Fixes for buildbot code refactoring #687
Conversation
This is a fix for MariaDB@c270d1a If the line exist is should actually append to the ls_filenames list so that packages property is populated. If packages property is empty, some unexpected behaviour can occur in BuildBot like in: https://buildbot.dev.mariadb.org/#/builders/17/builds/73 where this is exception is caused by workersrcs=util.Property("packages") being none Affected builders by this bug: - windows-packages - the packages property is needed for uploading the Packages on CI. - tarball docker - the packages property is for display only
This factory is not really used in any builder defined in master-docker-nonstandard-2. This can only create confusion among contributors searching for what does the factory do for any particular bintar builder. More, and one can update by mistake this one instead of contributing patches to the one in master-docker-nonstandard, one which is actually used for : - aarch64-debian-10-bintar - amd64-centos-7-bintar at the moment of writing this patch.
This is a fix for MariaDB@3d3722a Previously this step had a hasEco(step) condition which prevented the bintar builders from triggering any eco builders. Even with this bug, the eco builders are not triggered because the name of the parent builder doesn't match anything in the ECO_BUILDERS pattern. See: https://buildbot.dev.mariadb.org/#/builders/172/builds/303 See also ecoBuilders function which renders the ECO builder names at runtime based on parentbuildername. This bug, in fact, revealed another behaviour. Currently, the presence of this trigger in the Bintar factory is unnecessary, as there is no use case according to the ECO_BUILDERS list. Therefore, my proposal is to eliminate this step from the factory.
|
@RazvanLiviuVarzaru remember that build logs on DEV (but also true on PROD) are not keeped in-definitively. So, if you think that a commit message needs the error message to be more easy to understand (which is a very good practice IMO), make sure to directly paste it in the commit message instead of linking to an URL that will probably disappear in the long term. |
Right! |
Fixes for c270d1a
and 3d3722a
@cvicentiu please see each commit message for more details.
An open MDBF will greatly help me to amend the commit messages.