PR#163: De-hardcode py_SOVERSION in the files section - rpms/python3

Nice. Where did this bite you?

Metadata Update from @churchyard:
- Pull-request tagged with: merge - rebase - CI

4 years ago

vstinner commented 4 years ago

I never understood the SOVERSION thing. Who decides to increase SOVERSION?

Heads up, I'm gonna use this PR to test https://fedoraproject.org/wiki/Zuul-based-ci

recheck

Build failed.

rpm-scratch-build : FAILURE in 3m 00s
rpm-linter : SKIPPED
rpm-rpminspect : SKIPPED (non-voting)
rpm-test : SKIPPED

@tdecacqu Not sure why the build has failed, there is some weird error on https://fedora.softwarefactory-project.io/zuul/build/d67e8249ed4241f991449047e4811830

recheck

i'm not sure why it failed, the job worked recently for python-gear ( https://fedora.softwarefactory-project.io/zuul/builds?job_name=rpm-scratch-build ). I've requested a node hold, i'll try to debug more using ssh access.

Build failed.

rpm-scratch-build : FAILURE in 2m 40s
rpm-linter : SKIPPED
rpm-rpminspect : SKIPPED (non-voting)
rpm-test : SKIPPED

Ok, actually the weird error seems to be just mock failing to cleanup, the real issue seems to be error: Bad source: /builddir/build/SOURCES/Python-3.8.1.tar.xz: No such file or directory

And it seems to be caused by
error: line 674: Unclosed %if which prevent spectool to download the sources

Or something else I don't know. The procedure to download the source doesn't seems to work with the python3.spec file.

@churchyard would you see why this task doesn't work: https://pagure.io/zuul-distro-jobs/blob/master/f/roles/mock-srpm-build/tasks/main.yaml#_6 . If so, we could open a PR on zuul-distro-jobs, and test it here using Depends-On

That might be https://pagure.io/rpmdevtools/issue/38

But in order to test the Fedora environment, the job should not try to get sources from urls, it should use fedpkg sources or fedpkg srpm. That's what happens in Fedora. The URL might even be missing in some cases.

Edited 4 years ago by churchyard

Ok we'll merge this https://pagure.io/fedora-zuul-jobs-config/pull-request/20 to make use of fedpkg to build the srpm.

recheck

Build failed.

rpm-scratch-build : TIMED_OUT in 30m 31s
rpm-linter : SKIPPED
rpm-rpminspect : SKIPPED (non-voting)
rpm-test : SKIPPED

Our builds take usually up to 2 hours on arm, but if we can limit the architectures, it can be much faster.

Yes we can maybe start by limiting on x86_64 only. After all, on the CI we provide, AFAIK we won't be able to validate the build with functional tests (rpm-test job (tests/tests.yml)) on something else than x86_84.

I'll change the fedpkg-rpm-build role to handle a list of architectures in case of scratch build. And set x86_64 as the default arch in that list.

recheck

Build failed.

rpm-scratch-build : FAILURE in 21m 05s
rpm-linter : SKIPPED
rpm-rpminspect : SKIPPED (non-voting)
rpm-test : SKIPPED

It seems the build on koji for x86_64 failed on a test: https://fedora.softwarefactory-project.io/zuul/build/bc5d251efc9e4b00800ef7b613c0a58f/log/job-output.txt#220

FAIL: test_add_file_after_2107 (test.test_zipfile.StoredTestsWithSourceFile)

also note that the artifacts retrieval from koji in case of failing build is not working well. The fedpkg-build role (https://pagure.io/zuul-distro-jobs/blob/master/f/roles/fedpkg-build/tasks/main.yaml#_22) will need some improvements to make the koji build logs available in the zuul job result page.

That failure is known to us and it's a new problem in rawhide. Will get back to this once we fix it.

recheck