Philipp Kern <pkern@debian.org> writes: > On Fri, Jul 13, 2012 at 04:28:32PM +0200, Sébastien Villemot wrote: >> Please give back atlas 3.8.4-8 on i386. >> >> It already failed to build two times on i386 but yet I am convinced that >> it is buildable, because the diff with 3.8.4-7 (which built on i386) >> only affects maintainer scripts. Also note that version 3.8.4-8 built on >> all other release archs. Unfortunately, atlas exhibits random build >> failures as discussed in <87sjdw2tde.fsf@brouzouf.villemot.name>. >> >> gb atlas_3.8.4-8 . i386 > > Given back with gnashing teeth. Thanks for your help. Unfortunately it failed again, for the third time in a row on biber. This is probably a different issue than the known random failures (which are actually not so prevalent since atlas 3.8.4-8 built on all other arches at the first try). I successfully built atlas 3.8.4-8 in an i386 chroot on an amd64 machine, which is precisely the setup of biber IIRC. So I am rather clueless. Is there something special about biber that we should be aware of? Note that version 3.8.4-7 of atlas (which only differs from 3.4.8-8 in the maintainer scripts) built successfully on murphy, so it could be worth trying that specific buildd. Also note that 3.8.4-8 fixes two RC issues (not reported in the BTS) and implements a workaround for a 3rd one (#576972); so having it in Wheezy is valuable. For post-Wheezy, my hope is the latest major upstream release (3.10) will improve the situation. Best, -- .''`. Sébastien Villemot : :' : Debian Maintainer `. `' http://www.dynare.org/sebastien `- GPG Key: 4096R/381A7594
Attachment:
pgpPWO2AKupHq.pgp
Description: PGP signature