TeX 1995 July/TeX CD-ROM July 1995 (Disc 1)(Walnut Creek)(1995).ISO/tex-k/tex-k-archive.past/tex-k-archive.gz/tex-k-archive/[e-mail].…i-frankfurt.de_Fri May 6 16:58:25 1994.msg

Internet Message Format | 1994-10-11 | 4KB

Received: from riese.informatik.uni-frankfurt.de (riese.thi.informatik.uni-frankfurt.de) by cs.umb.edu with SMTP id AA20288
  (5.65c/IDA-1.4.4 for <[e-mail]>); Fri, 6 May 1994 09:34:16 -0400
Received: from lovelace.informatik.uni-frankfurt.de (lovelace.thi.informatik.uni-frankfurt.de) by riese.informatik.uni-frankfurt.de (4.1/THI-peleuck2.1)
    id AA01462; Fri, 6 May 94 14:58:25 +0200
Date: Fri, 6 May 94 14:58:25 +0200
From: [e-mail] (Peter Dyballa)
Content-Transfer-Encoding: binary
Content-Type: text/plain; charset=ISO-8859-1
Message-Id: <[e-mail]>
Received: by lovelace.informatik.uni-frankfurt.de (4.1/SMI-4.1)
    id AA12182; Fri, 6 May 94 14:59:06 +0200
To: [e-mail]
Subject: Kpathsea 1.8 and dvipsk 5.55a

Hello!

I had problems with dvipsk 5.519b when PKPATH was not defined. I told that Karl
Berry who wrote I should try to use dvipsk 5.55a, the recent version.

I did, and the success is still the same: when PKPATH is empty dvips does not
find any PK file (in a reasonable time).

I debugged dvipsk with xxgdb and came soon to line 722 in resident.c which is
reached by checkenv(1), line #829 in dvips.c. Now I present you what xxgdb
states:

GDB 4.7, Copyright 1992 Free Software Foundation, Inc...
(xxgdb) define redo
symbol-file dvips
exec-file dvips
end
(xxgdb) define aredo
symbol-file afm2tfm
exec-file afm2tfm
end
(xxgdb) (xxgdb) Breakpoint 1 at 0x1ef80: file resident.c, line 722.
(xxgdb) run
config path:   /usr/local/lib/texmf/dvips
</usr/local/lib/texmf/dvips/config.ps(r)> succeeded
</usr/local/lib/texmf/dvips/config.pscsads(r)> succeeded

Breakpoint 1, checkenv (which=1) at resident.c:722
(xxgdb) print pkpath
$1 = 0xe9a68 "/usr/local/tex/lib/fonts/PK/LaserJetIV//"
(xxgdb) print tfmpath
$2 = 0x4d930 "/usr/local/tex/lib/fonts/TFM//"
(xxgdb) print vfpath
$3 = 0x4d958 "/usr/local/tex/lib/fonts/VF//"
(xxgdb) print pictpath
$4 = 0xc220 "/usr/local/lib/texmf/texinput//"
(xxgdb) print headerpath
$5 = 0x441f0 "/usr/local/tex/lib/dvips"
(xxgdb) print figpath
$6 = 0x44218 "/usr/openwin/share/images/PostScript"
(xxgdb) next
(xxgdb) next
(xxgdb) next
(xxgdb) next
(xxgdb) next
(xxgdb) next
(xxgdb) print pkpath
$7 = 0xf7ffffdd ""
(xxgdb) print tfmpath
$8 = 0xf7ffff8e ""
(xxgdb) print vfpath
$9 = 0xf7ffffb7 ""
(xxgdb) print pictpath
$10 = 0xf7fffc4e ".:/usr/local/tex/lib/texinput//:/home/local/tex/inputs:/home/riese/thi/pete/texmacros"
(xxgdb) print headerpath
$11 = 0x441f0 "/usr/local/tex/lib/dvips"
(xxgdb) print figpath
$12 = 0xf7fffc4e ".:/usr/local/tex/lib/texinput//:/home/local/tex/inputs:/home/riese/thi/pete/texmacros"
(xxgdb)

The lines I'm stepping over are those in resident.c:

     INIT_PATH (pkpath, true, KPSE_PK_ENVS);
     INIT_PATH (tfmpath, true, KPSE_TFM_ENVS);
     INIT_PATH (vfpath, true, KPSE_VF_ENVS);
     INIT_PATH (pictpath, false, DVIPS_PICT_ENVS);
     INIT_PATH (headerpath, false, DVIPS_HEADER_ENVS);
     INIT_PATH (figpath, false, KPSE_TEX_ENVS);

-----------------------------------------------------------------------------
KPSE_PK_ENVS corresponds to "PKFONTS" or "TEXPKS"
KPSE_TFM_ENVS corresponds to "TFMFONTS" or "TEXFONTS"
KPSE_VF_ENVS corresponds to "VFFONTS" or "TEXFONTS"
=============================================================================
The conclusion is: dvips(k) gets configured correctly by parsing the config
files. Then it examines the process environment, heired from the parent shell,
and forgets almost every thing it learned before!

This isn't just ridiculous, this _is_ a bug!

Of course it's a good idea to be configured for an actual operation by the
environment, but in case the environment does not want to do that job, dvips(k)
does _not_ show a fail-safe behaviour but a fail-erroneous behaviour.

The INIT_PATH macro should leave the old contents of a string unchanged when
its corresponding environment variable is not set! This helps at least to save
space in the shell's environment and to spawn a new process

Greetings

Pete

Johann Wolfgang Goethe-Universitdt      Robert-Mayer-Stra_e 11-15
Fachbereich 20 - Theoretische Informatik  D 60325 Frankfurt am Main
(Department for Computer Science)      (Federal Republic of Germany)

email:[e-mail]