D27924: Tentative fix for bug 372305

[Jiří Paleček]

jpalecek added a comment.


  In D27924#628197 <https://phabricator.kde.org/D27924#628197>, @davidedmundson wrote:
  
  > Not performing any action till the next call makes sense, kinda like the delay in the first call after a swapBuffers(), but even so surely the order of events remains the same and we realloc before usage? It doesn't seem to make sense to me.
  
  
  That's true. I still see this more as a kludge than a real fix, since IMO that is up to Nvidia. Moreover, I've found that while it does reduce the amount of crashes about 10 times, it doesn't unfortunately fix them all (some happen when the computer is suspended in the middle of drawing, without a chance to do cleanup/`initGL` - however the unmapping of the memory which causes the crash happens some time later and pretty synchronously, although I'm not yet sure on which call it happens).
  
  > But this fix is clean and putting in an explicit fence on cleanup doesn't sound like a bad idea.
  
  I would rather wait on this, there's no need to be hasty. But if there are other comments on this, I'd like to read them (right now, I can't see anything).

REPOSITORY
  R108 KWin

BRANCH
  wip

REVISION DETAIL
  https://phabricator.kde.org/D27924

To: jpalecek, davidedmundson, #kwin, ekurzinger
Cc: kwin, Orage, cacarry, LeGast00n, The-Feren-OS-Dev, cblack, jraleigh, zachus, fbampaloukas, GB_2, mkulinski, ragreen, jackyalcine, iodelay, crozbo, bwowk, ZrenBot, ngraham, alexeymin, himcesjf, lesliezhai, ali-mohamed, hardening, romangg, jensreuterberg, abetts, sebas, apol, ahiemstra, mart
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kwin/attachments/20200318/60b56b9f/attachment-0001.html>