[webkit-reviews] review granted: [Bug 89901] [chromium] Respect memory needed for RenderSurfaces when reserving contents textures : [Attachment 150828] Patch

[bugzilla-daemon at webkit.org]

Adrienne Walker <enne at google.com> has granted Dana Jansens
<danakj at chromium.org>'s request for review:
Bug 89901: [chromium] Respect memory needed for RenderSurfaces when reserving
contents textures
https://bugs.webkit.org/show_bug.cgi?id=89901

Attachment 150828: Patch
https://bugs.webkit.org/attachment.cgi?id=150828&action=review

------- Additional Comments from Adrienne Walker <enne at google.com>
Thanks for testing on those pages.  They've OOMed us in the past (when we had
lower limits) and have had anecdotally had a lot of RS usage.  If they don't
really have much at all, then all the better.

Re: root surface vs. root layer.  Ah, quite right.  I misread that.  Just
rendering the root surface is an entirely different thing.  That makes sense
that you wouldn't want to reserve anything but the root surface in that case. 
I am totally swayed by arguments about simplicity here.

Re: requestLate.  Sure, it doesn't happen in priority order and I buy that for
render surface OOM, you wouldn't want anything else.  But for content OOM +
occlusion, maybe you would, even if you found out that you could at the wrong
time to make an optimal choice? Maybe that's just too much complication.  In
the future, I'm hoping impl-side painting will simplify this because we can
have SkPicture and occlusion information before we allocate or prioritize
textures.

R=me.  Let's see how this goes.