Workaround in WebView...

[Alexis Ménard]

On Wed, Mar 24, 2010 at 3:35 AM, Aaron J. Seigo <aseigo at kde.org> wrote:
> On March 22, 2010, Alexis Ménard wrote:
>> There are two crashes because of this workaround.
>>
>> One in QGV is fixed and is on its way to 4.6.3
>>
>> The other is reported :
>>
>> https://bugs.kde.org/show_bug.cgi?id=227639
>
> that looks a lot like a bug in QGraphicsView as well: it's crashing in the bsp
> tree indexing.

Well the workaround trigger something that's it's hard to expect :
asking the content of the scene before the first round in the event
loop.

>
>> It would be nice if the original writer of the workaround can help the
>> guys of QtWebkit. I'm kinda busy right now.
>
> i have.
>
>> Next time when adding a workaround please open a bug report in the
>> upstream project.
>
> i did. with code showing the problem, minutes after committing the work around
> to libplasma so at least plasma-desktop wouldn't crash horribly.
>
> nor was i the only one to do so: there's a near duplicate report that's also
> been triaged. it's not quite the same crash, but has the same underlying
> reason.
>
> since we're handing out advice to each other, let's try this: in future those
> working on webkit and/or qgraphicsview might wish to test things like "let's
> open a combobox on a qwebpage that's in a qgraphicsscene" before making a
> release?

Well sorry to say that you experienced bad bugs like this one but in
the other hand QGraphicsView is the most tested framework in Qt and we
did massive refactoring without _too much_ breakage. Plasma-Mobile
wouldn't exist without those changes. At each bug reported/fixed an
autotest is added and the coverage is quite good. Concerning WebKit I
can't talk, I was not part of the implementation. But if you've
reported the bug then my bad i apologize. I was not able to find it
that's why.

>
> the number of these kinds of very basic yet critical (e.g. crash causing)
> problems that are making it into Qt releases is discouraging.
>
> this one wasn't the result of some accidental interaction between two Qt
> modules (QtWebkit and QGraphicsView), this was the result of a very poor
> implementation of QGraphicsView support inside of QtWebkit. it is/was code in
> QtWebkit that specifically uses QGraphicsView and which was full of
> assumptions that was quite plainly and obviously (just from reading the code,
> not even from testing) _wrong_.
>
> if it were simply a matter of "the plasma guys put QtWebkit and QGraphicsView
> together in a way we didn't expect" i'd be more charitable about the
> situation. but this is a crash that was introduced by Qt hackers working on
> integrating two Qt modules.
>
> --
> Aaron J. Seigo
> humru othro a kohnu se
> GPG Fingerprint: 8B8B 2209 0C6F 7C47 B1EA  EE75 D6B7 2EB1 A7F1 DB43
>
> KDE core developer sponsored by Qt Development Frameworks
> _______________________________________________
> Plasma-devel mailing list
> Plasma-devel at kde.org
> https://mail.kde.org/mailman/listinfo/plasma-devel
>