FUDCon Lawrence 2013 Hackfests

While the presentations at FUDCon are always interesting, one of the big things which is possible at FUDCon is to collaborate with people while we're in the same location.

Fedora Test Automation

There hasn't been a whole lot of movement on the test automation front in Fedora as of late; mostly because the people who had been working on it have been consumed with other Fedora testing activities. I recently started a thread on autoqa-devel talking about future requirements and nice-to-have features for automation. We still haven't gotten very far with figuring out how to get to the test automation promised land but I was able to have several discussions that should help us get there.

One suggestion for a user-submitted test is static analysis of python extensions. Dave Malcom first proposed this last year but we still don't have support for running it inside AutoQA. He's still working on this test and will hopefully have enough code to start exploring AutoQA integration soon.

Lucas Rodrigues was also interested in talking about what autotest can do and what we would need to implement in order to get what we want out of it. There is at least interest upstream for much of what we want to do and while we haven't finished exploring everything, I'm planning to poke at a demo implementation as I have time.

Fedora QA Changes

One of the things that Adam and I wanted to do at FUDCon was to discuss some occasional topics that come up in Fedora QA: blocker tracking aliases and group policies.

A lot of this is already being discussed on the QA lists and I expect that the changes will be done soon. No more 'fxx-alpha-accepted' tracker aliases!

Blocker Tracking App

After the blocker presentation yesterday, it feels very much like the blocker process is quite a bit more difficult to understand than it should be. We went through, wrote up a set of requirements and got some help with an initial UI sketch from Emily Dirsh. I'll be posting more about this when I have an HTML mockup done but I'm expecting to have a first implementation in place for F19.

I also had a discussion with Dennis Gilmore about better methods for communicating the non-stable builds required for TC/RC composes during freeze. It's currently an ad-hoc method where either Adam or I add a comment to the rel-eng trac ticket and Dennis re-parses that into a format he can use. We already have some spin request stuff in the blocker tracking app and I think we could get something that works much better for both QA and rel-eng without too much effort. Whether this actually happens before F19 is going to depend on the actual release date and how quickly the other changes happen, though.

EFI in KVM/QEMU Virtual Machines

After finding some upgrade bugs specific to UEFI systems during F18, I decided to start looking into using OVMF with KVM/QEMU again. I've tried doing this in the past but was never able to actually get it working well enough to use for install or upgrade testing.

I've recently been able to get the OVMF firmware working on Fedora but wasn't able to make libvirt defined VMs use the firmware while loading - it would always fall back to SeaBIOS and ignore the loader parameters.

I had the chance to speak with Cole Robinson at FUDCon and it looks like the problem is limited to F17 - following the same process under F18 works without a hitch.

I'm planning to get some builds and better instructions together soon so that we can start doing UEFI testing in virtual machines. There are licensing issues which prevent including the firmware in Fedora but the code is distributable and the process for using the OVMF firmware isn't too difficult. I'll post more when I get more coherent instructions together if someone else doesn't get it done first.

Conclusion and Final Thoughts

Overall, I think that FUDCon Lawrence was very productive and useful for me. There is never enough time to get everything done but at least progress was made and a bunch of new ideas have come forward. Now we just have to get these new ideas done :)