Hi, On Mon, Dec 19, 2011 at 05:40:35PM +0000, Edd Barrett wrote: > I looked into what would need to be done to package separately. It still > needs metapost and luatex, which makes it a pain to separate.
It could just depend (run- and/or build-, whatever is needed) on the texlive packages containing the metapost and luatex stuff. > Making ls-R at fake time might be possible, but how would you deal with > different subsets being installed? First, the (currently) three texlive-texmf_* packages shouldn't share anything searched by libkpathsea (i.e. any of the directories in share/texmf-* that contain ls-R files). If -main, -full and -docs would use different hierarchies, every package could use its own static ls-R file created at fake time. I don't know how much changed in the web2c file search magic during the past 20 years, but I'd be surprised if this couldn't be done by tweaking texmf.cnf a little bit, even if it references some directories which only exist if certain packages like texlive-texmf_full are installed. > As long as we make sure external ports > only ever install into texmf-local, we can dump a "full" ls-R in for texmf > and texmf-dist and hash only texmf-local at install time when it changes? Yes, this should be possible. If all else fails, just run ls -R on texmf-local with an @exec entry. Or, if libkpathsea really needs the silly comment on top of the ls-R file: { echo '% ls-R -- filename database for kpathsea; do not change this line.'; ls -R %D/share/texmf-local } > %D/share/texmf-local/ls-R > Sadly, IIRC, it looks like context uses a single database file under > texmf-var, although i would have to check this (currently on a train). Be careful, trains are dangerous. Our leader kim jong il died on a train. Ciao, Kili