Bump tbb to build3 (Fixup for Mullvad Browser backports bustage).

Actually override build number (Fixup for Mullvad Browser backports bustage).

Fixup for Mullvad Browser backports bustage.

Mozilla uplifted Bug 1900421 to 115 to 128, so we have to switch to the
new GitHub repository immediately as well.

boklm authored 9 months ago and Pier Angelo Vendrame committed 8 months ago

Following @ruihildt's suggestions we update the desktop file:

- Added MimeType field (to have the browser listed as a possible default
  browser application)

- Removed `X-MullvadBrowser-ExecShell` which is not used

- Removed the "Security" category

- Added the Version (of the desktop entry specification we use)

- Added Terminal

After #41167, the portable Windows installer are not
generated anymore for Mullvad Browser. We should include the normal
installer (without `-portable-` or `-install-` in the filename) in
the download-*.json files.

And remove the `-install` from the other one.

This should make sure the script always bumps the build number in
rbm.conf if needed.

Until #41159 is fixed, disable building packages for alpha releases.

Update the browser build script to apply the Windows installer
localization.

Add a script to add the translated strings to languages.nsh.

Currently, our localization platform (Weblate) does not support NSIS
files. Therefore, we use another standard format, and, before creating
the installer, we inject the translated strings to an NSIS script.

Define the user facing strings with LangString in languages.nsh.

Otherwise, the Windows installer might show an error when opening.

Tor Browser requires SSE2, so a long time ago we added a check on the
Windows installer to warn users in case their machine did not support
this instruction set.
However, Windows 7 is the minimum OS we support, and from 2018 it also
requires SSE2. Moreover, 14.0 will require Windows 10.

We are going to add translations for the installer, and at this point
it makes more sense to remove this check, rather than asking
translators to translate a messages nobody will see in a few months
from now.

Also, add support for skipping languages when they are evaluated as
empty strings after being processed as templates.
We already process them for Japanese, but with this change we will be
able to enable/disable languages for specific channels.

Also, fix a typo in the "id" language code in the map we use to
associate language names to codes for NSIS.

They are needed for formatting APIs to work as expected.

See !967.

Use stretch's glibc and stretch-security's linux to maintain
compatibility with Debian stretch, add linux-aarch64 glibc patches to
avoid build failures