[chromium] REGRESSION(112286) Compositor initialization blocks for program compilation / linking
Created attachment 138727 [details] Patch
Created attachment 138744 [details] Patch
Comment on attachment 138744 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=138744&action=review > Source/WebCore/platform/graphics/chromium/TextureCopier.cpp:82 > + if (m_blitProgram->initialized()) is that supposed to be if (!initialized) ?
Oh, yes (interesting that no layout tests break)
(In reply to comment #4) > Oh, yes (interesting that no layout tests break) It'd just re-get the uniforms on ever ycopy, so it's inefficient but still correct. Can I trouble you to add an assert to ProgramBinding::initialize() to make sure we don't do this?
Comment on attachment 138744 [details] Patch Other than than the initialized() check, looks great. Thanks for writing a test. :)
(In reply to comment #5) > (In reply to comment #4) > > Oh, yes (interesting that no layout tests break) > > It'd just re-get the uniforms on ever ycopy, so it's inefficient but still correct. Can I trouble you to add an assert to ProgramBinding::initialize() to make sure we don't do this? No, since i'm not calling initialize() unless initialized() is true, I think it will never get the uniforms at all. The only uniform on the program is the s_texture on the fragment shader. So I have no idea WTF it was doing. Anyway, added ASSERT()s for both ways - calling initialize() too much or calling program() before initialize() and it seems good now.
Created attachment 138928 [details] Patch for landing
Comment on attachment 138928 [details] Patch for landing Clearing flags on attachment: 138928 Committed r115285: <http://trac.webkit.org/changeset/115285>
All reviewed patches have been landed. Closing bug.