Mozilla Status Board

mh+mozilla@glandium.org

Current Status - 2016-01-26

Done:

Reviewed: - bug 1232707, bug 1236014, bug 1235610, bug 1236577, bug 1237869, bug 1238896, bug 1214885, bug 1239207, bug 1239210, bug 1237863, bug 1230117, bug 1239539, bug 1239202, bug 1239672, bug 1155198, bug 1232707, bug 1074258, bug 1239789, bug 676828, bug 1153053, bug 1241771, bug 1242074, bug 1069556, bug 1144842, bug 1013882, bug 1242663, bug 1242722

Landed bugs:

  • bug 1239169: ./mach build faster fails on "no such file or directory"
  • bug 1239872: mach build faster fails with a traceback
  • bug 1237140: nss.symbols cannot include NSS_EXTRA_SYMBOLS_FILE: not in context
  • bug 1240657: [FasterMake] Race condition leads to missing "manifest interfaces.manifest" entries
  • bug 1240660: Remove support for "+" prefixes in jar manifests
  • bug 1240671: Move INSTALL_TARGETS and PP_TARGETS to misc instead of libs by default
  • bug 1233963: elfhack: Work around recent GNU gold behavior with segments starting before the first section they contain
  • bug 1240945: Add a --verbose option to mach package
  • bug 1240990: Define all backends in one place
  • bug 1241022: Rework affected tiers for the recursive make backend
  • bug 1241398: Misc mach build-backend improvements
  • bug 1241416: Create a tier for things happening before export
  • bug 1241421: Remove jar.mn with no JAR_MANIFESTS check in config/rules.mk
  • bug 1239217: Use "faster" traversal techniques for artifact-based builds aka hybrid build system
  • bug 1242074: Fixed some race condition in the hybrid build system
  • bug 1234130: Fix MSVC LNK4217 warning in hunspell

Other:

  • Sprinted for a release of git-cinnabar 0.3.0 https://github.com/glandium/git-cinnabar/wiki/Release-Notes:-0.3.0, followed by a brown paper bad release 0.3.1 https://github.com/glandium/git-cinnabar/wiki/Release-Notes:-0.3.1
  • Looked at integration of git-cinnabar with the script for mozreview in bug 1153053
  • Tested git-cinnabar against a broader range of mercurial repositories and found a bunch of corner cases it doesn't handle well, as well as some buffer overflow (thanks ASAN) and use-after-free (thanks rr) in the native helper.
  • No bug yet: Looked back at my PoC for a python configure, and restarted work on this. At this point, I have a shell bootstrap that sets up both autoconf and the python virtualenv before invoking the current configure.

Next:

  • More of the same

Coordination:

Still waiting for review: - bug 1238769: Add a -xpcshell option to Firefox

Older Posts

Member of Projects

Team Posts

erahm@mozilla.com 6 days ago

Done:

Monday 1/25/2016

  • Revived Windows partition
  • Added USS measurement to psutils for windows, worked out issues w/ the windows python toolchain not supporting c99
  • Tweaks to get Firefox measurements working
  • Looked into native options for getting the command line of other processes, there's really no good way

Tuesday 1/26/2016

  • Got Chrome measurements working
  • Worked out how to get a proper command line for other processes via wmic
  • Measurements for Chrome, Fx +1 , Fx +2, Fx +4, Fx +8, Fx +30
  • Initial attempt at IE, won't work, IE webdriver is pretty broken. Tried several workarounds.
  • Initial attempt at Edge, won't work, current Edge webdriver is lacking most features.
  • memshrink pre-triage

Wednesday 1/27/2016

Thursday 1/28/2016

Next:

  • Finally blog about e10s memory usage
  • Blog about Fx vs Cr memory usage
  • Polish atsy scripts (retest on Linux/OSX)
  • Take a look at remaining purge xpcomrt tasks

nfroyd@mozilla.com 6 days ago

Done:

  • Tracked down shutdown leaks in bug 1204784.
  • Activated Rust on 64-bit Windows nightly builds.
  • Started writing build system support for Rust cross-compiles (necessary for Android and some OS X builds)
  • Looked at some Rust packaging issues. Uploaded more things to tooltool.
  • Mind melds with our event loop and handling thereof.
  • Looked at an old "high cpu usage on Linux" bug.
  • Assisted with debugging some treeherder authentication issues, still unresolved.
  • Reviews. Almost done with PR_snprintf mentoring bugs.
  • Some UniquePtr conversion patches.
  • MemShrink meeting.
  • Hampered by household sickness.

Next:

  • More of the same.
  • C++ standard library needs seems to be cropping up more and more...

Coordination:

nnethercote@mozilla.com 6 days ago

Done:

Landed:

  • Bug 1144842 - Remove need for --enable-profiling with DMD on Mac and Windows. Eliminates a configuration footgun.
  • Bug 1241656 - Lazify AutoTaskDispatcher::mDirectTasks. Reduces heap churn by 10--15% on YouTube for Mac and Linux.
  • Bug 1239864 - Greatly improve our rect iterators.
  • Bug 1187137, bug 1187141, bug 1187144, bug 1187151 - nsBaseHashtable enumeration-to-iteration conversions.

Next:

e10s memory optimizations.

jseward@acm.org last week

Done:

First status report in way too long.

Memory cleanness: fixed * bug 1232326 - Uninitialised value use in AudioBufferInPlaceScale

Memory cleanness: misc checking/investigation/fix-verification of bclary/BugHunter discovered invalid mem accesses

Memory cleanness: other

  • Mochi runs w/ gcc -Os to match automation settings as V false-error behaviour is a concern. Tree looks remarkably clean.
  • Offer to help :cervantes with V Mochitest runs for FxOS / x86 / QEMU/KVM. (bug 1236830)

Race removal

  • bug 1141565 - TSan: data race ipc/chromium/src/base/histogram.cc:730 Accumulate. The current xpcshell tests "top race". Iterate patch to fix alleged memory leaks, and measure performance loss (none detectable). Finally landed.
  • bug 1136762 - TSan: data race xpcom/io/nsPipe3.cpp:1061 CloseWithStatus. The next most frequent race. Started to investigate.
  • Begin to wonder how many of these we are not really fixing "properly". Fixing them so as to cause TSan to be quiet is often to add locking at too fine a level of granularity. This was especially apparent in 1141565.

Automation with V:

  • bug 1229348 - Add a "valgrind-plain" suite to all_mochitest_suites.
  • Finalise and get r+ on patch that provides |mach mozharness mochitest-valgrind| but not yet landed.
  • Work on a patch to get this scheduled on TaskCluster, a slow and confusing process. Much help from ArmenZG. Can now push to try with "M-vg" syntax and it gets to the point where the run would start, except for the fact that Valgrind is not installed.

Other:

  • Red Hat 1283774 (V aspacemgr assertion regression in 3.11.0): cause to be fixed
  • V segfaults with Fx on F19: bisect the tree, turns out to be wierdness to do with sys_unshare. No fix.
  • Start to marshal bug fixes for a V 3.11.1 release.
  • CWW in Orlando

Mishaps: * Laptop died on New Year's day. Now fixed under warranty. No loss of data but a major loss of time (6 days) trying to assemble a replacement working scenario. I was unimpressed. A top-spec machine and it has been in use just a year.

Next:

  • Continue driving the automation stuff, since this is the most difficult and susceptible to blocking on A-Team involvement.
  • Continue working on TSan-detected races.
  • Push along a V 3.11.1 release.

mh+mozilla@glandium.org last week

Done:

Reviewed: - bug 1232707, bug 1236014, bug 1235610, bug 1236577, bug 1237869, bug 1238896, bug 1214885, bug 1239207, bug 1239210, bug 1237863, bug 1230117, bug 1239539, bug 1239202, bug 1239672, bug 1155198, bug 1232707, bug 1074258, bug 1239789, bug 676828, bug 1153053, bug 1241771, bug 1242074, bug 1069556, bug 1144842, bug 1013882, bug 1242663, bug 1242722

Landed bugs:

  • bug 1239169: ./mach build faster fails on "no such file or directory"
  • bug 1239872: mach build faster fails with a traceback
  • bug 1237140: nss.symbols cannot include NSS_EXTRA_SYMBOLS_FILE: not in context
  • bug 1240657: [FasterMake] Race condition leads to missing "manifest interfaces.manifest" entries
  • bug 1240660: Remove support for "+" prefixes in jar manifests
  • bug 1240671: Move INSTALL_TARGETS and PP_TARGETS to misc instead of libs by default
  • bug 1233963: elfhack: Work around recent GNU gold behavior with segments starting before the first section they contain
  • bug 1240945: Add a --verbose option to mach package
  • bug 1240990: Define all backends in one place
  • bug 1241022: Rework affected tiers for the recursive make backend
  • bug 1241398: Misc mach build-backend improvements
  • bug 1241416: Create a tier for things happening before export
  • bug 1241421: Remove jar.mn with no JAR_MANIFESTS check in config/rules.mk
  • bug 1239217: Use "faster" traversal techniques for artifact-based builds aka hybrid build system
  • bug 1242074: Fixed some race condition in the hybrid build system
  • bug 1234130: Fix MSVC LNK4217 warning in hunspell

Other:

  • Sprinted for a release of git-cinnabar 0.3.0 https://github.com/glandium/git-cinnabar/wiki/Release-Notes:-0.3.0, followed by a brown paper bad release 0.3.1 https://github.com/glandium/git-cinnabar/wiki/Release-Notes:-0.3.1
  • Looked at integration of git-cinnabar with the script for mozreview in bug 1153053
  • Tested git-cinnabar against a broader range of mercurial repositories and found a bunch of corner cases it doesn't handle well, as well as some buffer overflow (thanks ASAN) and use-after-free (thanks rr) in the native helper.
  • No bug yet: Looked back at my PoC for a python configure, and restarted work on this. At this point, I have a shell bootstrap that sets up both autoconf and the python virtualenv before invoking the current configure.

Next:

  • More of the same

Coordination:

Still waiting for review: - bug 1238769: Add a -xpcshell option to Firefox