Kathleen Brade authored 10 years ago and Pier Angelo Vendrame committed 1 year ago

Allow using NSS on all platforms for checking MAR signatures (instead
  of using OS-native APIs, the default on Mac OS and Windows).
  So that the NSS and NSPR libraries the updater depends on can be
  found at runtime, we add the firefox directory to the shared library
  search path on macOS.
  On Linux, rpath is used to solve that problem, but that approach
  won't work on macOS because the updater executable is copied during
  the update process to a location that can vary.

Firefox creates a profile as a fallback for old versions.
However, we do not need this, so we want to prevent Firefox from
creating the related directories.

We comment out the code, but should Mozilla be interested in a
compile-time flag, we could rewrite the patch to add one.

See also https://bugzilla.mozilla.org/show_bug.cgi?id=1770174.

henry authored 2 years ago and Pier Angelo Vendrame committed 1 year ago

Treat about:blank as the default home page and new tab page.

Avoid loading AboutNewTab in BrowserGlue.jsm in order
to avoid several network requests that we do not need.

Bug 41624: Disable about:pocket-* pages and about:firefoxview.

Bug 40144: Redirect about:privatebrowsing to the user's home

MOZ_SOURCE_URL is created by combining MOZ_SOURCE_REPO and
MOZ_SOURCE_CHANGESET.
But the code takes for granted that it refers to a Hg instance, so it
combines them as `$MOZ_SOURCE_REPO/rev/$MOZ_SOURCE_CHANGESET`.
With this commit, we change this logic to combine them to create a URL
that is valid for GitLab.
$MOZ_SOURCE_CHANGESET needs to be a commit hash, not a branch or a tag.
If that is needed, we could use /-/tree/, instead of /-/commit/.

Arthur Edelstein authored 8 years ago and Pier Angelo Vendrame committed 1 year ago

Bug 25660: Remove the "New Private Window" option

Change the language selector to be sorted by language code, rather than
name, and to display the language code to the user.

Bug 41372: Handle Japanese as a special case in preferences on macOS

Japanese is treated in a special way on macOS. However, seeing the
Japanese language tag could be confusing for users, and moreover the
language name is not localized correctly like other langs.

Bug 41378: Tell users that they can change their language at the first start

With multi-lingual builds, Tor Browser matches the user's system
language, but some users might want to change it.
So, we tell them that it is possible, but only once.

ma1 authored 2 years ago and Pier Angelo Vendrame committed 1 year ago

Bug 41434: Letterboxing, preemptively apply margins in a global CSS rule to mitigate race conditions on newly created windows and tabs.

This commit adds all the strings needed by following Base Browser
patches.

Matthew Finkel authored 3 years ago and Pier Angelo Vendrame committed 1 year ago

Bugzilla: https://bugzilla.mozilla.org/show_bug.cgi?id=1711084

Avoid regional OS locales if the pref
`intl.regional_prefs.use_os_locales` is false but RFP is enabled.

Igor Oliveira authored 7 years ago and Pier Angelo Vendrame committed 1 year ago

Many fonts have issues with their vertical metrics. they
are used to influence the height of ascenders and depth
of descenders. Gecko uses it to calculate the line height
(font height + ascender + descender), however because of
that idiosyncratic behavior across multiple operating
systems, it can be used to identify the user's OS.

The solution proposed in the patch uses a default factor
to be multiplied with the font size, simulating the concept
of ascender and descender. This way all operating
systems will have the same line height only and only if the
frame is outside the chrome.

Mike Perry authored 10 years ago and Pier Angelo Vendrame committed 1 year ago

It looks like these cases should only be invoked in the NSS command line
tools, and not the browser, but I decided to patch them anyway because there
literally is a maze of network function pointers being passed around, and it's
very hard to tell if some random code might not pass in the proper proxied
versions of the networking code here by accident.

Bugzilla: https://bugzilla.mozilla.org/show_bug.cgi?id=1433509

Alex Catarineu authored 5 years ago and Pier Angelo Vendrame committed 1 year ago

Bugzilla: https://bugzilla.mozilla.org/show_bug.cgi?id=1429838

This commit makes Firefox look for the default profile directory in a
directory relative to the binary path.
The directory can be specified through the --with-relative-data-dir.
This is relative to the same directory as the firefox main binary for
Linux and Windows.

On macOS, we remove Contents/MacOS from it.
Or, in other words, the directory is relative to the application
bundle.

This behavior can be overriden at runtime, by placing a file called
system-install adjacent to the firefox main binary (also on macOS).

Alex Catarineu authored 4 years ago and Pier Angelo Vendrame committed 1 year ago

This checks `privacy.spoof_english` whenever `setLocales` is
called from Fenix side and sets `intl.accept_languages`
accordingly.

Bug 40198: Expose privacy.spoof_english pref in GeckoView

The mechanism to choose the UI font does not play well with our
fontconfig configuration. As a result, the final criterion to choose
the font for the UI was its version.

Since we hardcode Arimo as a default sans-serif on preferences, we use
it also for the UI. FontConfig will fall back to some other font for
scripts Arimo does not cover as expected (we tested with Japanese).

Bug 41732: Use font.system.whitelist also on Linux as a defense-in-depth

Mike Perry authored 11 years ago and Pier Angelo Vendrame committed 1 year ago

This hack directly includes our preference changes in omni.ja.

Bug 18292: Staged updates fail on Windows

Temporarily disable staged updates on Windows.

Bug 18297: Use separate Noto JP,KR,SC,TC fonts

Bug 23404: Add Noto Sans Buginese to the macOS whitelist

Bug 23745: Set dom.indexedDB.enabled = true

Bug 13575: Disable randomised Firefox HTTP cache decay user tests.
(Fernando Fernandez Mancera <ffmancera@riseup.net>)

Bug 17252: Enable session identifiers with FPI

Session tickets and session identifiers were isolated
by OriginAttributes, so we can re-enable them by
allowing the default value (true) of
"security.ssl.disable_session_identifiers".

The pref "security.enable_tls_session_tickets" is obsolete
(removed in https://bugzilla.mozilla.org/917049)

Bug 14952: Enable http/2 and AltSvc

In Firefox, SPDY/HTTP2 now uses Origin Attributes for
isolation of connections, push streams, origin frames, etc.
That means we get first-party isolation provided
"privacy.firstparty.isolate" is true. So in this patch, we
stop overriding "network.http.spdy.enabled" and
"network.http.spdy.enabled.http2".

Alternate Services also use Origin Attributes for isolation.
So we stop overriding
"network.http.altsvc.enabled" and "network.http.altsvc.oe"
as well.

(All 4 of the abovementioned "network.http.*" prefs adopt
Firefox 60ESR's default value of true.)

However, we want to disable HTTP/2 push for now, so we
set "network.http.spdy.allow-push" to false.

"network.http.spdy.enabled.http2draft" was removed in Bug 1132357.
"network.http.sped.enabled.v2" was removed in Bug 912550.
"network.http.sped.enabled.v3" was removed in Bug 1097944.
"network.http.sped.enabled.v3-1" was removed in Bug 1248197.

Bug 26114: addons.mozilla.org is not special
* Don't expose navigator.mozAddonManager on any site
* Don't block NoScript from modifying addons.mozilla.org or other sites

Enable ReaderView mode again (#27281).

Bug 29916: Make sure enterprise policies are disabled

Bug 2874: Block Components.interfaces from content

Bug 26146: Spoof HTTP User-Agent header for desktop platforms

In Tor Browser 8.0, the OS was revealed in both the HTTP User-Agent
header and to JavaScript code via navigator.userAgent. To avoid
leaking the OS inside each HTTP request (which many web servers
log), always use the Windows 7 OS value in the desktop User-Agent
header. We continue to allow access to the actual OS via JavaScript,
since doing so improves compatibility with web applications such
as GitHub and Google Docs.

Bug 12885: Windows Jump Lists fail for Tor Browser

Jumplist entries are stored in a binary file in:
  %APPDATA%\\Microsoft\Windows\Recent\CustomDestinations\
and has a name in the form
  [a-f0-9]+.customDestinations-ms

The hex at the front is unique per app, and is ultimately derived from
something called the 'App User Model ID' (AUMID) via some unknown
hashing method. The AUMID is provided as a key when programmatically
creating, updating, and deleting a jumplist. The default behaviour in
firefox is for the installer to define an AUMID for an app, and save it
in the registry so that the jumplist data can be removed by the
uninstaller.

However, the Tor Browser does not set this (or any other) regkey during
installation, so this codepath fails and the app's AUMID is left
undefined. As a result the app's AUMID ends up being defined by
windows, but unknowable by Tor Browser. This unknown AUMID is used to
create and modify the jumplist, but the delete API requires that we
provide the app's AUMID explicitly. Since we don't know what the AUMID
is (since the expected regkey where it is normally stored does not
exist) jumplist deletion will fail and we will leave behind a mostly
empty customDestinations-ms file. The name of the file is derived from
the binary path, so an enterprising person could reverse engineer how
that hex name is calculated, and generate the name for Tor Browser's
default Desktop installation path to determine whether a person had
used Tor Browser in the past.

The 'taskbar.grouping.useprofile' option that is enabled by this patch
works around this AUMID problem by having firefox.exe create it's own
AUMID based on the profile path (rather than looking for a regkey). This
way, if a user goes in and enables and disables jumplist entries, the
backing store is properly deleted.

Unfortunately, all windows users currently have this file lurking in
the above mentioned directory and this patch will not remove it since it
was created with an unknown AUMID. However, another patch could be
written which goes to that directory and deletes any item containing the
'Tor Browser' string.  See bug 28996.

Bug 30845: Make sure default themes and other internal extensions are enabled

Bug 28896: Enable extensions in private browsing by default

Bug 31065: Explicitly allow proxying localhost

Bug 31598: Enable letterboxing

Disable Presentation API everywhere

Bug 21549 - Use Firefox's WASM default pref. It is disabled at safer
security levels.

Bug 32321: Disable Mozilla's MitM pings

Bug 19890: Disable installation of system addons

By setting the URL to "" we make sure that already installed system
addons get deleted as well.

Bug 22548: Firefox downgrades VP9 videos to VP8.

On systems where H.264 is not available or no HWA, VP9 is preferred. But in Tor
Browser 7.0 all youtube videos are degraded to VP8.

This behaviour can be turned off by setting media.benchmark.vp9.threshold to 0.
All clients will get better experience and lower traffic, beause TBB doesn't
use "Use hardware acceleration when available".

Bug 25741 - TBA: Add mobile-override of 000-tor-browser prefs

Bug 16441: Suppress "Reset Tor Browser" prompt.

Bug 29120: Use the in-memory media cache and increase its maximum size.

Bug 33697: use old search config based on list.json

Bug 33855: Ensure that site-specific browser mode is disabled.

Bug 30682: Disable Intermediate CA Preloading.

Bug 40061: Omit the Windows default browser agent from the build

Bug 40322: Consider disabling network.connectivity-service.enabled

Bug 40408: Disallow SVG Context Paint in all web content

Bug 40308: Disable network partitioning until we evaluate dFPI

Bug 40322: Consider disabling network.connectivity-service.enabled

Bug 40383: Disable dom.enable_event_timing

Bug 40423: Disable http/3

Bug 40177: Update prefs for Fx91esr

Bug 40700: Disable addons and features recommendations

Bug 40682: Disable network.proxy.allow_bypass

Bug 40736: Disable third-party cookies in PBM

Bug 19850: Enabled HTTPS-Only by default

Bug 40912: Hide the screenshot menu

Bug 41292: Disable moreFromMozilla in preferences page

Bug 40057: Ensure the CSS4 system colors are not a fingerprinting vector

Bug 24686: Set network.http.tailing.enabled to true

Bug 40183: Disable TLS ciphersuites using SHA-1

Bug 40783: Review 000-tor-browser.js and 001-base-profile.js for 102

We reviewed all the preferences we set for 102, and remove a few old
ones. See the description of that issue to see all the preferences we
believed were still valid for 102, and some brief description for the
reasons to keep them.

Firefox computes an internal network ID used to detect network changes
and act consequently (e.g., to improve WebSocket UX).
However, there are a few ways to get this internal network ID, so we
patch them out, to be sure any new code will not be able to use them and
possibly link users.

We also sent a patch to Mozilla to seed the internal network ID, to
prevent any accidental leak in the future.
Upstream: https://bugzilla.mozilla.org/show_bug.cgi?id=1817756

The URL query stripping service is enabled only in nightly builds,
still it is initialized and remote settings are downloaded.
This adds a condition that prevents the service from being initialized
if disabled.

Upstream Bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1812594
We should remove this patch if Mozilla decides that this is a problem
also for them, or if they do not but we decide to use the feature.

Georg Koppen authored 5 years ago and Pier Angelo Vendrame committed 1 year ago

Related Bugzilla: https://bugzilla.mozilla.org/show_bug.cgi?id=1428034

Do not include Normandy at all whenever MOZ_NORMANDY is False.

Alex Catarineu authored 5 years ago and Pier Angelo Vendrame committed 1 year ago

More concretely, SearchService.jsm 'hijack-blocklists' and
url-classifier-skip-urls.

Avoid creating instance for 'anti-tracking-url-decoration'.

If prefs are disabling their usage, avoid creating instances for
'cert-revocations' and 'intermediates'.

Do not ship JSON dumps for collections we do not expect to need. For
the ones in the 'main' bucket, this prevents them from being synced
unnecessarily (the code in remote-settings does so for collections
in the main bucket for which a dump or local data exists). For the
collections in the other buckets, we just save some size by not
shipping their dumps.

We also clear the collections database on the v2 -> v3 migration.

Arthur Edelstein authored 6 years ago and Pier Angelo Vendrame committed 1 year ago

Connections were observed in the catch-all circuit when
the user entered an https or http URL in the URL bar, or
typed a search term.