retitle 795300 Regression tests fail on several architectures severity 795300 normal tag 795300 +help thank you
On Fri, Aug 14, 2015 at 10:14:59AM +0200, chrysn wrote: > I'll next compare the build logs. Relevant architectures for that are > armhf, hurd-i386, i386, mips and powerpc. (Other architectures, like > arm64, just had completely odd failures like [2]). I've been comparing the following builds: Good: kfreebsd-i386 kfreebsd-amd64 s390x Bad: armhf hurd-i386 i386 mips powerpc Noted differences after some manual normalization: * Some binNMU version differences (completely to be expected) * m4_1.4.17-4 from ../archives: Yes: s390x mips i386 armhf No: powerpc kfreebsd-i386 kfreebsd-amd64 hurd-i386 (no correlation * Has libasan and libusan: Yes: powerpc i386 armhf No: kfreebsd-i386 kfreebsd-amd64 s390x mips hurd-i386 Those libraries can not be the *only* source of trouble * Architecture specific int size warnings (s390x vs powerpc, but in value.cc) Reverting the strict build test checks for now (because the packages, even with the build failures, work in the *regular* use case with hardware rendering). Tagging this as help because I'm a little out of ideas, and while this is something I'll discuss with upstream, they might again have a hard time accessing more exotic architectures. My next steps (apart from the current upload fixing the build failures by ignoring the test errors) will be reporting the issues to upstream and tracing down the discrepancies on amd64 of build-time success vs. runtime failure described in the last mail.
signature.asc
Description: Digital signature
