Debating future tinderbox work

I’ve been not working on the tinderbox lately because my “daily job” (which is not really daily) swamped me out badly. Since this week I’m going to London to take some days off, I’ll probably get back to the tinderbox after that.

For the next ride of the tinderbox, there is at least one thing that’s definitely going to be interesting: the new X11R7.5 release means that quite a bit of packages might not build at all since they don’t have the new includes fixed. I found one or two packages with such problems while doing Yamato’s root filesystem rebuild (after glibc 2.11 update).

There is another interesting idea that I should probably toy with: the way the tinderbox works, it tests all non-masked packages; by QA standards, those should not use the network at build time. During my world rebuild this night, network went offline, and one package failed since it tried to wget a piece of source from the network. And it’s not even the first one lately.

Thanks to the fact that my tinderbox uses containers I can easily isolate it out of the network so that it cannot access the network, and then make sure that the ebuilds trying to use the network get their access refused.

The other problem to cope with is the size of the logs and the fact that I still lack an analysis script and thus opening new batches of bugs requires a huge amount of work, especially when it comes to attaching the log and getting some information out of it.

Any suggestion on how to proceed with the tinderbox will definitely be welcome.