stark raving finkle

The Mobile team spent last week in not-so-sunny Toronto, pushing Firefox for Android closer to beta-level release criteria. We are getting very close. There is a lot of amazing work landing, especially in graphics. The Graphics team is working with Mobile to integrate GL support into the Gecko rendering system. It’s taken some time, but the results are really starting to look good in Nightly builds.

Mobile wasn’t the only team meeting in Toronto. The Firefox team also got together, and one of main features of the week was HACKING. Pure, seat-of-the-pants hacking. A few of the Mobile team were able to take part in the hack sessions too:

• Lucas created a nice prototype of the new Reader Mode in Firefox Mobile. He integrated some of the readability code, created a special viewer and a way to manage reading lists.
• Margaret has a work-in-progress of HTML5 context menu support working.
• Margaret also updated a patch to add support for the site identity UI.
• Sriram started work to support text selection in web page content.
• Jan Odvarko (DevTools) created a prototype of a remote network monitor tool working with Firefox on Android. With the new remote JS debugger, that’s two new remote development tools in the pipeline.

I was asked to give a presentation on the recent developments in Firefox Mobile for the Mozilla Vision 2012 conference in Tokyo. It gave me a chance to reflect a bit on what we did, why we did it and how things are going. The “what” in this case is rewriting the UI for Firefox Mobile using native Android widgets. The “why” can be summed up in the following goals:

• Faster start-up time
• Support for Flash
• Use less memory

We have been fairly “heads down” working on native Firefox for Android over the last 3 months. Re-writes are scary. It’s a race to re-implement existing functionality while adding all new bugs. We are finally to the point in the project where things are settling down. We are focused on stability issues, getting ready to release the native version to a larger audience. As for the primary goals, we have good news:

• Start-up time is many times faster than the XUL version. Launching via the icon is almost instantaneous.
• Flash is supported on Froyo and Gingerbread. We need to reverse engineer the changes made for Honeycomb and ICS. There is no documentation for this work.
• With a single-process, we have reduced memory quite a bit. We still can get killed in the background by Android – it’s supposed to do this – but when this happens, we start up very quickly and restore your session.
• We still support add-ons using the native UI!

I’m looking forward to being able to move forward again, designing/implementing new features on a solid foundation. We have plenty of new experiments and projects waiting to move forward.

Mozilla Vision 2012 was a great conference and very well attended. I’m very glad I had the chance to participate. Here is the link to my Google Docs presentation. It has a bit more details.

… and go grab a Nightly and see the changes for yourself!

The new native Android UI version of Firefox does not ship with a dedicated Error Console. Instead, all console messages are redirected to the Android system log – also known as logcat. If you have the Android SDK installed, you already have a way to view the logcat:

# show the complete log
adb logcat

# show only Firefox log messages
adb logcat | grep "Gecko"

# show only Firefox error console messages
adb logcat | grep "GeckoConsole"

The Android stock browser also does the same thing. If you don’t have the Android SDK installed, you can install an Android app, like aLogCat, to scan the log instead.

Firefox for Android does support console API. You can use the API to send data to logcat from your web pages:

// Outputs a message. See also: info, warn, error, debug
console.log("value=", x);

// Outputs a simple call stack from where the call is executed
console.trace();

// Outputs results of a timer
console.time("my-timer");
console.timeEnd("my-timer");

We have been talking a lot recently about the new native UI version of Firefox for Android. Built using native Android UI widgets, packed with major speed improvements and large memory reductions. We have not been working on the project very long, but we have been making solid progress. We decided it’s time to move the native builds into our nightly delivery channel.

The nightly channel is a snapshot of our very latest development work, before it is ready for widespread testing. This is our least stable update channel, and is not recommended for most users. Nightly will offer to update itself automatically, once a day, with the latest changes. If you have a nightly Firefox installed on your Android device, when it updates on November 23rd, you’ll be running the new native UI build.

The XUL-based version of Firefox had several years of development and the nightly builds have a relatively high level of stability. The new native UI builds are not that mature and you will find some OMGWTFBBQ flaws. We have a quick summary of known issues on the Nightly download page. I’ll call out a few of the big ones here:

• Sync is not supported yet: Sync is also being refactored for the move to native and isn’t ready yet.
• Panning and zooming have issues: Pages layout too wide for phones and zooming results in fuzzy content. These are our highest priority items and we are landing improvements everyday.
• Tablet UI is not ready: We are focused on phones and the tablet UI has not caught up. Tablet UI is not a high priority for the native UI project since we can fallback to shipping the XUL-based version on tablets until we have a solid native tablet UI.
• Bookmarks and History are saved to the system storage: We save bookmarks and history in the same storage that the stock Android browser uses. It provides tighter integration, but does mean the data is not locked away in a Firefox-only location. Passwords and other data are stored in a Firefox-only location.

If any of these issues make you uncomfortable or you are worried about the changes in stability, we urge you to move to the more stable Aurora channel. The Aurora channel is still the XUL-based Firefox you have been using.

We certainly hope you give the new nightly a chance and help us improve the quality and stability by filing bugs.

Note: If you have been using the native UI builds, especially on a tablet, we recommend uninstalling and re-installing the current nightly.

Firefox for Android, both the XUL and Native versions, currently requires an ARMv7 CPU. We are looking into adding support for ARMv6, but the ARMv7 limitation had one painful drawback: We could not run Firefox on the Android emulator since it only supported ARMv6.

This is no longer true! The latest version of the Android SDK ships with ARMv7 images for the emulator.

Christian Holler (decoder on IRC) was kind enough to make a wiki page with instructions on setting up the emulator with ARMv7 support. As Christian points out, Firefox does not run at top speed in the emulator, but it’s a heck of a lot better than no “desktop” options at all. Oh yeah, the new native Android UI builds have no desktop counterparts, so the emulator is the only way to run those builds on a desktop machine.

Yes! The emulator runs Ice Cream Sandwich (ICS). Yes! Fennec needs to support the ActionBar on ICS 🙂

You have probably seen some announcements and read some blog posts about Mozilla’s recent change in direction for Firefox Mobile on Android. We have decided to drop the XUL-based UI and re-build the application using native Android widgets. Here’s some of the rationale, from Johnathan’s newsgroup post:

• Startup – A native UI can be presented much faster than a XUL based UI, since it can happen in parallel with Gecko startup. This means startup times in fractions of a second, versus several seconds for a XUL UI on some phones.
• Memory Use – We believe a native UI will use significantly less memory.
• Responsiveness – A native UI has the potential for beautiful panning and zooming performance.

Another major change is a move away from multi-process architecture (e10s) as well. The web content process was great for stabiilty, keeping crashes from taking down the entire application, but it also increased our memory usage and created some performance issues. In the new application, Gecko is running in a separate thread, not a separate process.

Quite simply everything you see is a native Android widget. Even the web content is displayed in a native view, very similar to the multi-process layers system we previously used. This allows us to asynchronously pan and zoom the web content, without waiting for the browser to scroll or zoom the actual content.

Even though the UI is completely implemented in native widgets, there is still a lot of JavaScript around, it’s just not visible. JavaScript is the perfect binding language into the Gecko platform and we are still using it for many of the same tasks. If you have ever built a browser using XUL, take a look at the browser,js file and you will see some familiar code. Because we have such a strong JavaScript binding layer, we can support add-ons in much the same way as a XUL-based application. More on that in a future blog post.

A few more details on how these three systems interact can be found in this basic architecture document. There is some details on the simple messaging system we use to allow the JavaScript and Java systems to communicate too.

The current nightlies are very usable, and support Flash (yes they do!) – so go grab the APK and try it out yourself.

Firefox Mobile recently hooked up the same performance data telemetry system used by desktop Firefox. If you use the Nightly version of Firefox Mobile, you can start sending anonymous performance data back to Mozilla. We will use this data to improve future versions of Firefox Mobile.

It’s really easy to use and is completely anonymous. Once you enable the feature, everything happens in the background. To enable, go to the Feedback panel and make sure Send performance data is toggled on.

While you’re there, send us some feedback on Firefox Mobile as well.