Monday, 19 May 2014

Re: and upgrade crashes

On Sun, May 18, 2014 at 06:17:09PM +0200, Bjoern Michaelsen wrote:
> Hi,
> On Fri, May 16, 2014 at 03:21:36PM +0100, Matthew Paul Thomas wrote:
> > Oh dear. I've reported that as a bug. <>
> > [...]
> > Frequency is an absolute count. (Perhaps we should call it "Count"
> > instead?)
> thanks for filing the bug. I should have done that myself. My only lame excuse
> for not doing that is, given the other behaviour I saw, I wasnt sure if I
> possibly completely misinterpreted the thing. As for relabeling from frequency
> to count, Id support that, its a lot less confusing.
> > That's a shame. The database server is unreliable, but I don't
> > remember a case where it failed *and* the Web site didn't communicate
> > this with an error message.
> I probably just see more of those as LibreOffice[1] is a huge package, thus I
> trigger more demanding queries ...
> > Sure. And if the Software Updater developers think otherwise, you
> > should get in a room together until you agree on a solution. Don't
> > just shrug and blame each other from a distance. :-)
> Im not just shrugging, I have a few ideas on how to solve that (and a few other
> things) -- but I dont think we should avoid building local workarounds, esp.
> when we might reconsider the way we pack and deploy software in the next years
> anyway ... ;)
> Ill try to convey some of my lunatic ideas on the next sprint and check for
> feedback and how much they survive the scrutiny of my fellow engineers.
> > And specifically, we have no evidence that it's a problem for you.
> Well, there is still the mystery on why this particular issue wasnt too high
> ranked on 13.10, but is so on 14.04, with the bug unlikely having changed. No
> need to get to deep into this for just one issue (maybe e.g. audience of LTSes
> does use LibreOffice a lot more), but would be interesting if there are more
> cases like that.
> > On the contrary, it seems highly unlikely that anyone would write "crash
> > while installing a font" in a bug report if the button they'd actually
> > clicked was labelled "Start Upgrade".
> The scenario I assumed would be more tricky: its not click "start Upgrade" ->
> instant crash, rather click "start Upgrade" -> this dooms the running
> LibreOffice instance, but doesnt crash it immediately -> LibreOffice crashes
> way later on without any obvious trigger.
> > If Launchpad's bug statuses weren't so messed up, we could do the last
> > of these by highlighting those errors that were linked to a bug report
> > in whichever status meant "reproducible".
> Speaking of Launchpad bugs: I dont know of any way to search in
> for an Launchpad bug. I just checked the the Top 20
> stacktraces happening on LibreOffice and all but 4 never happened on the
> version currently in trusty-updates. I would love to close bugs not happening
> anymore, esp. if they a/ happened regulary/statistically significant in earlier
> versions b/ have no good reproduction scenario.

This will work for any Launchpad bug report that is a crash reported by
apport (tagged apport-crash).

I've also written a greasemonkey script that adds an "Errors" link in
Launchpad that will take you to the crash at If
anyone else is interested in that I will put it in the launchpad
greasemonkey scripts project.

Brian Murray