updates Meet the browser: Firefox Next

[Max Norris]

Firefox is really beginning to make strives - today I've been surfing using a e10s window without a single crash. Now all they need to do is move downloads/history/bookmark manager to a tab instead of a new window!

Ditto that sentiment, they're really making it look a lot better than it used to, as much as I like the flexibility the not-so-older versions still looked like a refugee from the late 90's. A few more bits to get revamped, although there is an addon that lets you move those specific windows you mentioned into a sidebar, not quite the same but still. I just hope they start putting a lot more effort into the under-the-hood stuff, give it a much needed kick in the pants for performance and new developments, progressing but taking wayyy too long.

[.stan]

New InContent based Default Browser bar and I must say I welcome this Chrome copy since it moves away from old stupid dialog.

 

 

Chrome is moving away from those info bars because they can easily be faked on a website. Chrome does info bubbles now.

[.stan]

Our favorite Australis theme creator blinky has updated his userstyle ( https://userstyles.org/styles/96014/firefox-australis-windows-8-appearance ) to match the latest mockups of Project Chameleon ( https://blog.mozilla.org/creative/2013/11/27/introducing-project-chameleon/ )

[+Neyht Subscriber²]

Indeed, I welcome it too as I really hated that dialog. Firefox is really beginning to make strives - today I've been surfing using a e10s window without a single crash. Now all they need to do is move downloads/history/bookmark manager to a tab instead of a new window!

 

How about the print window? Nothing about the print window looks good.

[BooBerry]

• Compiler (JIT)
  • bug 1041781: Landed native to bytecode mapping
  • bug 1054340: Landed post-native-to-bytecode-mapping fix to remove MPcOffset instructions

• Garbage Collection
  • bug 650161: Landed first cut of compacting GC implementation (build with --enable-gccompacting to try it out)
  • bug 995284: Made GC behvaiour deterministic in shell builds

• Front End and Other
  • bug 1052248: Tweak the goodAllocated heuristics for JS arrays
  • bug 1041688: acquired properties analysis working but has some regressions

Layout

• caret cleanup landed (bug 1048752)
• CSS transitions now work for style changes that trigger reframes (bug 625289)
• initial CSS Ruby reflow code landed, but still more work to do on ruby

Media

• Continue to prep OpenH264 and Screensharing for Fx33.
  • Major UI redesign slated for Fx35
  • Window list rescan uplifted to 33
  • Talky.io is experimenting with ScreenSharing in their beta pages
• Working a 2.0 blocker with QC
• Firefox Desktop has been giving the Loop team a hand on front-end issues
• GMP plugin sandboxing has landed

Necko

• landed bug 820391: we're finally getting DNS time-to-live (TTL) accurately (Windows-only for now).
• HTTP cache v2: some last bugs before we release: mostly tests but also a few real bugs (bug 1042192, bug 1054425, bug 660749)
• minor fixups for HTTP2: we're close to being able to pref it on.

Performance

• bug 1045108: Please review your Telemetry probes and make sure they're still useful. Also please set the expiry dates for your probes (e.g. "expires_in_version": "never", or "35" for Firefox 35) -- we'll be automatically marking unclaimed probes for expiry in Firefox 40.
• bug 1037494: Also provide email addresses for automatic regression notifications in the "alert_emails" field.

[The_Decryptor Veteran]

...

landed bug 820391: we're finally getting DNS time-to-live (TTL) accurately (Windows-only for now).

...

We don't actually have TTL info yet, that's coming later (That patch just moved them over from using the getaddrinfo API to the special Windows API, but didn't enable any of the new functionality)

Ideally Firefox would do DNS lookups itself (Similar to Chrome), but doing so would break certain features people rely on (Like editing the hosts file, or network computer name lookup, etc.) The standard APIs the system provides is kinda crap (No TTL info, the system can sort them wrongly like on OS X, etc.) On the upside it would provide better support for DNSSEC and stuff like SRV records.

[BooBerry]

We don't actually have TTL info yet, that's coming later (That patch just moved them over from using the getaddrinfo API to the special Windows API, but didn't enable any of the new functionality)

Ideally Firefox would do DNS lookups itself (Similar to Chrome), but doing so would break certain features people rely on (Like editing the hosts file, or network computer name lookup, etc.) The standard APIs the system provides is kinda crap (No TTL info, the system can sort them wrongly like on OS X, etc.) On the upside it would provide better support for DNSSEC and stuff like SRV records.

 

Thanks for that

 

 

what do you exactly mean by "Like editing the hosts file"

[The_Decryptor Veteran]

Wikipedia can explain it better than I can: http://en.wikipedia.org/wiki/Hosts_%28file%29

The API the OS provides for applications to map hostnames to IPs isn't just for DNS, it maintains a cache of previous lookups, and can also pull information from other sources (Such as Bonjour for *.local names, WINS for Windows boxes, the hosts file for custom overrides, etc.) alongside DNS. The problem is, the API kinda sucks, so applications that want proper DNS information need to query the DNS servers themselves (Or use the better API that Windows provides, etc.), but doing so completely bypasses all the other systems the OS provides for name lookup.

[BooBerry]

Wikipedia can explain it better than I can: http://en.wikipedia.org/wiki/Hosts_%28file%29

The API the OS provides for applications to map hostnames to IPs isn't just for DNS, it maintains a cache of previous lookups, and can also pull information from other sources (Such as Bonjour for *.local names, WINS for Windows boxes, the hosts file for custom overrides, etc.) alongside DNS. The problem is, the API kinda sucks, so applications that want proper DNS information need to query the DNS servers themselves (Or use the better API that Windows provides, etc.), but doing so completely bypasses all the other systems the OS provides for name lookup.

 

 

actually i meant was does it now use windows apis?

and so honors the hosts file better? as opposed to bypassing hosts allowed & blocked lists?

[The_Decryptor Veteran]

Oh, right :laugh:

 

It still uses the data from the hosts file because it's still using the system provided lookup APIs, just a slightly different one that gives them more info (The TTL among other things).

[BooBerry]

Oh, right :laugh:

 

It still uses the data from the hosts file because it's still using the system provided lookup APIs, just a slightly different one that gives them more info (The TTL among other things).

Thanks

would it result in better performance ?

 

& does anyone know if offmainthread are enabled yet[or stopping it] on linux?

and e10s are on schedule?

[The_Decryptor Veteran]

Performance should be exactly the same, regardless of the method they use.

 

OMTC on Linux depends on image surfaces (vs. X11 surfaces), I know some work for that recently landed but I think it's still disabled by default.

[BooBerry]

Adobe Flash Player 15.0.0.130 Beta includes improved support for browser zoom levels, an enhancement that?s specifically designed for Windows 8 devices with a touch-capable display.

Adobe says that this particular improvement has been specifically developed to solve issues experienced in the past when trying to change the zoom level of playing Flash content, with browsers on Windows 8 sometimes causing playback speed problems.

the new preview build brings a number of improvements for Chrome/Firefox users on Windows, making hardware video decoding enabled by default in this browser

Adobe says that this particular option will ?drastically? reduce the CPU usage and improve performance on supported video players. The same feature should also be introduced on Mac OS X computers in the coming months

[+Zlip792 MVC]

Small UI changes:

 

- Restart to Update was seems vague as per developer since people might get confused and think of it as system *OS* restart, so they changed that string to include channel name as well.

 

- Default browser control has been moved to General tab in Options.

 

- Array and string performance improvement are happening in JS engine and not to mention Bobby Holley is working as well to improve WebIDL JS based Bindings faster.

[The_Decryptor Veteran]

The DNS changes on Windows got backed out because they broke IPv6 connections.

[BooBerry]

• Bug 1054050 ? rsync should have a timeout in new vcs sync
• Bug 1054048 ? Should monitor lock files for new (mozharness) vcs sync
• Bug 1053742 ? Runner dies before starting buildbot due to hg failures and leaves slave doing nothing
• Bug 1053736 ? Sheriffs/community members should have access to all RelEng log files
• Bug 1050808 ? l10n repacks should retry on failure
• Bug 1050172 ? Dev services (e.g. hg.mozilla.org and git.mozilla.org) status page(s)
• Bug 1049530 ? vcs sync should avoid polling hg when possible
• Bug 1048971 ? Slave loan request for a bld-linux64-ec2 vm for pmoore
• Bug 1047501 ? Make b2g-manifests less fragile to changes which require mirroring
• Bug 1046047 ? Try jobs backing up - need to deprioritise ash

Other bugs

• Bug 1055596 ? Add command line options to web-platform-tests for crash reporting.
• Bug 1053960 ? fx-team branch of gecko-dev hasn?t synced in 8 hours
• Bug 1053197 ? Failures during install_module() are shown as ERRORs in the log even though the run succeeded
• Bug 1050109 ? stop jobs from hg?n mozharness/tools repos
• Bug 1049930 ? Create builds for Dolphin on mozilla-central, mozilla-inbound, b2g-inbound, fx-team, and mozilla-aurora
• Bug 1049711 ? Install treeherder-client on internal PyPI
• Bug 1049525 ? Tooltool timeouts do not produce a TBPL-compatible error message
• Bug 1047979 ? (foopy32) foopy32 problem tracking
• Bug 1047244 ? gaia-build jobs failing with ?build.test.js | Build GAIA from differece app list GAIA_DEVICE_TYPE=tablet? or ?multilocale.test.js | multilocale Integration tests ?before all? hook?
• Bug 1046694 ? Splitting Gij and Gip tests in several chunks
• Bug 1045675 ? Decommission foopy117 and associated tegras
• Bug 1045488 ? esr31 branch on https://github.com/mozilla/gecko-dev
• Bug 1044515 ? Legacy vcs sync not able to update legacy mapper quick enough for b2g_build.py, causing b2g builds calls to mapper to timeout
• Bug 1044391 ? Funsize should return correct MIME type
• Bug 1043938 ? Decommission foopy117 and associated tegras
• Bug 1043662 ? [Flame] Missing labels in Firefox Ringtones and Alerts
• Bug 1043530 ? Loan an ami-6a395a5a instance to Aaron Klotz
• Bug 1042128 ? Figure out Play Store dynamics for Android 2.2/ARMv6 EOL support releases
• Bug 1040887 ? Funsize returns errors out and returns an HTTP 500 when requested for a non-existent partial
• Bug 1040319 ? Ensure that Fennec builds from mozilla-esr31 have a buildID to allow for armv6/Android 2.2 users to update to mozilla-esr31 apks
• Bug 1040308 ? Trees closed due to constant slowness and unavailability of hg.mozilla.org
• Bug 1040013 ? No manual reconfigs: buildbot masters and foopies to update themselves
• Bug 1038478 ? releases/l10n/mozilla-beta/ repos are inconsistent across hgweb*
• Bug 1036998 ? hg web head synchronization taking excessively long
• Bug 1031378 ? Changes to manifests in b2g-manifest that reference repos that are not mirrored to git.mozilla.org should trigger devs/releng/sheriffs
• Bug 1029342 ? Request to mirror a FFOS repository
• Bug 1025788 ? [Flame] build tcpdump
• Bug 1023576 ? Deploy new vcs_sync into production with changes from bug 962853
• Bug 1022701 ? Fix and re-enable hg publisher
• Bug 1020613 ? vcs sync should only push tags/heads that have changed since last successful push
• Bug 1019165 ? install vs2013 across all our win64 machines and add a junction link for it in our GPO
• Bug 1017759 ? AWS region-local caches for https stuff
• Bug 1010126 ? Buildbot masters should only be able to run a single reconfig at once
• Bug 1007230 ? Timeouts during pip install (?Error running install of package, /builds/slave/test/build/venv/bin/pip install ?download-cache /builds/??, ?Error running install of package, /builds/slave/talos-slave/test/build/venv/bin/pip install ?download-cache??)
• Bug 1006136 ? The VM for NSS testing on Windows has turned unstable (nss-vm-win32-1)
• Bug 1004617 ? https://hg.mozilla.org/build/tools/file/tip/buildfarm/maintenance/end_to_end_reconfig.sh should also update bug status once they are in production
• Bug 978928 ? Reconfigs should be automatic, and scheduled via a cron job
• Bug 929336 ? permanent location for vcs-sync mapfiles, status json, logs
• Bug 927199 ? two new instances like vcssync{1,2}, puppetized
• Bug 851270 ? Don?t clobber the source checkout
• Bug 805515 ? (b-2008-ix-0092) b-2008-ix-0092 problem tracking
• Bug 795791 ? (b-2008-ix-0087) b-2008-ix-0087 problem tracking
• Bug 772959 ? [tracking bug] foopy-less setup for pandas
• Bug 753291 ? mock should fail if it doesn?t install a dependency

mozilla::pkix ships in Firefox!

[BooBerry]

Updated Firefox VR Builds  

I?d like to announce the third Firefox Nightly build with experimental VR support. Download links:

• Windows (64-bit) Download
• OS X (10.7+) Download

This build includes a number of fixes to CSS VR rendering, as well as some API additions and changes:

• Fixed CSS rendering (see below for more information)
• Support for DK2 via 0.4.1 SDK (extended mode only)
• Experimental auto-positioning on MacOS X ? when going fullscreen, the window should move itself to the Rift automatically
• hmd.setFieldOfView() now takes zNear and zFar arguments
• New API call: hmd.getRecommendedEyeRenderRect() returns the suggested render dimensions for a given eye; useful for WebGL rendering (see below)

The DK2 Rift must be in Extended Desktop mode. You will also need to rotate the Rift?s display to landscape. If tracking doesn?t seem to be working, stop the Oculus service using the Configuration Tool first, then launch Firefox.

CSS Rendering

Many issues with CSS rendering were fixed in this release. As part of this, the coordinate space when in fullscreen VR is different than normal CSS. When in fullscreen VR mode, the 0,0,0 coordinate location refers to the center of the viewport (and not the top left as is regular in CSS). Additionally, the zNear/zFar values specified to setFieldOfView control the near and far clipping planes.

The coordinate units are also not rationalized with CSS coordinates. The browser applies a per-eye transform in meters (~ 0.032 meters left/right, or 3.2cm) before rendering the scene; tthus the coordinate space ends up being ~1px = ~1m in real space, which is not correct. This will be fixed in the next release.

Here?s a simple example of showing 4 CSS images on all sides around the viewer, along with some text. The source includes copious comments about what?s being done and why.

Known issues:

• The Y axis is flipped in the resulting rendering. (Workaround: add a rotateZ() to the camera transform div)
• The initial view doesn?t face the same direction as CSS (Workaround: add a rotateY() to the camera transform div)
• Manual application of the HMD orientation/position is required.
• Very large CSS elements (>1000px in width/height) may not be rendered properly
• Units are not consistent when in VR mode

getRecommendedEyeRenderRect()

NOTE: This API will likely change (and become simpler) in the next release.

getRecommendedEyeRenderRect will return the rectangle into which each eye should be rendered, and the best resolution for the given field of view settings. To create an appropriately sized canvas, the size computation should be:

var leftRect = hmd.getRecommendedEyeRenderRect("left");var rightRect = hmd.getRecommendedEyeRenderRect("right");var width = leftRect.x + Math.max(leftRect.width + rightRect.x) + rightRect.width;var height = Math.max(leftRect.y, rightRect.y) + Math.max(leftRect.height, leftRect.height);

In practice, leftRect.x will be 0, and the y coordinates will both be 0, so this can be simplified to:

var width = leftRect.width + rightRect.width;var height = Math.max(leftRect.height, rightRect.height);

Each eye should be rendered into the leftRect and rightRect coordinates. This API will change in the next release to make it simpler to obtain the appropriate render sizes and viewports.

[BooBerry]

Today?s Nightly 34 build includes the bug 286355:

 

a profile unlocker for our Windows users. This should be very helpful to those users whose workflow is interrupted by a Firefox instance that cannot start because a previous Firefox instance has not finished shutting down.

 

 

Firefox 34 users running Windows Vista or newer will now be presented with this dialog box:

 

 

 

 

Clicking ?Close Firefox? will terminate that previous instance and proceed with starting your new Firefox instance.

 

Unfortunately this feature is not available to Windows XP users.

 

 

While this feature adds some convenience to an otherwise annoying issue, please be assured that the Desktop Performance Team will continue to investigate and fix the root causes of long shutdowns so that a profile unlocker hopefully becomes unnecessary.

The DNS changes on Windows got backed out because they broke IPv6 connections.

 

 

Any idea when it will re-land?

[The_Decryptor Veteran]

No idea, there hasn't been any activity on the bug since it was backed out.

[BooBerry]

No idea, there hasn't been any activity on the bug since it was backed out.

 

:/  thanks mate

[BooBerry]

The e10s team is asking everyone to dogfood their first milestone of e10s.
 

Chris Peterson wrote:The e10s team has completed our M1 milestone to fix major pain points for dogfood testing. Now it's time to dogfood!

To enable e10s, flip the "browser.tabs.remote.autostart" pref to true and restart Nightly. If you have any problems or questions, drop by the #e10s channel on IRC.

PLEASE read this list of known issues you are likely to hit:

https://etherpad.mozilla.org/e10s-known-issues

If you file e10s bugs, please include the word "e10s" in the summary and/or set the "tracking-e10s" flag to '?' to ensure your bugs get triaged.

For a list of tested add-ons (compatible, incompatible, and popular add-ons we'd like people to test), see:

http://arewee10syet.com

thanks!

chris

[Karl-Johan]

Anyone been able to run Java with a e10 build? I've tried both Java 7 and 8 and latest Firefox tinderbox build.

 

Java version doesn't display when checking for version. http://www.java.com/sv/download/installed.jsp

 

Another issue that's quite annoying is when clicking on links that open in a new tab. When I click a link on e10

it opens a empty tab and loads the new page in the current tab. Anyone else getting this?

[BooBerry]

Big changes coming with FX35 Nightly train on September 2nd with switch to Visual C++ 2013 compiler.

Some benefits from this change:
* No more linker OOM crashes. VS2013 includes a 64-bit toolchain for 32-bit builds, so the linker will no longer be limited to 4GB address space.
* The linker capacity opens the door for merging our binaries into libxul (like we do on the other platforms)
* More than 2x improvement in PGO build times
* Better language support

[+Zlip792 MVC]

Easy theme switching through Customization: http://msujaws.wordpress.com/2014/08/27/new-in-firefox-nightly-recommended-and-easy-theme-switching-through-customize-mode/

 

 

New Private Window content page following new simple design process:

[The_Decryptor Veteran]

It begins?