1200399 - The dynamic tool bar in Fennec will some times permanently hide

Status: RESOLVED FIXED

(Core :: Panning and Zooming, defect)

Description

[Randall Barker [:rbarker]]

When scrolling page http://jsbin.com/nekumuyonu/3 with C++APZ enabled in Fennec. The dynamic toolbar will hide, but scrolling down does not make it visible again.

Comment 1

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

I wasn't able to reproduce on the above jsbin, but I can reproduce on this page: https://people.mozilla.org/~kgupta/griddiv.html - just scroll the page down and then you won't be able to scroll back up.

Comment 2

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

On this page you shouldn't be able to scroll at all to begin with, so the computed page size is probably wrong, possibly coming from an incorrect CSS viewport.

Comment 3

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

At least one problem here is that the metrics being picked up by the Java code are for the subdocument rather than the parent document. i.e. in the ImmutableViewportMetrics on the Java side I'm seeing a page size of 2000x2000 which is the size of the scrollable div, not the size of the document.

Comment 4

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

Attachment: Patch

The problem here was pretty straightforward - Java was getting the wrong metrics because the C++ code was sending the metrics from every scrollable layer up, and whatever layer was processed last would be the one that Java ended up with. Instead we should only be sending the metrics for the root content document, because that's what the Java code cares about.

Comment 5

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

Comment on attachment 8656052 [details] [diff] [review]
Patch

This ain't quite right...

Comment 6

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

Attachment: Patch v2

Here's a better version. the mIsFirstPaint and mLayersUpdated flags are only used by Fennec so I moved them into the ifdef as well. On Fennec with the Java PZC we should never be executing this code anyway because we'll hit the "continue" condition at the top of the loop every time.

Comment 7

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/4c8fcc50b299

Comment 8

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/4c8fcc50b299

Comment 9

[Kevin Brosnan [Ex-Mozilla]]

I wonder if this will resolve bug 1006294, which I still see from time to time. Setting a see also so I can close out the older bug if I don't see it any more when C++ APZ makes it onto the trains.