HomeTube 2.10.0 fixes two user-reported issues — wrong audio track selection and runaway memory on fragmented downloads — and makes configuration more forgiving.
🎧 The right audio track, every time
Some videos ship several audio tracks, and the original language is not always the one your region gets by default. In that situation HomeTube could pick the dub — a video with a Spanish original came down in English (#69). Original-voice detection now correctly prefers the track marked original over the locale-based default, so VO_FIRST delivers what it promises.
🧠 Flat memory usage on fragmented downloads
Downloading from platforms that stream in small fragments (Dailymotion, HLS sources) could make memory usage grow throughout the download until the machine froze (#82). The log pipeline was keeping ever-growing copies of the session logs. Logs are now bounded and refreshed smartly — memory stays flat even on hours-long downloads, with the same live logs in the UI.
🛟 Configuration typos no longer fail silently
Wrote LANGUAGE_SECONDARIES instead of LANGUAGES_SECONDARIES? HomeTube now honors the misspelled variable (when the canonical one is unset) and logs a warning telling you exactly what to rename, instead of silently ignoring your preferences.
🧹 Under the hood
- Hardened release pipeline: releases are drafted automatically once tests and the multi-arch Docker build succeed, then curated and published by hand — like these notes. Process documented in docs/releasing.md.
- New regression tests for audio-track selection and configuration parsing.
Docker image: ghcr.io/egalitarianmonkey/hometube:2.10.0
Full changelog: v2.9.1...v2.10.0