787397 - Invalidation problem in gaia contacts app (with -moz-element and transform transitions)

Status: RESOLVED WORKSFORME

(Core :: Layout, defect)

Description

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

Vivien, Etienne: I talked to some Telefonica guys about this but I don't know their bugzilla handles.  If you could CC them here, that'd be great.  Thanks!

STR
 (1) Load "UI Tests" app
 (2) Choose "contacts"
 (3) "Insert fake contacts"
 (4) Get a cup of coffee, wait for it to finish
 (5) Open contacts app
 (6) Scroll up and down.  Should be 50+ fps.
 (7) Tap a contact to see details
 (8) Press "X" in top-left to return to main screen
 (9) Scroll up and down again

The panning fps drops quite a bit, to ~17 fps IIRC.  When transitioning from main screen to details screen, gaia takes a -moz-element capture of the main screen.  Aside: it does this to work around bug 779968 :(.  But anyway, when returning to the main screen, something about the <div> with -moz-element capture background causes bad invalidation behavior.

DLBI mostly fixes this.  However, the "Contacts" title at the top still invalidates after the transition to details view, even with dlbi.  Because this is a CSS gradient, performance suffers a bit.

Comment 1

[Ismael Gonzalez [:basiclines]]

Please CC me on this issue:
[e-mail]

Comment 2

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

I'm not in a position to test this atm, because I don't have a source of fake contacts.

Comment 3

[Ismael Gonzalez [:basiclines]]

@Chris there is an application called, "uitest" in Gaia, you can generate at least 500 contacts with it.

Comment 4

[alberto.pastor]

I can not see repaintings anymore in Contacts list after transitions. @Ismael, can you confirm?

Comment 5

[Ismael Gonzalez [:basiclines]]

It does not happen anymore in contact list section, but it still happening in contact details one!

Comment 6

[alberto.pastor]

I don't think that issue is related to this bug. I already have a Pull Request ready to stop repainting the image (using transform instead background-position). I need more time to check what's happening with the details, but I'll open a new bug if needed.