1809216 - (webgl-GLBlitHelperD3D-getD3D11-refcount) `GLBlitHelperD3D::getD3D11()` does not increment ANGLE device refcount

Status: RESOLVED FIXED

(Core :: Graphics: CanvasWebGL, defect)

Description

[Erich Gubler [:ErichDonGubler] (he/him)]

While working on 1753349, :jrmuizel, :jgilbert, and I discovered that Firefox has incorrectly foregone increasing the reference count for the D3D device retrieved from ANGLE via QueryDeviceAttribEXT in GLBlitHelper::GetD3D11(). The COM pointer returned should have its strong refcount incremented via IUnknown::AddRef, but it currently does not.

Comment 1

[BugBot [:suhaib / :marco/ :calixte]]

Set release status flags based on info from the regressing bug 1322746

Comment 2

[Erich Gubler [:ErichDonGubler] (he/him)]

Attachment: Bug 1809216: add missing refcount increment for `GLBlitHelperD3D::mD3D11` in `GLBlitHelperD3D::getD3D11` r=jgilbert,jrmuizel,#gfx-reviewers

Comment 3

[Erich Gubler [:ErichDonGubler] (he/him)]

A more detailed set of reproduction steps can be found here. Feel free to NI me here if anybody needs support in validating this.

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/01b132ada66e
add missing refcount increment for `GLBlitHelperD3D::mD3D11` in `GLBlitHelperD3D::getD3D11` r=jgilbert,jrmuizel,gfx-reviewers

Comment 5

[Erich Gubler [:ErichDonGubler] (he/him)]

The fix for this has landed in autoland, but it's unclear whether we wish to take additional effort to backport this. NI'ing the folks who will likely be making the call there.

Comment 6

[Ryan VanderMeulen [:RyanVM]]

Bugs get resolved when they're merged to m-c.

Comment 7

[Bob Hood [:bhood]]

Let's see how things react in Nightly to this change, and if all is well in a few days, then I say patch ESR. It won't be seeing this ANGLE update for a long time.

Comment 8

[Serban Stanca [:SerbanS]]

https://hg.mozilla.org/mozilla-central/rev/01b132ada66e

Comment 9

[Ryan VanderMeulen [:RyanVM]]

Go ahead and nominate this for ESR approval whenever.

Comment 10

[Erich Gubler [:ErichDonGubler] (he/him)]

:ryanvm: This patch isn't /quite/ a slam-dunk, since it hasn't been vetted really hard without the ANGLE rebase patches that also landed right after it. So, I'm not sure. Will consult with :jgilbert and :jrmuizel on this one.

Comment 11

[Erich Gubler [:ErichDonGubler] (he/him)]

:ryanvm: Since this is already in Fx 110, we'd like to see if there's any fall-out before trying to get it into ESR. If Fx 110 goes for a few weeks without there being an apparent regression, we think it'd be worth considering again.