Tor Browser 7.0a3 is released

Update (May 8 10:20 UTC): Mozilla released Firefox 52.1.1esr fixing among others a security bug which could lead to a potentially exploitable crash on the Windows platform. By default Tor Browser users are not affected as WebGL is put behind click-to-play placeholders, thanks to NoScript. Nevertheless, we are preparing a release later this week picking this fix up and, in addition to that, a lot of other improvements for our alpha series.

Update (Apr 24 8:36 UTC): Thanks to all for testing this alpha release so far. It turns out there are a number of issues that are affecting a lot of our alpha users. The following list should give an overview and help to avoid duplicate bug reports:

  • Tor Browser is crashing when opening/downloading files that need an external application to handle them. This is bug 21766.
  • Tor Browser is crashing on about:addons with the security slider set to "high" and does not show any preferences on about:preferences ticked. This issue is tracked in bug 21962.
  • The canvas prompt is not shown anymore in Tor Browser. This issue is tracked in bug 21778.
  • There is no sound on Linux systems without PulseAudio anymore. This is bug 1247056. Check this one out for Mozilla's reasoning behind dropping ALSA support.

Tor Browser 7.0a3 is now available from the Tor Browser Project page and also from our distribution directory.

This release features important security updates to Firefox.

This is the first alpha release which is based on Firefox ESR 52. We updated all of our patches that did not get upstreamed yet and made Torbutton and Tor Launcher multiprocess (e10s) compatible. After the first nightly build based on ESR52 went out we already fixed a number of bugs associated with this switch. But more remain, please help!

We hope having e10s and Mozilla's content sandbox enabled will be one of the major new features in the upcoming Tor Browser 7.0 series, both security- and performance-wise. While we are still working on the sandboxing part for Windows, both Linux and macOS have e10s and content sandboxing enabled by default in Tor Browser 7.0a3. There are already a number of bugs related to that on our radar which can be found on our bug tracker and which are tagged with the `tbb-e10s` keyword. If you find more, please report them!

The switch to Firefox ESR 52 raises the system requirements for Tor Browser on Windows and macOS. Computers running Windows and are not SSE2-capable are not supported anymore. On Apple computers with OS X < 10.9 Tor Browser won't run anymore either. Update (Apr 24 8:41 UTC): Only the browser part of Tor Browser is affected by these new constraints. If you are e.g. on Windows and are using the expert bundle or are extracting tor from Tor Browser it should run on any computer it used to run. The same holds for macOS with one exception: tor we ship in Tor Browser won't run on Apple computers with OS X 10.6 anymore either.

We updated our toolchains during the ESR transition as well. In particular we retired the old GCC-based one for our macOS cross-compilation and rely solely on clang/cctools now. As with previous releases building 7.0a3 is fully reproducible on all three supported platforms, even though we needed to deploy a last minute patch for Linux bundles this time.

Apart from switching to the new ESR and dealing with related issues we included a new Tor alpha (0.3.0.5-rc) and updated our NoScript (5.0.2) and HTTPS-Everywhere versions (5.2.14). The Sandboxed Tor Browser for Linux got updated to 0.0.6 making sure it is compatible with Firefox ESR 52.

As in Tor Browser 6.5.2 we provide a fix for Tor Browser crashing on github.com on Windows and for Twitter issues that got reported already a while ago. We update our security slider as well taking newer JIT preferences into account.

A note to Windows users: We signed the .exe files with a new codesigning certificate as the old one is about to expire. If there are issues with that new certificate, e.g. scary warnings showing up after downloading a Tor Browser .exe file and double-clicking on it, please let us know.

The full changelog since Tor Browser 7.0a2 is:

  • All Platforms
    • Update Firefox to 52.1.0esr
    • Tor to 0.3.0.5-rc
    • Update Torbutton to 1.9.7.2
      • Bug 21865: Update our JIT preferences in the security slider
      • Bug 21747: Make 'New Tor Circuit for this Site' work in ESR52
      • Bug 21745: Fix handling of catch-all circuit
      • Bug 21547: Fix circuit display under e10s
      • Bug 21268: e10s compatibility for New Identity
      • Bug 21267: Remove window resize implementation for now
      • Bug 21201: Make Torbutton multiprocess compatible
      • Translations update
    • Update Tor Launcher to 0.2.12
      • Bug 21920: Don't show locale selection dialog
      • Bug 21546: Mark Tor Launcher as multiprocess compatible
      • Bug 21264: Add a README file
      • Translations update
    • Update HTTPS-Everywhere to 5.2.14
    • Update NoScript to 5.0.2
    • Update sandboxed-tor-browser to 0.0.6
      • Bug 21764: Use bubblewrap's `--die-with-parent` when supported
      • Fix e10s Web Content crash on systems with grsec kernels
      • Bug 21928: Force a reinstall if an existing hardened bundle is present
      • Bug 21929: Remove hardened/ASAN related code
      • Bug 21927: Remove the ability to install/update the hardened bundle
      • Bug 21244: Update the MAR signing key for 7.0
      • Bug 21536: Remove asn's scramblesuit bridge from Tor Browser
      • Add back old MAR signing key to not break updating Tor Browser stable
      • Add `prlimit64` to the firefox system call whitelist
      • Fix compilation with Go 1.8
      • Use Config.Clone() to clone TLS configs when available
    • Update Go to 1.7.5 (bug 21709)
    • Bug 21555+16450: Don't remove Authorization header on subdomains (e.g. Twitter)
    • Bug 21887: Fix broken error pages on higher security levels
    • Bug 21876: Enable e10s by default on all supported platforms
    • Bug 21876: Always use esr policies for e10s
    • Bug 20905: Fix resizing issues after moving to a direct Firefox patch
    • Bug 21875: Modal dialogs are maximized in ESR52 nightly builds
    • Bug 21885: SVG is not disabled in Tor Browser based on ESR52
    • Bug 17334: Hide Referer when leaving a .onion domain (improved patch)
    • Bug 3246: Double-key cookies
    • Bug 8842: Fix XML parsing error
    • Bug 16886: 16886: "Add-on compatibility check dialog" contains Firefox logo
    • Bug 19192: Untrust Blue Coat CA
    • Bug 19955: Avoid confusing warning that favicon load request got cancelled
    • Bug 20005: Backport fixes for memory leaks investigation
    • Bug 20755: ltn.com.tw is broken in Tor Browser
    • Bug 21896: Commenting on website is broken due to CAPTCHA not being displayed
    • Bug 20680: Rebase Tor Browser patches to 52 ESR
    • Bug 21917: Add new obfs4 bridges
    • Bug 21918: Move meek-amazon to d2cly7j4zqgua7.cloudfront.net backend
  • Windows
    • Bug 21795: Fix Tor Browser crashing on github.com
    • Bug 12426: Make use of HeapEnableTerminationOnCorruption
    • Bug 19316: Make sure our Windows updates can deal with the SSE2 requirement
    • Bug 21868: Fix build bustage with FIREFOX_52_0_2esr_RELEASE for Windows
  • OS X
    • Bug 21723: Fix inconsistent generation of MOZ_MACBUNDLE_ID
    • Bug 21724: Make Firefox and Tor Browser distinct macOS apps
    • Bug 21931: Backport OSX SetupMacCommandLine updater fixes
    • Bug 15910: Don't download GMPs via the local fallback
  • Linux
    • Bug 21907: Fix runtime error on CentOS 6
    • Bug 21748: Fix broken Snowflake build and update bridge details
    • Bug 21954: Snowflake breaks the 7.0a3 build
    • Bug 15910: Don't download GMPs via the local fallback
  • Build system
    • Windows
      • Bug 21837: Fix reproducibility of accessibility code for Windows
      • Bug 21240: Create patches to fix mingw-w64 compilation of Firefox ESR 52
      • Bug 21904: Bump mingw-w64 commit to help with sandbox compilation
      • Bug 18831: Use own Yasm for Firefox cross-compilation
    • OS X
      • Bug 21328: Updating to clang 3.8.0
      • Bug 21754: Remove old GCC toolchain and macOS SDK
      • Bug 19783: Remove unused macOS helper scripts
      • Bug 10369: Don't use old GCC toolchain anymore for utils
      • Bug 21753: Replace our old GCC toolchain in PT descriptor
      • Bug 18530: ESR52 based Tor Browser only runs on macOS 10.9+
    • Linux
      • Bug 21930: NSS libraries are missing from mar-tools archive
      • Bug 21239: Adapt Linux Firefox descriptor to ESR52 (use GTK2)
      • Bug 21960: Linux bundles based on ESR 52 are not reproducible anymore
      • Bug 21629: Fix broken ASan builds when switching to ESR 52

You mean when you will be able to use Snowflake? If you have Tor Browser alpha this is available on Linux only for now. The Snowflake folks are working on getting it integrated into macOS and Windows bundles as well.

Anonymous

April 20, 2017

Permalink

Indeed it does seem that the effects are pretty noticeable! Youtube is much more smoother than it used to be on earlier version! :)

Anonymous

April 20, 2017

Permalink

Is the content process sandbox for Firefox 52 on Windows the same one that is used in Chromium?

Roughly, yes. Mozilla imports Google's code and adapts it to their needs. I am not sure whether the sandbox in Mozilla provides the same protections yet but the codebase is the same.

Anonymous

April 20, 2017

Permalink

When I went to about:addons and then "Get Add-ons" and I choosed uBlock to install it, it told me "Firefox blocked this site from attempting to install add-on" or something allong those lines

This doesn't happen with Tor Browser 6.5.2

So, I just tried with a clean 7.0a3 on a Linux machine. I opened about:addons and entered "ublock" in the "Search all add-ons"-checkbox and installed it that way. It worked as expected. Going to a website to test whether it actually does things confirmed that. How can I reproduce your problem?

That does not make a difference for me. Does setting "extensions.logging.enabled" to "true" give you some error messages in the browser console (Ctrl+Shift+J) that explain what is going on? Is that reproducible with a clean new 7.0a3?

FF allows installing addons from AMO & testpilot sites only. Try to install NoScript from its site, and you'll get the same error. Also the behavior of GetAddons is different for every Security Settings level (+ e10s on/off)!

I was able to install uBlock, but all it shows is the button to open the dashboard/settings. I can't see or use anything below enable/disable i/o button e.g. requests blocked, element picker, logger, block elements & media, etc.

Anonymous

April 20, 2017

Permalink

Since there is no mention of it I'd assume Thunderbird/IceW with TorBirdie is still not working in the 7.0x platform

Why not? I think on Windows it should work out of the box. On macOs and Linux we currently test unix domain sockets and I guess TorBirdy/Thunderbird need to get taught how to deal with that, yes. But I think it is reasonable to assume that this will be solved until that domain sockets feature reaches the stable channel.

Anonymous

April 20, 2017

Permalink

Please clarify your announcement about now requiring SSE2-capable processors : is it for the *browser* component only, or is it applicable to the core Tor proxy (viz, tor.exe on Windows 32 bits) ?

Anonymous

April 20, 2017

Permalink

About your note to Windows users: what exes do you sign? It is distributive only. No code signing found. And what could be wrong with it?

Anonymous

April 20, 2017

Permalink

19:23:17.896 XML Parsing Error: undefined entity
Location: jar:file:///x:/Tor%20Browser/Browser/omni.ja!/chrome/toolkit/content/mozapps/extensions/pluginPrefs.xul
Line Number 10, Column 3: 1 pluginPrefs.xul:10:3
19:23:17.932 TypeError: libLabel is null 1 PluginProvider.jsm:85:32

Anonymous

April 20, 2017

Permalink

Very crashy on windows. Managed to crash it from the hamburger menu's Customize feature and crashed again when trying to post a comment on this very page. Looks like some very strong improvements otherwise, keep up the great work

Do you have steps to reproduce any of your crashes? If so, please let us know them, so we have a chance to get those issues fixed in the next (alpha) release. Thanks.

Both crashes happened within minutes of a fresh install. I'm happy to go more in-depth, but when the crashes are that easy to experience I ponder the value of reporting when just a few minutes of experimenting with the UI and browsing around some sites such as the tor blog can provoke several. The browser totally freezed when typing a comment in the comment box of this blog, for one. Another crash I experience happened after clicking the hamburger menu, then clicking Customize, then dragging some icons to and from (and on either side of) the address bar. IIRC the crash happened after dragging the search bar off and placing it off the browser (to the dock of unused items just above the cut/copy/paste entries). I experienced another crash when opening about:config, and another crash when changing my download folder location from the Options screen. Also, many UI elements are missing throughout the browser (such as the little dots that indicate a selected option from the General section of the Options menu where the Downloads folder may be changed). Sorry if all of this is brief. I don't get any crashes in 7.0a2 or in the new 6.5.2. I greatly anticipate being able to use the new e10s features and hope a lot of this can be ironed out for 7.0a4. Platform is Win 10 64bit with creators update.

Thanks. We'll look into it and get back to you in case we have further questions. Special thanks for testing this on Windows as this is usually the platform that gets tested not so much by the developers yet we have the most users on it.

Okay, I tried quite a bit on my Windows 7 machine but I could not reproduce any of those crashes you are seeing. Could you give me some steps to reproduce at least some of them? Are they happening with a clean, new Tor Browser 7.0a3 (i.e. one freshly downloaded and without any modifications)?

Anonymous

April 20, 2017

Permalink

Warning: ‘nsIOService::NewChannel()’ deprecated, please use ‘nsIOService::NewChannel2()’ aboutTor.js:42:18
Warning: ‘nsIAboutModule->newChannel(aURI)’ deprecated, please use ‘nsIAboutModule->newChannel(aURI, aLoadInfo)’ NetUtil.jsm:357:15

Anonymous

April 20, 2017

Permalink

There are so many changes in FPI, so could you change Torbutton's INFOs about violations of it into warnings about it?

I see. There is a bunch of stuff that we'll need to disable for the next alpha I hope or latest for the stable release. Thus, I think this will get much better. Let's wait for that until we think about changing the log level.

Anonymous

April 20, 2017

Permalink

Antivirus AVG told me TOR was behaving suspicuously. I hit the wrong button, lost my TOR and cannot reinstall on Win 10. Merde!

Anonymous

April 21, 2017

Permalink

*All* Sound in Linux is broken now after update, :-(
I updated from the hardend version to
"7.0a3 (based on Mozilla Firefox 52.1.0) (64-bit)"

Hi Yawning
Yes,
It looks really bad at the moment, from the multiple bug reports and even articles on the net it seems that they are very firm on to this track now,

they do not want to listen to the users that ,without warning, lost their websound, the solution they tell is add pulseaudio and make your system bad or "then you have to switch distro" , many many of the users in this situation can or will not add on the pulseaudio-layers now,

Therefore users will probably look for another browser instead? But as far as I know noscript and more is only for firefox, so that is not an option for some users that still care not to run all kind of not so nice scripts by default.

Same on Gentoo after installing Pulseaudio. No more sound. Pulseaudio doesn't detect my Xonar card. All it shows is the dummy-sink.

Hmm

This is probably not the optimal way to go now for security, but if mozillafox really have decided that it must use a "man in the middle", is there some other easy way to use or make a small and more secure middle-ware, that now ONLY redirects firefox audio to use alsa as normal,

and maybe even without using JACK, ..if JACK is the normal system setup..
So if our program just pretend to be pulseaudio, what happens? can we then get sound back again without having all the bloaty scary downsides with pulseaudios insecurity and the other damages to the sound quality/latency problems in our systems?

(It seems so, as kxstudio have some option to run both JACK and Pulse in parallel (?) as I understand the solutions it will always be another program layer on top of alsa and not direct to alsa as usual, But why use pulseaudio at all then? Only because of the browser.. )

so if possible (maybe it have been done already by someone?), would this "minipulse" middleman be the second best alternative after having a future "pulsefree" Firefox version?

/Cw

Hi,
sorry I dont know much about that problem with firefox,
is that "worse" than a possible memory and audio-snooping trough pulseaudio?

But the point was more from a security perspective, as I see it.
How to get rid of as much unknown software in the middle as possible.
Not to install truckloads more of them :-)

/ Cw

Anonymous

April 21, 2017

Permalink

I can't download files throguh TOR in this version. The page gets blank before asking to save file, then I get the "window crashed" warning (and ALL windows actually crash).

Join the discussion...

This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.

8 + 8 =
Solve this simple math problem and enter the result. E.g. for 1+3, enter 4.