1083635 - improve layer-tree dumping

Status: RESOLVED FIXED

(Core :: Layout, defect)

Comment 1

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: Part 1: Remove nsDisplayZoom::Paint since it's dead code

Comment 2

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: Part 2: Inline nsDisplayList::PaintForFrame into PaintRoot

Comment 3

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: Part 3: Flush displaylist dump before calling PaintRoot

Comment 4

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: Part 4: Return LayerManager from nsDisplayList::PaintRoot, and dump its contents in nsLayoutUtils::PaintFrame

Comment 5

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: Part 5: ClientLayerManager::BeginTransactionWithTarget should assert if we're not going to take a snapshot

Comment 6

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: Part 6: Rename viewport to scrollport in ScrollFrameHelper::ComputeFrameMetrics

Comment 7

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: Create nsDOMWindowUtils::UpdateLayerTree and use it to flush layers in reftest content processes instead of doing a DRAW_WIDGET_LAYERS drawWindow. r=mattwoodrow

Comment 8

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: Part 4.5: Avoid requesting PAINT_WIDGET_LAYERS if we're in a content process

Comment 9

[Robert O'Callahan (:roc) (email my personal email if necessary)]

We're green with this:
https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=c45a80b4dd60

Comment 10

[Robert O'Callahan (:roc) (email my personal email if necessary)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/d223d7cd7e62
https://hg.mozilla.org/integration/mozilla-inbound/rev/dc63a7b6f073
https://hg.mozilla.org/integration/mozilla-inbound/rev/7e996612b3b1
https://hg.mozilla.org/integration/mozilla-inbound/rev/0af9b31d714c
https://hg.mozilla.org/integration/mozilla-inbound/rev/80142bab0f14
https://hg.mozilla.org/integration/mozilla-inbound/rev/346924e6e44d
https://hg.mozilla.org/integration/mozilla-inbound/rev/518530dc92c7
https://hg.mozilla.org/integration/mozilla-inbound/rev/d84207ef55e8

Comment 11

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/d223d7cd7e62
https://hg.mozilla.org/mozilla-central/rev/dc63a7b6f073
https://hg.mozilla.org/mozilla-central/rev/7e996612b3b1
https://hg.mozilla.org/mozilla-central/rev/0af9b31d714c
https://hg.mozilla.org/mozilla-central/rev/80142bab0f14
https://hg.mozilla.org/mozilla-central/rev/346924e6e44d
https://hg.mozilla.org/mozilla-central/rev/518530dc92c7
https://hg.mozilla.org/mozilla-central/rev/d84207ef55e8

Comment 12

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

Unfortunately this seems to break e10s plugin reftests. You can reproduce as follows:
  mach reftest --e10s dom/plugins/test/reftest/reftest.list

We're currently running e10s reftests on all checkins but they're hidden because of this problem. Rob, if it takes longer than a few hours to fix this, would it be okay if we back this change out?

Comment 13

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

Also, if it helps, this is the bad cset:

The first bad revision is:
changeset:   213937:80142bab0f14
user:        Robert O'Callahan <[e-mail]>
date:        Tue Nov 04 17:05:44 2014 +1300
summary:     Bug 1083635. Part 4.2: Create nsDOMWindowUtils::UpdateLayerTree and use it to flush layers in reftest content processes instead of doing a DRAW_WIDGET_LAYERS drawWindow. r=mattwoodrow

Comment 14

[Robert O'Callahan (:roc) (email my personal email if necessary)]

(In reply to Bill McCloskey (:billm) from comment #12)
> Unfortunately this seems to break e10s plugin reftests. You can reproduce as
> follows:
>   mach reftest --e10s dom/plugins/test/reftest/reftest.list
> 
> We're currently running e10s reftests on all checkins but they're hidden
> because of this problem.

If they're hidden because of this problem, why didn't they show up in https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=c45a80b4dd60 ?

> Rob, if it takes longer than a few hours to fix
> this, would it be okay if we back this change out?

OK I guess but let me try to fix it.

Comment 15

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

(In reply to Robert O'Callahan (:roc) (Mozilla Corporation) from comment #14)
> (In reply to Bill McCloskey (:billm) from comment #12)
> If they're hidden because of this problem, why didn't they show up in
> https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=c45a80b4dd60 ?

These tests are new. It takes about a day to get tests running on tbpl once we know they're green. We just got unlucky in that this bug regressed the tests during that short time. There was nothing you could have done to avoid this.

Comment 16

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: Part 7: Fix UpdateLayerTree to sync-decode im pending notifications

This fixes the e10s plugins issue.

Comment 17

[Robert O'Callahan (:roc) (email my personal email if necessary)]

https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=9ce4a794e3a1

Comment 18

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

Thanks!

Comment 19

[Robert O'Callahan (:roc) (email my personal email if necessary)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/5bebfbefb3c2

Comment 20

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/5bebfbefb3c2