texmf.cnf for BLU

Sebastian Rahtz sebastian.rahtz@computing-services.oxford.ac.uk
Sat, 11 Mar 2000 21:41:31 +0000 (GMT)

Thomas Esser writes:
 > This only makes sense, if you compile kpathsea with $VARTEXMF included
 > in the search path for texmf.cnf files (before everything else). This is
 > not the case in standard teTeX, so it is correct not to copy texmf.cnf
 > into the VARTEXMF tree.
 > It does not help to set TEXMFCNF inside some texmf.cnf file to something
 > that includes $VARTEXMF. For texmf.cnf files, the *compile time* path
 > is important (I think that we all agree on not to rely on environment
 > variables).
 > For TeX Live 5, it might be possible to do this change. One migth even
 > consider to include $HOMETEXMF.

Can you explain this a bit more? my build-time texmf.cnf says


If I changed that to


then the programs would search whatever was the value of VARTEXMF set
in the texmf.cnf at compile time, yes? if someone changes VARTEXMF at
install time, it does not have any effect, because at that stage it
has not read texmf.cnf.

Fabrice's proposal only works because he *does* set some environment

FPTEX = c:\Local\TeX
PATH = %FPTEX%\bin\win32;%PATH%
TEXMFCNF = %FPTEX%\texmf_local\web2c

and hard-wires in this texmf_local tree.

I am reluctant to follow this, because by default it forces the
original and the local trees to be on the same drive. If someone wants
to install the original on a read-only system, and have local
configuration on a writeable system, they are screwed unless they set
environment variables, which we do not want to do.

Staszek, since you are on the same system as me, can you describe
exactly how you would ideally like the TeX system a) compiled, and b)