kajisho5/ffmpeg-skill v3.1.0 on GitHub

Automated release: version and notes generated from pull requests merged since 3.0.0.

An ffmpeg failure names its cause when ffmpeg's message does: error.reason disk_full,
encoder_missing, filter_missing (details.name), input_incomplete (no moov atom) or
input_unreadable, with a hint. Kind and exit code are unchanged.
"No local speech-to-text engine found" is kind: missing_tool, reason: asr_engine_missing
(was kind: input).
CI spreads the suite over three machines per platform (tests/run_parallel.py --part I/N, which
deals the shards out by their measured cost and includes test_contract), so a run takes about
a third of the 26 minutes it took on one 4-core machine. The tests themselves are unchanged.

A tool killed with SIGKILL (an agent's time limit) left ffmpeg running, writing the output under
its final name after the caller had been told the run failed. On Linux the kernel now stops
ffmpeg with its parent (PR_SET_PDEATHSIG).
The in-place re-encoding tools (fit, denoise, color, audio, deinterlace, ...) ran the whole encode
a second time after any failure and blamed subtitle streams the source did not have; they
retry only when the source has subtitle or data streams and the failure is not one of the
causes above.
render.py templates with a loudness stage (reels, shorts, ...) failed on a video with no audio
after the picture stages had run, while the dry run said they would work. The stage is skipped
and recorded (loudness_no_audio).
Failures say why: ffmpeg error reasons, no orphan ffmpeg, no blind re-encode, silent templates (#389)
ci: spread the suite over three machines per platform (#388)

What's Changed

ci: spread the suite over three machines per platform by @kajisho5 in #388
Failures say why: ffmpeg error reasons, no orphan ffmpeg, no blind re-encode, silent templates by @kajisho5 in #389

Full Changelog: v3.0.0...v3.1.0