Bug#927313: parsinsert: probably broken on armhf, failing autopkgtests in Ubuntu
Hi Andreas, On Thu, Dec 05, 2019 at 07:13:38PM +0100, Andreas Tille wrote: > On Fri, Jul 26, 2019 at 02:52:34PM -0700, Steve Langasek wrote: > > > So your suggestion is that for future uploads we should run the test > > > written in Debian as autopkgtest as a test for the upstream code. > > Yes, this would catch the problem earlier and fail to build the package on > > architectures where it is broken. Then you could request the old binaries be > > removed from the archive. > I've implemented this and according to >https://buildd.debian.org/status/package.php?p=parsinsert > all architectures are passing. Am I missing something? Well, this package version appears to also build now in Ubuntu on arm64 and armhf but not on ppc64el: https://launchpad.net/ubuntu/+source/parsinsert/1.04-6 so it looks like there has perhaps been improvement in the cross-architecture compatibility of this package since the bug was initially filed. > > > > Note that these tests also fail on arm64, i386, and ppc64el in Ubuntu, > > > > suggestings the packages are also broken there, but none of these are > > > > regressions. > > > > > Thanks a lot for these hints > > > > My pleasure! > > :-) > > Kind regards > > Andreas. > > -- > http://fam-tille.de -- Steve Langasek Give me a lever long enough and a Free OS Debian Developer to set it on, and I can move the world. Ubuntu Developer https://www.debian.org/ slanga...@ubuntu.com vor...@debian.org signature.asc Description: PGP signature
Bug#927313: parsinsert: probably broken on armhf, failing autopkgtests in Ubuntu
Control: tags -1 unreproducible On Fri, Jul 26, 2019 at 02:52:34PM -0700, Steve Langasek wrote: > > > So your suggestion is that for future uploads we should run the test > > written in Debian as autopkgtest as a test for the upstream code. > > Yes, this would catch the problem earlier and fail to build the package on > architectures where it is broken. Then you could request the old binaries be > removed from the archive. I've implemented this and according to https://buildd.debian.org/status/package.php?p=parsinsert all architectures are passing. Am I missing something? > > > Note that these tests also fail on arm64, i386, and ppc64el in Ubuntu, > > > suggestings the packages are also broken there, but none of these are > > > regressions. > > > Thanks a lot for these hints > > My pleasure! :-) Kind regards Andreas. -- http://fam-tille.de
Bug#927313: parsinsert: probably broken on armhf, failing autopkgtests in Ubuntu
On Fri, Jul 26, 2019 at 10:28:31PM +0200, Andreas Tille wrote: > Hi Steve, > thanks a lot for this bug report. > On Wed, Apr 17, 2019 at 04:18:30PM -0700, Steve Langasek wrote: > > [...] > > > > (https://objectstorage.prodstack4-5.canonical.com/v1/AUTH_77e2ada1e7a84929a74ba3b87153c0ac/autopkgtest-disco/disco/armhf/p/parsinsert/20190206_231724_739fb@/log.gz) > > Investigation suggests this is a regression caused by toolchain changes that > > have resulted in a broken armhf binary build in 1.04-4: there are clearly no > > changes to the testsuite between -3 and -4, the -3 binary still passes the > > testsuite with current libraries, and a no-change rebuild of -3 fails the > > same way. > I need to admit that from a parsinerst maintainers point of view I have > no idea what to do. > > Since Debian does not run autopkgtests on !amd64, I would strongly recommend > > running these tests at build time as well, to avoid shipping broken binaries > > on other architectures. > So your suggestion is that for future uploads we should run the test > written in Debian as autopkgtest as a test for the upstream code. Yes, this would catch the problem earlier and fail to build the package on architectures where it is broken. Then you could request the old binaries be removed from the archive. > > Note that these tests also fail on arm64, i386, and ppc64el in Ubuntu, > > suggestings the packages are also broken there, but none of these are > > regressions. > Thanks a lot for these hints My pleasure! Cheers, -- Steve Langasek Give me a lever long enough and a Free OS Debian Developer to set it on, and I can move the world. Ubuntu Developer https://www.debian.org/ slanga...@ubuntu.com vor...@debian.org signature.asc Description: PGP signature
Bug#927313: parsinsert: probably broken on armhf, failing autopkgtests in Ubuntu
Hi Steve, thanks a lot for this bug report. On Wed, Apr 17, 2019 at 04:18:30PM -0700, Steve Langasek wrote: > [...] > > > (https://objectstorage.prodstack4-5.canonical.com/v1/AUTH_77e2ada1e7a84929a74ba3b87153c0ac/autopkgtest-disco/disco/armhf/p/parsinsert/20190206_231724_739fb@/log.gz) > > Investigation suggests this is a regression caused by toolchain changes that > have resulted in a broken armhf binary build in 1.04-4: there are clearly no > changes to the testsuite between -3 and -4, the -3 binary still passes the > testsuite with current libraries, and a no-change rebuild of -3 fails the > same way. I need to admit that from a parsinerst maintainers point of view I have no idea what to do. > Since Debian does not run autopkgtests on !amd64, I would strongly recommend > running these tests at build time as well, to avoid shipping broken binaries > on other architectures. So your suggestion is that for future uploads we should run the test written in Debian as autopkgtest as a test for the upstream code. > Note that these tests also fail on arm64, i386, and ppc64el in Ubuntu, > suggestings the packages are also broken there, but none of these are > regressions. Thanks a lot for these hints Andreas. -- http://fam-tille.de
Bug#927313: parsinsert: probably broken on armhf, failing autopkgtests in Ubuntu
Package: parsinsert Version: 1.04-4 Severity: important User: ubuntu-de...@lists.ubuntu.com Usertags: origin-ubuntu disco Dear maintainers, The parsinsert package passed its autopkgtests in Ubuntu on armhf as of 1.04-3; but with the 1.04-4 update, the tests now fail: [...] Rank Matches: _Precision__Recall__ Domain: 957 938 ( 98.01%) 938 957 ( 98.01%) Phylum: 957 83 ( 8.67%) 83 957 ( 8.67%) Class: 957 83 ( 8.67%) 83 957 ( 8.67%) Order: 9150 ( 0.00%)0 915 ( 0.00%) Family: 8760 ( 0.00%)0 876 ( 0.00%) Genus: 7060 ( 0.00%)0 706 ( 0.00%) Species: 530 ( 0.00%)0 53 ( 0.00%) Insert Time = 82, 42014 per hour Process Completed: 102 sec incorrect result autopkgtest [23:16:43]: test run-unit-test: ---] autopkgtest [23:16:46]: test run-unit-test: - - - - - - - - - - results - - - - - - - - - - run-unit-testFAIL non-zero exit status 1 [...] (https://objectstorage.prodstack4-5.canonical.com/v1/AUTH_77e2ada1e7a84929a74ba3b87153c0ac/autopkgtest-disco/disco/armhf/p/parsinsert/20190206_231724_739fb@/log.gz) Investigation suggests this is a regression caused by toolchain changes that have resulted in a broken armhf binary build in 1.04-4: there are clearly no changes to the testsuite between -3 and -4, the -3 binary still passes the testsuite with current libraries, and a no-change rebuild of -3 fails the same way. Since Debian does not run autopkgtests on !amd64, I would strongly recommend running these tests at build time as well, to avoid shipping broken binaries on other architectures. Note that these tests also fail on arm64, i386, and ppc64el in Ubuntu, suggestings the packages are also broken there, but none of these are regressions. -- Steve Langasek Give me a lever long enough and a Free OS Debian Developer to set it on, and I can move the world. Ubuntu Developer https://www.debian.org/ slanga...@ubuntu.com vor...@debian.org signature.asc Description: PGP signature