1127289 - DrawTargetTiled code paths in TiledContentClient is missing some of the synchronization code.

Status: RESOLVED INVALID

(Core :: Graphics: Layers, defect)

Description

[Nicolas Silva [:nical]]

This is most probably what's causing the R14 failure on the the emulator with DrawTargetTiled.

Comment 1

[Nicolas Silva [:nical]]

Attachment: Patch

Comment 2

[Sotaro Ikeda [:sotaro]]

nical, SyncObject derived class seems to exist only for D3D11. It is SyncObjectD3D11. Gralloc uses FenceHandle. How does the patch fix the problem for gralloc?

Comment 3

[Sotaro Ikeda [:sotaro]]

And emulator is run on ICS. ICS does not have sync object like android::Fence.

Comment 4

[Nicolas Silva [:nical]]

(In reply to Sotaro Ikeda [:sotaro] from comment #2)
> nical, SyncObject derived class seems to exist only for D3D11. It is
> SyncObjectD3D11. Gralloc uses FenceHandle. How does the patch fix the
> problem for gralloc?

Darnit! I thought it was a gralloc thing. Well I guess it doesn't fix the issue, although we should still take this patch since this is still missing from the DrawTargetTiled code path.

Comment 5

[Sotaro Ikeda [:sotaro]]

Comment on attachment 8556427 [details] [diff] [review]
Patch

Review of attachment 8556427 [details] [diff] [review]:
-----------------------------------------------------------------

Looks good.

Comment 6

[Nicolas Silva [:nical]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/a1d913968a8a

Comment 7

[Carsten Book [:Tomcat]]

sorry had to back this out for test failures like https://treeherder.mozilla.org/logviewer.html#?job_id=6161085&repo=mozilla-inbound

Comment 8

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/8a49d4f4bb25

Comment 9

[[:fabrice] Fabrice Desré]

https://hg.mozilla.org/projects/cypress/rev/73b3c1c8d3d9

Comment 10

[Ryan VanderMeulen [:RyanVM]]

So, is this the reason bug 1081349 et al are spiking again?

Comment 11

[Nicolas Silva [:nical]]

(In reply to Ryan VanderMeulen [:RyanVM UTC-5] from comment #10)
> So, is this the reason bug 1081349 et al are spiking again?

This patch only affects windows under a configuration that we don't ship yet, so I don't think this can be the reason for bug 1081349 to spike again.

Comment 12

[Nicolas Silva [:nical]]

(In reply to Nicolas Silva [:nical] from comment #11)
> This patch only affects windows under a configuration that we don't ship
> yet, so I don't think this can be the reason for bug 1081349 to spike again.

Actually I take that back. It can also affect other platforms. digging.

Comment 13

[Ryan VanderMeulen [:RyanVM]]

The patch that landed in bug 1129360 caused Android reftest failures, so we've gone ahead and backed this out while they're being investigated.
https://hg.mozilla.org/integration/mozilla-inbound/rev/37321f61780e

Comment 14

[Ryan VanderMeulen [:RyanVM]]

W.T.F.

The reftest failures that started on the push for bug 1129360 didn't go away when I backed it and this bug out! So I've relanded this bug on the hope that it makes the Android permafail go away, even if we're still stuck with with the B2G intermittents. *cry*

https://hg.mozilla.org/integration/mozilla-inbound/rev/7d87dc37c707

Comment 15

[Ryan VanderMeulen [:RyanVM]]

Yup, re-landing this fix made the Android permafails go away. *sigh*

Comment 16

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

https://hg.mozilla.org/mozilla-central/rev/7d87dc37c707

Comment 17

[Mason Chang [Inactive] [:mchang]]

Partial Backout - https://hg.mozilla.org/mozilla-central/rev/9cf76d4021ce

Comment 18

[Mason Chang [Inactive] [:mchang]]

Backout story - https://bugzilla.mozilla.org/show_bug.cgi?id=1130681#c21

Comment 19

[Nicolas Silva [:nical]]

The TiledCOntentClient code has changed a lot since, and I don't think this applies anymore.