2008/12/5 Callum Lerwick <seg@xxxxxxxxxx>: > Which is why I proposed automated regression tests, which would catch > this. I think you have all the necessary tools in hand to start exploring this sort of regression concept. If you want to regression test go right ahead by using rpm with the appropriate cmdline options to "downgrade" to a previous version. Do filesystem diffs after individual transations and mine the data. And keep track of how long these sort of tests take on average. Just make sure you regression test all possible pre/post and trigger scenarios for every package that has pre/post and trigger scripts. Triggers are a pain in the ass.. or so I'm told. > This is what is driving me nuts. If rpm -e of a package is always valid, I don't think you can actually assume that. Though it would be something you could start regression testing and show us if it is. Start F8 gold, snapshot the filesystem, update a package to the latest update, snapshot the filesystem, remove the package, snapshot the filesystem, install the gold version, snapshot the filesystem. Do some filesystem diffing and see if you end up where you started...for every self-consistent set of updates. -jef -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
