QSG_RENDER_LOOP woes

[Martin Gräßlin]

On Friday 09 May 2014 15:56:53 Sebastian Kügler wrote:
> Hey,
> 
> Since 30d2489930 in plasma-workspace, we're setting the QSG_RENDER_LOOP to
> "basic", if it's not specified otherwise by an env var. This has been done
> to prevent crashes in the threaded renderer, but unfortunately, it seems to
> come with its own set of problems: Plasma produces artifacts, high CPU load
> and grinds to a complete halt after a while.
> 
> I've tried setting the QSG_RENDER_LOOP to threaded, and that does indeed
> prevent those problems, but obviously, plasma-shell is more crashy with it.
> Now some of those crashes (at least one I've observed) is indeed in the
> scenegraph, but I've also seen other crashes (one in groupmanager of the
> taskbar, indirectly, not sure if that's related).
> 
> I think we might need to revisit this patch, and find a "stable" way
> forward. Just switching to basic renderloop doesn't seem to be the
> solution. :/

Please note that Qt itself switched it off for all of Mesa. The change now was 
more to only have one possible way of usage and preventing us hunting down 
non-reproducable bugs.

The crashes we see with threaded render loop cannot be fixed or worked around 
if I understand the Qt commit disabling it for all of Mesa correctly. Have a 
look at the plasma-devel archives, I sent a FYI mail some time back.

Cheers
Martin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.kde.org/pipermail/plasma-devel/attachments/20140509/b9d10b72/attachment.sig>