Electrolysis/Meetings/2016-08-25
From MozillaWiki
< Electrolysis | Meetings
« previous week | index | next week »
mconley
- bug 1290280 - [e10s] Tab crashes on startup
- This got uplifted to Aurora and verified last week.
- Seems to have cleared the FX_SESSION_RESTORE_RESTORE_WINDOW_MS regression on Aurora.
- There's been a similar improvement on Nightly, but the magnitude isn't as great. Going to wait for a few more days of data before filing a new bug to see if we can figure out what the difference is between Nightly and Aurora.
- Clearing fallout from bug 1261842 - Make initial browser remote sooner if we're defaulting to using remote tabs
- Solved very frequent orange bug 1290567 - "ERROR: GC found live Cell ... at shutdown" errors since bug 1261842
- Mostly solved bug 1294456 - [e10s] After opening a tab/window Marionette hangs if any command is send which executes in the listener.js frame script
jimicy
- LANDED bug 1261299 - [e10s] Mac OS X Services is not available for e10s tab content
- bug 1297539 - Handle eContentCommandPasteTransferable for remote targets
- Patches waiting on review
- Needed to be able to paste from Mac OS X Services
kanru
- bug 1264642 RESOLVED::FIXED Reduce the contiguous address space needed for StructuredClone serialization
- Landed
- bug 1269036 NEW:: crash in mozalloc_abort | NS_DebugBreak | mozilla::ipc::LogicError | mozilla::dom::PBrowser::Transition from ABORT: __delete__()d actor
- Landed
- bug 1051567 NEW:: [e10s] Crash in [@ mozilla::ipc::MessageChannel::OnChannelErrorFromLink]
- Mac can easily run out of file descriptors.
- bug 1296756 NEW:: Record number of open file descriptors in crash report
ting
- bug 1297314 NEW:: Crash in mozilla::ipc::FatalError | mozilla::dom::PContentParent::Write from ContentParent::RecvKeywordToURI()
- Landed
- bug 1145613 NEW:: Crash in [@ ReleaseData ] & [@ ReleaseData(void*, unsigned int)]
- added some other similar crashes with different signatures
- checked the crash annotations, am double checking the code around nsHttp::Host
mrbkap
- bug 1165309 - View source (and print preview) don't work in e10s-multi.
- Landed with a couple of followups needed.
shell
not attending today due to boring dental conflict - would totally prefer e10s meeting
- Telemetry to validate the system add-on is targeting the correct add-ons, then move ahead with finalizing inclusion of add-ons/e10s "experiment" in Release 49
- Estimation of Release population that would qualify (based off the 2.48% of Beta that met criteria) is 5-7 million new users qualifying for e10s in Release.
- There would be no throttling capabilities in 49 - so would go to all at once. First few weeks of Beta have shown no issues created by the add-on/e10s inclusion with Stability or Performance.
- Erin let Release Management know (thank you) and Shell will finish of details needed for Felipe and Dave Zeber and add-on developers to complete.
- After talking with Billm - he showed us how a few (5 according to telemetry) add-ons are calling CPOWs directly. This hadn't been expected. Assumed developers would use the Shims, so MPC:true doesn't block calling CPOWs.
- Plan to address
- whitelist the 5 existing that have so they don't break (video download helper is an example - so not small add-ons, can expand if list increases)
- change MPC:true to also block CPOW usage (in addition to shims)
- The code for the top 2 bugs is written.
- Action still needed / being worked on by add-on team:
- Add-on team outreach to devs that have done it ask them to fix it. we won't break them soon (because of whitelist). We will need them to change code, since the CPOWs will go away with Shims. If they are waiting on something from webextensions or SDK fixes in next couple of months - they can do that. But we left a gap that we're closing in 2017 for certain.
- docs? do we need to document that this won't work anymore? is it something we think more people will try?
- testing with the system add-on "experiment list" (since VDH is on both whitelist and experiment list)
- Plan to address
aklotz
- Windows a11y available on Nightly as of Tuesday!
- Still prefed off until initial QA pass is done and any showstoppers are dealt with.
- bug 1297549 Arrow keys will not read the page with a11y and e10s - Behaving differently between in-proc (ie, injected DLL) and out-of-proc a11y clients.
- Next steps:
- Re-enabling some tests that were disabled for landing
- Performance optimization