On Thu, Mar 19, 2020 at 10:16:32AM +0100, Paul Gevers wrote: >... > On 25-02-2020 15:56, Michael Banck wrote: >... > > On Mon, Feb 24, 2020 at 08:57:40PM +0100, Paul Gevers wrote: > >> https://ci.debian.net/data/autopkgtest/testing/arm64/b/bagel/4258642/log.gz > >> > >> running test case 'hf_sto3g_relfci_gaunt'... terminate called without an > >> active exception > >> /tmp/autopkgtest-lxc.q3t4znog/downtmp/build.F1y/src/debian/tests/testsuite.sh: > >> line 85: 15487 Aborted BAGEL test/${testname}.json > > >> ${testname}.out > >> /tmp/autopkgtest-lxc.q3t4znog/downtmp/build.F1y/src/debian/tests/testsuite.sh: > >> fork: retry: Resource temporarily unavailable > >> FAILED. > > > > This looks like the arm64 authobuilder has not enough memory or > > something? > > autopkgtest don't run on the autobuilders, but I the message is of > course still relevant. The arm64 workers have plenty of memory, so if > the memory is exhausted, the test should implement a check and fail > gracefully if there isn't enough to run the test. Interestingly, in the > logs I check I see different tests and different amount of tests fail. > Are the tests run in parallel? Could this be turned off (at least to see > if things become more stable) or the amount of parallelism be reduced?
The results from accidental builds in recent months look good: https://ci.debian.net/packages/b/bagel/unstable/arm64/ Considering the fact that noone ever understood what the problem was, could you re-enable the autopkgtest to see whether any problems are still present? > Side note which warrants it's own bug; bagel (at least the autopkgtest) > needs an update due to the new mpich (seen both on arm64 and amd64): > running test case 'hf_sto3g_relfci_coulomb'... BAGEL: error while > loading shared libraries: libmpichcxx.so.12: cannot open shared object > file: No such file or directory >... This was a bug in libmpich12 that was fixed/reverted in April. > Paul cu Adrian