Intel graphics driver test day tomorrow (Thursday March 12th)

Source

This is a big one, folks!

The Intel graphics driver is one of the most important in a distribution, as many many people use Intel-based graphics hardware. There are extensive changes coming to the intel driver in Fedora 11 - see the feature page for some details. Tomorrow - apologies for the short notice - we will be having a Test Day for the intel driver, in #fedora-qa on Freenode IRC as usual (see how to use IRC). You will, unfortunately, need Rawhide to test - there's too many changed components to backport to Fedora 10 for testing. Rawhide isn't installable today. It may work tomorrow, but you'd be best off installing Fedora 10 or Fedora 11 Alpha and then updating to Rawhide via yum (see this known issue for upgrading to Rawhide from Alpha).

We really really need wide testing of the new Intel driver stuff for 11, to make sure it works on all (or at least most...) hardware. If you have an Intel video adapter, please ideally come out to the test day tomorrow, but even if you can't, take a look at the Wiki page, run through the tests listed there, and submit your results - it's all documented on the page and it should be easy to step through. We need you to help make sure this is working well. Thanks a lot.

Comments

macemoneta wrote on 2009-03-11 22:01:
There's no stability test? X locks-up up to several times a day here (on G45/X4500HD, bug 473347). There's no test case for video tearing either. Is that just a known issue at this point? For the glxgears test, should the test be considered a failure if the reported frame rate is significantly below F10's frame rate? For example, rawhide is currently running about 50% of the F10 frame rate for me. It just seems like a test day may be a little premature for code in this state.
katzj wrote on 2009-03-11 22:14:
This is the perfect example of a test day where someone building a livecd and making it available can help a lot -- then people just have to download and try it out without removing the install they already have
adamw wrote on 2009-03-11 22:42:
katzj: yep, I agree, and it would be lovely to be able to do that. At present, though, Rawhide's screwed and live CD generation is apparently screwed, so it doesn't stand a high chance of happening :) This slot opened up very late - Friday, I think it was - so we've been scrambling to get an event to fill the slot. I was already working with Dave Airlie and others to plan Intel and ATI test days, and they wanted to have the Intel one relatively early so things wouldn't be too frozen for them to make significant changes in response to problems discovered during testing, so they wanted to have the Intel one in this slot. We've been scrambling to get everything together for it since then, it's all a bit more last-minute than it usually would be for one we know is happening well in advance. macemoneta: see above - this is being organized at the request of the devs, they feel this is a good time to be having the testing. Stability tests are a bit difficult to codify and run as test cases. A test case should be a concrete procedure with a clear end result, which doesn't really work with stability. How many crashes in what timeframe constitutes a 'pass', and how many constitutes a 'fail'? In general I'd sort of expect people who run the code and notice instability to pipe up about it. That doesn't seem to be a problem, usually. :) Video tearing - if you mean really *video*, it's implicit in the video playback test, although perhaps I'll check with Kristian if they're expecting tear-free playback, and make it explicit if so. Frame rate - that would be a good thing to ask in the IRC channel (this is why test days are interactive :>). Additionally, glxgears is a terrible benchmark that tells you nothing useful. Get numbers from some proper 3D application - a decently intensive game, or at least ppracer, or something.
katzj wrote on 2009-03-12 12:45:
Live image built... http://katzj.fedorapeople.org/rawhide-i686-20090312.0.iso