By default GN produces a build with all of the debug assertions enabled (is_debug=true) and including full debug info (symbol_level=2). Setting symbol_level=1 will produce enough information for stack traces, but not line-by-line debugging. Setting symbol_level=0 will include no debug symbols at all. Either will speed up the build compared to full symbols.
This is done to avoid driver specific issues and restores the previous
behaviour. Like before video acceleration can be enabled without having
to rebuild Chromium.
This will additionally install the following files:
libEGL.so libGLESv2.so
libVkICD_mock_icd.so libvk_swiftshader.so libvulkan.so
libEGL.so and libGLESv2.so are required to fix our ANGLE support.
The rest should help with the Vulkan support (currently an experimental
feature that is disabled by default).
This is required for certain URIs that require launching external
programs (e.g. mailto:, magnet:, or irc:) or setting the default browser
via xdg-settings.
Fix#96897 and fix#92751.
This reverts commit 5da66561d1.
It seems the chromium build now unconditionally tries to enable ozone
(even though we disable it), causing the build to fail (as we only
provide xkbcommon when enabling Ozone):
```
configuring
ERROR at //build/config/linux/pkg_config.gni:103:17: Script returned non-zero exit code.
pkgresult = exec_script(pkg_config_script, args, "value")
^----------
Current dir: /build/chromium-87.0.4252.0/out/Release/
Command: python /build/chromium-87.0.4252.0/build/config/linux/pkg-config.py xkbcommon
Returned 1.
stderr:
Package xkbcommon was not found in the pkg-config search path.
Perhaps you should add the directory containing `xkbcommon.pc'
to the PKG_CONFIG_PATH environment variable
No package 'xkbcommon' found
Could not run pkg-config.
See //ui/events/ozone/layout/BUILD.gn:12:3: whence it was called.
pkg_config("xkbcommon") {
^------------------------
See //chrome/test/chromedriver/BUILD.gn:273:15: which caused the file to be included.
deps += [ "//ui/events/ozone/layout" ]
^-------------------------
builder for '/nix/store/2dqhrd2qzyms078wnvwv6ays53ppvgc2-chromium-unwrapped-87.0.4252.0.drv' failed with exit code 1
cannot build derivation '/nix/store/4iyhgzsmpx80v75hvk1jycwzanw4z5dn-chromium-dev-87.0.4252.0.drv': 1 dependencies couldn't be built
```
update.nix was a huuuuge hack, abusing checksum collisions, etc., and
was extremely difficult to read and maintain, especially because
values from update.nix were also used in the derivations themselves!
I've replaced this with an implementation in Python, which I chose for
readability. Rather than generating Nix, I chose to
generate JSON, since Python can do that in the standard library and
Nix can read it.
I also set update.py as an updateScript, so Chromium can now
automatically be updated!
Fixes: https://github.com/NixOS/nixpkgs/issues/89635
Note: The following might also need to be updated:
substituteStream(): WARNING: pattern '/usr/share/xcb' doesn't match anything in file 'ui/gfx/x/BUILD.gn'
I didn't look into this yet but IIRC M86 will finally have a flag for
Linux to enable VA-API. So we shouldn't need
enable-video-acceleration-on-linux.patch anymore.
But we likely need to update enable-vdpau-support-for-nvidia.patch
when/before M86 hits the stable channel if we want to keep VDPAU
support.
Ok, so I was about to update the patch (didn't apply anymore) when I
also started looking at it's usage and realized that
NIX_CHROMIUM_PLUGIN_PATH_ (and other substrings) only appears in the
patch itself (i.e. it seemed like we don't need this patch anymore).
Turns out that we have this patch since 2014 (1b84fbf0ca) and it was
only ever used for NIX_CHROMIUM_PLUGIN_PATH_WIDEVINE (and from the log
it isn't clear if/when or how well that worked). But in 2019 that last
usage got removed (545d58a1ef) so we should be able to safely drop this
patch now :) \o/
(I just wanted to note that as it seemed somewhat of a funny story :D
But there is of course nothing wrong with it.)
Git history (git log --oneline -S NIX_CHROMIUM_PLUGIN_PATH_):
7205bd64a3 ungoogled-chromium: init at 81.0.4044.92-2
545d58a1ef chromium: fix widevine
cd3283f921 chromium: 67.0.3396.99 -> 68.0.3440.75
72d7b5ddb1 chromium: fix nix_plugin_paths for 68+
7a3a16dd80 chromium: Remove plugin paths patch for version 50
79d18eb604 chromium: Update dev channel to v52.0.2743.10
c7a3645e7b chromium: Remove stuff for versions <= v51
8b97ca270e chromium: Update all channels to latest versions
b9093f1c64 chromium: Updates, fixes#11492471cdd15e2 chromium: Update beta and dev channels.
5c6aa391fc chromium: Cleanup old patch and update stable
af54ddf8b6 chromium: Drop plugin_paths patch for old versions.
6a8afa4bb3 chromium: Fix plugin_paths patch for version 44.
0aad4b7ee4 chromium: Update all channels to latest versions.
1b84fbf0ca chromium: Allow env vars for passing plugin paths.
This is more robust than depending on the channel, though the version
should only matter if the configuration phase fails.
This also switches to the intended version for `chromium` which should
be higher since M85 is in the stable channel.
Thanks `@volth` for pointing this out.
The script that runs chromium calls tr from coreutils - however
it just assumed that coreutils are in PATH.
With missing coreutils chromium did still launch (at least with
d433839007 applied) but emitted
`line 15: tr: command not found` error messages.
Mark chromiumDev as broken since the build requires LLVM 11 which is not
yet in Nixpkgs (due to the lack of an RC, see #93324). Build error:
clang (LLVM option parsing): Unknown command line argument '-basic-aa-recphi=0'. Try: 'clang (LLVM option parsing) --help'
clang (LLVM option parsing): Did you mean '--basicaa-recphi=0'?
ninja: build stopped: subcommand failed.
The bin script that runs chromium calls out to gnugrep - but gnugrep is
missing as a runtime dependency of the chromium package. I found this
out when I was trying to put it in a docker image.
After making `ffmpeg` point to the latest `ffmpeg_4`, all packages that
used `ffmpeg` without requiring a specific version now use ffmpeg_3
explicitly so they shouldn't change.
This provides the browser flag #enable-webrtc-pipewire-capturer, which
adds support for screensharing on Wayland via xdg-desktop-portal.
The browser flag is disabled by default until a user enables it. At
least one other major distribution (Arch) enables this compile time
option, and so I believe it should be safe to enable by default.
This is also needed to support xdg-desktop-portal-wlr which was added in
https://github.com/NixOS/nixpkgs/pull/83485.
The configuration phase was failing due to:
```
configuring
ERROR at //BUILD.gn:1376:5: Unknown function.
filter_exclude([ "$root_build_dir/foo" ],
^-------------
```
This can e.g. save around 150k lines of unnecessary log messages which
take up around 66% of the total lines (based on a log of 80.0.3987.100):
29527 warning: unknown warning option '-Wno-bitwise-conditional-parentheses'; did you mean '-Wno-bitwise-op-parentheses'? [-Wunknown-warning-option]
29527 warning: unknown warning option '-Wno-builtin-assume-aligned-alignment' [-Wunknown-warning-option]
29527 warning: unknown warning option '-Wno-deprecated-copy'; did you mean '-Wno-deprecated'? [-Wunknown-warning-option]
29527 warning: unknown warning option '-Wno-final-dtor-non-final-class'; did you mean '-Wno-abstract-final-class'? [-Wunknown-warning-option]
29527 warning: unknown warning option '-Wno-implicit-int-float-conversion'; did you mean '-Wno-implicit-float-conversion'? [-Wunknown-warning-option]
This makes it possible to use chromium headless with WebGL
(e.g. for webdriver tests) without having to rebuild from source.
The upstram default is to enable, thus simply removing our disabling switch.
Also fixes#41918.
This reverts commit 02ca096a85.
The clang-format binary is now included in the release tarball [0] and
we don't need to add it anymore. This fixes the following build error:
ln: failed to create symbolic link 'buildtools/linux64/clang-format': File exists
However, the build still fails a bit later in the configure phase.
[0]: https://bugs.chromium.org/p/chromium/issues/detail?id=1052503
The webrtc code suffered from a race condition when used
with Pulseaudio. This lead to audio input breaking every
couple of minutes during a webrtc session.
The fix from 1d961a4c6d was unfortunately incomplete. This commit fixes
the second error [0] that occurs later during the build (the Python
script third_party/blink/renderer/bindings/scripts/generate_bindings.py
requires buildtools/linux64/clang-format).
[0]: https://github.com/NixOS/nixpkgs/pull/81016#issuecomment-591572821
I don't really have the hardware resources nor time to do this properly,
but I'll try to keep a watch on Chromium (updates, PRs, and issues)
until we've found a new team [0].
Testing will be performed on a best effort basis (no guarantees :o).
I've also briefly documented the current maintainer
roles/responsibilities and added `meta.longDescription`.
[0]: https://github.com/NixOS/nixpkgs/issues/78450
TredwellGit
Michael Weiss
Gabriel Ebner
taku0
Tethys Svensson
Florian Klink
Alyssa Ross
Matthew Bauer
Samuel Gräfenstein
f4814n
Mario Rodas
Griffin Smith
Domen Kožar
Andrew Childs
Doron Behar
James Ravn
Michael Reilly
Artemis Tosini
Niklas Hambüchen
Peter Simons
taku0
worldofpeace
Ivan Kozik
Graham Christensen