1120753 - Fallback to software vsync if hardware vsync fails to initialize on b2g

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[Mason Chang [Inactive] [:mchang]]

+++ This bug was initially created as a clone of Bug #1119850 +++

If hardware vsync fails to initialize, fallback and use software vsync. This is required for the b2g-emulators which have no HwcComposer but still use gfxAndroidPlatform. In the cases where vsync + compositor aligned vsync are enabled but the HwcComposer fails to initialize, we won't be able to composite anything.

Comment 1

[Mason Chang [Inactive] [:mchang]]

Requesting blocking-b2g 2.2 as this is required to enable silk by default, bug 987532.

Comment 2

[Mason Chang [Inactive] [:mchang]]

Attachment: Fallback to Software Vsync

Comment 3

[Mason Chang [Inactive] [:mchang]]

Attachment: Fallback to Software Vsync. v2

Rebased on master. Also found a bug on a flame device where we initialize the gfxPlatform before we initialize the HwcComposer vsync callbacks. Hardware vsync wouldn't be enabled yet and so we'd fallback to software vsync. This version initializes the hardware vsync callback when we create the HwcComposer instead of when we initialize the HwcComposer. Initialization of the HwcComposer occurs once we have a glcontext, which is later in the startup process.

Comment 4

[Kartikaya Gupta (email:[e-mail])]

Comment on attachment 8549034 [details] [diff] [review]
Fallback to Software Vsync. v2

Review of attachment 8549034 [details] [diff] [review]:
-----------------------------------------------------------------

r? mwu for HwcComposer2D changes.

::: gfx/thebes/VsyncSource.h
@@ -45,5 @@
>        // Large parts of Gecko assume TimeStamps should not be in the future such as animations
>        virtual void NotifyVsync(TimeStamp aVsyncTimestamp);
>  
>        nsRefPtr<RefreshTimerVsyncDispatcher> GetRefreshTimerVsyncDispatcher();
> -

nit: whitespace deletion

@@ -62,5 @@
>    };
>  
>    void AddCompositorVsyncDispatcher(CompositorVsyncDispatcher* aCompositorVsyncDispatcher);
>    void RemoveCompositorVsyncDispatcher(CompositorVsyncDispatcher* aCompositorVsyncDispatcher);
> -

nit: whitespace deletion

Comment 5

[Mason Chang [Inactive] [:mchang]]

Attachment: Fallback to Software Vsync. v3

Carrying r+, updated without whitespace deletions.

Comment 6

[Mason Chang [Inactive] [:mchang]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/a3d8163ee939

Comment 7

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

https://hg.mozilla.org/mozilla-central/rev/a3d8163ee939

Comment 8

[Mason Chang [Inactive] [:mchang]]

Comment on attachment 8549141 [details] [diff] [review]
Fallback to Software Vsync. v3

Approval Request Comment
[Feature/regressing bug #]: Required for silk, bug 987532 
[User impact if declined]: Once silk is enabled by default, mochitests will fail to execute.
[Describe test coverage new/current, TBPL]: Without this patch, all tests will fail with silk enabled by default since hardware vsync does not work on the emulators.
[Risks and why]: Low - Currently only used if silk is enabled, which is currently prefed off.
[String/UUID change made/needed]: None

Comment 9

[Mason Chang [Inactive] [:mchang]]

Comment on attachment 8549141 [details] [diff] [review]
Fallback to Software Vsync. v3

From https://bugzilla.mozilla.org/show_bug.cgi?id=1121065#c8, asking for b2g37 approval instead. Approval request in comment 8.

Comment 10

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-b2g37_v2_2/rev/ba6975fbdf68