Compare commits

..

2 commits

Author SHA1 Message Date
Deivid Soto
89236f13b5 docs(roadmap): hueco #3 3c closed (capability negotiation) + TTFF diagnosis 2026-05-31 12:48:50 +02:00
Deivid Soto
957d499658 feat(stream): device-aware remux (HEVC/AV1 + non-aac audio) + TTFF timers
Hueco #3 / 3c (CLI). NewRemuxSource now copies the video for any
browser-decodable codec: h264, or HEVC/AV1 when the web says the device
decodes them (caps). HEVC is muxed with -tag:v hvc1 (Apple requirement),
and non-aac audio (ac3/eac3/dts) is transcoded to aac while the video is
still copied (ActionRemuxAudio) — this covers the very common h264+ac3 mkv.

Startup instrumentation for time-to-first-frame diagnosis:
- remux branch logs [probe=.. spawn=..]
- transcodeSource logs 'first fMP4 bytes after ..' (ffmpeg → first output)
- serveGrowing logs reads that block >250ms (client seeking ahead of the
  live edge) + the first read's offset vs produced/estimated size.

Verified: caps gate (hls without caps, remux with), hvc1 retag (ffprobe of
the /stream output = hevc/hvc1), HEVC playback confirmed on a real iPhone
Safari over Tailscale. LAN timeline: probe 16ms, spawn 1ms, first byte
201ms, no serveGrowing blocks.
2026-05-31 12:44:12 +02:00
5 changed files with 93 additions and 5 deletions

View file

@ -54,7 +54,8 @@ El path HLS **siempre re-encoda** (incluso mp4 h264/aac ya compatible). `DecideA
(passthrough/remux) existe pero muerto en el path browser. Sin negociación por (passthrough/remux) existe pero muerto en el path browser. Sin negociación por
capacidades del dispositivo. Sin ABR multi-bitrate. capacidades del dispositivo. Sin ABR multi-bitrate.
Diseño por fases (3a direct-play / 3b remux fMP4 / 3c capability-negotiation / 3d ABR) Diseño por fases (3a direct-play / 3b remux fMP4 / 3c capability-negotiation / 3d ABR)
en el estado abajo. **Fases 3a + 3b CERRADAS** (smoke e2e en browser); 3c/3d pendientes. en el estado abajo. **Fases 3a + 3b + 3c CERRADAS** (smoke e2e, incl. HEVC en iPhone
Safari real); 3d (ABR) pendiente, baja prioridad.
### Hueco #4 — Pre-transcode (transcode-on-download) 🔵 DISEÑADO (ver estado abajo) ### Hueco #4 — Pre-transcode (transcode-on-download) 🔵 DISEÑADO (ver estado abajo)
Al completar una descarga/import, transcodificar/remuxar en background para que el Al completar una descarga/import, transcodificar/remuxar en background para que el
@ -398,6 +399,38 @@ binario local para el smoke, **sin publicar nada**. `DIRECT_PLAY_MIN_VERSION = 0
- Listener `loadedmetadata {once:true}` del attach nativo no se limpia - Listener `loadedmetadata {once:true}` del attach nativo no se limpia
explícitamente en cleanup (idempotente, impacto nulo). explícitamente en cleanup (idempotente, impacto nulo).
**Fase 3c CERRADA 2026-05-31** (capability-negotiation, alcance ampliado):
- CLI (`feat/unarr-agent` 957d499): `NewRemuxSource` copia el vídeo para cualquier
codec decodificable: h264, o HEVC/AV1 si el dispositivo lo declara. HEVC se muxea
con `-tag:v hvc1` (Apple lo exige). Audio no-aac (ac3/eac3/dts) se transcodifica a
aac copiando el vídeo (`ActionRemuxAudio`) → cubre el muy común **h264+ac3 mkv**.
- WEB (`feat/unarr-brand` b0681d99): player sondea `canPlayType` (`detectDeviceCaps`)
y envía `{hevc,av1}` en el POST; `decidePlayMethod(p, caps)` device-aware:
HEVC/AV1 → `remux` solo si el dispositivo decodifica; audio no-aac ya no fuerza
`hls`. Tests caps actualizados (10).
- **Smoke e2e:** caps gate (sin caps→`hls`, con caps→`remux`); h264+ac3 remux
reproduce en Chrome (audio transcodeado, vídeo copiado); retag verificado por
ffprobe (`codec_name=hevc`, `codec_tag_string=hvc1`); **HEVC reproduce en iPhone
Safari real (Tailscale) — confirmado por el usuario.** ✓
- **Caveat:** playback HEVC en Apple no se puede smokear en este host (Chrome-Linux
no decodifica HEVC; Mac-mini Safari por SSH bloqueado por TCC: Automation +
Screen Recording necesitan click GUI). Verificado vía iPhone del usuario.
**Diagnóstico time-to-first-frame (2026-05-31)** (instrumentación en 957d499:
timers `probe`/`spawn`, `first fMP4 bytes after`, `serveGrowing blocked`):
- Agente NO es el cuello: probe 1698ms, spawn 1194ms, primer byte fMP4 ~201ms,
**0 bloqueos** en `serveGrowing` (LAN ni remoto). Remux `-c copy` completo de un
fichero de ~780MB en ~16s (limitado por lectura NAS).
- `moov` al frente (empty_moov OK) → el player no busca metadata al final.
- Cliente (Chrome/LAN): POST→primer request ~480ms (sobre todo carga de página).
- **1ª reproducción lenta = warm-up de red (Tailscale); 2ª/3ª rápidas** (confirmado
por el usuario). No es un problema de código.
- Player YA da feedback (fases `loading-meta`/`probing-transport`/`playing` +
overlay "Preparando…" + spinner de buffering + mensaje stuck >10s). El "sin
feedback" del test fue por usar URL cruda (sin UI), no el flujo real.
- **Conclusión:** sin optimización de código necesaria. Arranque garantizado-instante
= **hueco #4 (pre-transcode)**: dejar el remux/encode hecho antes del play.
--- ---
### Hueco #4 — Pre-transcode (transcode-on-download) ### Hueco #4 — Pre-transcode (transcode-on-download)

View file

@ -624,6 +624,7 @@ func runDaemonStart() error {
// over /stream — no video re-encode, no HLS. The web decided this from // over /stream — no video re-encode, no HLS. The web decided this from
// scan metadata + version gate; we still need ffmpeg (copy uses it). // scan metadata + version gate; we still need ffmpeg (copy uses it).
if sess.PlayMethod == "remux" { if sess.PlayMethod == "remux" {
tStart := time.Now()
probeCtx, cancelProbe := context.WithTimeout(ctx, 15*time.Second) probeCtx, cancelProbe := context.WithTimeout(ctx, 15*time.Second)
probe, perr := engine.ProbeFile(probeCtx, tcRuntime.FFprobePath, filePath) probe, perr := engine.ProbeFile(probeCtx, tcRuntime.FFprobePath, filePath)
cancelProbe() cancelProbe()
@ -631,6 +632,7 @@ func runDaemonStart() error {
log.Printf("[stream %s] remux probe failed: %v", agent.ShortID(sess.SessionID), perr) log.Printf("[stream %s] remux probe failed: %v", agent.ShortID(sess.SessionID), perr)
return return
} }
tProbe := time.Now()
remuxCtx, remuxCancel := context.WithCancel(ctx) remuxCtx, remuxCancel := context.WithCancel(ctx)
src, serr := engine.NewRemuxSource(remuxCtx, filePath, probe, tcRuntime.FFmpegPath, sess.FileName) src, serr := engine.NewRemuxSource(remuxCtx, filePath, probe, tcRuntime.FFmpegPath, sess.FileName)
if serr != nil { if serr != nil {
@ -642,7 +644,13 @@ func runDaemonStart() error {
// cancel stops the ffmpeg copy; SetGrowingFile/ClearFile also Close() // cancel stops the ffmpeg copy; SetGrowingFile/ClearFile also Close()
// the source, so the temp file is always cleaned up. // the source, so the temp file is always cleaned up.
playerSessionRegistry.add(sess.SessionID, func() { remuxCancel(); streamSrv.ClearFile() }) playerSessionRegistry.add(sess.SessionID, func() { remuxCancel(); streamSrv.ClearFile() })
log.Printf("[stream %s] remux (copy) → fMP4: %s", agent.ShortID(sess.SessionID), filepath.Base(filePath)) // Startup timing (TTFF diagnosis): probe = ffprobe on the source;
// spawn = ffmpeg launch + tmp setup. First-fMP4-byte is logged by the
// source itself; serveGrowing logs any client read that blocks waiting
// for ffmpeg to catch up.
log.Printf("[stream %s] remux (copy) → fMP4: %s [probe=%v spawn=%v]",
agent.ShortID(sess.SessionID), filepath.Base(filePath),
tProbe.Sub(tStart).Round(time.Millisecond), time.Since(tProbe).Round(time.Millisecond))
go func() { go func() {
rctx, cancel := context.WithTimeout(ctx, 10*time.Second) rctx, cancel := context.WithTimeout(ctx, 10*time.Second)
defer cancel() defer cancel()

View file

@ -863,6 +863,7 @@ func (ss *StreamServer) serveGrowing(w http.ResponseWriter, r *http.Request, src
buf := make([]byte, 256*1024) buf := make([]byte, 256*1024)
off := start off := start
firstRead := true
for { for {
if explicitEnd >= 0 && off > explicitEnd { if explicitEnd >= 0 && off > explicitEnd {
return return
@ -870,7 +871,19 @@ func (ss *StreamServer) serveGrowing(w http.ResponseWriter, r *http.Request, src
if r.Context().Err() != nil { if r.Context().Err() != nil {
return // client disconnected / request cancelled return // client disconnected / request cancelled
} }
readStart := time.Now()
n, err := src.ReadAt(buf, off) n, err := src.ReadAt(buf, off)
// TTFF diagnosis: a read that blocks means the client asked for bytes the
// remux hasn't produced yet (a seek ahead of the live edge, or the very
// first read before ffmpeg's init lands). Log it so a slow start is
// attributable to "waiting on ffmpeg" vs network/decoder.
if waited := time.Since(readStart); waited > 250*time.Millisecond {
log.Printf("[stream] serveGrowing read off=%d blocked %v (produced=%d est=%d)",
off, waited.Round(time.Millisecond), src.Size(), src.EstimatedSize())
} else if firstRead {
log.Printf("[stream] serveGrowing start off=%d (produced=%d est=%d)", start, src.Size(), src.EstimatedSize())
}
firstRead = false
if n > 0 { if n > 0 {
toWrite := n toWrite := n
if explicitEnd >= 0 { if explicitEnd >= 0 {

View file

@ -5,6 +5,7 @@ import (
"errors" "errors"
"fmt" "fmt"
"io" "io"
"log"
"os" "os"
"path/filepath" "path/filepath"
"strings" "strings"
@ -172,8 +173,21 @@ func newTranscodeSource(
// before sending PlayMethod="remux". The browser plays the result progressively // before sending PlayMethod="remux". The browser plays the result progressively
// via byte-range. Caller MUST Close() it (kills ffmpeg + removes the temp file). // via byte-range. Caller MUST Close() it (kills ffmpeg + removes the temp file).
func NewRemuxSource(ctx context.Context, srcPath string, probe *StreamProbe, ffmpegPath, displayName string) (GrowingSource, error) { func NewRemuxSource(ctx context.Context, srcPath string, probe *StreamProbe, ffmpegPath, displayName string) (GrowingSource, error) {
opts := TranscodeOpts{Action: ActionRemux, FFmpegPath: ffmpegPath} // Audio: copy when already AAC; otherwise transcode to AAC (ActionRemuxAudio).
return newTranscodeSource(ctx, srcPath, probe, ActionRemux, opts, displayName) // Either way the VIDEO is copied — the expensive part is never re-encoded.
// This lets remux cover the very common h264+AC3/DTS mkv case (hueco #3 / 3c),
// not just h264+AAC.
action := ActionRemux
if probe != nil && probe.AudioCodec != "" && probe.AudioCodec != "aac" {
action = ActionRemuxAudio
}
opts := TranscodeOpts{Action: action, FFmpegPath: ffmpegPath}
// HEVC muxed into MP4 must carry the hvc1 tag or Apple/Safari won't decode
// it (hueco #3 / 3c). h264 (avc1) needs no override.
if probe != nil && probe.VideoCodec == "hevc" {
opts.VideoTag = "hvc1"
}
return newTranscodeSource(ctx, srcPath, probe, action, opts, displayName)
} }
// signalNotify wakes any goroutine blocked in ReadAt. Non-blocking: if a // signalNotify wakes any goroutine blocked in ReadAt. Non-blocking: if a
@ -209,6 +223,13 @@ func (t *transcodeSource) watchSize(ctx context.Context) {
} }
current := stat.Size() current := stat.Size()
if current > t.size.Load() { if current > t.size.Load() {
if t.size.Load() == 0 && current > 0 {
// TTFF diagnosis: how long from ffmpeg spawn to the first
// fMP4 bytes (init + first fragment) landing — the floor on
// when /stream can serve anything playable.
log.Printf("[stream] %s first fMP4 bytes after %v (%d KB)",
t.name, time.Since(t.startedAt).Round(time.Millisecond), current/1024)
}
t.size.Store(current) t.size.Store(current)
t.signalNotify() t.signalNotify()
} }

View file

@ -27,6 +27,11 @@ type TranscodeOpts struct {
SourceHeight int // probed source height — used to derive a sane H.264 level SourceHeight int // probed source height — used to derive a sane H.264 level
StartSeconds float64 StartSeconds float64
FFmpegPath string FFmpegPath string
// VideoTag forces the output stream's codec tag on a copy remux. HEVC muxed
// into MP4 must carry the `hvc1` tag (not the default `hev1`) or Safari /
// Apple devices refuse to decode it. Empty = leave ffmpeg's default. Only
// applied on copy actions (passthrough/remux); a real re-encode sets its own.
VideoTag string
} }
// Transcoder wraps a long-running ffmpeg child process whose stdout streams // Transcoder wraps a long-running ffmpeg child process whose stdout streams
@ -222,8 +227,16 @@ func buildFFmpegArgs(filePath string, opts TranscodeOpts) []string {
switch opts.Action { switch opts.Action {
case ActionPassthrough, ActionRemux: case ActionPassthrough, ActionRemux:
args = append(args, "-c:v", "copy", "-c:a", "copy") args = append(args, "-c:v", "copy", "-c:a", "copy")
// HEVC → MP4 needs the hvc1 tag for Apple/Safari (hueco #3 / 3c).
if opts.VideoTag != "" {
args = append(args, "-tag:v", opts.VideoTag)
}
case ActionRemuxAudio: case ActionRemuxAudio:
args = append(args, "-c:v", "copy", "-c:a", "aac", "-b:a", coalesce(opts.AudioBitrate, "192k")) args = append(args, "-c:v", "copy")
if opts.VideoTag != "" {
args = append(args, "-tag:v", opts.VideoTag) // HEVC → hvc1 for Apple
}
args = append(args, "-c:a", "aac", "-b:a", coalesce(opts.AudioBitrate, "192k"))
case ActionTranscodeVideo: case ActionTranscodeVideo:
videoCodec := opts.HWAccel.FFmpegVideoCodec("h264") videoCodec := opts.HWAccel.FFmpegVideoCodec("h264")
args = append(args, "-c:v", videoCodec) args = append(args, "-c:v", videoCodec)