to, 2006-01-26 kello 03:02 +0200, Lars Wirzenius kirjoitti: > ma, 2006-01-16 kello 00:45 +0100, Bart Martens kirjoitti: > > I've had a look at your patches, and I didn't feel comfortable with the > > idea of adding a generated list of parent directories. I think it's > > better to prevent that the forgotten directories become forgotten. The > > patch below does that. > > I've now run piuparts with a dpkg using this patch against approximately > 1200 packages that used to pass piuparts testing, and only three failed > and none of those failures were related to dpkg. This was without > piuparts minimizing the chroot (which triggered the bug for me), because > that's how things worked before. I conclude that the patch doesn't make > dpkg broken in usual usage.
Blah, I made a mistake in my test setup, and it didn't used the patched version of dpkg, but the one from sid. So that test was about as useful as Visual Basic for writing a kernel. However, I did something else instead, to actually test the patched dpkg. I took all the packages that had passed piuparts, and ran the simple install-remove test on them and this time I really did use the patched dpkg. These tests all worked; about 6600 packages without trouble, and a few more hundred with trouble unrelated to dpkg (e.g., packages no longer existing in sid). > I'll now see how well things work with the minimzing, i.e., to really > start excercising Bart's patch. I'll report tomorrow how things work. I next ran piuparts on the packages that had failed due to the dpkg directory bug, and they all passed (except for two that failed for unrelated reasons). In this test, piuparts did not minimize the chroot. My third test was to re-run the second test *with* the minizing. This also ran without dpkg-related trouble, but did unconver a few bugs with missing dependencies (i.e., it was working exactly as it should). Summary: as far as I can see, the experimental version of dpkg plus Bart's patch to fix this bug work quite well. I'm looking forward to seeing them in unstable. -- The most difficult thing in programming is to be simple and straightforward. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
