Pier Angelo Vendrame authored Dec 06, 2023 and boklm committed Dec 07, 2023

We enabled base browser builds before Mullvad Browser was completely
created.
But now Mullvad Browser has some code paths that are not tested in
Base Browser, therefore it makes more sense to build MB instead of BB.

Fix a syntax error in previous commit,
695bb8ef.

With tor-browser-build#40737 the nightly/ directory is now prefixed
with the $projectname, and with tor-browser-build#40742 the makefile
rule to generate the incrementals is now $projectname-incrementals-nightly.

Following tor-browser-build#40439 and tor-browser-build#40067, the rbm
target to build for macos was changed.

At the same time we also change the name and directory of the macos
builds, using `macos` instead of `osx-x86_64`.

Add @dan, @henry, @ma1 to and remove @sysrqb from nightly build emails.

Since we now have multi-locale builds, keeping old builds uses less
space. We are also not currently doing builds every day since building
takes more than one day. So I think we can increase the number of days
of builds we keep (previously only one).

Archive the `ALL` bundles instead of `en-US`, since we now do
multi-locale bundles.

We also archive base-browser builds.

With tor-browser-build#40501, the name of the targets changed. We
also add the base-browser builds there.

Closes #40041

Pier Angelo Vendrame authored Feb 28, 2022 and richard committed Mar 31, 2022

In our Firefox patch set we have a test that assures that preferences
are kept in memory, instead of saving them to the disk or other
storage.
However, usually we check this kind of functionality here, not on
the Firefox/tor-browser code.

Other tests in that tor-browser commit should already be covered.

Pier Angelo Vendrame authored Mar 01, 2022 and richard committed Mar 31, 2022

Added a test to check that fingerprinting can only detect the fonts we
bundle with Tor Browser.
The code is readapted from TorZilla Print:
https://arkenfox.github.io/TZP/tests/fontlists.html
The list should be updated to match the bundled fonts. Also, we ship
different fonts in different platforms, so the test is enabled only for
Linux.

Added also a test to detect which glyph the browser can render and
their size. The code was originally developed by David Fifield and
Serge Egelman:
https://www.bamsoftware.com/talks/fc15-fontfp/fontfp.html#demo
The expected result of the test comes directly from a run on Tor
Browser, and we want it to be persistent, rather than having some
deisred metrics to match.

We remove anything related to gk while doing so. It should not be needed
anymore.

Closes: #40039.

The bridge test was not ending, because the bridge specified in the
configuration was down. With this new bridge, the test passed as
expected.

Canged fp_navigator and fp_useragent to test with the correct version.

The screen dimensions test was failing because letterboxing is disabled
on about:pages (I checked with Richard, and this is the indended
behavior).
Therefore, the test is now run on a TPO page, and it passes.

Updated the settings test, to check for new settings (see
tor-browser!215), and not to check anymore for deprecated and removed
settings (see tor-browser#40177).

As a result of these settings changes, some DOM objects are now
exposed:
* pointer events
* gamepads
* applicationCache
* visualViewport
However, Tor Browser already contains mitigatins against their use for
fingerprinting (e.g., gamepads do not work, Mozilla added some
protections to pointer events for Bugzilla#1363508, letterboxing should
prevent fingerprinting on visualViewport, cache storage is 0 bytes,
etc...).
Some other DOM objects are just new (e.g., clientInformation).

Finally, HTTPS everywhere test failed because now Firefox redirects to
HTTPS if available, even when HTTPS-Only Mode is not enabled.
To detect if HTTPS everywhere is actually the responsible of the
redirect, we need to run it with dom.security.https_first_pbm set to
false.

If the option test_post is defined and pointing to a subroutine, then
this subroutine will be run after each test.

We will be using this option to run an rsync after each Tor Browser build.

This should be closer to how it's done by tor-launcher and should
help us catch issues like tor-browser#40282.