home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
The Datafile PD-CD 3
/
PDCD_3.iso
/
tex
/
texsrc2
/
Src
/
PROBLEMS
< prev
next >
Wrap
Text File
|
1993-02-21
|
12KB
|
375 lines
Some generic problems you might encounter (for other machine-specific
problems, see MACHINES):
* Conflicts between the declarations of read/malloc/free in the flex
scanner and include files. Solution: use the distributed
lex.yy.c-dist, generated with flex, which uses a flex.skel slightly
modified by karl.
* For DOS, you may have to change the definitions of WORDS_BIGENDIAN and
DOS in lib/c-auto.h. Here are some notes:
From: Richard Mansfield <melvin@math.psu.edu>
Date: Mon, 9 Nov 92 14:24:10 -0500
Make the changes in site.h. I don't know if I got the default
directories to work. TEXPOOL seemed to work fine (using /), but I had trouble
with TEXINPUTS. After minimal attempts to get it right, I punted
and decided to use environmental variables. In the variables I use
\ and ; as a separator as in DOS. Thinking about in some more as I write
this, my troubles could have been caused by not correctly reading tex.pool
although I don't know.
My method for getting the source files was to do
make TeX > tex.log
on the SUN and then editing tex.log to become tex.bat. I then carried
the .c and .h files home on a floppy and compiled them at home.
DOS gcc does have yacc and bison, but I did not use them.
If you want to see my exact set up. ftp to ftp.math.psu.edu,
look in directory pub/melvin/pctex and get the files read.me
and tex.zip.
* GNU sed 1.06 cannot handle the sed script in mf/convert; neither can
1.07. 1.08 and later fix it.
* GNU Make 3.57 does not pass arguments to shell scripts (like convert),
under system V. 3.56 does not have this problem, and neither do later
versions.
* On an i386 machine, GNU C 1.36 does not compile tex5.c correctly when
making a 32-bit TeX. 1.37 fixes this bug.
* On the AT&T 3b2, under System V release 3.2, cc thinks that
eject_penalty (which is -1000, defined at line 3216), is a very large
positive number. oc@vmp.com is developing a port for the 3b2; he has
contributed some (incomplete) patches; they are below.
* On a Sun 3 running 3.x, cc cannot compile mf2.c or mf4.c, because the
preprocessor puts a # directive at the end of line 1811 (or
thereabouts), instead of at the beginning. You have to manually move
it.
* On a Sun 386i running 4.0.1, cc cannot compile Metafont with -O, because it
can't compile x = x + y correctly. Don't use -O, or don't use cc.
(From morgan@ics.uci.edu.)
* Make loses heavily if you are using a networked filesystem, and the
server and the local machine have different times. You might have
this problem if `initex' gets made twice, for instance.
* On a *86 running Xenix, if you use flex you may have to add the
following to c-auto.h (although it seems to karl that configure should
be able to handle this). From wally@lpds.sublink.org.
#ifndef FLEX_SCANNER /* Flex treat unput('\n') differently from
what Xenix (and probably others) lex does.
I don't remember the problem but the
following redefinition let compile TeX
correctly -wp10/16/92. */
#undef DECLARE_YYTEXT
#define DECLARE_YYTEXT extern char yytext[];
#undef unput
# define unput(c) {yytchar=(c);if(yytchar=='\n')yylineno--;else*yysptr++=yytchar;}
#endif /* FLEX_SCANNER */
* On a 386 machine running 386/ix, cc cannot compile itex.c, according
to Reiner Petersen <reiner@db0tui62.bitnet>. Here is an ed script to
get around the problem (alternatively, you can get gcc, and compile with
that):
:
# Reiner Petersen (reiner@coma.uucp)
# Technical University of Berlin
#
ed - $1 << end-of-edscript
!echo change:
/ register integer for_end; i = 1 ; for_end = 31 ; if ( i <= for_end) do
+1,.+2p
-2,.c
register integer for_end;
register char j; /* 386/ix cc bug workaround rp */
i = 1 ;
for_end = 31 ;
if ( i <= for_end) do
j = i;
xchr [ j ] = chr ( i ) ;
while ( i++ < for_end ) ;
.
!echo to:
-8,.p
w $1
end-of-edscript
If you want, you can put this into the pipeline in ./tex/convert.
* undump doesn't work under 386/ix. Here is a diff to undump.SYS_V.c to
fix this, again contributed by reiner:
*** undump.SYS_V.c Mon Apr 20 00:57:23 1987
--- undump.c Fri Aug 19 17:55:27 1988
***************
*** 1,7 ****
/*
* undump - resurrect a core file into a running program.
*
! * for UNIX System V on a 3Bx
* that uses the Common Object File Format
*
* Author:
--- 1,7 ----
/*
* undump - resurrect a core file into a running program.
*
! * for UNIX System V on a 3Bx or Interactive 386/ix
* that uses the Common Object File Format
*
* Author:
***************
*** 9,14 ****
--- 9,19 ----
* New York University
* Tue Mar 3 13:18:25 EST 1987
*
+ * Additions for 386/ix:
+ * Reiner Petersen (reiner@coma.uucp)
+ * Technical University of Berlin
+ * Thu Aug 18 13:40:20 MET DST 1988
+ *
* Adapted from:
* Spencer Thomas's undump and the file unexec.c in GNU emacs
*/
***************
*** 15,22 ****
--- 20,33 ----
#include <sys/param.h>
#include <sys/types.h>
+
+ #ifndef ISC_386ix /* not availible in ISC 386/ix */
+
#include <sys/psw.h>
#include <sys/pcb.h>
+
+ #endif /* ISC_386ix */
+
#include <sys/signal.h>
#include <sys/fs/s5dir.h>
#include <sys/user.h>
***************
*** 37,46 ****
--- 48,66 ----
struct scnhdr dsc;
struct scnhdr bsc;
+ #ifdef ISC_386ix
+ struct scnhdr csc; /* Comment section used by ISC 386/ix */
+ #endif /* ISC_386ix */
+
long bias;
long lnnoptr;
long text_scnptr;
long data_scnptr;
+
+ #ifdef ISC_386ix
+ long comment_scnptr;
+ #endif /* ISC_386ix */
+
long symlocptr;
main(argc, argv)
***************
*** 89,94 ****
--- 109,118 ----
dsc = sc;
else if (strcmp(sc.s_name, ".bss") == 0)
bsc = sc;
+ #ifdef ISC_386ix
+ else if (strcmp(sc.s_name, ".comment") == 0)
+ csc = sc;
+ #endif /* ISC_386ix */
***************
*** 97,103 ****
if (u.u_exdata.ux_tsize != aout.tsize ||
u.u_exdata.ux_dsize != aout.dsize ||
u.u_exdata.ux_bsize != aout.bsize)
! fprintf("mismatch between %s and %s sizes\n", a_out_name,
core_name);
exit(1);
--- 121,128 ----
if (u.u_exdata.ux_tsize != aout.tsize ||
u.u_exdata.ux_dsize != aout.dsize ||
u.u_exdata.ux_bsize != aout.bsize)
! fprintf( stderr, "mismatch between %s and %s sizes\n",
! a_out_name, core_name);
exit(1);
***************
*** 122,128 ****
--- 147,159 ----
bsc.s_paddr = bsc.s_vaddr = aout.data_start + aout.dsize;
bsc.s_size = aout.bsize;
bsc.s_scnptr = 0L;
+ #ifndef ISC_386ix
bias = dsc.s_scnptr + dsc.s_size - lnnoptr;
+ #else
+ comment_scnptr = csc.s_scnptr;
+ csc.s_scnptr = dsc.s_scnptr + dsc.s_size;
+ bias = csc.s_scnptr + csc.s_size - lnnoptr;
+ #endif /* ISC_386ix */
if (fh.f_symptr > 0L)
fh.f_symptr += bias;
***************
*** 139,149 ****
--- 170,186 ----
Perror("ds write");
if (fwrite(&bsc, sizeof(bsc), 1, nfp) != 1)
Perror("bs write");
+ if (fwrite(&csc, sizeof(csc), 1, nfp) != 1)
+ Perror("cs write");
fseek(nfp, (long)text_scnptr, 0);
copy(afp, nfp, aout.tsize);
fseek(cfp, off, 0);
fseek(nfp, (long)data_scnptr, 0);
copy(cfp, nfp, size);
+ #ifdef ISC_386ix
+ fseek(afp, (long)comment_scnptr, 0);
+ copy(afp, nfp, csc.s_size);
+ #endif /* ISC_386ix */
copy_syms(afp, nfp);
fclose(nfp);
fclose(afp);
These patches for the 3b2 machine are meant to serve as a starting point
only, as they are for pre-2.992 TeX. If you work on this, contact oc@vmp.com.
#
# 3b2patch.sh, to fix web2c for 3B2s
# for TeX 2.9*
#
# First, make sure we have EXTRA.c and ITEX.c that are the
# same as the distribution versions of extra.c and itex.c
# si