1180899 - MOZ_ASSERT(result) at gfx/layers/apz/src/APZCTreeManager.cpp:1526

Status: RESOLVED FIXED

(Core :: Panning and Zooming, defect)

Description

[Gregor Wagner [:gwagner]]

Attachment: bt

On current trunk during monkey testing with debug gecko.

Comment 1

[Timothy Nikkel (:tnikkel)]

What are the steps to reproduce? I don't know what monkey testing is.

Comment 2

[Gregor Wagner [:gwagner]]

The monkey testing framework is a script that generates random input events on B2G: https://github.com/mozilla-b2g/B2G/blob/master/scripts/run-monkey.sh

I believe it happened during FTU when it opened a link in a new window but I don't have a real STR.
I could add a logging patch and see if it reproduces.
Botond knows how to run this script. Maybe he can reproduce.

Comment 3

[Botond Ballo [:botond]]

I can try. ni'ing myself so I don't forget.

Comment 4

[Gregor Wagner [:gwagner]]

I found STR:

Run the FTU (First time usage app)
In the geolocation panel open the privacy link
Open Marketplace and start touching the screen.

Comment 5

[Botond Ballo [:botond]]

Attachment: b2g-ftu-marketplace.png

(In reply to Gregor Wagner [:gwagner] from comment #4)
> I found STR:
> 
> Run the FTU (First time usage app)
> In the geolocation panel open the privacy link
> Open Marketplace and start touching the screen.

When I try following this STR, this is what I see when I got to Marketplace. Not sure what's up with that...

Comment 6

[Botond Ballo [:botond]]

Tried again today, that broken-looking screen is gone, but the Marketplace thinks I'm not connected to the Internet (even though I am, and browsing works...).

Comment 7

[Gregor Wagner [:gwagner]]

I can still repro on current trunk with debug gecko:
https://www.youtube.com/watch?v=ThL4JgM0BPY

Comment 8

[Botond Ballo [:botond]]

(In reply to Gregor Wagner [:gwagner] from comment #7)
> I can still repro on current trunk with debug gecko:
> https://www.youtube.com/watch?v=ThL4JgM0BPY

Thanks, that helped me repro it! (The trick was to tap quickly after pressing "Marketplace" (without waiting for the resulting page to load), and to tap near the rocketbar area.)

Comment 9

[Botond Ballo [:botond]]

Attachment: APZC tree dump with layers-id-roots annotated

Attached is an APZC tree dump with the layers-id-roots annotated:

It seems to be in order, i.e. each layers id has a layers-id-root APZC. I would expect the breadth-first search on the line preceding the assertion to find one of these. I'll have to debug it to see why it doesn't.

Comment 10

[Botond Ballo [:botond]]

(In reply to Botond Ballo [:botond] from comment #9)
> Created attachment 8632372 [details]
> APZC tree dump with layers-id-roots annotated
> 
> Attached is an APZC tree dump with the layers-id-roots annotated:
> 
> It seems to be in order, i.e. each layers id has a layers-id-root APZC.

Actually, I overlooked that there's a second RefLayer there with no APZC. The hit testing is a layer in that RefLayer's subtree, and then we're failing to find a layers-id-root APZC for it.

Comment 11

[Botond Ballo [:botond]]

There is a bug in the logic added in bug 1168629 for ensuring that there's always at least one layer with metrics for the root scroll frame or root element, where on a subsequent paint the metrics added on the previous paint can be clobbered.

Comment 12

[Botond Ballo [:botond]]

Attachment: MozReview Request: Bug 1180899 - Do not clobber frame metrics on the root layer added during a previous paint if there is no other frame metrics for the root scroll frame. r=tn

Bug 1180899 - Do not clobber frame metrics on the root layer added during a previous paint if there is no other frame metrics for the root scroll frame. r=tn

Comment 13

[Botond Ballo [:botond]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=831cfdbf0704

Comment 14

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/3bc7de3b8891

Comment 15

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

https://hg.mozilla.org/mozilla-central/rev/3bc7de3b8891