917060 - Column layout flicker

Status: RESOLVED WORKSFORME

(Core :: Layout, defect)

Description

[Mats Palmgren (inactive)]

STR:
1. load the URL
2. wait 10-15 seconds

ACTUAL RESULTS
The main article text flicker every 5-10 seconds or so.
It's in a -moz-column-count:2 layout according to Inspector.

PLATFORMS AND BUILDS TESTED
Bug occurs in Firefox Nightly 26.0a1 (2013-09-12) on Linux64
Bug does not occur in Firefox 23.0.1 on Linux64
Bug does not occur in my local beta (24) and aurora (25) debug builds.
Bug does not occur in Chrome.

Comment 1

[Alice0775 White]

Regression window(m-c)
Good:
http://hg.mozilla.org/mozilla-central/rev/3d20597e0a07
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:26.0) Gecko/20100101 Firefox/26.0 ID:20130811002840
Bad:
http://hg.mozilla.org/mozilla-central/rev/f057fca09627
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:26.0) Gecko/20100101 Firefox/26.0 ID:20130812022940
Pushlog:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=3d20597e0a07&tochange=f057fca09627


Regression window(m-i)
Good:
http://hg.mozilla.org/integration/mozilla-inbound/rev/9813b854b713
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:26.0) Gecko/20100101 Firefox/26.0 ID:20130810082041
Bad:
http://hg.mozilla.org/integration/mozilla-inbound/rev/072fffc8a0d1
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:26.0) Gecko/20100101 Firefox/26.0 ID:20130810095641
Pushlog:
http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=9813b854b713&tochange=072fffc8a0d1
Regressed by 072fffc8a0d1	Mats Palmgren — Bug 883568 - Call UpdateOverflow() in the scroll frame base class to actually update the overflow areas. r=roc

Comment 2

[Lukas Blakk [:lsblakk] use ?needinfo]

How common is this layout?  Can QA test out this bug on more frequently accessed sites to help find a testcase and to confirm user impact?

Comment 3

[Mihai Morar, (:MihaiMorar)]

I can easy reproduce this issue on Windows 7 x64 too using Latest Nightly 27. I think this is more related to JS because in the Error Console there is an JS error displayed while loading 

http://www.svd.se/kultur/kultursvepet-16-september_8522326.svd?sidan=2

Error: Permission denied to access property 'toString' this error is also related to Bug 783260 

I would have said NPAPI could be also involved while handling certain content types but in this case there are no audio/mp3 or other content types used. Removing for now qawanted. Lukas if there is anything else I can investigate here please let me know.

Comment 4

[Mats Palmgren (inactive)]

Re: comment 3.  Can you still reproduce this in the latest Nightly?
(it works for me on Linux64)

Comment 5

[Mihai Morar, (:MihaiMorar)]

(In reply to Mats Palmgren (:mats) from comment #4)
> Re: comment 3.  Can you still reproduce this in the latest Nightly?
> (it works for me on Linux64)

I am not able to reproduce this issue using Latest Nightly 28 on Windows 7 x64

Mozilla/5.0 (Windows NT 6.1; WOW64; rv:28.0) Gecko/20100101 Firefox/28.0 BuildID: 20131106030200

Marking as WFM based on my investigation done here and per Comment 4.

Comment 6

[Mats Palmgren (inactive)]

Attachment: Testcase

I can reproduce the problem with this testcase in Beta and Aurora,
but not in Nightly on Linux64.

Comment 7

[Mats Palmgren (inactive)]

The problem was fixed on m-c in this interval:  2013-10-17-03-02-01 -- 2013-10-18-03-02-06
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=423b9c30c73d&tochange=4e7d1e2c93a6

Comment 8

[Mats Palmgren (inactive)]

Bug 926706 is what fixed it.  I've verified that applying the patch locally
in a mozilla-beta build fixes it.

Comment 9

[Mats Palmgren (inactive)]

I think column layout is still quite rare on the web in general, and the specific
conditions that triggers this bug seems like it would be even more rare.
That said, when it does occur it's quite disturbing, as the URL shows.

So I think our best option for Aurora/Beta is to land bug 926706 and bug 928607
there.  I think that would be straightforward and low-risk.  Alternatively,
not do anything and let the fixes ride the trains.  Backing out bug 883568
seems like it would be riskier.

That said, I think there is still an underlying bug in column reflow here
in that reflowing it multiple times leads to different results.

roc, what do you think about the risk of landing bug 926706 and bug 928607
on branches?

Comment 10

[Mats Palmgren (inactive)]

Actually, it's only Beta (26) that is affected, my Aurora build wasn't fully up-to-date
when I tested it the first time.  (bug 926706 and bug 928607 is already in 27)

Comment 11

[Mats Palmgren (inactive)]

Attachment: Reduced testcase

Comment 12

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

(In reply to Mats Palmgren (:mats) from comment #9)
> roc, what do you think about the risk of landing bug 926706 and bug 928607
> on branches?

They're straightforward patches and there has been bake time. Let's try and get them on beta.

Comment 13

[Lukas Blakk [:lsblakk] use ?needinfo]

I've approved both uplifts in bug 926706 and bug 928607 - lets keep this bug tracking to confirm this works as expected.

Comment 14

[Mats Palmgren (inactive)]

I've landed those on Beta(26).  It fixed the problem in my local mozilla-beta build.
I filed bug 936652 to followup on the possible underlying bug in column layout.

Comment 15

[Mats Palmgren (inactive)]

We should try to make a reftest from the attached testcase.