Mozilla Status Board

mh+mozilla@glandium.org

Current Status - 2015-01-27

Done:

  • Reviewed bug 1123386, bug 866301, bug 1119072, bug 1089363, bug 1123990, bug 1122926, bug 1126017, bug 870366
  • Feedback on bug 1123237
  • Still fighting with the regressions on android with jemalloc3. I found different things on desktop, but none of them are actually making a difference on android. Tested different things with autophone on try, none with any sort of outcome so far, but the last round might get me something interesting, considering what I'm seeing on a nexus S and a nexus 4 locally. Fingers crossed.

Next:

  • More of the same.

Older Posts

Member of Projects

Team Posts

mh+mozilla@glandium.org 3 days ago

Done:

  • Reviewed bug 1123386, bug 866301, bug 1119072, bug 1089363, bug 1123990, bug 1122926, bug 1126017, bug 870366
  • Feedback on bug 1123237
  • Still fighting with the regressions on android with jemalloc3. I found different things on desktop, but none of them are actually making a difference on android. Tested different things with autophone on try, none with any sort of outcome so far, but the last round might get me something interesting, considering what I'm seeing on a nexus S and a nexus 4 locally. Fingers crossed.

Next:

  • More of the same.

erahm@mozilla.com 4 days ago

Done:

  • Monday off
  • Solidified marionette port of AWSY
    • Added handling of marionette exeptions
    • Removed greenlight dependencies (pure marionette-client now)
    • Extensive testing
    • Moved over to marionette branch, cleaned up history, pushed PR
  • Helped pinpoint e10s marionette issue when retrieving a window list

Next:

  • AWSY
    • Land marionette port
    • Start working out how to handle multiprocess data for e10s
  • j2me.js
    • perf testing w/ latest code
    • look into possible platform issue playing AMR audio

nnethercote@mozilla.com 4 days ago

Done:

  • Bug 1100738 - Avoid many heap allocations under gfxPangoFontGroup::GetBaseFont(). Speeds up rendering of several complex PDF maps on Linux by up to 2.5x, and improved CanvasMark on Linux by 12.5%.
  • Bug 1122322 - Fix obscure crash in the worker memory reporter.
  • Bug 1120510 - Fix an assertion failure in the WebGL memory reporter.
  • Bug 1123151 - Privatize PLDHashTable::ops. Makes PLDHashTable safer and easier to use.
  • Bug 1123527, bug 1124539 - Fixed numerous minor things found by cppcheck.
  • Bug 1123533 - Clarify thread handling in NS_StackWalk on Windows. A small refactoring.
  • Bug 1123576 - Measure TypeNewScript in the JS memory reporter. A small coverage improvement.
  • Bug 1124476 - Avoid double-counting documents shared between windows (a rare case).
  • Bug 1124545 - Avoid creating the mergedFeatures hash table when possible. Avoids lots of heap allocations on text-heavy pages.

nfroyd@mozilla.com 4 days ago

Done:

  • Wrote stack of patches for bug 1108750, moving most unified compilation stuff out of the recursive make backend.
  • Watched Air Mozilla talks on our graphics stack and layout-y bits, hoping to get up to speed to help debug power-chewing webpages.
  • Talked to BenWa about graphics stuff.
  • Wrote some simple rendering testcases based on aforementioned talk with BenWa. Talked with BenWa some more about results.
  • Checked out energia and related required software. Internet comments suggest some of the components required are broken on newer OS X/Windows versions. :(
  • Looked at Helgrind patches from Ben Turner. None of them are applicable to modern TSan. :(
  • Checked whether soon-to-be-released clang 3.6 + TSan works properly with Firefox. It does!
  • Reviews
  • MemShrink meeting
  • 1:1 with Nick
  • Monday was a holiday; took Friday off instead to participate in yearly event of eyeball-bleeding amounts of computer game playing over the weekend with friends.

Next:

  • More of the same
  • Off Thursday and Friday

jseward@acm.org 4 days ago

Done:

  • V/Memcheck: Peered at inlined instrumentation a bit more, considered how to write templates for 16- and 8-bit loads.
  • V: Realised it might be easy to significantly speed up reg-alloc, which is the slowest part of the JIT.
  • V: initial triage of new bugs over the past couple of months
  • bug 1122375 (fixes for gcc-4.9 -O2 valgrind FPs): land
  • bug 1056864 (related to 1122375) came back to life alas, made new patch.
  • Start to run B2G on V on Nexus 5 in earnest. Start filing bugs:
  • Fixed up V's ION ioctl handling to behave sanely on b2g runs.
  • Looked at the dreaded Nuwa and finally understood it enough to mark it up for Valgrind. Filed bug 1125091 (Make Nuwa be Valgrind-friendly).
  • Noted that even Nexus 5 w/ 2GB ram but no swap runs into OOM problems on V startups. Plan to return to Flame, which although much slower, at least has a swap enabled kernel.

Now in a position to perform more or less noise-free runs of B2G startups on Valgrind \o/

Next:

  • Make slides for talk at V devroom at Fosdem: https://fosdem.org/2015/schedule/event/valgrind_inlining/
  • Keep filing V B2G bugs. Try to get some of them fixed! Keep chasing remaining V bugs arising from Desktop/Mochitest runs.
  • More bug triage and some fixing for V. There are some high-ish priority bugs un-fixed.