REGRESSION(250984@main): [Linux] DrmFourCCFormatToGLInternalFormat() change breaks framebuffer blitting

• Component: ANGLE
• Status: RESOLVED FIXED

[Zan Dobersek]

ANGLE update at 250984@main brought in the following change:
36f5e6ce5 Vulkan: Fix DRM to GL format conversion
https://chromium-review.googlesource.com/c/angle/angle/+/3624139
https://chromium.googlesource.com/angle/angle/+/36f5e6ce550870261998b77987d01c4a4e34e620

An XRGB8888 DMABuf is used as output for ANGLE execution. DrmFourCCFormatToGLInternalFormat(), called from the DmaBufImageSiblingEGL constructor, now returns a gl::Format object for the GL_BGRX8_ANGLEX internal format.

This trips up the GL_BlitFramebufferANGLE() call, during the validation of attachments, with gl::Format::EquivalentForBlit() returning false for the read and draw attachments. Read attachment at that point has the GL_RGB8 sized internal format, but the draw attachment (with the DMABuf underneath) sticks to the GL_BGRX8_ANGLEX format.

EquivalentBlitInternalFormat() helper maps GL_RGBX8_ANGLE to GL_RGB8 -- possibly it should do the same for GL_BGRX8_ANGLEX.

[Zan Dobersek]

Reported upstream:
https://bugs.chromium.org/p/angleproject/issues/detail?id=7452

[Kenneth Russell]

The ANGLE bug indicates this has already been fixed upstream. We should roll forward ANGLE into WebKit again to pick up this and other fixes.

[Radar WebKit Bug Importer]

<rdar://problem/96127632>

[Zan Dobersek]

Resolved with the update in #242055.