752994 - Fixed position element doesn't scroll with viewport

Status: RESOLVED WORKSFORME

(Core :: Layout, defect)

Description

[Jeff Balogh (:jbalogh)]

Attachment: pagely before and after scrolling

STR:

1. Load page.ly.
2. Scroll down to the bottom.
3. Scroll back up to the top.

Watch the page header flash multiple colors and textures before it figures out the right rendering.

Bonus: sometimes it gets stuck with a white background.

Screenshot left: stable image after page load
Screenshot right: temporary jagged rendering after scrolling down and up. Eventually it comes back to the stable image

Comment 1

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

Getting stuck with the white background seems to happen as soon as I scroll over to the right a little bit. This is really weird, and I have no idea what might be causing this. The page itself uses responsive design, position:fixed elements, and gradients. Craziness!

Comment 2

[Johnathan Nightingale [:johnath]]

Release blocking+. Thanks, Jeff - you've been filing some really helpful bugs.

Comment 3

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

One problem here is a fixed position background with an offset location; this seems to move around as the user pans, which isn't intended.  Testcase is at http://conduit.bitops.com/~vladimir/misc/fixed-pos-bg.html .  When you zoom in/out, note that the "circle" gradient in the image starts at the same place.  Zooming in/out gets it confused, and sometimes the background color bleeds through.  I've made it red in the testcase so that it shows clearly.  Some of the weird rendering could be because as we render different tiles, the background gets offset to the origin of the tile.

You can see this same effect in desktop firefox, actually; if you open that URL, zoom in a few times, then use the scrollbar to scroll left/right... the image stays in the same spot.

Comment 4

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Hm, no, ignore that; apparently it's correct because the viewport size changes.

Comment 5

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

I don't see the white, but I do see the background color instead of the background image a bunch.  Guess - we only get the background image painted in one tile due to viewport shenanigans, and it's not painted in the others because for whatever reason it's decided that it's not visible.  So we just get the background color.

Comment 6

[Brad Lassey [:blassey] (use needinfo?)]

qawanted to retest with fixed position content fixed (test on trunk)

Comment 7

[Aaron Train [:aaronmt]]

Testing on trunk (20120524, Galaxy Nexus) this is better than before but there are still issues. Panning under portrait and landscape paints black content, and is still kind of slow

Comment 8

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Black or dark blue?  Dark blue is the background colour of the top image, so it sounds like you might be seeing the fixed-position background issue not covering the entire element like I mentioned before?

Comment 9

[Aaron Train [:aaronmt]]

Attachment: Nightly (20120524) - Screenshot

Nightly (20120524) - Screenshot

Black when panning (Galaxy Nexus, 4.0.4).

Comment 10

[Erin Lancaster [:elan]]

tagging as layout for now.

Comment 11

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Note; this still happens with an 05-31 build from m-c, so it's unaffected by bug 607417 (which we weren't sure if it would affect this or not).

Comment 12

[Joe Drew (not getting mail)]

Morphing this bug based on the actual behaviour we see.

Comment 13

[Boris Zbarsky [:bzbarsky]]

So is this about fixed background attachment, or about fixed positions?  Those aren't the same thing....

Comment 14

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

In general this behaves much better on recent nightlies, but the radial gradient background at the top of the page still doesn't render fully when you're zoomed out.

Comment 15

[Mark Finkle (:mfinkle) (use needinfo?)]

Chris - Is this fixed? Or can it be duped to a different bug.

Comment 16

[Chris Lord [:cwiiis]]

The page in question works fine for me, but Vlad's test page doesn't - that's a bit annoying, as I'd have expected it to be fixed by bug 786502... Not sure if this should be closed - the title bug has been fixed and the previously broken site works for me, but Vlad's test-case shows a situation where the fixed-pos stuff should work and it doesn't :/

Perhaps a new bug should be opened specifically for Vlad's test-case and this should be marked as fixed/worksforme?

Comment 17

[Kevin Brosnan [Ex-Mozilla]]

Marking this bug works for me and opening a new bug for vlad's test case.