Summary: | app-text/xetex-0.995 cannot build because of some problem with how it sets up its configuration. | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Ferris McCormick (RETIRED) <fmccor> |
Component: | Current packages | Assignee: | Joshua (RETIRED) <joslwah> |
Status: | RESOLVED INVALID | ||
Severity: | normal | CC: | tex |
Priority: | High | ||
Version: | 2007.0 | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- |
Description
Ferris McCormick (RETIRED)
2007-10-17 01:10:29 UTC
I can narrow it down as to dates, because I believe something upgraded within this time bracket destroyed the build: xetex built on 25 Sept --- good (that's the good example in that very long description). xetex built on 5 Oct --- bad (missing .fmt files, SB1000, amd64) (In reply to comment #0) > On 25 Sept. 2007, xe[la]tex worked well. And if I have it packages up and look > for a critical file, everything is happy: > > anaconda All # tar tjvf xetex-0.995.tbz2 | grep xelatex.fmt > -rw-r--r-- 0/0 3047342 2007-09-25 14:06 var/lib/texmf/web2c/xelatex.fmt > There is a problem here, or was, but it is not xetex's. The root problem is in pkg_preinst with the texhash (or a bit beyond), which can hang. This, I believe, is a timing gate problem between bash-3.2_p17 & glibc-2.5-r4 and is not reproducible with consistency. You have to abort out of it, and unfortunately that can leave an existing xetex or tetex in a horrible state. I have seen: missing .fmt or .cnf files; an empty texmf.cnf; a need to do a complete rehash; with a texhash which has forgotten which directories it cares about. For fun, I have rebuilt bash against glibc-2.6.1 and made sure that all of tex is in reasonable shape. Also, I have been playing with the proposed xetex-0.995.ebuild replacement in Bug 191815. So far, on liasis (another amd64 where the problem originally showed up yesterday), I've now been able to 5 working xetex s in a row successfully. Yesterday that would have been impossible because of a cascading problem traced back to some sort lock in a tetex utility. Just about everything in the original description is wrong because my testing methods on anaconda were inconnrect at best and destructive at worst. So, the Summary does describe a valid problem, but for now I cannot duplicate it. In any case, xetex is the victim, not the culprit, and I am marking this bug invalid for now. |