1041031 - Microsoft Skype Lync's windowed plugin has Z-order problems with video starting in Firefox 30 and fixed in Firefox 33

Status: RESOLVED INCOMPLETE

(Core Graveyard :: Plug-ins, defect)

Description

[Chris Peterson [:cpeterson]]

Yovka Pencheva, test lead of Lync Web Experience team in Skype, reported a plugin problem that seems to be a regression in Firefox 30 but fixed in Nightly 33:

We have a browser plugin for Microsoft Lync web application which uses Windowed API and renders video/desktop sharing in OS-level windows.  We have not experienced any issues with z-order and layers with Firefox prior 30/31 version.  Starting Firefox 30 however, Lync plugin video/desktop sharing windows stay always on the top of all other windows/iframes.  This breaks our user experience in certain use cases.

We have tried latest nightly build (firefox-33.0a1.en-US.win32.installer.exe) and noticed the behavior changed back again as it was in old versions of Firefox (prior 30), we don’t have issues with plugin windows staying always on top.

Have you done some intentional change in FF 30 related to Windowed API? Or is this an accidental change/bug which got fixed in FF 33?  If yes, do you plan to backport it to FF 31, as it looks FF 33 goes out in mid of Oct.

Comment 1

[Chris Peterson [:cpeterson]]

Yovka says Beta 31 is broken and Nightly 33 is fixed. We need to verify whether Aurora 32 is also fixed or needs a fix uplifted from 33.

Comment 2

[:Gavin Sharp [email: [e-mail]]]

Chris, can Yovka test Beta 32 and comment here with the results?

Comment 3

[Chris Peterson [:cpeterson]]

OK. I emailed Yovka at Skype about testing Beta 32.

Comment 4

[Yovka Pencheva]

Hi Gavin, Chris,

Yes, the issue is reproducible with FF 32 beta.  We are trying now to identify the 33 build where the issue is fixed using mozregression tool.  
I will also try to setup test page which can be used without Lync account. 

Thanks,
Yovka

Comment 5

[Yovka Pencheva]

You can reproduce the issue using this Lync meeting URL  
https://join.microsoft.com/meet/yopenche/P79R8BV2?sl= 

Login as guest user (just put any name), check Install Lync Web App plugin checkbox, click Join the meeting link.  The Lync web plugin will be installed (you might need to allow some actions).  
Once you join the meeting, start your video (see StartVideo.png).
Once you have your video running, click Exit Meeting for ex, this will popup modal dialog which will be displayed below the video frame (see VideoAboveModalDlg.png).

We started to experience this with FF30, and we found this fixed back in FF33.

Comment 6

[Yovka Pencheva]

Attachment: StartVideo.png

Comment 7

[Yovka Pencheva]

Attachment: VideoAboveModalDlg.png

Comment 8

[Hugh Eland]

We have found the issue is solve since version 33.0a1(2014-06-24).

Comment 9

[:Gavin Sharp [email: [e-mail]]]

Thanks for testing! Could you provide "Source" URLs from about:buildconfig in the first "fixed" build and last "affected" build?

Comment 10

[Hugh Eland]

Here is information about first version which bug was fixed
*************************************************************************************************************
about:buildconfig
Build Machine
B-2008-IX-0003
Source
Built from https://hg.mozilla.org/mozilla-central/rev/e86b84998b18
Build platform
target
i686-pc-mingw32
Build tools
Compiler Version Compiler flags
cl 1600 -TC -nologo -D_HAS_EXCEPTIONS=0 -W3 -Gy -wd4244 -wd4819 -we4553
cl 1600 TP -nologo -D_HAS_EXCEPTIONS=0 -W3 -Gy -wd4251 -wd4244 -wd4345 -wd4351 -wd4482 -wd4800 -wd4819 -we4553 -GR DNDEBUG -DTRIMMED -Zi -UDEBUG -DNDEBUG -O1 -Oi -Oy
Configure arguments
--enable-crashreporter --enable-release --enable-update-channel=nightly --enable-update-packaging --enable-jemalloc --enable-require-all-d3dc-versions --with-google-api-keyfile=/c/builds/gapi.data --enable-warnings-as-errors --enable-signmar --enable-profiling --enable-js-diagnostics

*********************************************************************************************************************************************
And here is the build information about last version which bug was exists
*********************************************************************************************************************************************

about:buildconfig
Build Machine
B-2008-IX-0084
Source
Built from https://hg.mozilla.org/mozilla-central/rev/366b5c0c02d3
Build platform
target
i686-pc-mingw32
Build tools
Compiler Version Compiler flags
cl 1600 -TC -nologo -D_HAS_EXCEPTIONS=0 -W3 -Gy -wd4244 -wd4819 -we4553
cl 1600 TP -nologo -D_HAS_EXCEPTIONS=0 -W3 -Gy -wd4251 -wd4244 -wd4345 -wd4351 -wd4482 -wd4800 -wd4819 -we4553 -GR DNDEBUG -DTRIMMED -Zi -UDEBUG -DNDEBUG -O1 -Oi -Oy
Configure arguments
--enable-crashreporter --enable-release --enable-update-channel=nightly --enable-update-packaging --enable-jemalloc --enable-require-all-d3dc-versions --with-google-api-keyfile=/c/builds/gapi.data --enable-warnings-as-errors --enable-signmar --enable-profiling --enable-js-diagnostics

Comment 11

[Chris Peterson [:cpeterson]]

Thanks, Hugh!

Based on comment 10, this bug was fixed in this pushlog between Nightly 33 build 2014-06-21 to 2014-06-23:

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=366b5c0c02d3&tochange=e86b84998b18

* kats: could your fix for bug 1021085 have also fixed this Skype plugin's video rendering on Windows? Bug 1021085 touches gfx code and is in the "unregression range". (We're looking for an upliftable fix.) However, your fix was uplifted from Nightly 33 to Aurora 32 and this Skype problem is reproducible in Beta 32 and not Aurora 33.

Comment 12

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

I think it's unlikely my changes would have affected this. They should only have affected B2G and Fennec since those are the platforms where tiling and progressive painting is enabled.

Comment 13

[Chris Peterson [:cpeterson]]

Hugh, I don't see any obviously code changes that might have fixed the plugin problem between those two Nightly builds.

If you are able to run the mozregression script [1] to reproduce the bug, you can bisect the "inbound builds" which have fewer code changes. (The "inbound builds" are the incremental Firefox builds during the day between the official Nightly releases.) That will help is pinpoint the exact bug fix so we can merge it to Firefox Beta 32. You can

  mozregression --good=2014-06-21 --bad=2014-06-23


[1] http://mozilla.github.io/mozregression/

Comment 14

[Hugh Eland]

Sorry - we went back and tested again - the testers had found another issue masking this one - now we have correctly identified the last and first good builds.

Here is information about first version which bug was fixed
**************************************************************************************************************
bout:buildconfig
Build Machine
B-2008-IX-0095
Source
Built from https://hg.mozilla.org/mozilla-central/rev/776f6d341b3f
Build platform
target
i686-pc-mingw32
Build tools
Compiler Version Compiler flags
cl 1600 -TC -nologo -D_HAS_EXCEPTIONS=0 -W3 -Gy -wd4244 -wd4819 -we4553
cl 1600 TP -nologo -D_HAS_EXCEPTIONS=0 -W3 -Gy -wd4251 -wd4244 -wd4345 -wd4351 -wd4482 -wd4800 -wd4819 -we4553 -GR DNDEBUG -DTRIMMED -Zi -UDEBUG -DNDEBUG -O1 -Oi -Oy
Configure arguments
--enable-crashreporter --enable-release --enable-update-channel=nightly --enable-update-packaging --enable-jemalloc --enable-require-all-d3dc-versions --with-google-api-keyfile=/c/builds/gapi.data --enable-warnings-as-errors --enable-signmar --enable-profiling --enable-js-diagnostics

*********************************************************************************************************************************************
And here is the build information about last version in which bug was exists
*********************************************************************************************************************************************
about:buildconfig
Build Machine
B-2008-IX-0136
Source
Built from https://hg.mozilla.org/mozilla-central/rev/bdac18bd6c74
Build platform
target
i686-pc-mingw32
Build tools
Compiler Version Compiler flags
cl 1600 -TC -nologo -D_HAS_EXCEPTIONS=0 -W3 -Gy -wd4244 -wd4819 -we4553
cl 1600 TP -nologo -D_HAS_EXCEPTIONS=0 -W3 -Gy -wd4251 -wd4244 -wd4345 -wd4351 -wd4482 -wd4800 -wd4819 -we4553 -GR DNDEBUG -DTRIMMED -Zi -UDEBUG -DNDEBUG -O1 -Oi -Oy
Configure arguments
--enable-crashreporter --enable-release --enable-update-channel=nightly --enable-update-packaging --enable-jemalloc --enable-require-all-d3dc-versions --with-google-api-keyfile=/c/builds/gapi.data --enable-warnings-as-errors --enable-signmar --enable-profiling --enable-js-diagnostics

So - the previous data was incorrect.  My apologies for this.

Comment 15

[Chris Peterson [:cpeterson]]

Here is the new pushlog for the fix range in comment 14:

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=bdac18bd6c74&tochange=776f6d341b3f

* tn: do you think your fix for layer bug 977831 could have also fixed this plugin Z-order problem? This bug seems to have been fixed, but we're trying to isolate the fix so we can uplift it from Aurora 33 to Beta 32.

Comment 16

[Timothy Nikkel (:tnikkel)]

No, I don't think bug 977831 can be involved. It is only a fix for code that is not enabled anywhere right now (and when it's enabled it will only be used on b2g and maybe fennec).

Comment 17

[Timothy Nikkel (:tnikkel)]

Nothing jumps out at me in that pushlog to blame. Can we narrow this down more using inbound builds?

Comment 18

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

Given that we have already fixed this issue in 33, I don't think has the same priority for 32. I'm dropping tracking bug will gladly consider an uplift request.

Comment 19

[Chris Peterson [:cpeterson]]

We never identified the regression point or fix point, but this bug is apparently not reproducible in Beta 33, which will be released next week.