Bug#494348: [epiphany-browser] edit->personal data->Passwords -> crash

[Ondrej Certik]

On Fri, Aug 15, 2008 at 6:21 PM, Sam Morris <sam at robots.org.uk> wrote:
> On Fri, 2008-08-15 at 18:15 +0200, Ondrej Certik wrote:
>> On Fri, Aug 15, 2008 at 5:48 PM, Sam Morris <sam at robots.org.uk> wrote:
>> > On Fri, 2008-08-15 at 17:38 +0200, Ondrej Certik wrote:
>> >> > Please can you provide the stack trace gathered by bug-buddy when
>> >> > Epiphany crashes?
>> >> >
>> >> > You will need to install gnome-dbg and xulrunner-1.9-dbg and
>> >> > epiphany-browser-dbg for this data to be useful.
>> >>
>> >> Unfortunately, I cannot reproduce this bug anymore. Yes, I know that
>> >> my bug report is useless without a stack trace. But I just wanted to
>> >> report it anyways, because these segfaults are a reason why I switched
>> >> back to iceweasel. It's rock stable, hasn't crashed not even once for
>> >> a week. Epiphany was crashing several times a day.
>> >
>> > That's certainly no problem, it is definately useful to know about
>> > crashes even if they don't come with backtraces. :)
>> >
>> > I too believe this bug is fixed, or at least, does not occur any
>> > more--as I used to hit crashes in the cookie/password manager quite
>> > often, but don't any more.
>>
>> Right. But why do you think epiphany is so unstable after so many
>> years of development? Is it because we screwed something up in Debian
>> (like we use some different versions of libraries that just segfault
>> with epiphany) or is epiphany just unstable everywhere?
>> I think production programs should not just segfault.
>
> Most of the bugs I run into seem to be tricky corner cases caused by the
> horrific gecko embedding API, or bugs in gecko itself. But maybe are in
> epiphany itself. Many more are caused by plugins (Flash or the Totem
> plugin).
>
> At the end of the day, the software shouldn't crash, but the developers
> need more manpower to identify crashes and fix them. This is even harder
> if the developers can't reproduce the bug because we use a different
> version of xulrunner than they do, or we patch it with some odd patches
> that they don't use, or we use a different architecture to them, etc,
> etc.

I see. So things should improve after moving to webkit. Yes, it's true
that sometimes it crashes due to my flash plugin, but this password
crash seems unrelated.

Ok, thanks for enlightening me the issues and I'll keep reporting
bugs, preferably with a stacktrace.

Ondrej