![](https://secure.gravatar.com/avatar/565476157eb3895a9c928ffce26d988d.jpg?s=120&d=mm&r=g)
Hi all, On 05/23/2012 04:46 AM, Norbert Preining wrote:
On Mi, 23 Mai 2012, Khaled Hosny wrote:
AFAIK, context does not use texmf.cnf anymore, it used its own texmfcng.lua file. Anyway, kpathsea is not used by context at all, it has its own lua based file finder.
Hmm, if this is so, that is quite brain damaged.
It made perfect sense when at some point a few years ago kpathsea needed literally seconds to initialize itself. And later, Hans added functionality to this own code that is not present in kpathsea. Anyway...
context should at least use kpse to find texmfcnf.lua!
To help with that, Hans, Mojca and me had a bit of email exchange, and we came up with the following solution: luatex now has (as of 5 minutes ago) a new lua function in the kpse library: kpse.default_texmfcnf() which returns the value of the C preprocessor macro DEFAULT_TEXMFCNF without initializing anything else from kpathsea. The code for that (just a few lines actually) has been submitted to TL repository as well as the trunk of the luatex repository. Hans can then use the output of that function to find any texmfcnf.lua that lives alongside texmf.cnf, now using the hardcoded built-in from the executable as opposed to the hardcoded value in mtxrun that has been used upto now. Norbert, will that fix the problem? (we hope so) Of course, it means another TL rebuild and another commit of a new and better context release .. Best wishes, Taco