Platform/2010-02-02: Difference between revisions

From MozillaWiki
Jump to navigation Jump to search
 
(18 intermediate revisions by 11 users not shown)
Line 32: Line 32:


= GFX Update =
= GFX Update =
* Several projects are still ongoing:
** Cairo update & merge
** Direct2D/DirectWrite patch reviews and updates
** Windows out-of-process plugins work - there's one large outstanding issue that Jim is working on.
** Core Animation plugin support attempts - three independent attempts to support them, each with their own issues. Benoit is still working.
** Fennelectrolysis shared memory work hopes to wrap up this week.
For more, see [[Platform/GFX/2010-Feb-1|our weekly meeting page]].


= Layout Update =
= Layout Update =
* Found Windows "exit code 1" orange. Need to keep working on orange bugs. How do we get people working on them?
* Fullscreen API proposed in WHATWG mailing list; please investigate. [http://lists.whatwg.org/htdig.cgi/whatwg-whatwg.org/2010-January/024860.html email thread starts here] and [http://lists.whatwg.org/htdig.cgi/whatwg-whatwg.org/2010-January/024872.html here]


= Content Update =
= Content Update =
* Landed HTML5 pushState (Justin Lebar, Jonas)
* Worked on branch back porting (mrbkap, jst)
* Atom table patches ready and up for review (Jonas)
* Attended the Velocity summit. Feedback collected at [[Performance:Velocity2010]] (Jonas)
* Investigating what we need from electrolysis to do our security notifications right, for the lock icon etc (Honza)
* Got large parts of content process event handling done, tests included (smaug)
* HTML5 parser related about:blank work, and mochitest investigation (hsivonen)
* CPOWs are getting easier and easier to use (bnewman)


= Platform-specific Support Update =
= Platform-specific Support Update =


= JS =
= JS =
* nanojit work winding down for the moment
* new baseline compiler starting up


= Electrolysis =
= Electrolysis =
* Plugins
** Windows hangs edge cases still bedeviling: jimm on it (today?)
** Two most obvious crashes dealt with today ({{bug|542248}}, {{bug|543376}}
** One more topcrash diagnosed, perhaps dealt with today {{bug|542263}}
** UI needs to land
** status for 1.9.3a still ambiguous: either off-by-default or manipulate prefs so that only flash and silverlight are on by default
* Fennec
** Benedict starting on webprogresslistener
** smaug has event listeners almost ready, might need CPOW
** benjamn has CPOWs, need review... can somebody other than bsmedberg review this?
** romaxa working on Qt
** joe, memory-mapped drawing?
* Jetpack
** jdm working on multi-process jetpack with CPOW support


= Startup Performance =
= Startup Performance =
'''Summary'''
* Read the latest [http://autonome.wordpress.com/2010/01/30/firefox-startup-performance-jan-29/ weekly update].
* View [http://graphs.mozilla.org/dashboard/snapshot/ snapshot results] for startup and all other tests.
'''Priorities'''
<table id="ts" border="1" cellspacing="0" cellpadding="2">
<tr>
<td class="s0">Estimated win</td>
<td class="s1">Bug #</td>
<td class="s1">Summary</td>
<td class="s1">Owner</td>
<td class="s1">Status</td>
<td class="s1">Notes</td>
</tr>
<tr>
<td class="s2">~8%</td>
<td class="s3">{{bug|525013}}</td>
<td class="s4">Investigate a more static build configuration of Firefox</td>
<td class="s4">Joel Reymont</td>
<td class="s4">in progress</td>
<td class="s4">The main body of work is in the patch now, waiting on review from Ted. Joel's working various build-related cleanup, with the help of Brad Lassey and Ben Smedberg. Experiment Windows tryserver build available [http://build.mozilla.org/tryserver-builds/jreymont@mozilla.com-1265128477 here].</td>
</tr>
<tr>
<td class="s2">up to 25%</td>
<td class="s3">{{bug|514083}}</td>
<td class="s4">Per-file HFS+ compression on Mac OSX 10.6</td>
<td class="s4">Joel Reymont</td>
<td class="s4">in progress</td>
<td class="s4">Apple developers on the darwin list recommended pre-compressing our builds, or not compressing at all. There are issues with code-signing and the update process. This is getting more complicated.</td>
</tr>
<tr>
<td class="s2">10%</td>
<td class="s3">{{bug|520309}}</td>
<td class="s4">Startup cache: replacement for fastload cache</td>
<td class="s4">Ben Hsieh</td>
<td class="s4">in progress</td>
<td class="s4">Ben's got a patch working now, and showing ~10% improvement in startup time. AWESOME.</td>
</tr>
<tr>
<td class="s2">60%</td>
<td class="s3">{{bug|503483}}</td>
<td class="s4">Turn on --enable-faststart for Firefox by default</td>
<td class="s4">Dietrich</td>
<td class="s4">dodgy</td>
<td class="s4">I got it working on Win7, and unsurprisingly, loading Firefox's core at user-login time meant that the first load of the app was very fast. However, Rob Strong brought up a bunch of problems on multi-user systems that might block turning this on by default. In the short term, we'll likely just enable some mechanism for testers to try it out.
</td>
</tr>
<tr>
<td class="s2">TBD</td>
<td class="s3">{{bug|513149}}</td>
<td class="s4">Speed up CSS parsing</td>
<td class="s4">Zack Weinberg</td>
<td class="s4">in progress</td>
<td class="s4">Zack has a series of small patches. Some help, most hurt (see [https://bug513149.bugzilla.mozilla.org/attachment.cgi?id=424821 chart]). Work ongoing.</td>
</tr>
<tr>
<td class="s2">&gt;10%</td>
<td class="s3">{{bug|512584}}</td>
<td class="s4">Super fast paths for Components.classes and Components.interfaces</td>
<td class="s4">Taras Glek</td>
<td class="s4">in progress</td>
<td>Taras is traveling, says he'll close it out when he returns.</td>
</tr>
</table>
'''More'''
* Rob Campbell [http://antennasoft.net/robcee/2010/02/01/mac-optimized-builds-of-firefox/ did research on Mac build optimizations].
* More details on the [https://wiki.mozilla.org/Firefox/Projects/Startup_Time_Improvements project page].
* Join us on IRC in [http://irc.mozilla.org/#startup #startup].


= Security =
= Security =


= Breakpad =
= Breakpad =
* DWARF CFI support submitted to Google for review. 6 of 12 patches approved; waiting for review on others.
* .eh_frame support {{bug|543542}}
** .eh_frame is a minor variant of DWARF CFI that the Linux C++ implementation uses to walk the stack to find exception handlers. Since it effectively provides the same stack-walking information as DWARF CFI, the toolchain omits DWARF CFI if the object file has .eh_frame data. Expecting to submit patches to Google for review today or tomorrow.
* DWARF debugging information: contributors showing up to extend functionality.


= Tree Management =
= Tree Management =
* switching from rev2 to rev3 talos hardware
** when is the best time?


= Roundtable =
= Roundtable =
==  Firebug ==
* {{bug|542203}} - fixed in Yslow, diff on-hand
* {{bug|542120}} - crash freq. increasing
* {{bug|541390}} - still unverified, but likely a freeze from the past
=== other highlights ===
* getfirebug.com refresh released
* 1.5.1 scheduled for next week

Latest revision as of 21:04, 2 February 2010

« previous week | index | next week »

Notices / Schedule

Gecko 1.9.3a1

Blocker Report

Browser / Front End

Behold, a mere sampling of the mighty awesome that the Firefox team hath wrought:

Lorentz

  • UI for crashed, out-of-process plugins basically done bug 538910

Firefox.Next

See the active projects page for the complete list of active work by the Firefox team.

GFX Update

  • Several projects are still ongoing:
    • Cairo update & merge
    • Direct2D/DirectWrite patch reviews and updates
    • Windows out-of-process plugins work - there's one large outstanding issue that Jim is working on.
    • Core Animation plugin support attempts - three independent attempts to support them, each with their own issues. Benoit is still working.
    • Fennelectrolysis shared memory work hopes to wrap up this week.

For more, see our weekly meeting page.

Layout Update

  • Found Windows "exit code 1" orange. Need to keep working on orange bugs. How do we get people working on them?
  • Fullscreen API proposed in WHATWG mailing list; please investigate. email thread starts here and here

Content Update

  • Landed HTML5 pushState (Justin Lebar, Jonas)
  • Worked on branch back porting (mrbkap, jst)
  • Atom table patches ready and up for review (Jonas)
  • Attended the Velocity summit. Feedback collected at Performance:Velocity2010 (Jonas)
  • Investigating what we need from electrolysis to do our security notifications right, for the lock icon etc (Honza)
  • Got large parts of content process event handling done, tests included (smaug)
  • HTML5 parser related about:blank work, and mochitest investigation (hsivonen)
  • CPOWs are getting easier and easier to use (bnewman)

Platform-specific Support Update

JS

  • nanojit work winding down for the moment
  • new baseline compiler starting up

Electrolysis

  • Plugins
    • Windows hangs edge cases still bedeviling: jimm on it (today?)
    • Two most obvious crashes dealt with today (bug 542248, bug 543376
    • One more topcrash diagnosed, perhaps dealt with today bug 542263
    • UI needs to land
    • status for 1.9.3a still ambiguous: either off-by-default or manipulate prefs so that only flash and silverlight are on by default
  • Fennec
    • Benedict starting on webprogresslistener
    • smaug has event listeners almost ready, might need CPOW
    • benjamn has CPOWs, need review... can somebody other than bsmedberg review this?
    • romaxa working on Qt
    • joe, memory-mapped drawing?
  • Jetpack
    • jdm working on multi-process jetpack with CPOW support

Startup Performance

Summary

Priorities

Estimated win Bug # Summary Owner Status Notes
~8% bug 525013 Investigate a more static build configuration of Firefox Joel Reymont in progress The main body of work is in the patch now, waiting on review from Ted. Joel's working various build-related cleanup, with the help of Brad Lassey and Ben Smedberg. Experiment Windows tryserver build available here.
up to 25% bug 514083 Per-file HFS+ compression on Mac OSX 10.6 Joel Reymont in progress Apple developers on the darwin list recommended pre-compressing our builds, or not compressing at all. There are issues with code-signing and the update process. This is getting more complicated.
10% bug 520309 Startup cache: replacement for fastload cache Ben Hsieh in progress Ben's got a patch working now, and showing ~10% improvement in startup time. AWESOME.
60% bug 503483 Turn on --enable-faststart for Firefox by default Dietrich dodgy I got it working on Win7, and unsurprisingly, loading Firefox's core at user-login time meant that the first load of the app was very fast. However, Rob Strong brought up a bunch of problems on multi-user systems that might block turning this on by default. In the short term, we'll likely just enable some mechanism for testers to try it out.
TBD bug 513149 Speed up CSS parsing Zack Weinberg in progress Zack has a series of small patches. Some help, most hurt (see chart). Work ongoing.
>10% bug 512584 Super fast paths for Components.classes and Components.interfaces Taras Glek in progress Taras is traveling, says he'll close it out when he returns.

More

Security

Breakpad

  • DWARF CFI support submitted to Google for review. 6 of 12 patches approved; waiting for review on others.
  • .eh_frame support bug 543542
    • .eh_frame is a minor variant of DWARF CFI that the Linux C++ implementation uses to walk the stack to find exception handlers. Since it effectively provides the same stack-walking information as DWARF CFI, the toolchain omits DWARF CFI if the object file has .eh_frame data. Expecting to submit patches to Google for review today or tomorrow.
  • DWARF debugging information: contributors showing up to extend functionality.

Tree Management

  • switching from rev2 to rev3 talos hardware
    • when is the best time?

Roundtable

Firebug

other highlights

  • getfirebug.com refresh released
  • 1.5.1 scheduled for next week
Retrieved from "https://wiki.mozilla.org/index.php?title=Platform/2010-02-02&oldid=199043"