action #60029

OBS builds of os-autoinst fail for SLE12SP5 but not for SLE12SP3-4

Added by okurz 5 months ago. Updated 2 months ago.

Status:BlockedStart date:19/11/2019
Priority:NormalDue date:
Assignee:mkittler% Done:


Category:Concrete Bugs
Target version:Done


#1 Updated by mkittler 5 months ago

  • Status changed from New to In Progress
  • Assignee set to mkittler

I disabled netcdf support in ffmpeg and it builds now. As far as I know it is only used for the "sofalizer" filter which we don't need (the current version of ffmpeg uses libmysofa for this). OpenCV is currently building against the new ffmpeg. That means we could actually remove the failing hdf and netcdf packages if everything works in the end.

#2 Updated by mkittler 5 months ago

  • Target version set to Current Sprint

It builds now for x86_64.

However, for aarch64 and ppc64le there's an additional problem which seems to be a product/OBS bug:

If these architectures are required in production we could workaround this by also disabling libva support because I doubt we make use of libva ("Video Acceleration (VA) API for Linux") within openQA.

Should I remove netcdf and related packages such as hdf from our SLE12 devel repo? These are no longer used and mostly failing.

#3 Updated by okurz 5 months ago

By now all our production instances for o3 and osd run Leap and not SLE so we would not even need the repos for that purpose. You can of course remove packages which are neither needed by os-autoinst nor openQA itself.

#4 Updated by mkittler 4 months ago

  • Status changed from In Progress to Resolved
  • Target version changed from Current Sprint to Done

Then I'm closing the issue here because it is not worth adding a workaround for it and we also still have the Bugzilla ticket. I removed the unused hdf and netcdf packages.

#5 Updated by okurz 4 months ago

  • Status changed from Resolved to In Progress

I am confused now. Build failures do not go away by closing a ticket. You certainly did a step in the right direction to fix the build and a valid alternative might be to just remove SLE repos from devel:openQA completely when we agree that this is the right way to go – I personally don't think so – but as long as os-autoinst fails to build for SLE12SP5 (currently for aarch64 and ppc64le) I would not consider the ticket done. Did I misunderstand you?

#6 Updated by mkittler 4 months ago

  • Status changed from In Progress to New

I wouldn't remove the SLE 12 SP5 repo because I have referenced it in the Bugzilla ticket as an example. I've just closed the ticket because I thought one ticket for the issue is enough and our failing build is just a symptom of the issue tracked in If you want to keep it open although we don't even use the repo I'm fine with that but I've set it back to 'New' because I'm currently not working on it anymore besides keeping track of

#7 Updated by okurz 4 months ago

  • Status changed from New to Blocked

ah, I see. For this we can use "Blocked". Why not "Resolved"? Because someone might set the bug in bugzilla to "RESOLVED FIXED" or maybe also "WONTFIX" which is all acceptable for the bug per se but it does not necessarily fix the problem for us which is what the subject line tells: As long as os-autoinst does not build on SLE12SP5 for our usual architectures I see the as not "Resolved". Of course there are always alternative so we could delete the repos, do not support SLE12 and call that our "Resolution" as well :)

#8 Updated by okurz 2 months ago

discussion in bug still ongoing …

Also available in: Atom PDF