1.5.1 RC3 Issues / unstability

This thread for unstability issues that cannot be reproduced easily.

1)
Crashes and freezes:
Twice during this testing period i have left the machine render over night to find out that in the morning the rc3 had died.
This morning it had thrown an exception. This is quite nasty since MLT takes time to render

2)
One scene i had wouldn't render on presets #7 or #8, died in the mesh initialization phase on exception. Cannot reproduce it anymore and that scene used i cannot zip and send. Worked fine with #6 though.

[giannis]

now application crash after nights rendering.

i come up, klick on the KT i left rendering over night and it disapears.

The time-stamp on the temp file was just seconds earlier than that. Somehow it seems that the screensaver (displays off) and waking from that killed the application.

These things that take the whole night to reproduce are kinda tedious to track down

[giannis]

i know... is there a scene that more or less presents such problem
in a determintic mannel and as soon as possible? If you have found one,
well you know...

i'll try to tick of the "render in background" box this night, we'll see if it affects anything.

(i'm persuming here that it affects the thread priorities)

[giannis]

Oh, ok.. I understand now.

So those of us with the ol' single cores its probably not a good idea then?

(still in background mode)
Hmm again the render had frozen during the night. this seems to be same issue.

got some new info out of it this time. The render was frozen again but the dialogs were working so i tried to save the image via the save button but it failed to write the image to the disk. got the "wrong extension" error message, but the image size was 0.

While doing that i noticed something, KT tried to write the temp_bmlt.jpg in the same directory, resulting in size 0 image and the application crash after that.

This time the crash was controlled and it showed the dialog about sending the scene.

Giannis, can you compile a debug build for me that does not have as much traces as the one you run? We could try to single out the problem on a bit wider front.

[giannis]