1042052 - The TextureClients on white and on back should always be of the same type.

Status: RESOLVED FIXED

(Core :: Graphics: Layers, defect)

Description

[Nicolas Silva [:nical]]

In rare occasions (most likely when running out of video memory) we may fail to allocate a TextureClient in the non-tiled ContentClient. When this happens we try to fallback to a BufferTextureClient (backed by either plain memory or shmem). This can cause the two component alpha planes to not be of the same type in the compositor side and cause one to be a BigImageIterator while the other isn't. It would be needlessly complicated to support this on the compositor side so we should always make sure both planes are of the same type on the client side.

Comment 1

[Nicolas Silva [:nical]]

Attachment: Add TextureClient::CreateSimilar and use that in ContentClient for component alpha

try push: https://tbpl.mozilla.org/?tree=Try&rev=bae7aed0999c

Comment 2

[Sotaro Ikeda [:sotaro]]

Comment on attachment 8460313 [details] [diff] [review]
Add TextureClient::CreateSimilar and use that in ContentClient for component alpha

Review of attachment 8460313 [details] [diff] [review]:
-----------------------------------------------------------------

Yeah, this could simplify the situation. "review+" if tryserver test is passed.

Comment 3

[Nicolas Silva [:nical]]

try push: https://tbpl.mozilla.org/?tree=Try&rev=17f60d6a825b

Comment 4

[Nicolas Silva [:nical]]

fixed the windows build error, windows-only try push: https://tbpl.mozilla.org/?tree=Try&rev=07d20210f5f8

Comment 5

[Nicolas Silva [:nical]]

successful windows try push: https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=c28cd7aba86d

Comment 6

[Nicolas Silva [:nical]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/780188e2dab7

Comment 7

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/780188e2dab7