- Fix FDK-AAC C shim: use proper OUT_BITSTREAM_DATA=3 constant and INT types
- Add frame-aligned PCM accumulation buffer to AAC encoder for clean output
- Add fdkaac_encode and fdkaac_close to C shim (all FDK-AAC calls in C)
- Implement crossfade between tracks (3s overlap, 2s fade-in/out)
- Tune buffer: 90% drain sleep for encoding headroom, 64KB burst-on-connect
- Add FFI bindings for new shim functions
- Rewrite README.org with full architecture docs and integration plan
- harmony-backend.lisp: Audio pipeline connecting Harmony to encoders
- Uses :dummy drain to capture PCM without audio output
- Reads from packer bip-buffer via request-read/finish-read API
- Encodes PCM to MP3 via LAME and feeds to stream server
All three systems compile cleanly:
cl-streamer, cl-streamer/encoder, cl-streamer/harmony
- Architecture document outlining Icecast/Liquidsoap replacement
- Core streaming server with ICY metadata protocol support
- Thread-safe ring buffer for audio data
- Mount point abstraction with metadata updates
- Multi-client connection handling
This is experimental groundwork for integrating with Harmony/cl-mixed
and playlisp/parsector for a pure CL streaming solution.
Glenn Thompson