Electrolysis/Meetings/2016-05-12
From MozillaWiki
< Electrolysis | Meetings
Contents
jimm
- bug 1271978 (slow script telemetry patches to detect notification delays) - landed
- bug 1272116 (investigate if terminating hung script browsers causes the loss of notification counts)
- looks like may lose data here in non-e10s, although I don't think it's going to be large, the timeout is only ten seconds
- bug 1266575 (M9, click event is not dispatched to <select> when clicking on opened popup) - haven't had a chance to look at this yet.
gw280
- bug 1114647 - handed off to sandboxing guys now. Jed Davis has taken over, and waiting on Bob Owen's patches in bug 1035125. I have some review feedback from glandium for the main process rename patch, but going to wait until the sandboxing patches are ready before addressing them.
- Working on a shared (per-process instead of per-tab) TextureClientPool right now.
tchou
- bug 1245068 RESOLVED::FIXED [e10s] Ctrl+mouse wheel to change zoom level doesn't work if web page is not focused
- landed
- bug 1267329 ASSIGNED:: CreateMinidumpsAndPair doesn't record memory information
- submitted a patch for reviewing
kanru
- bug 1264642 NEW:: Reduce the size of PBrowser::Msg_AsyncMessage IPC messages
- try result with lots of leaks
- rebuild with valgrind enabled, trying to find the leaks.
- fixed one buffer overflow and one memory leak, push to try again
- meeting with :billm to discuss about rebasing onto bug 1262671
- Filed bug 1271887 to not normalize template types for IPC::ParamTraits
- IPC::ParamTraits<T>::{Read,Write} will have different signature for each type T
cervantes
- bug 1259480 NEW:: [e10s] FatalError crash in PContentChild::Read from PContentChild::SendReadPrefsArray
- try to implement dry run of message deserialization.
- try to implement message deserialization in ipdl code generator
- bug 1258317 REOPENED:: crash in mozilla::ipc::TransferHandleToProcess (PBackground related)
- pt2 patch has concerns for shared memory. Checking if it can be addressed easily for landing.
- working on the fallback mechanism on access denied in DuplicateHandle()
- filed bug 1271601 (blocking bug 1258317) for gracefully fail IPC transports
- bug 1271601 NEW:: Handling IPC transport failures more gracefully
bobowen
- bug 1035125 - On Windows, plugin-container.exe is linked against the sandbox_s library twice - patches reviewed and some changes up in response to glandium's review. Reasonable chance of landing next week, now that the VS2015 problem looks like it is resolved.
mconley
- bug 1247137 - New tab page briefly flashes when closing tab
- Fixed earlier this week, uplifted today.
- bug 1261842 - Make initial browser remote sooner if we're defaulting to using remote tabs
- Try build still shows that this is going to yield some pretty nice wins
- This was supposed to be done earlier this week, but I'm getting bogged down with nsWindowWatcher clean-ups. Hoping to have final patches up this week, or early next week at the latest. (I know I said that last week, but I'm over the hump now).
- bug 1166351 - Unable to access new tabs and unable to close any tab
- This is my Good Samaritan bug, and it's being a pretty big pain. I've found a solution, I think - but the assertion I've added is still hit in some cases. Still trying to figure out why.
- The good news is that I'm learning how to use rr, which is nice.
- Lots and lots of reviews.
gabor
- bug 1247420 - (m9 1255138 - Resizing windows from JS is not supported in e10s mode) - Had some extra manual testing on it to see if it does not regress some mixed screen related bugs. Landed.
- bug 1262661 - (Many large SessionStore:update messages) - Landed the trivial fix on the message queue (after some struggle). Implemented and testing two more optimisations, one for history and one for session storage. After a some testing my hope is that in most of the cases these optimisation will fix the problem, and they are simple enough.
felipe
- System add-on will need to update via GoFaster - preparing for that
- bug 1239552 - Marionette-e10s is permafail on beta
Neil
- bug 1101975 M9, menus overriding accesskeys, finally checked in
- bug 1272116 M9, select dropdown and invalid form popups not appearing at the right place inside a transformed frame, checked in
- Working on Linux drag and drop issues
- Jimmy is back, working on P1s
mrbkap
- bug 1267844 - m9 - Infinite throbber shown when moving tab to another window. I have a patch up that might need to go through a couple of review cycles but should be in soon.
- Fixing more bugs up next!
cpeterson
- e10s release criteria:
- https://wiki.mozilla.org/Electrolysis/Release_Criteria#Graphics_Performance
- Still waiting for Beta 47 data for stability, slow scripts, and user engagement.
- Coordinating with module owners to sign off remaining tests disabled for e10s:
blassey
- bug 1269998 - Prompt users with pending crash reports to submit them, landed, bounced
- bug 1270553 - Allow users to opt-in to auto submitting crash data
ddurst
- almost done with 1271450 (and 1265015) -- data is backfilled from 4/27 to yesterday (5/11). Scheduled for daily run.
- Meetings (with Uptime group) & reading (now on fhr-dev list).
elan
- We are meeting our milestones, thank you!
- Focusing on what the release plan is; suddenly 47 is being discussed as being a viable option. Two steps
- Extend the Beta Experiment
- GO/NO-GO for GA
chutten
- e10s_experiment.ipynb for the first two weeks of Beta47 experiment is up for review.
- Looking for changes since the previous
aklotz
- Testing rough cut of a11y+e10s on Windows. I was hoping to have a demo-able build yesterday but no luck. Problems due to...
- bug 1272146 - Custom object IDs for NotifyWinEvent/WM_GETOBJECT: The way we do a11y notifications causes a11y to always execute against the root a11y COM object, passing a child id as a parameter, instead of directly resolving the COM object for the remote document. I want to fix this but I'm encountering some friction.
ksteuber
- bug 1263774 (Memory reports sent with content crash reports) - Patch revised and seeking re-review
- bug 1206133 - A good number of tests are broken by this patch. I am working on fixing them.
haik
- bug 1266484 crash in libsystem_kernel.dylib@0x16db6 when using Pinboard bookmarklet - Approved for uplift to 48
- bug 1165945 Better handle a failure to start a content child process. - Working on this one, adding telemetry whenever starting a content process fails
mccr8
- bug 1270247 - Investigating deserialization crash for NetAddr. Turned up an unexpected case that is likely causing crashes.
- bug 1268662 - More work on reducing crashes from sending large prefs.
- filed: bug 1272018 - Use shared memory to send images for drag and drop.
RyanVM
- Some new Ash permafails hit recently. Trying to chase down owners for those.
- New Mac/Windows capacity coming online soon!
Round table
General note: If you land something that hasn't gone through triage that impacts e10s, please mark it tracking+ and P1 so our uplift searches can find it.
Crash Rates
We still have a modest 25% (main process + content process) crash rate regression to solve.
- Crash rate comparison on Presto
- spread is trending down over the lifetime of the beta experiment.
- Out of approximately 2 million users, 116,107 crash pings, of which 85,652 are in shutdown!
- We have e10s specific crashes we can look into fixing to help bring this down
- We know we have elevated OOM rates when running e10s so working reduced fragmentation and reducing large allocations can help.
Crashstats
bug e10s-oom and bug e10s-crashes
- (unowned) bug 1259480 - [e10s] FatalError crash in PContentChild::Read from PContentChild::SendReadPrefsArray
- #32 crash under e10s, e10s specific
- crashes while deserializing preferences
- related: bug 1268662 - OOM crash in Pickle::WriteBytes from replying to ReadPrefsArray (Send prefs in chunks bug)
- (gabor) bug 1262661 - Many large SessionStore:update messages
- (billm) bug 1262671 - IPC::Message data need not be contiguous
- (mccr8) bug 1268662 - OOM crash in Pickle::WriteBytes from replying to ReadPrefsArray (Send prefs in chunks bug)
- (ksteuber) bug 1263774 - No memory reports in content process crash reports
- (unowned) bug 1266517 - OOM | large crash in MessageChannel::OnMessageReceivedFromLink()
- message queue resize ooms with large IPC message backlogs
- (ting) bug 1267329 - CreateMinidumpsAndPair doesn't record memory information
- (unowned) bug 1272018 - Use shared memory to send images for drag and drop
- (unowned) bug 1182927 - crash at [@ shutdownhang | NtCreateFile ]
- browser hang while writing out shutdown crash reports
- (edwin) bug 1244883 - crash in RtlEnterCriticalSection | PR_EnterMonitor | mozilla::layers::ImageContainer::ClearImagesFromImageBridge
- waiting on diagnositc data
- (unowned) bug 1272151 - Crash in IPCError-browser | (msgtype=0x220003,name=PBlob::Msg_PBlobStreamConstructor) Processing error: message was deserialized
- failure in blob constructor related to http