Hello, On trečiadienis 19 Sausis 2011 11:18:34 Julien Cristau wrote: > On Tue, Jan 18, 2011 at 02:25:34 +0200, Modestas Vainius wrote: > > Hello, > > > > via unstable: > > > > 1) kde (Depends: kde-full) and kde-core (Depends: kde-standard) > > transitional metapackages in src:meta-kde. And yes, without preinst > > hack, only appropriate debian/control changes. Hopefully I don't regret > > this (Julien knows what I'm talking about). > > I guess you've tested upgrades with that, and they go at least as well > as without those packages :) > ack. Yes, I tested. Will upload today. > > 2) krunner #607974 src:kdebase-workspace. The upstream patch > > http://tinyurl.com/6e6b24l is supposed to fix the crash. There is no time > > to confirm the fix 100% as the crash, while pretty common, is random and > > hard to reproduce reliably. The crash no longer occurs in KDE SC 4.5 (at > > least there no new reports against it upstream) which this patch is > > included in. > > > > I have already experienced the crash more than a dozen times. Whenever > > krunner goes down, it's really annoying as krunner is one of the primary > > UI interfaces for launching new applications in KDE. > > would delaying this to r1 allow to confirm the fix? It's tricky to reproduce so we can never be sure. We can only "feel" that it's fixed. For example, I tried really hard for an hour or so yesterday but I could not reproduce it. But then, when I was really in hurry, it happened to me 3 times in (almost) a row doing nothing else but trying to run kcalc! That was Murphy's law biting me very hard. Such indications are pretty typical for threading related issues. Given that everything the patch does is moves some code from the worker thread to the main GUI thread, it is very likely to be right one. After all, the code was not touched again in KDE 4.5.x cycle (past 6 months) so it should not be buggy. Therefore, I think that waiting for r1 won't do us more good as far as this bug is concerned. There are more testers in unstable rather than future stable-proposed-updates anyway and I'm definitely moving to fresher KDE soon after .0 release. I think I will upload to unstable anyway but you may choose not to unblock it (or delay unblock for some more time) if you still have some considerations. > > via t-p-u: > > > > 4) src:krusader #604196 There are reports that the included fix for this > > grave bug is wrong. Too bad I would need to do a t-p-u upload in order > > to include the supposedly right fix ( > > http://websvn.kde.org/?view=revision&revision=1169424 + > > http://websvn.kde.org/?view=revision&revision=1169519 ). I will test it > > as best as I can. > > maybe r1 as well. First of all, some background on the issue: 1) Before 1:2.2.0~beta1-2 was uploaded, upstream [1] and me [2] had some considerations about the fix because it apparently introduced a new regression as serious as a crash [3]. To make things worse, #604196 was later opened saying that the patch didn't actually fix the bug in the end. 2) QCoreApplication::processEvents() [4] is considered to be a dangerous function when called manually. It messes up control flow in unpredictable ways as it triggers processing of the event loop right away. It may even dead lock an application in the worst case scenario (event loop processing from within unfinished event handler, oops). Who knows, maybe #604196 is because of event loop deadlock. The new patch apparently nails and fixes the real cause of the hang-on-quit. Of course if there is release process related reasons to wait (e.g. it's simply too late for t-p-u at this point), we can delay it. But if we could fix it now, we should do it (it's -1 RC bug anyway, karma++ :-)). [1] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=587842#22 [2] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=587842#29 [3] http://groups.google.com/group/krusader- devel/browse_thread/thread/8f5482e3a4b2c08e/5a4495887ef76c9a see quoted text of the first message in that thread (and remaining messages). [4] http://doc.qt.nokia.com/stable/qcoreapplication.html#processEvents -- Modestas Vainius <modestas@vainius.eu>
Attachment:
signature.asc
Description: This is a digitally signed message part.