Bug 66862 - [chromium] Renderer crashes if compositor runs out of texture memory
Summary: [chromium] Renderer crashes if compositor runs out of texture memory
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: New Bugs (show other bugs)
Version: 528+ (Nightly build)
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Nobody
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2011-08-24 10:18 PDT by Iain Merrick
Modified: 2011-08-24 15:10 PDT (History)
5 users (show)

See Also:


Attachments
gdb backtrace (6.63 KB, text/plain)
2011-08-24 11:36 PDT, Iain Merrick
no flags Details
Patch (1.61 KB, patch)
2011-08-24 14:48 PDT, Iain Merrick
no flags Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Iain Merrick 2011-08-24 10:18:19 PDT
To reproduce: load http://diden.net/test/tex.html and move the slider to the right. The renderer reliably crashes at around 60. (That might be because of the hard-coded texture limit of 64MB. The usage reported by the page may not be accurate, though, because of tiling.)

Possibly related to TextureManager refactoring? https://bugs.webkit.org/show_bug.cgi?id=64772

The old behavior was just to skip drawing some layers, but that wasn't really ideal either.
Comment 1 Iain Merrick 2011-08-24 10:31:47 PDT
It's an assert:

./gpu/command_buffer/client/../client/gles2_implementation_autogen.h(287): GPU_DCHECK(textures[i] != 0 (0)) failed.
Comment 2 James Robinson 2011-08-24 11:18:14 PDT
Stack?
Comment 3 Iain Merrick 2011-08-24 11:36:46 PDT
Created attachment 105029 [details]
gdb backtrace 

Looks like we're just deleting 0? I think that's kosher so maybe the DCHECK isn't needed. No crash in release builds.
Comment 4 James Robinson 2011-08-24 12:44:20 PDT
Yeah, I've been assuming that deleting texture 0 is safe.  It's quite easy to avoid calling deleteTexture() on zero if the underlying stack is unhappy with it, though.
Comment 5 Iain Merrick 2011-08-24 14:48:05 PDT
Created attachment 105072 [details]
Patch
Comment 6 Iain Merrick 2011-08-24 14:48:57 PDT
Turns out we got stricter back in June - http://code.google.com/p/chromium/issues/detail?id=85268
Comment 7 James Robinson 2011-08-24 14:52:56 PDT
Comment on attachment 105072 [details]
Patch

Ah, OK.  I wasn't aware of that change (or more likely just forgot about it).
Comment 8 WebKit Review Bot 2011-08-24 15:10:10 PDT
Comment on attachment 105072 [details]
Patch

Clearing flags on attachment: 105072

Committed r93743: <http://trac.webkit.org/changeset/93743>
Comment 9 WebKit Review Bot 2011-08-24 15:10:14 PDT
All reviewed patches have been landed.  Closing bug.