Bug 243819
| Summary: | webgl/2.0.y/deqp/functional/gles3/negativetextureapi.html assertion failure | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Kimmo Kinnunen <kkinnunen> |
| Component: | WebGL | Assignee: | Dan Glastonbury <djg> |
| Status: | RESOLVED FIXED | ||
| Severity: | Normal | CC: | dino, kbr, kkinnunen |
| Priority: | P2 | Keywords: | InRadar |
| Version: | Other | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
| See Also: | https://bugs.webkit.org/show_bug.cgi?id=242614 | ||
| Bug Depends on: | |||
| Bug Blocks: | 222812 | ||
Kimmo Kinnunen
webgl/2.0.y/deqp/functional/gles3/negativetextureapi.html assertion failure
The validation is not run when pixels == nullptr
Due to bug 242614
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Kimmo Kinnunen
<rdar://98347913>
Kimmo Kinnunen
The issue is that validation is not done when user passes pixels=null, where as the teximage3dbase expect that the validation is done.
WebGL defines pixels=null means pixels=sufficiently large 0 filled pixels.
The philosophical rabbit hole:
Current WebGL implementation naturally "optimizes" so that it doesn't create the sufficiently large 0 filled pixels buffer, rather that it just passes pixels=nullptr down.
However, it also validates other arguments based on pixels. The validation shoudn't be done this way, it should be done based on the GL type, since pixels isn't always there.
The other philosophical rabbit hole:
the ASSERT_UNUSED(ok) is the smell that reveals the bug. Instead, the validation should return the computed properties, if the validation passes. This way the computation is done once
Dan Glastonbury
Pull request: https://github.com/WebKit/WebKit/pull/3263
EWS
Committed 253543@main (592cd219941f): <https://commits.webkit.org/253543@main>
Reviewed commits have been landed. Closing PR #3263 and removing active labels.