We found an issue affecting our testing since last week. After analyzing, we found it would be fine if we switched off e10s. So, we look back on patches, blogs, and bugs. We suspect the issue might be from the fix[1][2] of last week, but we are not sure. We point out this bug just because that it's the most recent fix on graphic related code and almost matches the exact day we started to see this bug. Please also see the attached video for your reference. [3] Environment: Windows 7 only STR: 1. install a recent Firefox Nightly 2. make sure e10s is on (default on) 3. load a google document page you already have or any complex pages like yahoo usa homepage Expected result: The screen should be normal with all buttons and content show up. Actual result: There are missing buttons, missing contents, and strange contents. [1] https://blog.nightly.mozilla.org/2016/12/20/these-weeks-in-firefox-issue-7/ [2] https://bugzilla.mozilla.org/show_bug.cgi?id=1323586 [3] https://drive.google.com/file/d/0B8iJiMtqUSFRUGE0NGZzd3FQTm8/view

After doing mozregression, I have found the suspected root cause might be from bug 1325199. (Mike Conley, sorry for my mistake!) 2017-01-04T11:23:37: INFO : application_buildid: 20161222010952 2017-01-04T11:23:37: INFO : application_changeset: 66b31b9d9a1a636ef8855e578867bdb34c435546 2017-01-04T11:23:37: INFO : application_display_name: Nightly 2017-01-04T11:23:37: INFO : application_id: {ec8030f7-c20a-464f-9b0e-13a3a9e97384} 2017-01-04T11:23:37: INFO : application_name: Firefox 2017-01-04T11:23:37: INFO : application_remotingname: firefox 2017-01-04T11:23:37: INFO : application_repository: https://hg.mozilla.org/integration/mozilla-inbound 2017-01-04T11:23:37: INFO : application_vendor: Mozilla 2017-01-04T11:23:37: INFO : application_version: 53.0a1 2017-01-04T11:23:37: INFO : platform_buildid: 20161222010952 2017-01-04T11:23:37: INFO : platform_changeset: 66b31b9d9a1a636ef8855e578867bdb34c435546 2017-01-04T11:23:37: INFO : platform_repository: https://hg.mozilla.org/integration/mozilla-inbound 2017-01-04T11:23:37: INFO : platform_version: 53.0a1 2017-01-04T11:23:39: INFO : [GPU 6492] WARNING: pipe error: 109: file c:/builds/moz2_slave/m-in-w64-000000000000000000000/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346 2017-01-04T11:23:45: INFO : [Parent 7664] WARNING: pipe error: 109: file c:/builds/moz2_slave/m-in-w64-000000000000000000000/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346 2017-01-04T11:23:47: INFO : Narrowed inbound regression window from [3fe0ff6b, 96cf05fa] (4 revisions) to [66b31b9d, 96cf05fa] (2 revisions) (~1 steps left) 2017-01-04T11:23:47: DEBUG : Starting merge handling... 2017-01-04T11:23:47: DEBUG : Using url: https://hg.mozilla.org/integration/mozilla-inbound/json-pushes?changeset=96cf05faacfe6ba130553323bde9f0bb80298c8a&full=1 2017-01-04T11:23:47: DEBUG : Found commit message: Bug 1325199: Only read system parameters for ClearType when we receive a paint event. r=jrmuizel MozReview-Commit-ID: EjQyCcdWXEc 2017-01-04T11:23:47: INFO : The bisection is done. 2017-01-04T11:23:47: INFO : Stopped The full log is attached.

Bas, your fix for gfxDWriteFont bug 1325199 appears to have caused an e10s painting regression that is blocking the Hasal team's Quantum performance profiling.

I had just updated to 20170103 build from 20161206. Initially my tabs were fine. This morning they are all quite unusable, including new tabs. Only new non-e10s windows display correctly, confirming comment 0.

(In reply to Chris Peterson (out until January 2) [:cpeterson] from comment #2) > Bas, your fix for gfxDWriteFont bug 1325199 appears to have caused an e10s > painting regression that is blocking the Hasal team's Quantum performance > profiling. I will have a look, but that seems really, really strange. At best that should accidentally disable cleartype where we don't want to. But I don't see how that could cause things to disappear.

(In reply to Wayne Mery (:wsmwk, NI for questions) from comment #3) > I had just updated to 20170103 build from 20161206. Initially my tabs were > fine. > This morning they are all quite unusable, including new tabs. > Only new non-e10s windows display correctly, confirming comment 0. turns out my example was only reproduced when using firefox via remote desktop - as described in bug 1325743