Electrolysis/Meetings/2016-03-31

From MozillaWiki
Jump to: navigation, search


« previous week | index | next week »


kanru

  • bug 1258312 (crash in mozalloc_abort | NS_DebugBreak | mozilla::ipc::FatalError | mozilla::dom::PBrowserChild::FatalError | mozilla::dom::PBrowserChild::OnMessageReceived) - requested review for diagnose patches

ting

  • bug 1197913 (M9, Moving the cursor outside the <select> drop-down list doesn't keep the last hovered item highlighted) - landed
  • bug 1255968 (P1, Interruptible reflow is broken in e10s) - under review
  • bug 1259512 (P1, significantly higher rates of OOM crashes in the content process of Firefox with e10s than in the main process of non-e10s) - checking jemalloc profiling and stats report mechanism

cervantes

  • bug 1253575 (crash in mozalloc_abort | NS_DebugBreak | mozilla::ipc::GeckoChildProcessHost::OpenPrivilegedHandle) - investigating
  • bug 1257486 (Add additional memory annotations to content process crash reports) - under review

chutten

  • Plugin Jank: BLOCKED_ON_PLUGIN_MODULE_INIT_MS is being reported by too few e10s-enabled clients' parent processes to get a bead on whether it might be good or bad. According to t.m.o, we're only getting submissions from builds 0321, 0322, and 0324... and not many submissions, at that (some tens of thousands).

gsvelto

  • All the mochitests under browser/base/content/test/general have a fix, not all the fixes have landed yet though
  • I've moved to fixing the tests under dom/base starting with the ones affecting the clipboard, progress is slow since I'm unfamiliar with this code and the first one I tackled turned out to be a real issue in Firefox when running with e10s enabled
    • bug 1258857 - (Enable test_copypaste.html with e10s) - Patch ready to fix an issue when serializing the clipboard contents for transfer over IPC, still have to fix the remaining tests

haik

  • bug 1248683 [e10s] Find bar sometimes gets keypresses in the wrong order when system is under heavy load - mostly root caused, getting some help on how to fix it

mconley

  • bug 1174770 - tpaint regressions (3%-33%) in e10s mode compared to non-e10s mode
    • 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
      • This should take care of a big chunk
    • bug 1258465 - Only nsIWebBrowserFocus->activate a TabChild once it has composited
      • This should almost close it out. We might just be hovering around the 5% threshold after this one.
  • bug 1186585 - 28.2% - 48.5% tps talos regression with e10s enabled
    • bug 1251939 - IPDL::PLayerTransaction::RecvUpdateNoSwap takes too long with e10s during TPS talos runs
      • This seems to be a big culprit - we're spending a lot of time doing the layer transaction from the child.
      • jrmuizel suspects we're page-faulting when allocating memory on the GPU.
      • Meeting with billm, jimm, gw280 today to plan attack on this one
  • bug 1226564 - [e10s] CACHE V2 telemetry probe regressions
    • I'm reasonably sure this isn't a big deal - I think we're dealing with an unwarm cache.

felipe

gabor

  • bug 1247420 - (m9 1247420 - Hovering over a select popup inside ul elements breaks hover css selector on list elements) - Landed. Will request for an uplift soon.
  • bug 1255138 - (m9 1255138 - Resizing windows from JS is not supported in e10s mode) - Baku is half way done with it but he wouldn't mind some help so I'm looking into it. It seems like that after we get the content area size and convert back and forth between CSS pixels and dev pixels we get a bigger size in the e10s case for some reason.
  • bug 1253575 - (crash in mozalloc_abort | NS_DebugBreak | mozilla::ipc::GeckoChildProcessHost::OpenPrivilegedHandle) - Looking into it with Cervantes Yu. In most of the case probably there is a dll injection from various security softwares that might cause the handle to reject the open call. We don't think we should try to open a handle for the process anyway since we already have one handle that is ignored...
  • bug 1258663 - (Annotate system call failure regarding IPC transports) - just some review about landing some additional crash annotation when some process handle manipulation fails.

mrbkap

  • bug 1257790 - Loop text is unselectable with the mouse - investigated, turned out to be a toolkit bug. Attached a patch.
  • bug 1174036 - Find gets stuck when hidden textarea in the page - wrote a patch & checked it in (with bonus fix for an intermittent problem my mochitest found).
  • bug 1257861 - Bad regression in 3rd party cookie blocking code - wrote a patch & checked it in.
  • bug 1258921 - m9 - Can't remove logins via the autocomplete dropdown - wrote a patch (green on try) & waiting for review.
    • I'm seeing a related assertion on try, will debug it this week.
  • More reviews and test fixes.

jimm

  • bug 1148978 (M9, Flash plugins fail to paint correctly when open sidebar/fullscreen) - fixed.

elan

aklotz

  • bug 1261107 Working on landing core COM marshaling and proxy IPC code
  • Opening a discussion with nvaccess (an open-source a11y client) about proposed changes

gw280

  • bug 1114647 - hoped to land this on Monday but when I ran a final try run it showed a tp failure which I can't reproduce locally. I've been investigating this but not getting anywhere.
  • bug 1260507 - WARP being slower. I have landed a patch to disable WARP on nightly now, so we probably don't need to worry about WARP anymore.
  • bug 1251939 - RecvUpdateNoSwap taking a long time. mconley already talked about this, but it's basically due to a copy-on-write page allocation. Likely fix is to keep long-lived textures in a texture pool.
  • bug 1255841 - disabled option elements not working in e10s. fixed, landed and uplifted (to aurora only, not beta).
  • bug 1237769 - XP with e10s and d3d9 layers. We have a problem here, and upon consultation with milan we're going to go ahead and disable e10s if acceleration is requested on XP. if acceleration isn't requested, we can go ahead and have e10s/basic.

blassey

  • bug 1184739 P1 - Blob URLs as favicons don't work, landed
  • bug 1164931 P1 - The e10s slow script notification bar keeps spontaneously disappearing and reappearing in cycles, waiting for review
  • bug 1164286 P1 - The page load stop and refresh buttons do nothing when the E10S slow script dialog is being shown, got review
  • bug 1257554 m9 - Remove restore on demand checkbox, landed