1191961 - With APZ enabled some reftests fail due to scrollframe activation in the test page but not the reference page

Status: RESOLVED FIXED

(Core :: Panning and Zooming, defect)

Description

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

With APZ enabled, some of the Linux R-e10s tests fail because the test page activates a scrollframe but the reference page does not. This results in a layerization difference that causes a few pixel differences. We can force layerization in the pages using "will-change: scroll-position" to make sure both test and reference pages end up with the same layerization.

Comment 1

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

Attachment: Patch

Comment 2

[:kip (Kearwood Gilbert)]

Comment on attachment 8644554 [details] [diff] [review]
Patch

Review of attachment 8644554 [details] [diff] [review]:
-----------------------------------------------------------------

Looks good

Comment 3

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

Thanks. tn, are you ok with this or would you prefer fuzzing it?

Comment 4

[Timothy Nikkel (:tnikkel)]

I'm fine with this.

Comment 5

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/24f1bbabe995

Comment 6

[Botond Ballo [:botond]]

Do we know why a layerization difference causes rendering differences? It that something worth addressing?

Comment 7

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

If there's a layerization difference the final pixels are produced by a different codepath - in one case layout rasterizes it all together and in the other case the compositor backend is what combines the pixels. Assuming that the pixel differences are only off by small amounts and in areas where layers are being combined due to alpha I (which has been the case so far) I think it's probably not worth worrying about.

Comment 8

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/24f1bbabe995