Mozilla Status Board

mh+mozilla@glandium.org

Current Status - 2014-11-14

Done:

  • Reviewed bug 1095934, bug 1095436, bug 1094141, bug 1095990, bug 788974, bug 1088488, bug 1098135, bug 1071296, bug 1097979, bug 1074711, bug 1073662, bug 1093218
  • bug 1096651: Avoid using random bits when determining SSE3/SSE4 availability for the JIT. Landed.
  • bug 818922: Add bidirectional method calls with replace-malloc library
  • bug 1097506: Pass a JSONWriteFunc to dmd::AnalyzeReports instead of a JSONWriter. Landed.
  • bug 1097507: Make libxul independent of libdmd when DMD is enabled
  • bug 1087245: Moved PR_Lock removal from mozilla::Mutex to bug 1097513 and killed the dead lock by avoiding using a lock entirely.
  • Discussed bug 1091912 with Luke.
  • bug 1098967: LogAlloc has mangled output with e10s on Windows
  • Move my Windows AWS instance to the new IAM, which made me try chocolatey, which really rocks to install things like MSVC.
  • Tried enabling DMD on opt builds but somehow, on Windows, stack traces are broken for some reason.
  • bug 1098537: Looked at how to get rid of httpd.js as a XPCOM component, which mostly works, except that many things, including mozharness, are expecting the file to be in the components directory even when it's not used as a component and that it's in at least 3 different places in the test package.

Next:

  • Move last AWS instance to new IAM.
  • Really really finish build system blog post.
  • More blogging.

Older Posts

Member of Projects

Team Posts

erahm@mozilla.com 6 days ago

Done:

  • Looked into AWSY bustage
    • Got linux box setup, openvpn configured, etc
    • Determined that upgrading from mozmill 1.5 -> 2.0 would fix our issues
    • Fixed API incompatibility issues from upgrading to 2.0
    • Relaunched AWSY, started backfilling
  • Filed follow up for TypedArray crashes, repro'd on b2g w/ my test case
  • Fixed test failures and landed j2me.js fs.stat caching to avoid ~75% of IndexedDB accesses

Next:

  • Look into getting e10s enabled on AWSY
  • Test fixes for TypedArray crashes
  • Look into memory usage of j2me.js

Coordination:

  • Review from wtc

nfroyd@mozilla.com 6 days ago

Done:

  • nsRefPtr.h cleanup patches, bug 1095541, bug 1096501.
  • Some telemetry histogram bugfixes, bug 1097762.
  • Removed some RefCounted<T> usages from the tree, bug 1097734, bug 1097740.
  • Discovered thread safety issue from doing the above. Looked into putting RefCounted<T> things under interdiction.
  • Mentored XPCOM cleanup bug, bug 683786.
  • Cleaned up some rr code, #1369, #1371.
  • Discovered there's only three failing rr tests after patches from #1360, found lingering 32-bit-isms cauing them to fail, but fixing those 32-bit-isms trigger other test failures.
  • Watched bug reports come in for 64-bit rr, including nasty things like #1370 and #1372. Good to know people are using it!
  • Talked rr things with khuey.
  • Made ChaosMode in Gecko more fine-grained, bug 1099251. Still talking through this with roc.
  • Looked briefly at e10s performance in tests, wound up improving mach because of it, bug 1097215
  • Looked at fixing the gross way the reftest harness builds, wrote patches in bug 1098537, inspired glandium to fix it himself, discovered it was a dupe of bug 948278, which has a patch (by glandium, no less) with f? that's almost a year old...
  • Reviews.
  • MemShrink meeting.
  • 1:1 with Nick.
  • Flu shot to ward off MozFlu.

Next:

  • More of the same.
  • Rebase rr patches in #1360 over monkeypatching changes in #1372, possibly debug issues and/or split patches up to try and get some of them in faster.
  • Outstanding reviews.
  • PTO from Thursday through U.S. Thanksgiving.

Coordination:

nnethercote@mozilla.com last week

Done:

Landed:

  • Use a segmented array in SnowWhiteKiller and CollectWhite (bug 1094564 and bug 1096624). Avoids some large and mostly-unused allocations during cycle collection; the CollectWhite case is complicit in 1% of OOM crashes.
  • Move bloattable.pl and bloatdiff.pl into tools/bloatview/ (bug 1089441).
  • Initialize some more fields in ShadowLayerForwarder when Valgrind is enabled (bug 1098185). Fixes one e10s+Valgrind issue.
  • Use a smaller initial IdSet size in jsiter.cpp:Snapshot() (bug 1098618). A small heap churn win.

Other:

Next:

  • More Valgrind+e10s work.
  • More cumulative heap profiling.

Coordination:

  • Waiting on NSS reviews and checkins.

mh+mozilla@glandium.org last week

Done:

  • Reviewed bug 1095934, bug 1095436, bug 1094141, bug 1095990, bug 788974, bug 1088488, bug 1098135, bug 1071296, bug 1097979, bug 1074711, bug 1073662, bug 1093218
  • bug 1096651: Avoid using random bits when determining SSE3/SSE4 availability for the JIT. Landed.
  • bug 818922: Add bidirectional method calls with replace-malloc library
  • bug 1097506: Pass a JSONWriteFunc to dmd::AnalyzeReports instead of a JSONWriter. Landed.
  • bug 1097507: Make libxul independent of libdmd when DMD is enabled
  • bug 1087245: Moved PR_Lock removal from mozilla::Mutex to bug 1097513 and killed the dead lock by avoiding using a lock entirely.
  • Discussed bug 1091912 with Luke.
  • bug 1098967: LogAlloc has mangled output with e10s on Windows
  • Move my Windows AWS instance to the new IAM, which made me try chocolatey, which really rocks to install things like MSVC.
  • Tried enabling DMD on opt builds but somehow, on Windows, stack traces are broken for some reason.
  • bug 1098537: Looked at how to get rid of httpd.js as a XPCOM component, which mostly works, except that many things, including mozharness, are expecting the file to be in the components directory even when it's not used as a component and that it's in at least 3 different places in the test package.

Next:

  • Move last AWS instance to new IAM.
  • Really really finish build system blog post.
  • More blogging.

jseward@acm.org last week

Done:

  • bug 788974 (make LUL unwinding work with elfhack): made a final patch, landed.
  • bug 788974: Redid the original build system patch that removes most of the --disable-elf-hack flags.
  • V: sent out CFP for talks at Valgrind devroom at Fosdem 2015
  • V: hacked up initial patch for "avoid the 10%-15% performance loss caused by the recent need to use --vex-iropt-register-updates=allregs-at-mem-access". Is blocked by bug 1091912, though.
  • V: OSX 10.10: install Yosemite, many fixes to get V working, back and forth with a couple of brave souls dogfooding it. Can run Firefox now.
    • V bug 340725 (Incorrect decoding of vpbroadcast{b,w} reg,reg forms): fallout from the V/Yosemite work.
  • Continued with Valgrind runs of Mochitests, harvesting and fixing bugs. We are now very close to being completely Valgrind-clean on Mochitests. The last time that was the case was just before the release of Firefox 4.
    • bug 1092333: nsDOMStringMap::NamedDeleter doesn't always write its out-parameter: fixed.
    • bug 1093823: ThreadLocal bool yields valgrind warning: commented
    • bug 1096054: Uninitialised value use in Interpret(JSContext*, js::RunState&): analysed
    • bug 1078211: Uninitialised value use in DoUnpremultiplicationCalculation_SIMD: proposed fix
  • bug 1041962: build to fail with musl libc with: error: 'basename' was not declared in this scope: reviewed

Next:

  • Wrap up V/Yosemite work, write a blog post so people can use it, update MDN how-to-run-Fx-on-V docs accordingly.
  • Try to get bug 788974 through review/landing.
  • Decide on when to do V 3.10.1.
  • Get back (finally!) to B2G on V on Flame.
  • Continue with V/Mochi runs.