Mozilla Status Board

mh+mozilla@glandium.org

Current Status - 2014-10-24

Done:

Next:

  • I declare next week a blog week

Older Posts

Member of Projects

Team Posts

mh+mozilla@glandium.org today

Done:

Next:

  • I declare next week a blog week

nfroyd@mozilla.com 4 days ago

Done:

  • Fixed subtle bug in how rr recorded clone(2), #1340
  • Fixed syscall-entering-checking for rr, #1342
  • Fixed clone(2), select(2), and how exits work on x86-64, #1343
  • Fixed getgroups(2) and related UID/GID syscalls, #1344
  • Fixed rr's gdb stub for restarting x86-64 syscalls, #1346
  • Fixed some 32-bit-isms in rr tests, #1347
  • Fixed a few networking syscalls and other miscellanies, #1348
  • Banged head against problems with an rr test for accept(2), only to have roc point out it was because I'd mis-categorized some syscalls from earlier patches. Hooray for an expanded understanding of how rr works.
  • Reviews, mostly a big patch-bomb from glandium concerning the Android linker.
  • Investigated some nasty threading races, still thinking of a good solution.
  • Finalized flights for workweek.
  • MemShrink meeting.
  • 1:1 with Nick.

Next:

  • More 64-bit rr stuff. Local tree has ~94% tests passing, big things left to fix are:
    • recvmsg(2)
    • msgrcv(2) and friends
    • sendfile(2) (dealing with 32-bit vs. 64-bit)
    • Some lingering fallout from x86 vs. x86-64 TLS differences.

nnethercote@mozilla.com 5 days ago

Done:

Landed:

  • Removed lots of ancient and unused code from tools/ (bug 1080291, bug 1081776, bug 1081809, bug 1081812, bug 1081792, bug 1081777, bug 1082554). Mostly stuff that no-one has touched a decade or more.
  • Use a better buffer growth strategy during XDR encoding (bug 1084114). Avoids ~30 MiB of realloc churn at start-up.
  • Fixed an assertion failure in Factory::GetDirect3D10Device() when memory reporters run (bug 1083624). Found after I (finally!) set up a Windows VM, which was surprisingly pain-free.

Other: started reworking all our memory tools documentation. https://wiki.mozilla.org/Performance/MemShrink/Tools is the new entry point. Lots more to add.

Next:

  • More memory tool documentation clean-up.
  • Finish up DMD improvements.
  • Some fix-ups for new problems found with DMD.

Coordination:

Waiting on reviews for nine patches in five bugs, which is more than normal :)

jseward@acm.org last week

Done:

  • Talk w njn, over WebRTC (eventually) (hooray)
  • Registered for the Portland Dec meeting
  • Investigate Fx-on-V-on-OSX109 hangs. Eventually conclude it is somehow related to one thread telling another thread to go to sleep.
  • Get back to building B2G.
  • Tune V amd64 JIT pipeline (%rflags spechelper functions) for Fx as a workload, mostly to try and reduce the incidence of false positives
  • File 1078211 (Uninit value use in DoUnpremultiplicationCalculation_SIMD)
  • Look more at Fx+e10s on V. File 1078275 (TargetConfig::orientation is used uninited)
  • Lots of looking at bug 788974 (Don't disable elfhack [...] checking)
  • Deal with gkw/Irving mail re cycles in Callgrind profiles
  • V coordination (OSX patches, other patches)
  • V: write up design proposal for more aggressively inlined Memcheck instrumentation.
  • Deal with bug 1077031 (uninit value(s) involving IsFloatLiteral)

Next:

  • Continue with bug 788974 (Don't disable elfhack [...] checking)
  • Get back to V/Mochi runs, + more e10s
  • Deal with increasing backlog of bug reports for V 3.10.0

erahm@mozilla.com last week

Done:

  • Landed bug 806819, dealt with fallout
    • Fixed strange XP-only runtime issue (removed file from unified building)
  • Landed bug 1072605 - gfx/2d/Logging.h checks for PR_LOGGING before including prlog.h
  • Bug 1073578 - Make handling logging modules thread safe
    • Up for review, not holding my breath
  • Landed bug 1074387 - Double-counting of memory, probably in webaudio
    • Confirmed it was double-counting w/ a DMD run
    • Track down to a shared set of HRTFDatabases
  • Bug 1080859 - Make logcat console log messages one line if possible
    • Up for review, janx seemed happy
  • New coverity run, started triaging
    • Filed bugs for 55 coverity defects and cleared several FPs
    • Most egregious MemShrink issue: bug 1081260 - |js::MallocProvider| alloc functions leak memory after calling |onOutOfMemory|
    • Sketchy code generation: bug 1058252 - Uninitialised value use in DOMProxyHandler::delete_

Next:

  • Wrap up NSPR portion of logging
  • Coverity triage
  • Look into DMD double-reporting in CSS

Coordination:

  • Review from wtc
  • Review from dhylands