Snappy June 14 Minutes

Actions

• lmandel to follow up about enabling dev tools in webrt
  • From Dave Camp
    • Myk also started this discussion recently
    • Nothing currently in flight to make this happen
    • Remotable tools (debugger, profiler?) should be reasonbly easy to get working on WebRT with remote protocol work
    • Working with product to re-evaluate priority of WebRT developer tools
• gavin to find an owner for persona slowness

Incoming

• Firefox does very poorly on the IE chalkboard test https://ie.microsoft.com/testdrive/Performance/Chalkboard/

Projects

Fix cache - hurley

Results from the past week

• (No voice/video update this week)
• Investigations into possible regressions caused by bug 722034. bug 762576 indicates regression may not actually exist and telemetry may be misleading

Todo this week

• Coming up with plans/ordering for moving the rest of the blocking cache ops off the main thread

Profiler - jrmuizel/BenWa/Ehsan (and more)

Results from the past week

• bug 758741: Modified Symbolication Server to support using local symbols in BenWa's profiler http://blog.mozilla.org/vdjeric/2012/05/30/profiling-with-the-built-in-gecko-profiler-and-local-symbols-on-windows/

Todo this week

• Added automated diagnosis - Make the profiles easier to read and lets us annotate know issues: http://benoitgirard.wordpress.com/2012/06/08/making-profiling-easier-automated-profiler-diagnostics/
• Turning Framepointer on m-c
• Profiler Add-on has been ported to Thunderbird by mconley (missing some features), patch should land soon.
• Looking to port to the B2G soon.

Nondestructive chromehang - vlad

Results from the past week

• Analyzed chrome hang reports http://blog.mozilla.org/vdjeric/2012/06/08/cache-plugin-font-operations-most-common-in-chrome-hang-reports/
• Discussed Metrics setting up regular symbolication of hang reports

Peptest - mcote

Results from the past week

• bug 737886: Met with Gilbert and Christina last Friday. They want to see more detailed data in order to characterize it and make intelligent decisions about forecasting methods. Specifically, they'd like to see the length of every event loop iteration for a test run or two, rather than just those above 50 ms. This is going to delay implementation of forecasting and regression warnings, but I think it's important to do this right, rather than attempt to fix it later like we are doing with talos.

Todo this week

• Give detailed data to analytics.

GC pause reduction - billm

Results from the past week

• bug 761739 - Actually do incremental GCs
• bug 743396 - Don't unmark non-collected compartments during GC
• bug 742841 (needed for some CC speed up work)

CC pause reduction - smaug, mccr8 (meta bug 698919)

Results from the past week

• no update

Firefox shutdown - rafael

Results from the past week

• bug 722243: Made changes to cancellable query patch with mak's comments, waiting on his 2nd review

Areas of focus

https://bugzilla.mozilla.org/show_bug.cgi?id=748598