ci: add Logic AU input regression smoke and clean vendor-fork workflows - #4
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3cd1306fa2
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| voice_id, | ||
| channel: group as u8, | ||
| note, | ||
| velocity: params.velocity / 127.0, |
There was a problem hiding this comment.
When a host uses the extended MusicDeviceStartNote API, MusicDeviceNoteParams::mVelocity is already normalized to [0, 1]; dividing it by 127 makes a full-velocity note arrive at the plugin as approximately 0.0079, producing nearly silent instruments. The validation above likewise accepts the wrong [0, 127] range, and the smoke host masks this by sending the nonstandard value 96.0; validate against 1.0 and forward the value directly.
Useful? React with 👍 / 👎.
| .map(|layout| { | ||
| u32::from(layout.main_input_channels.is_some()) | ||
| + layout.aux_input_ports.len() as u32 | ||
| }) | ||
| .unwrap_or(0), |
There was a problem hiding this comment.
For a mono-only, surround-only, or otherwise non-stereo plugin, the wrapper still constructs n_channels as 2, so this lookup finds no layout and reports zero input elements before the host has had a chance to negotiate a format. Hosts commonly query the element count and current stream format before setting either one; such units therefore appear to have no input bus (and StreamFormat property info is also rejected) despite advertising a valid layout. Seed the initial channel count/topology from the first AUDIO_IO_LAYOUTS entry rather than assuming stereo.
Useful? React with 👍 / 👎.
1 participant
Summary
AudioBufferList.mDatareplacementMakeConnectionRoot cause covered
The historical Logic silence regression could return
noErrwhile producing silence when the wrapper dropped the validAudioTimeStampor read its scratch buffers after the host replacedmDatapointers with zero-copy source buffers. The new host reproduces both behaviors against the bundled Gain AU and requires non-silent output.Validation
scripts/au_logic_input_smoke.sh(macOS): passed withcallback_status=0,render_status=0,rendered_samples=16384,silent=falseswiftc -O scripts/au_logic_input_smoke/LogicPullHost.swiftbash -n scripts/au_logic_input_smoke.shgit diff --checkThe branch still contains no changes to the pre-existing untracked
.DS_Store,.openswarm/, oropenswarm.jsonartifacts.