[GPU Process] Unique RenderingResourceIdentifiers Part 7: Migrate RemoteResourceCache to QualifiedRenderingResourceIdentifier

[Myles C. Maxfield]

Migrate RemoteResourceCache to QualifiedRenderingResourceIdentifier

[Radar WebKit Bug Importer]

<rdar://problem/84010479>

[Myles C. Maxfield]

Created attachment 440567 [details]
Patch

[Said Abou-Hallawa]

Comment on attachment 440567 [details]
Patch

This patch looks okay. But I am surprised that RemoteImageBuffer currently has two identifiers. One is from the base class: ConcreteImageBuffer::m_renderingResourceIdentifier and the other is RemoteImageBuffer::m_renderingResourceIdentifier. The later was added in r269682. Luckily they hold the same value. But I think this needs to be cleaned-up.

[Wenson Hsieh]

(In reply to Said Abou-Hallawa from comment #3)
> Comment on attachment 440567 [details]
> Patch
> 
> This patch looks okay. But I am surprised that RemoteImageBuffer currently
> has two identifiers. One is from the base class:
> ConcreteImageBuffer::m_renderingResourceIdentifier and the other is
> RemoteImageBuffer::m_renderingResourceIdentifier. The later was added in
> r269682. Luckily they hold the same value. But I think this needs to be
> cleaned-up.

On trunk, RemoteImageBuffer's `m_renderingResourceIdentifier` is a WebKit::QualifiedRenderingResourceIdentifier while the base class just has a regular RenderingResourceIdentifier.

[Myles C. Maxfield]

(In reply to Said Abou-Hallawa from comment #3)
> Comment on attachment 440567 [details]
> Patch
> 
> This patch looks okay. But I am surprised that RemoteImageBuffer currently
> has two identifiers. One is from the base class:
> ConcreteImageBuffer::m_renderingResourceIdentifier and the other is
> RemoteImageBuffer::m_renderingResourceIdentifier. The later was added in
> r269682. Luckily they hold the same value. But I think this needs to be
> cleaned-up.

Yeah. Personally I think it's a mistake to put the identifiers inside the objects themselves, but I do understand why it's expedient. Philosophically, I think the objects themselves should have no idea that they're being shared across processes, and the process sharing code should be out-of-band. I think that 1) not everyone on the team agrees with this, though, and 2) even if they did, doing so is out of scope for this patch.

[Myles C. Maxfield]

Committed r283955 (242806@main): <https://commits.webkit.org/242806@main>

[Myles C. Maxfield]

This broke the build.

[Myles C. Maxfield]

Broke the build

[Myles C. Maxfield]

Created attachment 440882 [details]
Patch for committing

[Myles C. Maxfield]

Committed r283968 (242815@main): <https://commits.webkit.org/242815@main>