We had a build today that did not re-use its components when we expected
it to. I looked at the logs, but didn't have any information to help
figure out why this happened.
These log statements should help for the next time.
The checksum of the build.log file changes every time because it is (in
part) based on the timestamp that the file was written (at the time of
the test run).
Change of configuration to enable logging by default caused the test to
fail because mocked object had no id. This change fixes the test and
sets up cleanup for logs
Multiple build threads would try to write to the same distgit-clone-wrapper
location, which could result in ETXTBSY when one thread tried to execute
a script that another was writing to or vice-versa.
Instead of using a wrapper script, use a sh -c invocation for distgit_get.
https://pagure.io/fm-orchestrator/issue/592
There's no good reason to have artifacts in both the thread-specific results directory
and in the main results directory, so just move them across instead of copying them.
https://pagure.io/fm-orchestrator/issue/591
createrepo_c recursively looks for RPMs in the target directory and
subdirectories, meaning that it was finding RPMs in the thread-specific
directories. Pass in an explicit list of RPMs at the toplevel.
https://pagure.io/fm-orchestrator/issue/591
If the koji_enable_content_generator is configured to True koji instance
has to be configured as follows:
koji call addBType module
koji grant-cg-access <mbs-user> module-build-service --new
Without the above configuration MBS will start failing the builds
Both pungi and dnf use the new multidocument format, with the
old one not really being supported by anything at this point.
Let's be compatible.
This change requires modulemd-1.2.0+.
Signed-off-by: Petr Šabata <contyk@redhat.com>
