readPixels, and readPixels into a pixel buffer object, must be updated / added for WebGL 2.0 support.
Taking this.
This seems like it should be straightforward; one piece of additional WebGL 2.0 specific validation had to be added to ANGLE. However, I'm now seeing strange behavior where GL_PACK_SKIP_ROWS seems to be being ignored by the OpenGL driver for glReadPixels calls both into client memory, and into the pixel pack buffer. Still investigating.
Created attachment 404511 [details] Patch
Comment on attachment 404511 [details] Patch Not requesting review yet - want to test the upstreamed ANGLE patch first.
Note that there are important steps to take when updating ANGLE. See http://trac.webkit.org/wiki/UpdatingANGLE
Created attachment 404515 [details] Patch
While debugging layout test failures associated with this patch, bugs were discovered in the earlier addition of pixel unpack buffer support added in Bug 208875. Fixing those bugs in this patch as well so that more tests can fully pass.
Created attachment 404583 [details] Patch
It turns out that the bug fixes to the texture upload paths were actually needed to prevent regressions in WebGL 2.0 conformance tests that would otherwise have been introduced with the other correctness fixes in this patch.
Comment on attachment 404583 [details] Patch If this version of the patch passes tests, I think it's ready to land. Setting cq+.
Comment on attachment 404583 [details] Patch Undoing CQ - investigating layout test changes on bots.
Created attachment 404591 [details] Patch
Made a mistake rebaselining the layout tests in the last patch. Two more tests progress with this patch. CQ'ing.
Committed r264536: <https://trac.webkit.org/changeset/264536> All reviewed patches have been landed. Closing bug and clearing flags on attachment 404591 [details].
<rdar://problem/65745897>