[Kde-pim] Re: Frequent crashes in EntityTreeModel and other library models

[David Jarvie]

On Tue, November 23, 2010 8:05 am, Stephen Kelly wrote:
> David Jarvie wrote:
>
>> Hi Steve,
>>
>> I attach two example calendars which caused startup crashes. I
>> configured
>> a single Active Alarms calendar as the file /tmp/x.ics. The only other
>> calendar configured was an Archived Alarms calendar, /tmp/arch.ics,
>> which
>> was empty. I've attached the KAlarm config file just in case it matters
>> (although the collection IDs in it would need changing).
>>
>> Attachment x1.ics: crashed multiple times in succession on startup
>> before
>> finally starting successfully. I started KAlarm after the "daily" alarm
>> was due, but before the "display" alarm was due. On one occasion I ran
>> KAlarm under gdb, but the crash trace was useless. In no case did I get
>> a
>> crash dialog.
>>
>> Attachment x2.ics: also crashes most times on startup, again without a
>> crash dialog. This is actually the updated x1.ics calendar after the
>> "daily" alarm triggers. Both alarms were due the next day when KAlarm
>> started. Again, running under gdb the crash trace was useless.
>>
>> If KAlarm is built in debug mode (as I'm sure it will be), you may find
>> the --test-set-time command line option useful to fake the system time.
>> To
>> replicate what I did (after setting up the Akonadi calendars), start
>> KAlarm as follows:
>>
>> kalarm --test-set-time "2010-11-22-22:45 UTC"
>>
>> Thanks,
>> David.
>>
>
> It crashes on start up for me without my Qt patch. After patching it does
> not crash on start. Without the patch it should crash every time. There's
> an
> assert that gets hit.
>
> I presume you don't build your own Qt, your Qt is in release mode, and
> asserts are compiled out, and you're getting a segfault somewhere else
> later
> (sometimes).
>
> Anywhere you have reimplemented filterAcceptsColumn, comment out the
> implementation and return true (if that doesn't have any serious side-
> effects). Then see if you get any crashes anywhere else in the app.

Thanks, Steve. That's reasssuring to know, although I just hope that the
Qt bug fix actually gets applied in time for KDE 4.6 - it's not
immediately obvious how to avoid the filterAcceptsColumn calls, although
there might be a way round them if I think about it.

I did compile Qt, but rather stupidly not in debug mode. I've been
thinking for a while that I should recompile it for debug - this should
give me the motivation to get round to doing it.

-- 
David Jarvie.
KDE developer.
KAlarm author - http://www.astrojar.org.uk/kalarm

_______________________________________________
KDE PIM mailing list kde-pim at kde.org
https://mail.kde.org/mailman/listinfo/kde-pim
KDE PIM home page at http://pim.kde.org/