Re: Win32 - invalid cast GdkPixmap to GdkWindow in gdk_event_translate

[Maarten Bosmans <mkbosmans gmail com>]

Just a note: plain text emails are preferred on this list.

See answers inline below.

2011/5/16 Ryan Krumins <ryan krumins gmail com>:
> Hi Guys,
>
> I have a fairly substantial Gtk application im working with that is
> experiencing some seemlingly randomly timed, but consistent in
> symptoms, crashes on Windows.
>
> The application is run on the following platforms:
> Linux Gtk 2.4
> Linux Gtk 2.10
> Solaris Gtk 2.4
> Windows Gtk 2.4
>
> The issue occus usually after the application has been running for a few
> hours, but has happed within 10-15 minues of starting up before. The warning
> below first printed out (not that I have elevated invalid cast to a CRITICAL
> in this build and dumped the stack on any CRITICALs to try and locate the
> issue):
>
>
> ** (application): CRITICAL **: invalid cast from `GdkPixmap' to `GdkWindow'
>
> The stack trace for this is below (Windows Gtk 2.4.14):
[snip]
>
>
> We are considering the obvious option of moving to a newer version of Gtk on
> windows in the hope that will fix the issue, but to be prefectly honest, we
> have found it difficult to get a version for windows that doesnt contain a
> show stopping bug. Our approach there has been to repair the bugs and submit
> patches once everything is up to speed, but in the meantime, we still have
> some resources to continue looking at this issue as the preferred solution.

What's there to consider? Just try it with a newer version and see if
the problem is fixed there. That should take too long, right?

Besides, that version of Gtk is ancient, so even if you find the issue
and make a patch, it either is already in a newer version or the patch
most likely will not apply to the current git tree.

In general it is true that Gtk on non-linux/X11 platforms is lagging a
bit. So the more people/projects we have that use the latest version
on Windows and submit patches for things that are not up to par, the
better.

> We have had issues with the use of focus-out handlers before but have always
> been able to find the offending code/programmer through valgrind on linux.
> This problem, obviously eludes our focus-out handler search.
>
> Would anyone be able to provide some insight into where we could focus our
> efforts ?

You could try downloading the precompiled windows binaries from the
latest version before the csw (client side windows) branch was merged.
I think that would be Gtk 2.16.

If that all works out, you can try moving up to the most recent (and
last) Gtk2 series: 2.24.

> Could the use of GtkFixed and its realize-on-pack operation cause some sort
> of race condition here ?


Maarten