Repository navigation
Releases
Releases · renesenses/oaat
v0.2.8
Highlights
Linux endpoint: no native DSD announced on a DAC that refuses it
- With
--dsd, when the DAC'shw:device refuses the native DSD format (DSD_U32_BE), the endpoint no longer announces native DSD64: it announces no native DSD at all, and the Tune server converts DSD to PCM. Seen on the RME ADI-2 DAC, which only offers S32_LE to Linux (#62, #63). - The same applies when the probe cannot run at startup (device busy or absent) and when the output is not a
hw:device. The log says so:native DSD not supported by device, DSD will be converted to PCM by the server.
Linux endpoint: a device that refuses a stream no longer blocks playback
- The audio device is now opened before the endpoint answers the format proposal. A device that refuses the stream, or does not answer within 3 s, is reported to the controller with a
format_rejectthat carries the device error, instead of aformat_acceptfollowed by silence. The endpoint goes back to idle, and the next track plays (#62, #63). - Stopping no longer waits forever on a stuck device (15 s at most) (#63).
- Protocol unchanged.
Not yet supported: DoP for DACs without a native DSD format (tracked in #62).
Full Changelog: v0.2.7...v0.2.8
What's Changed
- fix(endpoint): pas de DSD natif annoncé quand le DAC le refuse, une ouverture refusée est signalée (FormatReject) (#62) by @renesenses in #63
- chore(release): v0.2.8 by @renesenses in #64
Full Changelog: v0.2.7...v0.2.8
v0.2.7
Highlights
Linux endpoint: automatic DAC detection
- Without
--audio-device, the endpoint now really uses the DAC it detects: a USB DAC first (recognised by its USB-Audio driver), then an I2S card whose name says DAC, never HDMI. Before this, the detectedcard N:line was silently replaced bydefault(#60). - A device name refused by the filter is now logged as a WARN with the name received (#60).
- A
card N: …line from the device list (web UI) is translated toplughw:CARD=<id>,DEV=<n>(hw:for native DSD) (#60).
Native DSD
- With
--dsd, the endpoint probes the DAC'shw:device at startup (DSD_U32_BE, stereo) and announces the highest rate accepted without a gap, up to DSD512 (RFC §5.6), inhello_ackand in the mDNS TXT record. It used to always announce DSD64. If the probe fails, it announces DSD64 and logs it (#60). - Protocol unchanged.
Not yet tried on a real USB DAC, nor in DSD512.
Full Changelog: v0.2.6...v0.2.7
v0.2.6
Highlights
endpoint_release
ALSA fixes
- Discover the
Playback Volumecontrol on USB DACs such as SMSL (#6). apply_volumeno longer corruptsPcmF32le: float samples were reinterpreted asi32(#21, #22).- Native ALSA output without ffmpeg or aplay, with FLAC decoded in-process (#44, #45, #48).
Transport and ring buffer
- Validate the UDP
payload_lenagainst the bytes actually received (#25). - Scale the ring buffer with the sample rate to prevent underruns at high rates (#26).
- Count
CpalOutputunderruns and report them instream_stats(#28, #49).
Clock
- Saturate timestamp conversions to prevent silent wraparound on extreme offsets (#35, #36).
- Use a separate sequence counter for clock sync so it no longer creates FEC gaps (#37, #38).
- Reject outlier RTT samples after warm-up, using a median/MAD threshold (#27, #47).
- One monotonic clock per process, anchored once on wall-clock time (#34, #50).
- Injectable time source and exposed OAAT clock measurements for tests (#14).
FLAC and aplay
- Idle periods of aplay are no longer counted as underruns (#12, #41).
- Bounded FLAC stream decoder that decodes the frames after init (#32, #43).
- aplay is respawned with backoff after an unexpected exit (#33, #42).
Daemon mode
endpoint --daemonnow forks before the tokio runtime starts: the control port, the web UI and SIGTERM work again (#52, #54).
Specification
- RFC version 0.4.0 (Draft):
endpoint_releaseandhello_ack.features(#55).
Licensing
- The specification is now under the Community Specification License 1.0 (#40).
- The reference implementation is now under Apache 2.0 (#20).
CI
Security
- The endpoint status page controls (
POST /api/release,POST /api/device) are now restricted: they listen on loopback by default, cross-origin requests are refused, a JSON content type is required, and an optional shared token allows deliberate remote control. Local controllers that send noOriginheader keep working. See GHSA-34qc-hmv4-hmg2. Affected:oaat-endpoint<= 0.2.5. Upgrade to 0.2.6.
Thanks
Many thanks to our external contributors: @spockfish (endpoint_release, and coordination on the advisory), @Xavijol, @another-Kin and @jprobbe.
Full Changelog: v0.2.5...v0.2.6
v0.2.5
What's Changed
- feat(alsa): native DSD_U32_BE playback, opt-in via --dsd — Xavier Joly by @renesenses in #4
- endpoint: enlarge aplay ALSA buffer 0.5s → 2s (stop XRUN micro-dropouts) by @renesenses in #5
- fix(endpoint): reclaim aplay buffer headroom + surface underruns by @renesenses in #10
- fix(lint): clippy stable récent sur le code Linux-only by @renesenses in #11
Full Changelog: v0.2.4...v0.2.5
v0.2.4
What's Changed
- PTS-scheduled playback, drift servo, and FEC parity filtering by @renesenses in #1
- FEC end-to-end, bit-perfect output path, RFC 0.3.0 by @renesenses in #2
- ci(release): publier aussi des noms d'assets sans version (URLs latest stables) by @renesenses in #3
New Contributors
- @renesenses made their first contribution in #1
Full Changelog: v0.2.3...v0.2.4