Panic mode
Today’s big surprise news at work was that we were going to make a sort of Fedora Core 6 test 3.5. Test 3 had a lot of installation problems, but we got most of those sorted out. However, we want to make sure they get extra testing before FC6 final goes out so it was decided to release this thing that’s not called test 4. Of course, I didn’t find out about this until I got to work today. And even after I got there and made some fixes, it was too late because the tree had been built the night before. I love our communication sometimes.
The announcement was made by our very own Will Woods:
However, that didn’t stop us from going into full panic mode trying to get last minute fixes in for the real release in a couple weeks. I first got bogged down working on a RHEL4 errata for system-config-kickstart. Those things are always a real pain to write up, but thankfully the new errata tool is a lot more simple than the old one. Finishing that off got me through to lunch time. Then after that, I had to work on a pair of crazy firstboot bugs. They’re only happening when starting up with rhgb, and there’s been a lot of work going on there the past couple days so I didn’t get too deep into it. I tried an X update to see if that would fix things but no luck. Hopefully the X and desktop guys will come up with something next week.
After that, I had a pykickstart traceback to fix. I introduced this problem yesterday when I built a new pykickstart to enable an option I’d missed back when I wrote the thing. We only use the option in the loader, but pykickstart still needs to understand it or it’ll throw out the kickstart file as invalid. So I just had to fix up some syntax there and make my tests less dumb.
For the big finish, I started working on various kickstart bugs that Will found in his automatic tests. The first was a stupid one I broke a couple days ago which made kickstart installs always graphical. Well, that’s clearly wrong but it was also quite simple to fix. The second was a lot more tricky. It took us an hour or so to track down the exact set of parameters that was causing tracebacks on text mode kickstart installs. It turns out that if you have the xconfig or monitor commands in your kickstart file and do a text install via HTTP or FTP, you get a traceback. That’s because there aren’t any X libraries on the minimal stage 2 image we use for HTTP and FTP installs, so we don’t set some variables properly. The fix for this is as much policy as it is technical, so I came up with a fix and bounced it over to Jeremy to see what he thought. That carried me all the way through to 6pm, and then I decided to leave.
So it was a pretty crazy day at work looking for the answers to those several problems. I didn’t even get into the other giant stack of bugs I’ve got, the other RHEL errata, or the shared inbox of anaconda bugs. Something tells me it’s going to be a busy couple more weeks leading up to the real FC6 release. But now it’s the weekend, so I’m not allowed to think about work anymore.