l! ( w c_addr -- )	gforth l_store_be
+""Store the bottom 32 bits of @i{w} at @i{c_addr} in big endian format.""
+c_addr[0] = w >> 24;
+c_addr[1] = w >> 16;
+c_addr[2] = w >> 8;
+c_addr[3] = w;
+
+le-w! ( w c_addr -- )	gforth w_store_le
+""Store the bottom 16 bits of @i{w} at @i{c_addr} in big endian format.""
+c_addr[1] = w >> 8;
+c_addr[0] = w;
+
+le-l! ( w c_addr -- )	gforth l_store_le
+""Store the bottom 32 bits of @i{w} at @i{c_addr} in big endian format.""
+c_addr[3] = w >> 24;
+c_addr[2] = w >> 16;
+c_addr[1] = w >> 8;
+c_addr[0] = w;
+
+be-uw@ ( c_addr -- u )	gforth w_fetch_be
+""@i{u} is the zero-extended 16-bit big endian value stored at @i{c_addr}.""
+u = (c_addr[0] << 8) | (c_addr[1]);
+
+be-ul@ ( c_addr -- u )	gforth l_fetch_be
+""@i{u} is the zero-extended 32-bit big endian value stored at @i{c_addr}.""
+u = (c_addr[0] << 24) | (c_addr[1] << 16) | (c_addr[2] << 8) | (c_addr[3]);
+
+le-uw@ ( c_addr -- u )	gforth w_fetch_le
+""@i{u} is the zero-extended 16-bit little endian value stored at @i{c_addr}.""
+u = (c_addr[1] << 8) | (c_addr[0]);
+
+le-ul@ ( c_addr -- u )	gforth l_fetch_le
+""@i{u} is the zero-extended 32-bit little endian value stored at @i{c_addr}.""
+u = (c_addr[3] << 24) | (c_addr[2] << 16) | (c_addr[1] << 8) | (c_addr[0]);
+
+\+64bit
+
+x! ( w c_addr -- )	gforth x_store
+""Store the bottom 64 bits of @i{w} at 64-bit-aligned @i{c_addr}.""
+*(UOctabyte *)c_addr = w;
+
+ux@ ( c_addr -- u )	gforth u_x_fetch
+""@i{u} is the zero-extended 64-bit value stored at 64-bit-aligned @i{c_addr}.""
+u = *(UOctabyte *)c_addr;
+
+sx@ ( c_addr -- n )	gforth s_x_fetch
+""@i{u} is the sign-extended 64-bit value stored at 64-bit-aligned @i{c_addr}.""
+n = *(Octabyte *)c_addr;
+
+be-x! ( w c_addr -- )	gforth b_e_x_store
+""Store the bottom 64 bits of @i{w} at @i{c_addr} in big endian format.""
+c_addr[0] = w >> 56;
+c_addr[1] = w >> 48;
+c_addr[2] = w >> 40;
+c_addr[3] = w >> 32;
+c_addr[4] = w >> 24;
+c_addr[5] = w >> 16;
+c_addr[6] = w >> 8;
+c_addr[7] = w;
+
+le-x! ( w c_addr -- )	gforth l_e_x_store
+""Store the bottom 64 bits of @i{w} at @i{c_addr} in big endian format.""
+c_addr[7] = w >> 56;
+c_addr[6] = w >> 48;
+c_addr[5] = w >> 40;
+c_addr[4] = w >> 32;
+c_addr[3] = w >> 24;
+c_addr[2] = w >> 16;
+c_addr[1] = w >> 8;
+c_addr[0] = w;
+
+be-ux@ ( c_addr -- u )	gforth b_e_u_x_fetch
+""@i{u} is the zero-extended 64-bit big endian value stored at @i{c_addr}.""
+u = (((Cell)(c_addr[0]) << 56) |
+     ((Cell)(c_addr[1]) << 48) |
+     ((Cell)(c_addr[2]) << 40) |
+     ((Cell)(c_addr[3]) << 32) |
+     ((Cell)(c_addr[4]) << 24) |
+     ((Cell)(c_addr[5]) << 16) |
+     ((Cell)(c_addr[6]) << 8) |
+     ((Cell)(c_addr[7])));
+
+le-ux@ ( c_addr -- u )	gforth l_e_u_x_fetch
+""@i{u} is the zero-extended 64-bit little endian value stored at @i{c_addr}.""
+u = (((Cell)(c_addr[7]) << 56) |
+     ((Cell)(c_addr[6]) << 48) |
+     ((Cell)(c_addr[5]) << 40) |
+     ((Cell)(c_addr[4]) << 32) |
+     ((Cell)(c_addr[3]) << 24) |
+     ((Cell)(c_addr[2]) << 16) |
+     ((Cell)(c_addr[1]) << 8) |
+     ((Cell)(c_addr[0])));
 
 \+
 \+
-
 \g peephole
 
 \+peephole
@@ -2870,6 +2658,36 @@ a_addr = groups;
 
 \+
 
+\g primitive_centric
+
+\ primitives for primitive-centric code
+\ another one is does-exec
+
+abi-call	( #a_callee ... -- ... ) gforth-internal abi_call
+/* primitive for compiled ABI-CODE words */
+abifunc *f = (abifunc *)a_callee;
+Float *fp_mem = fp;
+sp = (*f)(sp, &fp_mem);
+fp = fp_mem;
+
+;abi-code-exec ( #a_cfa ... -- ... ) gforth-internal semi_abi_code_exec
+/* primitive for performing ;ABI-CODE words */
+Float *fp_mem = fp;
+semiabifunc *f = (semiabifunc *)DOES_CODE1(a_cfa);
+Address body = (Address)PFA(a_cfa);
+sp = (*f)(sp, &fp_mem, body);
+fp = fp_mem;
+
+lit-execute	( #a_addr -- )	new	lit_execute
+/* for ;code and code words; a static superinstruction would be more general, 
+   but VM_JUMP is currently not supported there */
+#ifndef NO_IP
+ip=IP;
+#endif
+SUPER_END;
+VM_JUMP(EXEC1((Xt)a_addr));
+
+
 \g static_super
 
 ifdef(`STACK_CACHE_FILE',