Meeting Notes Meetings notes from the Mozilla community

Need To Know

(Release and system issues that may impact engineering this week.)

Notices/Schedule (lsblakk/sylvestre/lmandel)

Next Merge: October 13, 2014		Next Release: October 14, 2014
Trains
Central: 35	Aurora: 34	Beta: 33	Release: 32

• 32.0.1 updates disabled today, maybe 32.0.2…we’ll see
  • candidate driver: bug 1063052 Firefox 32 startup crash in nsFrame::BoxReflow

• Beta4 desktop releases today, mobile tomorrow
• Beta5 (desktop) go to build Thursday
• Beta6 (desktop and mobile) go to build Monday

Build Changes (gps)

(Build changes of which engineers should be aware.)

RelEng (catlee)

(Repo, test, and other information for engineers from the release engineering team.)

Upcoming Outages/Upgrades

(System outages/upgrades and tree closures that impact engineering.)

• Saturday, September 20, is the Tree Closing Window. See bug 1065514 for details. hg.mozilla.org will be offline for a period of time, so trees will be closed. This is likely to be from 0900-1300PT, but time will be confirmed on Wednesday. Look for the notice with final times later this week.

Quality Programs

(An opportunity to hear about status with the various quality programs that do not have a formal team structure.)

MemShrink (njn)

• No update.

Stability (kairo)

• We just halted 32.0.1 updates due to bug 1063052.
• Beta still has GFX issues on both desktop (OMTC/D3D11) and Android (AlphaBoxBlur bug 1028802).
• Prompt for e10s on Nightly triggered us seeing a lot of e10s crashes, we need a lot of work there until we’ll be able to ship anything there or even turn on by default.

Team Stand-ups

(In <2 mins, what did your team accomplish last week, on what is your team working on this week, and on what, if anything, is your team blocked? No questions during the stand-ups. All questions should be asked during the roundtable.)

A*Team (jgriffin)

<Read Only>

Blog and newsgroup posts:

• [armenzg] Running TBPL jobs locally with HTTP authentication and related dev.platform post
• [gbrown] Running my own Autophone
• [wlach] Hacking on the Treeherder front end: refreshingly easy
• [AutomatedTester] Marionette mailing list

Desktop Platform (bsmedberg)

<Read Only>
Mac V2 signing work is our top priority and we hope to have that on nightly next week.

Developer Tools (prouget/bgrins)

• New Timeline devtool has landed. See post and screenshot: https://groups.google.com/d/msg/mozilla.dev.gaia/IT-M42uGT40/_cV31uMY9fcJ
• Announced Safari/Chrome protocol support in Firefox devtools: https://hacks.mozilla.org/2014/09/firefox-tools-adapter/
• Hacks post with release notes for 34: https://hacks.mozilla.org/2014/09/webide-storage-inspector-jquery-events-iframe-switcher-more-firefox-developer-tools-episode-34/ (WebIDE, Storage inspector, jQuery events, iframe switcher)

DOM (jst/overholt)

• Ben Turner landed major refactoring of IndexedDB to use PBackground (bug 994190)
• successful mini Service Worker shindig last week; plans still at https://etherpad.mozilla.org/offline

Electrolysis (e10s) (blassey)

• Please help dogfood e10s in Nightly!
  • Just flip the “browser.tabs.remote.autostart” pref to true and restart Nightly.
  • Known issues: https://wiki.mozilla.org/Electrolysis#What_to_Expect
• e10s bugs fixed in the last week: http://is.gd/ksQL9N
• To get more testing from Nightly users:
  • bug 1064885 – Added e10s opt-in prompt in Nightly
  • bug 1064886 – Added “Enable E10S” checkbox in Firefox Preferences pane
  • Disable e10s if a11y (bug 1047076) or safe mode bug 1063848 are enabled (to avoid known issues, for now)
  • We might enable e10s for all Nightly users (for just one day) to get broader testing during Nightly 35. So now is a good time to report any critical e10s bugs you are seeing.

Firefox Desktop (gavin)

Summary of Work In Progress

• You can see the status of the current iteration (35.1) in progress here

Summary of Recent Landings

• Team landed 42 bugs over the last week, to close out the 35.1 iteration.
• Details of landings from the past week can be seen here

Firefox Mobile (snorp/blassey/mfinkle)

Work In Progress

• Browser toolbar refresh bug 1058909
• WebRTC OpenGL on Roku – fixing and tweaking
• Moz Stumbler currently off due to start time regression bug 1064677
• Janus https://wiki.mozilla.org/Mobile/Janus
  • investigating performance issues
  • looking into <picture> element injection
  • Adding TCP/socket stats

You can find more on upcoming feature plans in the[roadmap]

Landings from the past week

Nightly (35)

• Lightweight theme preview is broken ( Bug 887755 )
• Synchronous initialization of search service when handling open search tag ( Bug 1017903 )
• [Roku] Removing an entry from history will only take effect after restarting the Roku app ( Bug 1025923 )
• crash in java.lang.NullPointerException: at org.mozilla.gecko.home.PinSiteDialog$3.onFocusChange(PinSiteDialog.java) ( Bug 1041448 )
• crash in mozilla::layers::TileClient::GetBackBuffer(nsIntRegion const&, mozilla::layers::TextureClientPool*, bool*, bool) ( Bug 1041744 )
• about:home is unresponsive to hardware menu button on returning to main activity on 2.3 devices ( Bug 1058027 )
• Regression: Back button icon missing in menu on some small screen (?) hardware (?) devices ( Bug 1062338 )
• “Learn More” link for the Mozilla Location Service is visible on Beta/Release despite MLS setting not available ( Bug 1063518 )
• crash in webrtc::ViEInputManager::NumberOfCaptureDevices() ( Bug 1063547 )
• [roku] update wording in the public roku channel app ( Bug 1064222 )
• Search activity startup crash @ java.lang.IllegalArgumentException: Couldnt find search engine for identifier: yahoo at org.mozilla.search.providers.SearchEngineManager.createEngine(SearchEngineManager.java) ( Bug 1065123 )
• crash in java.lang.NoSuchMethodError: android.os.Bundle.getString at org.mozilla.gecko.preferences.GeckoPreferences.setupPreferences(GeckoPreferences.java) ( Bug 1065531 )
• Regression: Downloading is broken ( Bug 1066671 )

Firefox OS Devices/Porting (ericchou)

<Read Only>

• Gonk team (led by Thomas Tsai & Shawn Ku)
  • Tarako: Spice/Intex feedback issues
  • Dolphin: 9/4 MP. 9/26 on the shelf.
  • Woodduck: Meta bug
  • Flame: KK + v1.4/v2.0 porting

• Bluetooth (members: Shawn Huang, Ben Tian, Jamin Liu, Jocelyn Liu)
  • BLE API discussion on dev-webapi
  • Shawn and Jamin will participate in NFC workweek from 9/15 ~ 9/19 and will give two talks related to BT/NFC applications.
  • PTS test for Firefox OS 2.1 (current aurora) has begun. Jocelyn is the owner this time and will co-work with TPE QA.

• Media Playback (members: Bruce Sun, Blake Wu, Star Cheng)
  • bug 1064376 – [KK] Regression in playing few clips on v2.1 – landed (2.1+)
  • bug 1039182 – Add getGrapicBuffer interface to MediaCodecProxy – landed
  • bug 1036849 – Video part in MP4 DASH on B2G does not work – WIP

• Device Storage (members: Alphan Chen, Eden Chuang)
  • bug 1060196 – Should forbid ejecting SD card when user is playing music which is stored in SD card – feedback+ (2.1+)

Firefox OS RIL (htsai)

<Read Only>

1. https://wiki.mozilla.org/TPE_RIL_team/2014-09-16

GFX (milan)

• Proposed high level priorities for graphics for the next few releases (e.g., 35-38) are being discussed (and possibly modified) by product and execs. The proposed list top five – let’s see if it changes:
  • E10S support
  • B2G memory usage and performance
  • APZ on B2G: Project Silk support, OOP support, “fake overscroll”
  • WebGL Desktop performance
  • WebGL Mobile performance

JS (naveed)

<Read Only>

• Compiler (JIT)
  • bug 1062869: Add ability to execute recover instructions ahead of bailouts. This improves Scalar Replacement for arguments given to multi-basic-block inlined functions. This also one step forward to “eager” escape analysis.
  • bug 894596: Patch to allow inlining global variable values. Investigating possible performance regressions associated with the patch.

• Garbage Collection
  • bug 1020751 and bug 1022794: Enabled GGC for B2G

• Front End and Other
  • SIMD lands in Nightly. Firefox demoed running JS-SIMD on stage at IDF 2014. Dan Gohman’s JS-SIMD blog post
  • bug 1064451: eliminated the function-call in every MOZ_ASSERT expansion

Media (mreavy)

<Read Only>

• Fixed problem with Webrtc and H.264 under very high packet loss (bug 1054624)
• Resolved Screensharing issues on Windows due to multiple threads (MediaManager, D3D9 in gfx/compositor) which made windows calls without running on a Windows messageloop – thanks jimm!
• working on policy for additions to screensharing whitelist
• ScreenSharing now *must* occur from an https: site for security/privacy reasons

Necko (dougt/jduell)

<Read Only>

• No update this week.

Performance (vladan)

Update next week

Roundtable

(Comments and questions that arise during the course of the meeting or otherwise do not have a section.)

<Read only beyond this point>

Mailing List Threads

(Threads that are likely to be of interest to engineering from various mailing lists.)

Good Reads

(Links to blog posts, books, videos, etc. that you think will be of interest to others.)

• The Progress Principle: “Don’t Let Your Huge Goal Distract You from Small Wins”
• Raymond Chen’s Old New Thing: “The history of Win32 critical sections so far”

irc #planning Log From This Meeting

11:02 cpeterson: https://wiki.mozilla.org/Platform/2014-09-16
11:05 bsmedberg: blassey: jimm: I'll be following up with you later about the e10s crashes
11:05 blassey: ok
11:06 lmandel: bsmedberg: I see readonly update about v2 sigs. Do you want to tell people what that means?
11:06 cpeterson: e10s content crashes are up about 100x since the e10s opt-in prompt
11:06 bsmedberg: lmandel: not really, do you think I need to?
11:06 lmandel: bsmedberg: Only if you think people will be impacted in their work.
11:07 lmandel: bsmedberg: We should probably call out when this is done so that people know.
11:07 KaiRo: hmm, do I only voip badly in those larger meetings? I didn't hear complaints in the smaller ones I'm usually in
11:07 bsmedberg: I think rstrong will be sending an email around to dev.platform when the current oranges/issues are solved before we actually flip things over
11:07 lmandel: KaiRo: You were fine in the channel meeting.
11:07 lmandel: bsmedberg: k
11:08 KaiRo: lmandel: ok, maybe the added traffic of more people on the meeting tips it over or so
11:09 KaiRo: btw, "Firefox is not starting now" seems to be a quite common issue with e10s
11:09 armenzg_mtg: cpeterson, I will mention it here
11:09 armenzg_mtg: I tried e10s on Linux Desktop nightly recently and it was a pretty awful experience
11:10 armenzg_mtg: enabling it by default even for one day will be pretty terrible for nightly testers
11:10 KaiRo: snorp: did you get a crash reporter at all for that non-starting thing with e10s?
11:10 snorp: KaiRo: yeah, standby, will paste
11:11 snorp: https://crash-stats.mozilla.com/report/index/c5ef1fca-fde0-44c8-95ef-680052140916
11:11 KaiRo: snorp: ok, good to know, startup crashes sound like the first thing we need to fix
11:11 snorp: yes.
11:11 armenzg_mtg: cpeterson, what is the preferred way to give feedback about e10s?
11:11 KaiRo: xpc::AddonWindowOrNull
11:11 KaiRo: sounds like fun with add-ons
11:12 cpeterson: armenzg_mtg: you can mention e10s problems in #e10s or just file a bug. Be sure to include the word "e10s" in your bug summary so our e10s triage queries will find it!
11:13 dveditz: armenzg: there's a link to file bugs at https://wiki.mozilla.org/Electrolysis#Contributing
11:13 armenzg: cpeterson, dveditz thank you gentlemen - I will dog food again
11:13 dveditz: also descriptions of known issues and links to known bugs for dupe finding
11:13 dveditz: on that page
11:13 dveditz: I mean a link to a query
11:14 KaiRo: snorp: can you file this one? sounds like this specific signature only exists today
11:14 snorp: KaiRo: https://bugzilla.mozilla.org/show_bug.cgi?id=1068163
11:14 cpeterson: armenzg: TBH, if you had e10s crash problems, you might want to wait a day before dogfooding again for some recent fixes to make their way to mozilla-central and then Nightly builds.
11:15 KaiRo: snorp: awesome, thanks. the signature seems to be all only today's build. fun.
11:15 snorp: KaiRo: yay
11:15 armenzg: cpeterson, I'm good with crashes - IIRC I was having responsiveness issues
11:15 cpeterson: armenzg: oh, that problem has not been fixed yet <:)
11:15 snorp: heh
11:15 snorp: cpeterson: what causes it
11:16 snorp: shouldn't it be *more* responsive
11:16 armenzg: cpeterson, k thanks
11:16 armenzg: snorp, in some cases :P
11:16 cpeterson: snorp: I don't know the cause, but there is likely some IPC bottleneck between parent and content process.
11:17 cpeterson: snorp: the browser should be more responsive (because its UI thread is not blocked on content as much), but all tabs run in the same process that can block each other.

---