Electrolysis/Meetings/2016-03-10

From MozillaWiki
Jump to: navigation, search

« previous week | index | next week »

mconley (Will not be present at meeting)

  • Request: Can we record these meetings for the folks in the Taipei office who can't make it? <== Yes, I filed REQ0038893 to make this happen until June (elan).
  • Spent most of my week either reviewing test patches from the Desktop team, or working on tpaint
  • I'm away today at the D&I Working Group meetup in Palo Alto.
  • Perf P2 bug 1174770 - tpaint regressions (3%-33%) in e10s mode compared to non-e10s mode
    • I have a plan for how to attack this. Tentative results show us beating non-e10s on OS X. Investigating what needs to be further done for Windows and Linux to get us over the edge, and then I can polishing and landing the work
    • bug 1254669 - Tab progress listener doesn't know that browser starts at blank after remoteness flip
    • bug 1254865 - remote-browser.xml's browser-child.js should not send a sync message on load
    • bug 1251032 - Have ContentParent send RenderFrameInfo down when responding to the CreateWindow sync message
    • (No bug filed yet) - Have gBrowserInit.delayedStartup on e10s only run once content has been painted and composited
  • Perf P1 bug 1226564 - [e10s] CACHE V2 telemetry probe regressions
    • Just started to piece this together. I'm going to need to talk to jduell or mayhemer to see how much we care about the remaining probes / whether or not they matter for e10s.

bobowen

  • bug 1189846 - Print Edit 15.10 and bug 1236015 - Windows 10 print to PDF bug - nearly have something working for bug 1189846 - sorting out print settings a bit as part of that, which should help with the other bug.

chutten

Plugin Jank Release Criteria - For correctness/completness' sake need to aggregate all Flash responses together, but prelim results show no problems.

bug 1226564 - apparently the cache probes we've been looking at might not be the most relevant. Added different ones as recommended to the "Big Analysis of Everything E10s" notebook (this one) so we can keep an eye on them. Currently pending :rvitillo's review, but prelim results look better with e10s than without.

Somehow my name was added as being responsible for the memory criteria so I did some poking around. tp5o's Private Bytes is presently a valid measure in e10s as it sums parent+child measurements. It's a little weird, though, in that there are twice as many measurements coming out of the e10s test run than the non-e10s test run. Tracking it on bug 1250169.

jimm

  • bug 1041599 (M9, Broken audio session volume controls on Windows) - Currently broken in e10s and non-e10s. Have a wip that improves things but still have some issues remaining with shutdown.

felipe

  • Handling things for the new experiment on beta46
  • New experiment released
  • Writing down test plan for staged rollout code
  • Tests

mrbkap

  • bug 1113196 - m8 - top level loads have bogus security info - checked in.
 * bug 1253792 and bug 1240246 - two intermittent fallout bugs fixed this week.
  • A bunch of test-related stuff (enabling tests and fixing intermittent oranges) and reviews.
  • bug 1186139 - m9 - still need to fix a test. Should have a patch today.
  • Got a couple of bugs uplifted.

Neil

  • Fixed a few more tests and enabled some that worked as is
  • Working on some intermittent oranges related to the newtab tests that were recently enabled

RyanVM

  • Ongoing greening up of various platforms not currently covered in CI and enabling more passing test suites.
  • Discussions with various teams about the costs of maintaining support for older OSX/Windows versions.
  • Formal test plan still under review.

cpeterson

  • e10s release criteria dashboard
    • jank, page load, gfx perf, and startup/shutdown time metrics are all green!
    • stability, scrolling, UI smoothness, and memory usage metrics are red.
    • need more data for engagement, slow scripts, and plugin jank metrics.

aklotz

  • bug 1236108 (OOM annotations for content processes) - Landed, waiting for preliminary results on Nightly, if everything looks good then I'll proceed with uplift
  • a11y + e10s on Windows - Kickoff meeting with a11y folks this afternoon at 2PM Mountain, 4PM Eastern. 1PM pacific ;-)

bsmedberg

gw280

  • bug 1114647 - Content process rename. Spent ages debugging a weird content process spawning issue with the patch on Linux. It's now been fixed and I think this is ready to land pending review.
  • bug 1213432 - Very confused as to what exactly is going on here. The bug has a lot of data on it now, but the tl;dr is in non-e10s, WARP is slower than Basic. In e10s, WARP is also slower than Basic, but the delta is larger. Milan has escalated to see if WARP is a configuration that is a valuable to us given that Basic is so much faster than WARP in this particular testcase.
  • bug 1176019 - Keep around layers. Currently working on a patch, should have a good plan for how to do this.

billm

  • Looking into tab switch regressions (not too seriously though).
  • Landed fix for intentional IPC crashes.
  • Helping out with some Hello-related issues.

gabor

  • Had some time to do some research related to the multiple content processes project. Working on a Wiki page for it, almost finished.
  • Helped out some folks from the Berlin office with various problems.
  • PCookieService bugs still on hold.

kanru

  • bug 1233723 - Looking into how to reproduce. Possible related to loop extension. We can fix the bad PBrowserChild::Send__delete__ call, but that will hide the real problem.
  • Filed bug for accessing PII data on crash-stats
  • Filed some bugs and attached to e10s-crashes

cervantes

  • bug 1253575 - Submitted patch for annotating the crash report

tchou

  • bug 1249226 - RESOLVED::FIXED Leaked nsFrameLoader which entrains nsInProcessTabChildGlobal
  • bug 1251150 - top crash in DecrementVisibleCount - updated what I've done to the bug, nothing I can do if it is a memory corruption