On Wed, Aug 10, 2011 at 02:47:51PM -0700, Brian Murray wrote: > On Thu, Aug 04, 2011 at 07:28:24PM +0200, Sebastien Bacher wrote: > > On jeu., 2011-08-04 at 07:36 -0700, Brian Murray wrote: > > Do you have datas that showing that most of the bugs are coming from > > stable versions? I would rather think that most of the useful technical > > feedback is coming from unstable versions (we do turn apport off on > > stable series as well), stable user tend to report feedback (things they > > like or not, design issues, etc) rather than bugs than benefit from > > apport informations usually. > > With regards to the volume of bug reports - since Maverick has been > released on 2010-10-10 there have been 8572 bugs reported using apport. > The same number for Lucid since 2010-04-29 is 17,949. I'd tell you the > number for Natty but Launchpad times out. However, with the two data > points we have I think there are enough bugs reported about stable > releases using apport to justify an SRU for an updated package hook in > most situations.
I can attest we do get a lot of bug reports post-release; X crashes and freezes in the release are of particular interest to me - they're high priorities to do SRUs for if we can pinpoint the problem and if it's sufficiently widespread. But Seb hits on a key phrase - "useful technical feedback is coming from unstable versions". Post-release the volume of reports goes up but quality seems to go way down, so the value of post-release bug reports is a lot less to me than the pre-release ones. Still, with a well crafted apport hook you hardly need the user to write anything (and in fact I notice with GPU freezes, many people don't). Usually we just need to know answers to a few basic questions; I'm experimenting with having the apport hook ask those questions multiple choice, since few users think to provide the answers upfront - so far seems to be working well. Like Brian mentioned, I also am trying to build logic into the apport hooks to detect situations where reports would not be wanted (hardware we don't support, etc.), and have apport kick out early in those cases. (Perhaps having apport give the user some helpful direction, if it can be done without becoming too irritating.) In my mind the only issue with leaving apport automatic bug collection turned on post-release is that it could result in excessive volumes of dupe bug reports, and that's why I tend to favor the idea of aggregating them in a crash database. Bryce -- ubuntu-devel mailing list ubuntu-devel@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel
