livekit-libwebrtc: 125-unstable-2025-07-25 -> 137-unstable-2026-03-12, enable x11 and pipewire, zed-editor: 0.226.5 -> 0.228.0

[jrobsonchase]

Update livekit-libwebrtc to version 137. Also enable x11 and pipewire.

This is needed for use of libwebrtc's DesktopCapturer in Zed (zed-industries/zed#42670).

Things done

• Built on platform:
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• Tested, as applicable:
  • NixOS tests in nixos/tests.
  • Package tests at passthru.tests.
  • Tests in lib/tests or pkgs/test for functions and "core" functionality.
• Ran nixpkgs-review on this PR. See nixpkgs-review usage.
• Tested basic functionality of all binary files, usually in ./result/bin/.
• Nixpkgs Release Notes
  • Package update: when the change is major or breaking.
• NixOS Release Notes
  • Module addition: when adding a new NixOS module.
  • Module update: when the change is significant.
• Fits CONTRIBUTING.md, pkgs/README.md, maintainers/README.md and other READMEs.

---

Add a 👍 reaction to pull requests you find important.

[jrobsonchase]

Buncha new dependencies needed for the rtc_use_pipewire=true and rtc_use_x11=true change.

Are those options always something that we want? Or should they be configurable, producing multiple flavors of libwebrtc?

[niklaskorz]

Is this required for darwin builds too?

[jrobsonchase]

Oops, that should be mac, not linux64. But otherwise, yes - these directories are no longer extant by default.

Fixed.

[niklaskorz]

nixpkgs-review result

Generated using nixpkgs-review.

Command: nixpkgs-review pr 478907
Commit: 28a9fef3416b3dde8876ff24d3f0db0822577bd9

---

aarch64-darwin

❌ 4 packages failed to build:

• livekit-libwebrtc
• livekit-libwebrtc.dev
• zed-editor
• zed-editor.remote_server

---

Error logs: `aarch64-darwin`

livekit-libwebrtc

patching file third_party/dav1d/BUILD.gn
applying patch /nix/store/lnfqqcqmgqipa4iblpv93d0gw3qf4zws-0001-shared-libraries.patch
patching file BUILD.gn
patching script interpreter paths in build/toolchain/apple/linker_driver.py
build/toolchain/apple/linker_driver.py: interpreter directive changed from "#!/usr/bin/env python3" to "/nix/store/2z6flh9lzw7bq5z7vs8bl49qavi7p80q-python3-3.13.11-env/bin/python3"
Running phase: updateAutotoolsGnuConfigScriptsPhase
@nix { "action": "setPhase", "phase": "updateAutotoolsGnuConfigScriptsPhase" }
Updating Autotools / GNU config script to a newer upstream version: ./third_party/nasm/autoconf/helpers/config.sub
Updating Autotools / GNU config script to a newer upstream version: ./third_party/breakpad/breakpad/autotools/config.sub
Updating Autotools / GNU config script to a newer upstream version: ./third_party/nasm/autoconf/helpers/config.guess
Updating Autotools / GNU config script to a newer upstream version: ./third_party/breakpad/breakpad/autotools/config.guess
Running phase: configurePhase
@nix { "action": "setPhase", "phase": "configurePhase" }
gn flags: is_debug=false rtc_include_tests=false target_os=\"mac\" target_cpu=\"arm64\" treat_warnings_as_errors=false rtc_enable_protobuf=false rtc_include_tests=false rtc_build_examples=false rtc_build_tools=false rtc_libvpx_build_vp9=true enable_libaom=true use_dummy_lastchange=true is_component_build=true enable_stripping=true rtc_use_h264=true rtc_use_h265=true use_custom_libcxx=false use_custom_libcxx_for_host=false use_llvm_libatomic=false use_libcxx_modules=false use_rtti=true mac_deployment_target=\"14.0\" rtc_enable_symbol_export=true rtc_enable_objc_symbol_export=true rtc_include_dav1d_in_internal_decoder_factory=true clang_use_chrome_plugins=false use_lld=false clang_base_path=\"/nix/store/8yy15sk3826gb00zp44xxrm8by9i3ia0-clang-wrapper-21.1.7\"
Generating compile_commands took 45ms
Done. Made 2251 targets from 400 files in 3918ms
Running phase: buildPhase
@nix { "action": "setPhase", "phase": "buildPhase" }
build flags: -j6 :default api/audio_codecs:builtin_audio_decoder_factory api/task_queue:default_task_queue_factory sdk:native_api sdk:default_codec_factory_objc pc:peer_connection sdk:videocapture_objc sdk:mac_framework_objc
ninja: error: '../../../../../../../store/8yy15sk3826gb00zp44xxrm8by9i3ia0-clang-wrapper-21.1.7/lib/clang/21/lib/darwin/libclang_rt.osx.a', needed by 'obj/third_party/protobuf/libprotoc_lib.a', missing and no known rule to make it

[jrobsonchase]

Oh, hm. I think this will break the zed build until the PR that requires it gets merged... bit of a chicken-and-egg problem.

I think we'll have to get the PR merged on the zed side with an override, and then do the zed and libwebrtc version bumps simultaneously.

[jrobsonchase]

Temporarily targeting zed-editor at the branch that's primed to use this new libwebrtc version.

Setting this PR back in draft until that merges, but the work needed to update zed after it does is already done and verified (at least by me on linux).

[kubkon]

I have recently fast-forwarded Zed's dependency on livekit/rust-sdks and have included your changes from this PR in Zed proper (the commit: zed-industries/zed@641ef1f). FWIW I have made you a co-author @jrobsonchase. Anyhow, what we, NixOS users as Zed, have recently noticed is that linking libwebrtc as an SO with -fvisibility=default and rtc_use_pipewire = true causes the resultant libwebrtc.so to export pipewire symbols that you would normally expect in libpipewire.so:

$ ~/d/zed @0d215f250e• ❱ nm -D /nix/store/3wnqb79nwzh5gk7wkvx9l1sns8i9c5g5-livekit-libwebrtc-137-unstable-2025-11-24-dev/lib/libwebrtc.so 2>/dev/null | grep -i "pw_\|pipewire" | head -20
00000000007cec20 W pw_context_connect
00000000007cec30 W pw_context_connect_fd
00000000007cec00 W pw_context_destroy
00000000007cec10 W pw_context_new
00000000007cea50 W pw_core_disconnect
00000000007cea90 W pw_get_library_version
00000000007cea80 W pw_init
00000000007cea60 W pw_loop_destroy
00000000007cea70 W pw_loop_new
00000000007ceaa0 W pw_properties_new_string
00000000007cec40 W pw_proxy_destroy
00000000007ceab0 W pw_stream_add_listener
00000000007ceac0 W pw_stream_connect
00000000007ceae0 W pw_stream_dequeue_buffer
00000000007ceaf0 W pw_stream_destroy
00000000007cead0 W pw_stream_disconnect
00000000007ceb40 W pw_stream_get_node_id
00000000007ceb50 W pw_stream_get_state
00000000007ceb00 W pw_stream_new
00000000007ceb10 W pw_stream_queue_buffer

This in turns causes a null pointer dereference when rodio/cpal is trying to initialize a default audio device (input or output) since the loader seems to use libwebrtc.so exported pipewire stubs as pipewire symbols when dlopening pipewire ALSA plugin. Given that these are not yet resolved in libwebrtc.so - correct me if I'm wrong here but it seems that webrtc patches as the pointers in said stubs when we first request pipewire usage in libwebrtc - we deref address of 0x0 and crash.

My best take on this so far is to surgically hide the problematic symbols via a version script, but perhaps there's a better way? Looking forward to your thoughts on this!

[jrobsonchase]

If you look at the next line in that patch, it's also adding additional linker args for some other shared libraries that it depends on. I wonder if those libraries were in a similar situation, and adding pipewire there might be a better approach?

Mostly a shot in the dark here - I wasn't involved with the original build of this in nixpkgs. @niklaskorz or @WeetHet would probably have a better idea.

For now though, I'll go ahead and pull in your version script!

[jrobsonchase]

I'll fix the commit history here in a bit. The bigger problem is that when I merge master, it now fails to build 🫤

I'll push that up now that everything else seems to evaluate/build in its current state.

Seems like it's failing to find pipewire symbols now?

[jrobsonchase]

desktop-app/tg_owt#167 Seems to be the same issue. Trying a build with the linked patch now.

[niklaskorz]

Works fine on darwin, but a header included by zed-editor fails to find gio from glib eventhough it's in the zed-editor buildInputs:

zed-editor>   cargo:warning=In file included from /nix/store/wrx5m9x5z1bndfg85w6fiv1w6fr4zhjz-livekit-libwebrtc-137-unstable-2026-03-12-dev/include/modules/desktop_capture/desktop_capture_metadata.h:15,
zed-editor>   cargo:warning=                 from /nix/store/wrx5m9x5z1bndfg85w6fiv1w6fr4zhjz-livekit-libwebrtc-137-unstable-2026-03-12-dev/include/modules/desktop_capture/desktop_capturer.h:26,
zed-editor>   cargo:warning=                 from ./include/livekit/desktop_capturer.h:20,
zed-editor>   cargo:warning=                 from src/desktop_capturer.cpp:17:
zed-editor>   cargo:warning=/nix/store/wrx5m9x5z1bndfg85w6fiv1w6fr4zhjz-livekit-libwebrtc-137-unstable-2026-03-12-dev/include/modules/portal/xdg_session_details.h:14:10: fatal error: gio/gio.h: No such file or directory
zed-editor>   cargo:warning=   14 | #include <gio/gio.h>
zed-editor>   cargo:warning=      |          ^~~~~~~~~~~
zed-editor>   cargo:warning=compilation terminated.

Edit: that's an issue in the webrtc-sys crate which tries to be "smart" by using the system's absolute glib path instead of pkg-config...

https://github.com/livekit/rust-sdks/blob/e96d7c2adbff9ad7f9ed6519b776c9557ee65a24/webrtc-sys/build.rs#L428-L452

Edit 2: fixed by patching in the correct header paths

[niklaskorz]

nixpkgs-review result

Generated using nixpkgs-review.

Command: nixpkgs-review pr 478907
Commit: 048cd370e88a865e9907bf51153d2f2ef854d866

---

x86_64-linux

✅ 5 packages built:

• livekit-libwebrtc
• livekit-libwebrtc.dev
• zed-editor
• zed-editor-fhs
• zed-editor.remote_server

[niklaskorz]

nixpkgs-review result

Generated using nixpkgs-review.

Command: nixpkgs-review pr 478907
Commit: 048cd370e88a865e9907bf51153d2f2ef854d866

---

aarch64-darwin

✅ 4 packages built:

• livekit-libwebrtc
• livekit-libwebrtc.dev
• zed-editor
• zed-editor.remote_server