bug#5074:

[Lennart Borgman <lennart.borgman@gmail.com>]

On Wed, May 26, 2010 at 11:33 PM, Juanma Barranquero <lekktu@gmail.com> wrote:
> On Wed, May 26, 2010 at 20:20, Lennart Borgman
> <lennart.borgman@gmail.com> wrote:
>
>> I think it may be wrong closing this bug. You can not do anything with
>> this information alone, but please do not hide it by just closing it.
>
> But this bug report has *no* useful information, other than the fact
> that a non-reproducible crash happened perhaps during a call to
> url-retrieve-synchronously. Or perhaps during redisplay. There's not
> even a backtrace. It would be difficult to relate any possible fix to
> this bug.

Yes, sorry. I will try change my habit, always running Emacs under a debugger.

>> There was just a new bug report about url-retrieve-synchronously (bug
>> 6274). I suspect it can be related.
>
> Why? You reported a crash, #6274 talks about a temporary freeze.

Because all information about the function may be related. A bad call
to the system functions may result in a bad state resulting in a crash
or a hang. But ... since it is on different platforms with different
API:s you and Glenn are probably right here. There is probably no
connection between 6274 and this bug report.

However FYI I still have a lot of crashes/freezes. Emacs hangs and is
difficult to kill from Windows Task Manager, but it does not loop.
Seems to be low level. I suspect four sources at the moment: url
related (most suspected for hangs), frame related (very suspected for
crashes and a bit for hangs) and the current changes to the display
code (a little bit suspected for crashes). And my own changes for
frame handling (quite suspected, but they are very recent, last month
only). I will try to get back traces.

>> I also reported a hang in the same function in bug 5103 and another
>> hang calling it in a timer in bug 5372.
>
> Well, then perhaps these two bugs are really related to #6274 (it's
> difficult to say, because as usual your bug reports are pretty sparse,
> to say the least). But in any case, how are they related to #5074?
>
>    Juanma
>