Comment 16 for bug 80551

Hello Alex,

Your hypothesis sounds good and takes care of all the problems that we
have seen with this bug. I'll try using F-spot's slideshow feature and
see if I can trigger the crash. According to everything we've seen so
far, though, my SCIM should be crashing, and yet it isn't. I installed
fglrx so I could set up Beryl, and I haven't experienced problems in
switching over. If I need to type in Korean, I can: 너무 이상하다.
(translation: very strange). In fact, I just tried the F-spot slideshow
and it neither it nor SCIM crashed. Perhaps I just set it up in the
right order? Used the right set of instructions this time? I don't
know, but I'll take working over not working anyday.

Cheers,
John

On Thu, 2007-05-17 at 12:51 +0000, Alexander Hunziker wrote:
> I *think* the pattern is the following. Please everyone read and verify
> :-) My hypothesis is that it's always the same bug, not two bugs.
>
> * Gizmo: linked to libstdc++.so.5, while SCIM is linked to
> libstc++.so.6, so this is the "classical" bug that can be circumvented
> by using scim-bridge that uses sockets to communicate with the program.
> Whatever that means... :-)
>
> * f-spot and Openoffice are linked to the same version of libstdc++, yet
> they crash. Now here's my assumption: they use both SCIM *and* some
> OpenGL routines. f-spot uses OpenGL for it's slideshow and OpenOffice I
> guess uses it for some of the drawing routines. Now, it was observed
> that ATI's fglrx seems to be affecting the issue as well. And guess
> which libstdc++ it is linked to...? It's libstd++.so.5. In short I think
> if programs trigger use of certain routines from the ATI graphics driver
> *and* use scim, they crash, so, it's the same bug as the one crashing
> applications that are simply linked to libstdc++.so.5.
>