Hi Andreas, (adding firstname.lastname@example.org to cc: and keeping full quote for the benefit of those reading this via that list...) On Sonntag, 21. Juni 2015, Andreas Beckmann wrote: > (please keep me Cc:ed, I'm not subscribed) > > Hi, > > while your are doing rebuilds to check reproducibility, maybe you could > also check that the packages build in a path that contains all chars > that are allowed in version numbers. IIRC I had seen failures to do so > twice quite recently, one was beignet, the other I forgot. > > You usually run into such problems when doing backports or stable > updates since the version number suddenly gains '+' and/or '~'. > And then you need to work around this for backports/stable .... > > As I understand you don't test reproducability with changing the build > directory (seems to be embedded in too many places), so this change > would need to be done for both rebuilds. > > Putting something like this in an A hook script (after > reproduciblebuilds_user) seems to work fine with pbuilder: > > mkdir -p /tmp/build > mv /tmp/buildd /tmp/build/pkg-1:2.3+4~5-6 > ln -s build/pkg-1:2.3+4~5-6 /tmp/buildd > > it does not break pbuilder's assumption to build in /tmp/buildd, but all > attempts to get the absolute buildroot path will contain 1:2.3+4~5-6 while I do think this is a worthwhile QA test, I don't think we should test this on the reproducible.debian.net infrastructure, for two reasons: a.) I dont think we should hunt each and every bug type, just because we have a tool to make many bugs visible - because these other bugs distract from our primary mission, which is reproducible builds. b.) in the future we will want to do rebuilds in the same pathes as done on the buildds and I doubt they will use such "funny" pathes. Also Debian QA already has a tool to do such tests: archive wide rebuilds as documented in https://wiki.debian.org/qa.debian.org/ArchiveTesting - so I think that's the right tool for this job. cheers, Holger
Description: This is a digitally signed message part.