Merge textchat into shell #918

Closed
charles wants to merge 0 commits from shell-sync-textchat into shell
Owner

Merges textchat into shell: the Windows hardware encoder off again (#888), DeepFilterNet late-frame and warm-up fixes (#912, #914), late video frames after a stream ends (#909), screen-share labels (#879), keeping the display awake (#913/#915), the packaging check and today's CI fixes.

Conflicts and how they were joined

  • apps/fluxer/src/runtime.rs: the builder keeps both sides' fields and methods: the shell's shell, instance, window_host and shortcuts, plus textchat's keep_awake. They all default to inert or nothing in Runtime::new, and at shutdown both the window-host task and the keep-awake task are aborted. The keep-awake follower and the shell/shortcuts runtimes start side by side from the same snapshots.
  • crates/fluxer-platform/src/lib.rs and the fluxer-core and fluxer-platform READMEs: kept both sides' text. The platform README had three duplicated dependency paragraphs from earlier merges; they are now one.
  • CI: takes textchat's Windows MIN_FREE_GB: 45 and the Launchpad retry loop (ci.yaml and release.yaml). shell is still left out of ci.yaml's push branches and the release-start condition, so pushes to shell do not build nightlies.

Fixes beyond the conflicts (the audit of "tests and --devserver never touch the desktop")

  • Link opening: Desktop::new used to fall back to the system's opener. Now it opens nothing, and main passes desktop_rt::SystemLinkOpener. Test: the_default_link_opener_opens_nothing.
  • --devserver used to get ShellPlatform::system() and ShortcutsPlatform::system(), so it made a tray icon and could reconcile the real autostart entry or grab shortcuts. main now uses ShellPlatform::for_config and ShortcutsPlatform::for_config, which give inert() with --devserver. Tests: a_devserver_run_leaves_the_desktop_alone and a_devserver_run_grabs_no_shortcut.

Keep-awake and the shell window (#916): the two compile and run side by side. app.shell.window.visible is now available in the core (crates/fluxer-core/src/shell/window.rs), so #916 can use it. The "a hidden window releases the lock" rule was not added here; it stays follow-up #916. The docs now say that the field exists but is not read yet.

Windows: the Windows CI job compiles these for the first time on this branch: crates/fluxer-platform/src/keep_awake/{mod,win32}.rs (with the shell's windows features: Power and Threading come from the workspace dependency), apps/fluxer/src/main.rs (joined; its cfg(windows) AUMID and console code are unchanged), and apps/fluxer/src/runtime.rs (the Windows opener, unchanged and now reached only through SystemLinkOpener).

Gate: FLUXER_SCREENSHOT_TESTS=1 dbus-run-session -- cargo xtask ci ends with ==> xtask ci: all steps passed (livekit-interop skipped: FLUXER_LIVEKIT_TESTS not set).

🤖 Generated with Claude Code

Merges `textchat` into `shell`: the Windows hardware encoder off again (#888), DeepFilterNet late-frame and warm-up fixes (#912, #914), late video frames after a stream ends (#909), screen-share labels (#879), keeping the display awake (#913/#915), the packaging check and today's CI fixes. **Conflicts and how they were joined** - `apps/fluxer/src/runtime.rs`: the builder keeps both sides' fields and methods: the shell's `shell`, `instance`, `window_host` and `shortcuts`, plus textchat's `keep_awake`. They all default to inert or nothing in `Runtime::new`, and at shutdown both the window-host task and the keep-awake task are aborted. The keep-awake follower and the shell/shortcuts runtimes start side by side from the same snapshots. - `crates/fluxer-platform/src/lib.rs` and the `fluxer-core` and `fluxer-platform` READMEs: kept both sides' text. The platform README had three duplicated dependency paragraphs from earlier merges; they are now one. - CI: takes textchat's Windows `MIN_FREE_GB: 45` and the Launchpad retry loop (ci.yaml and release.yaml). `shell` is still left out of ci.yaml's push branches and the release-start condition, so pushes to `shell` do not build nightlies. **Fixes beyond the conflicts** (the audit of "tests and `--devserver` never touch the desktop") - Link opening: `Desktop::new` used to fall back to the system's opener. Now it opens nothing, and `main` passes `desktop_rt::SystemLinkOpener`. Test: `the_default_link_opener_opens_nothing`. - `--devserver` used to get `ShellPlatform::system()` and `ShortcutsPlatform::system()`, so it made a tray icon and could reconcile the real autostart entry or grab shortcuts. `main` now uses `ShellPlatform::for_config` and `ShortcutsPlatform::for_config`, which give `inert()` with `--devserver`. Tests: `a_devserver_run_leaves_the_desktop_alone` and `a_devserver_run_grabs_no_shortcut`. **Keep-awake and the shell window (#916)**: the two compile and run side by side. `app.shell.window.visible` is now available in the core (`crates/fluxer-core/src/shell/window.rs`), so #916 can use it. The "a hidden window releases the lock" rule was not added here; it stays follow-up #916. The docs now say that the field exists but is not read yet. **Windows**: the Windows CI job compiles these for the first time on this branch: `crates/fluxer-platform/src/keep_awake/{mod,win32}.rs` (with the shell's `windows` features: Power and Threading come from the workspace dependency), `apps/fluxer/src/main.rs` (joined; its `cfg(windows)` AUMID and console code are unchanged), and `apps/fluxer/src/runtime.rs` (the Windows `opener`, unchanged and now reached only through `SystemLinkOpener`). Gate: `FLUXER_SCREENSHOT_TESTS=1 dbus-run-session -- cargo xtask ci` ends with `==> xtask ci: all steps passed` (livekit-interop skipped: FLUXER_LIVEKIT_TESTS not set). 🤖 Generated with [Claude Code](https://claude.com/claude-code)
A 3440x1440 monitor shared at 1080p, 1440p and Source is sent as
1920x802, 2560x1070 and 3440x1440 (fit into the setting's 16:9 box,
aspect kept, no padding) and arrives as 1920x800, 2560x1068 and
3440x1440. The viewer's pill named those by the closest usual height,
"720p", "1080p" and "1440p", one step below each setting, which read
as a downgrade that did not happen.

- Viewer's pill: a 16:9 stream keeps the closest-height name (as the
  original); any other shape is named by its size.
- Sharer's own pill: what is sent (the share's EncodeParams), not the
  setting's preset, as the original's formatScreenShareTargetLabel; a
  share held to 720p30 for want of a hardware H.264 encoder now says so.
- Logs: INFO "voice: screen share size" at publish; WARN when the
  software H.264 clamp lowers the setting.
- sender_consistency SC_SETTING / run.sh --chain --setting: the capture
  is asked for the preset's box and published at fit_to_preset of what
  it delivers, for measuring sizes end to end.

Refs #876
Refs #589

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Windows: turn the hardware encoder off again (#888)
All checks were successful
ci / Windows (x64) (push) Successful in 23m0s
ci / Linux (x86_64) (push) Successful in 1h4m47s
ci / Start the release build (push) Successful in 32s
d01fcb1f11
Nightly 2026.1118.377 was the first build with fluxer-voice/mf-encoder on (#873). On real
Windows machines the client started slowly, hung when a screen share started, and the
stream was blocky. The feature is no longer turned on by the client's manifest or by
xtask package; the code stays, behind the feature, for #229.

Refs #888, #229

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0161qMQMsCyJhjYw4gqa6e4K
The release of d01fcb1 (run 382) failed on the Windows runner: the packaged client's
--check-noise-suppression reported 'this computer takes 250 ms to clean each 10 ms of
sound, too slow to keep up' — one frame held up past the client's limit on a loaded VM.
That says something about the machine, not the package. The check is now tried three
times, and a machine that is still too slow gives a warning; every other failure still
fails the package.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0161qMQMsCyJhjYw4gqa6e4K
Merge pull request 'Screen share: labels show the real size of non-16:9 streams and what is really sent (#876)' (#879) from fix-876-share-resolution-and-bars into textchat
All checks were successful
ci / Windows (x64) (push) Successful in 22m24s
ci / Linux (x86_64) (push) Successful in 1h4m39s
ci / Start the release build (push) Successful in 33s
6a5c70b7b9
Remote video frames and room events reach the runtime on two channels, read
by two tasks. A frame the session passed on just before its track was
unpublished could be taken in after `TrackUnpublished` had removed the track
from the `FrameBus`; `VideoRouting::frame` then put it back, keyed by the
gone sid, and nothing took it off again: the store never unsubscribes a
track that is no longer published, so the participant stayed "wanted". The
stale frame stayed on the bus until the streamer left or the room was left,
and showed under a "Stream ended" overlay of the same sid.

`VideoRouting` now remembers the sids unpublished in this room and refuses
their frames (a new `TrackPublished` of the sid lifts it; leaving the room
forgets them), and a participant that leaves is no longer wanted, so its
late frames stay off too.

The store side was right: the watch ended after REPUBLISH_GRACE every time.
`watched_stream_ending_stops_watching_by_itself` now checks the bus at the
moment the store has dropped the watch instead of polling it, and also
checks that the first, republished track left the bus.

Refs #908

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Merge pull request 'A late frame of an unpublished track stays off the frame bus (#908)' (#909) from fix-908-flaky-stream-end-test into textchat
Some checks failed
ci / Windows (x64) (push) Failing after 24m16s
ci / Start the release build (push) Has been cancelled
ci / Linux (x86_64) (push) Has been cancelled
3822f03de2
CI: Windows drops the build cache below 45 GB free (was 18)
Some checks failed
ci / Linux (x86_64) (push) Failing after 5m46s
ci / Windows (x64) (push) Successful in 35m40s
ci / Start the release build (push) Has been skipped
6467aa88e7
Run 385 started warm with 27.8 GB free on R:, the debug target dir grew to 120 GB during
the run and link.exe failed with the disk full (LNK1140, LNK1180). The Windows job has no
size cap on the target dir, only the free-space floor, which was too low for one warm run's
growth.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0161qMQMsCyJhjYw4gqa6e4K
The loader warmed DeepFilterNet up with silent frames, but libDF returns before
the networks for silence, so the first frame of sound paid for tract's first
runs (per-session, per-thread scratch and packing, first touch of the kernels).
The warm-up now runs a noisy tone through all three networks three times on
the thread and instance that clean the call, then restores the stream state
(DfTract::stream/set_stream, added to the vendored libDF).

The chain waits at most 8 ms for a frame. A frame not back by then goes out as
the unprocessed input delayed by DeepFilterNet's 3 frames, so the sound stays in
order with no added latency; the late result is dropped and DeepFilterNet
carries on. It is given up after 50 late frames in a row or an average above
8 ms over 200 frames. Scripted-worker tests cover a late first frame, a late
frame mid-stream, sustained slowness, a slow average and a worker that never
answers. The packaged check runs once (no more three attempts) and measures
the noise only on processed frames.

Refs #910

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Merge pull request 'DeepFilterNet: a late frame goes out unprocessed instead of ending it; real warm-up (#910)' (#912) from fix-910-deepfilter-late-frame into textchat
Some checks failed
ci / Linux (x86_64) (push) Failing after 5m30s
ci / Windows (x64) (push) Failing after 12m7s
ci / Start the release build (push) Has been skipped
66034edd71
CI: retry the Launchpad PPA install for ten minutes
Some checks failed
ci / Windows (x64) (push) Failing after 12m8s
ci / Start the release build (push) Has been cancelled
ci / Linux (x86_64) (push) Has been cancelled
86f2a14fbb
Runs 386 and 388 failed within minutes on '503 Service Unavailable' from
ppa.launchpadcontent.net while installing libpipewire-0.3-dev; nothing was built.
The install in ci.yaml and release.yaml now retries once a minute, ten times.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0161qMQMsCyJhjYw4gqa6e4K
The original desktop app holds Electron's prevent-display-sleep blocker
while "streaming priority" is acquired, which only the own screen share
does: from setEnabled(true) (the capture and its system picker) until the
share stops, fails, is cancelled or the room disconnects; kept across a
source switch. Watching a stream, a camera or a call never acquire it, and
the web client does not use the Screen Wake Lock API. Flaxative had no
inhibitor at all.

- fluxer-core: selectors::keep_display_awake, true while the own share
  is Starting, Live or switching its source.
- fluxer-platform: keep_awake::{KeepAwake, Inhibition}; on Linux the
  portal's Inhibit (idle), then org.freedesktop.ScreenSaver, then
  org.gnome.SessionManager, the inhibition owning its bus connection; on
  Windows a power request (display and system required); a fake.
- apps/fluxer: keep_awake_rt follows the snapshots and holds at most one
  inhibition on a thread of its own, released when the selector turns
  false, at shutdown and when the runtime task ends. main passes the
  system's; the runtime keeps nothing awake by default, so tests never
  reach the desktop.

Refs #913

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The chain waited for a frame with recv_timeout(8 ms). A timed blocking wait
ends on the system timer tick, 15.6 ms by default on Windows (nothing in the
app raises the resolution), so the engine thread could be held twice the
budget and frames counted as in time that were not; Windows CI run 388 failed
the sustained-slowness test that way. The wait now polls the result channel
against Instant (spin, then yield) and declares a frame late exactly when
Instant passes the deadline, on every system.

The scripted-worker tests no longer time anything: a scripted wait behind a
Wait seam decides which frames are late and only then releases them from the
worker, so every assertion is exact (paths, counts, reasons). Poll itself is
tested on what it guarantees (a frame that is there is taken; late only from
the deadline; the thread ending is seen).

Refs #910

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Merge pull request 'DeepFilterNet: lateness judged by the clock, wait without a timer; deterministic tests (#910)' (#914) from fix-910-deepfilter-windows-timing into textchat
All checks were successful
ci / Linux (x86_64) (push) Successful in 1h8m49s
ci / Windows (x64) (push) Successful in 1h19m8s
ci / Start the release build (push) Successful in 33s
1e2dbf6dda
In the original, Chromium holds its own display wake lock while a visible
video plays, so users see the display stay awake while watching a stream
or with a camera tile in the call, even though the fork never asks for it.
The selector now holds while any of: the own share runs (as before), a
watched stream's video arrives or is expected (in `watching`, not Failed;
an Ended tile counts only while the watch lasts), or a camera is shown
(the own `self_video`; remote cameras are not received on this branch).
A voice-only call keeps nothing awake. The reason string no longer names
only sharing.

No window-visibility condition yet: the core has none on this branch; a
hidden window should release it once the desktop shell lands.

Refs #913

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Refs #913

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Merge pull request 'Keep the display awake while sharing, watching or showing a camera (#913)' (#915) from fix-913-keep-screen-awake into textchat
Some checks failed
ci / Linux (x86_64) (push) Failing after 23m31s
ci / Windows (x64) (push) Successful in 23m36s
ci / Start the release build (push) Has been skipped
66ddd6280f
Brings in textchat since the last sync: the Windows hardware encoder off
again (#888), DeepFilterNet late-frame and warm-up fixes (#912, #914), late
video frames after a stream ends (#909), screen-share labels (#879), keep
the display awake (#913/#915), the packaging check and CI fixes.

Conflicts:
- apps/fluxer/src/runtime.rs: both sides added builder fields and
  methods; the runtime keeps the shell's ports (shell, instance,
  window_host, shortcuts) and textchat's keep_awake, all defaulting to
  inert / nothing in Runtime::new; at shutdown both the window-host task
  and the keep-awake task are aborted.
- crates/fluxer-platform/src/lib.rs, READMEs: both sides' text joined
  (the platform README's duplicated dependency paragraphs collapsed into
  one).
- CI: textchat's Windows MIN_FREE_GB 45 and Launchpad retry loop, while
  `shell` stays out of ci.yaml's push branches and the release start.

Beyond conflicts, so a test or a --devserver run never touches the
session's desktop:
- links: Desktop::new no longer falls back to the system's opener; main
  passes desktop_rt::SystemLinkOpener (test
  the_default_link_opener_opens_nothing).
- --devserver: main takes ShellPlatform/ShortcutsPlatform::for_config,
  inert with --devserver (no tray icon, autostart or desktop entry, no
  shortcut grab; tests a_devserver_run_leaves_the_desktop_alone,
  a_devserver_run_grabs_no_shortcut).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Author
Owner

Merged into shell by hand (the merge API returns 405); the commits are in the branch.

Merged into `shell` by hand (the merge API returns 405); the commits are in the branch.
charles closed this pull request 2026-11-18 22:36:37 +00:00

Pull request closed

Sign in to join this conversation.
No description provided.