Bug 238813

Summary: [GTK][WPE] Regression after r292025, in GraphicsContextGLANGLE if the dmabuf backed is not created we need to bail out
Product: WebKit Reporter: Alejandro G. Castro <alex>
Component: ANGLEAssignee: Alejandro G. Castro <alex>
Status: RESOLVED DUPLICATE    
Severity: Normal CC: clord, cmarcelo, dino, ews-watchlist, extinstall.mstoolsai.jupyter, kbr, kkinnunen, kondapallykalyan, luiz, zan, zdobersek
Priority: P2    
Version: WebKit Nightly Build   
Hardware: Unspecified   
OS: Unspecified   
Attachments:
Description Flags
Patch none

Alejandro G. Castro
Reported 2022-04-05 09:54:10 PDT
We are causing issues to the tests since r292025, there are still the situation we have with xvfb support but we will address that in a different patch.
Attachments
Patch (2.09 KB, patch)
2022-04-05 10:07 PDT, Alejandro G. Castro 		no flags
DetailsFormatted DiffDiff
View All Add attachment proposed patch, testcase, etc.
Alejandro G. Castro
Comment 1 2022-04-05 10:07:24 PDT
Created attachment 456713 [details] Patch
Zan Dobersek
Comment 2 2022-04-06 05:51:54 PDT
Comment on attachment 456713 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=456713&action=review > Source/WebCore/platform/graphics/texmap/GraphicsContextGLTextureMapperANGLE.cpp:139 > + if (!result.iterator->value) > + return madeCurrent; > + And how are the actual contents of the WebGL-backing framebuffer composited in this case? Cause you're not rendering anything in the bo, and that's what's to be displayed.
Alejandro G. Castro
Comment 3 2022-04-06 06:44:01 PDT
(In reply to Zan Dobersek from comment #2) > Comment on attachment 456713 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=456713&action=review > > > Source/WebCore/platform/graphics/texmap/GraphicsContextGLTextureMapperANGLE.cpp:139 > > + if (!result.iterator->value) > > + return madeCurrent; > > + > > And how are the actual contents of the WebGL-backing framebuffer composited > in this case? Cause you're not rendering anything in the bo, and that's > what's to be displayed. The idea here is to avoid calling GL functions using a null image to avoid errors when we do not have the backing buffer, I don't think we need to create a solution for this situation better than this. The fallback is to use the texture that is later send to the compositor copying an ImageBuffer. We could add a log here if that feels that can help to make sure people know they are in this situation, because it is true sometimes you can end up here and not know about it. Thanks for the comment!
Alejandro G. Castro
Comment 5 2022-07-12 12:19:49 PDT
That commit broke the previous implementation of the fallback mode that was the reason it failed. We have added a new implementation in the bug 242108. *** This bug has been marked as a duplicate of bug 242108 ***
Note You need to log in before you can comment on or make changes to this bug.