> @Sam,
>
> I just tested this branch and it works.
>
> if you like i can still test r3131 to see what the actual problem was, but it
> would appear your changes in this branch work.
>
> cheerz
It's probably worth mentioning as it would be a needle in the haystack in the diff due to the speedy revert
635 if (!d->decorated)
636 return FALSE;
637
There was a race where a window could have its decoration resources freed by the time compiz got back to us about whether or not it was ok to re-update decorations.
As a side note, this particular part of the code has always suffered from such race conditions which can lead to random crashes. It is worth looking into a way that we can fix this, but like I said earlier, I fear that larger architectural changes may be required.
> @Sam,
>
> I just tested this branch and it works.
>
> if you like i can still test r3131 to see what the actual problem was, but it
> would appear your changes in this branch work.
>
> cheerz
It's probably worth mentioning as it would be a needle in the haystack in the diff due to the speedy revert
635 if (!d->decorated)
636 return FALSE;
637
There was a race where a window could have its decoration resources freed by the time compiz got back to us about whether or not it was ok to re-update decorations.
As a side note, this particular part of the code has always suffered from such race conditions which can lead to random crashes. It is worth looking into a way that we can fix this, but like I said earlier, I fear that larger architectural changes may be required.