1089382 - When the sidebar is position:fixed, some of its contents will disappear

Status: VERIFIED FIXED

(Core :: Layout, defect)

Description

[Luís Miguel [:quicksaver] (not reading bugmail)]

Attachment: OmniSidebar 1.5a1

Last good build: firefox-35.0a1-2014-09-24
First bad build: firefox-35.0a1-2014-09-25

STR:

1. Install OmniSidebar xpi package attached
(1.1. Bookmarks sidebar should open, if not, open the sidebar, by clicking the window margin or hitting Ctrl+B)
2. Open add-ons manager in the sidebar (click the sidebar title for a dropdown of options to open)
3. Make sure you're in a pane that you can scroll
4. Undock the sidebar (button next to the sidebar close button), so it appears over the webpage instead of next to it
5. Un-maximize the window (normal mode)
6. Try scrolling the add-ons list

For example, if I try scrolling my add-ons list, it will disappear. This issue does not occur if the window is maximized, which makes me think this may not be something I can fix within the add-on. Also, I can reproduce this in Beta, which makes it kind of urgent! It has always worked correctly in previous versions of firefox, including the current release (33). The issue also doesn't happen if the sidebar is docked (not position:fixed), or if I disable pref layers.offmainthreadcomposition.enabled (requires restart).

[Tracking Requested - why for this release]: If my suspicion that I can't fix this within the add-on is correct, the add-on will break in Firefox 34 and there's nothing I can do about it. (I hope I'm using the tracking flag correctly for this, please let me know if not.)

Comment 1

[Luís Miguel [:quicksaver] (not reading bugmail)]

Attachment: Sidebar empties when scrolled

Comment 2

[Luís Miguel [:quicksaver] (not reading bugmail)]

Bug 1062100 was the only bug in that range I could find (unless I missed any others?) that was also uplifted to aurora at the time (34). I don't have enough knowledge to be sure if it really is the cause for this though. roc, since you worked on it, do you think it could be?

Comment 3

[Mats Palmgren (inactive)]

Which revisions were the builds you tested built from?
(load about:buildconfig in each build to see its rev)

Comment 4

[Luís Miguel [:quicksaver] (not reading bugmail)]

Last good build: firefox-35.0a1-2014-09-24 1e2993c99323
First bad build: firefox-35.0a1-2014-09-25 1735ff2bb23e

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=1e2993c99323&tochange=1735ff2bb23e

I forgot to mention before, in the screenshot you can see that the scrollbar's thumb also disappears, but that's probably unrelated: bug 1089380.

Comment 5

[Luís Miguel [:quicksaver] (not reading bugmail)]

Revised regression range: https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=435732392989&tochange=1735ff2bb23e

Bug 1062100 is the only one in that range that was uplifted to aurora and not to beta at the time.

Bug 1005322, bug 1059573, bug 1061435, bug 1062483, bug 1068394, bug 1069857 and bug 1071704 were also uplifted to aurora, but they seem specific to b2g/FxOS so I don't think they would cause this.

Bug 1039993, bug 1062920 and bug 1070764 were also uplifted to aurora at the time, but don't seem relevant here.

Comment 6

[Luís Miguel [:quicksaver] (not reading bugmail)]

Confirmed this was caused by bug 1062100: https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=9f9b3c194a76&tochange=1bdf7f5d1a6d

Roc, what are your thoughts?

Comment 7

[Mats Palmgren (inactive)]

FYI, the convention is to mark a regression bug as blocking the bug that caused it.

Comment 8

[Lawrence Mandel [:lmandel] (use needinfo)]

Jet - Can you please review and see about finding an owner?

Comment 9

[Luís Miguel [:quicksaver] (not reading bugmail)]

Ditto from bug 1089380 comment 18: I've been trying to come up with a more reduced test-case, but for example just setting position:fixed to #sidebar-box in a clean profile doesn't seem to trigger this bug by itself. Apparently it's the whole combination of CSS properties that the add-on is setting that causes the bug, and there are quite a few of them, I haven't been able to find out which ones are actually relevant...

Please let me know if you do need a more reduced test-case and I will try again.

Comment 10

[Robert O'Callahan (:roc) (email my personal email if necessary)]

I don't see this on Linux. I guess I'll have to try it at home on Windows.

Are you sure about the regression range? It sounds like this could be a dup of bug 1089380 otherwise.

Comment 11

[Luís Miguel [:quicksaver] (not reading bugmail)]

I'll confirm that if it also doesn't happen for me on linux.

And that's a yes about the regression range, this issue started much later than bug 1089380.

Comment 12

[Luís Miguel [:quicksaver] (not reading bugmail)]

(In reply to Luís Miguel [:Quicksaver] from comment #11)
> I'll confirm that if it also doesn't happen for me on linux.
> 
> And that's a yes about the regression range, this issue started much later
> than bug 1089380.

It could be a combination of both of course, their symptoms are very similar and so fixing bug 1089380 could still also fix this.

Comment 13

[Luís Miguel [:quicksaver] (not reading bugmail)]

Confirmed this in Windows, Mac and Linux. To reproduce needs these preferences set:
layers.acceleration.disabled=false
layers.offmainthreadcomposition.enabled=true

In addition in Linux also had to set:
layers.acceleration.force-enabled=true

Comment 14

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Was this fixed by the fix for bug 1089380?

Comment 15

[Luís Miguel [:quicksaver] (not reading bugmail)]

Yep, bug 1089380 fixed it, tested in the latest inbound.

I'm not marking this a duplicate though, as like I said it appeared only after bug 1062100 was applied. Please change the status if you think it's otherwise appropriate.

Comment 16

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Thanks!

Doesn't really matter whether we mark this as a dup or not.

Comment 17

[Lawrence Mandel [:lmandel] (use needinfo)]

Marking 34 and 35 as fixed now that bug 1089380 has landed on both branches.

Comment 18

[Luís Miguel [:quicksaver] (not reading bugmail)]

Following STR in comment 0, verified fixed on latest beta, aurora (developer ed.), and nightly. (Fixed by bug 1089380)

Comment 19

[[SV Manager] Florin Mezei, QA (:FlorinMezei)]

(In reply to Luís Miguel [:Quicksaver] from comment #18)
> Following STR in comment 0, verified fixed on latest beta, aurora (developer
> ed.), and nightly. (Fixed by bug 1089380)

Marking as Verified so we get it off our radar. Thanks Luis!