888311 - remove LayerManagerOGL

Status: RESOLVED DUPLICATE of bug 924403

(Core :: Graphics: Layers, defect)

Description

[Andreas Gal :gal]

With mac using OMTC its obsolete now.

Comment 1

[Nick Cameron [:nrc]]

(In reply to Andreas Gal :gal from comment #0)
> With mac using OMTC its obsolete now.

As much as I would love to see LayerManagerOGL gone, we're not quite ready yet - we still use it on Linux. Getting OMTC turned on there by 'default' (even HWA is not default) is in the pipeline, but I'm not sure if anyone is working on it just yet. I believe Vlad just got OGL working again on Windows too, not sure if that uses LayerManager or compositor.

Comment 2

[Matt Woodrow (:mattwoodrow)]

I don't think we need to worry too much about linux or windows.

We don't have LayerManagerOGL enabled by default on linux, so transitioning people that have it forced on via a pref to OMTC+OGL doesn't seem like a big issue.

Having windows OGL working is nice, but nobody is using it, so we shouldn't block on that.

Comment 3

[Matt Woodrow (:mattwoodrow)]

The main problem is that we have OMTC disabled on OSX 10.6

A whole bunch of test suites crash on tinderbox if we enable this. All deep within driver code.

I'm assuming that this is a driver bug, but we managed to avoid it with LayerManagerOGL, so we could probably figure something out for OMTC.

If someone has a 10.6 machine and wants to take a look, that would be great.

Comment 4

[Nick Cameron [:nrc]]

(In reply to Matt Woodrow (:mattwoodrow) from comment #2)
> I don't think we need to worry too much about linux or windows.
> 
> We don't have LayerManagerOGL enabled by default on linux, so transitioning
> people that have it forced on via a pref to OMTC+OGL doesn't seem like a big
> issue.
>

I agree, what I meant was we need to make sure that OMTC+OGL on Linux works almost as well as just OGL before we go and delete LayerManagerOGL. I don't think we should screw our Linux users too much for the sake of code removal. Hopefully that isn't a lot of work.
 
> Having windows OGL working is nice, but nobody is using it, so we shouldn't
> block on that.

Hopefully OGL Windows will work as well OMTC as MT. I expect Vlad had reasons for making it work and we should make sure we still address those before doing this.

Comment 5

[Markus Stange [:mstange]]

(In reply to Matt Woodrow (:mattwoodrow) from comment #3)
> The main problem is that we have OMTC disabled on OSX 10.6
> 
> A whole bunch of test suites crash on tinderbox if we enable this. All deep
> within driver code.

Here's a try run which tested BasicCompositor OMTC on 10.6:
https://tbpl.mozilla.org/?tree=Try&rev=ddc1948b06dd
We crash there, too, but not in driver code. Does this help or do the crashes look completely different? Do you have a link to a try run with OGL OMTC on 10.6?

> If someone has a 10.6 machine and wants to take a look, that would be great.

I have one, I can have a look next week.

Comment 6

[Andreas Gal :gal]

10:21:27  WARNING -  PROCESS-CRASH | file:///Users/cltbld/talos-slave/test/build/tests/reftest/tests/content/canvas/crashtests/789933-1.html | application crashed [@ 0x7fffffe012f0]
10:21:27     INFO -  Crash dump filename: /var/folders/Hs/HsDn6a9SG8idoIya6p9mtE+++TI/-Tmp-/tmpnOo9iz/minidumps/4735EE86-2582-4E6F-A482-448E97C9751D.dmp
10:21:27     INFO -  Operating system: Mac OS X
10:21:27     INFO -                    10.6.8 10K549
10:21:27     INFO -  CPU: amd64
10:21:27     INFO -       family 6 model 23 stepping 10
10:21:27     INFO -       2 CPUs
10:21:27     INFO -  Crash reason:  EXC_BAD_ACCESS / KERN_INVALID_ADDRESS
10:21:27     INFO -  Crash address: 0x67dee080
10:21:27     INFO -  Thread 17 (crashed)
10:21:27     INFO -   0  0x7fffffe012f0
10:21:27     INFO -      rbx = 0x000000011846f828   r12 = 0x0000000000009c40
10:21:27     INFO -      r13 = 0x0000000196000000   r14 = 0x0000000000000000
10:21:27     INFO -      r15 = 0x00000001180da400   rip = 0x00007fffffe012f0
10:21:27     INFO -      rsp = 0x000000011c0802a8   rbp = 0x000000011c0802a8
10:21:27     INFO -      Found by: given as instruction pointer in context
10:21:27     INFO -   1  XUL!mozilla::gfx::SourceSurfaceCG::InitFromData(unsigned char*, mozilla::gfx::IntSizeTyped<mozilla::gfx::UnknownUnits> const&, int, mozilla::gfx::SurfaceFormat) [SourceSurfaceCG.cpp:ddc1948b06dd : 130 + 0x12]
10:21:27     INFO -      rip = 0x0000000102a18846   rsp = 0x000000011c0802d0
10:21:27     INFO -      Found by: stack scanning
10:21:27     INFO -   2  XUL!mozilla::gfx::DrawTargetCG::CreateSourceSurfaceFromData(unsigned char*, mozilla::gfx::IntSizeTyped<mozilla::gfx::UnknownUnits> const&, int, mozilla::gfx::SurfaceFormat) const [DrawTargetCG.cpp:ddc1948b06dd : 188 + 0x14]
10:21:27     INFO -      rip = 0x0000000102a199e1   rsp = 0x000000011c080310
10:21:27     INFO -      Found by: stack scanning
10:21:27     INFO -   3  XUL!mozilla::layers::TextureSourceBasic::Lock() [BasicCompositor.cpp:ddc1948b06dd : 56 + 0x5]
10:21:27     INFO -      rip = 0x00000001026eb23e   rsp = 0x000000011c080350
10:21:27     INFO -      Found by: stack scanning
10:21:27     INFO -   4  XUL!gfxMatrix::TransformBounds(gfxRect const&) const [gfxMatrix.cpp:ddc1948b06dd : 107 + 0x7]
10:21:27     INFO -      rip = 0x00000001026cc3b7   rsp = 0x000000011c080360
10:21:27     INFO -      Found by: stack scanning
10:21:27     INFO -   5  XUL!mozilla::layers::ImageHostSingle::Composite(mozilla::layers::EffectChain&, float, mozilla::gfx::Matrix4x4 const&, mozilla::gfx::PointTyped<mozilla::gfx::UnknownUnits> const&, mozilla::gfx::Filter const&, mozilla::gfx::RectTyped<mozilla::gfx::UnknownUnits> const&, nsIntRegion const*, mozilla::layers::TiledLayerProperties*) [ImageHost.cpp:ddc1948b06dd : 88 + 0x5]
10:21:27     INFO -      rip = 0x00000001027192d2   rsp = 0x000000011c080380
10:21:27     INFO -      Found by: stack scanning
10:21:27     INFO -   6  XUL!mozilla::layers::Layer::CalculateScissorRect(nsIntRect const&, gfxMatrix const*) [Layers.cpp:ddc1948b06dd : 628 + 0xa]
10:21:27     INFO -      rip = 0x000000010272c284   rsp = 0x000000011c0803e0
10:21:27     INFO -      Found by: stack scanning
10:21:27     INFO -   7  XUL + 0x171926f
10:21:27     INFO -      rip = 0x0000000102719270   rsp = 0x000000011c080438
10:21:27     INFO -      Found by: stack scanning
10:21:27     INFO -   8  XUL!mozilla::layers::CanvasLayerComposite::RenderLayer(nsIntPoint const&, nsIntRect const&) [CanvasLayerComposite.cpp:ddc1948b06dd : 94 + 0x20]
10:21:27     INFO -      rip = 0x00000001026f34c5   rsp = 0x000000011c080450
10:21:27     INFO -      Found by: stack scanning