Electrolysis/Meetings/2016-02-11

From MozillaWiki
Jump to: navigation, search


« previous week | index | next week »

felipe

  • Landed my last M8, waiting for aurora approval
  • Final QA phase of second experiment
  • Second experiment sampling:
 ** 90% will match "sampling" criteria
 ** for non-addon users (40%), we will split them in two
 ** for add-on users, we will ship the experiment with a likelyhood of 15%
  • Final numbers:
 ** Control group: 20% of beta
 ** Experiment group: 20% of beta
 ** Addon group: 8% of beta
  • Question about blocking webextensions/themes

mconley

  • M9 bug 1055464 - "Warn me when websites try to redirect or reload the page" broken with e10s
    • Fixed
    • Apparently responsible for a 5.08% sessionrestore (linux64) regression. See bug 1247100
  • M9 bug 1246115 - "This isn't a web forgery…" never appear for webpages detected as forgery
    • Have a patch, writing tests
  • M9 bug 1243720 - [e10s] Firefox doesn't respect colors of high contrast mode for web content (Windows)
    • Fixed
  • bug 1245833 - Tab crash reports have empty comments
    • Fixed.

gabor

  • bug 1179735 (p1 perf, - tp5o_scroll regression) - Not a lot of progress. It turns out my simulation was not correct, I should have disabled vsync, I'm trying again.
  • bug 1182637 (p1 perf, - e10s is jankier than non-e10s) - Closed the bug for now
  • bug 1232638 - (p1 perf, -IPDL::PCookieService::RecvGetCookieString causing janks) - I looked into the possible solutions, and it's a lot harder than I originally thought. The janks happening because reading the database for cookies is slow (maybe the other related operations as well with the strings), and there is no way to do that work off the main thread while we're blocking the content process (the request has to be sync) Looking at the data though, I still don't think this is P1, this issue is responsible for about 3% of the chrome process janks, while we don't know anything about almost 50% (pesudo stack). As it turns out the pseudo stack enhancing patch did not help...
  • some initial multiprocess project planning
  • IPC learning
  • questions: why do we want to turn off e10s for web extensions and themes?

jimm

  • bug 1232181 (M8, Replace plugin windows with static window captures when scrolling) - wip
  • bug 1241060 (?, Flash content disappears when scrolling adjacent div) - fixed.


gw280

  • bug 1213432 - D3D11 WARP is slower scrolling on fb than basic layers. Also slower than non-e10s. Been working on this, gecko profiler didn't show much of any use so I'm currently profiling using Concurrency Visualizer. In the process I'm also implementing additional support for CV in Gecko (profile markers, etc)

mrbkap

  • Not too much to report. Working on the same things as last week.

chutten (e10s responsiveness)

  • bug 1223780 reran my jank threshold analysis using the latest experimental dataset. Looks like, even with addons, parent jank as measured by EVENTLOOP and INPUT_EVENT_RESPONSE is release-worthy.
  • Working with :felipe on planning out beta 45 experiment #2 which will be helpful for getting numbers from a release-like population.
  • Running addon analyses to see which are slow so the addons teams can reach out and product can decide how many and what kinds of lists we'll need for e10s post-release.

poiru (e10s stability)

build ID             non-e10s    e10s-parent   e10s-content
20160204142810         16.764          9.058         15.741
Retrieved from "https://wiki.mozilla.org/index.php?title=Electrolysis/Meetings/2016-02-11&oldid=1117688"