Skip to content

[webOS 5.6.2 / k6lp] Playing error 100 after accepted Load on 4K H.264 direct play #954

Open

Description

@GLinnik21

Follow-up to #665 (the 601 Load refusal, fixed in #697). Reported by @profvalente in #665 (comment).

What happens

LG 60UN7310, webOS 5.6.2 (k6lp, K6LP_ATSC, BOARD_DV_1ST), 0.9.0-nightly-20261020 (3882f6e). A 4K H.264 High@L5.1 MKV (3840x2160, ~30 fps, AC3) at Original/Auto quality fails with tv_pipeline. The 601 is gone: Load is accepted at 4K@30, then the pipeline throws error 100 after feeding starts.

decision: part=directplay ... -> DIRECT PLAY
load: v=H264 a="AC3" fps=30.303 ... max=3840x2160@30
SMP: Load returned ok=1
ff: seek begin 839s
rebase: first post-seek keyframe pts=839000000000 -> pts_shift=-839000000000
feed v#1..#4, a#1..#4 reply=O
smp_cb type=18 num=100 str=Playing error

Known

  • The official Plex app direct-plays the same file on the same TV.
  • On 20261016, 1080p/20 Mbps and other reduced qualities (server transcode) play fine.
  • This attempt resumed at 839s; a start from 0:00 has not been tried.
  • The submitted log also contains 0.8.0 requests (4K@50/60 with 601), so only the first 20261020 attempt counts as evidence for the nightly.
  • Startup still logs the devcaps fallback to the 49SM9000PLA profile (observation, not a proven cause).
  • A second title (The Girl Who Leapt Through Time) reportedly fails too; its codec, resolution, fps and error code are unknown.

Needed

  1. Same file started from 0:00 (no resume) on a clean 20261020 install, with the log.
  2. ffprobe of the second title and its error code.

Labels: bug, playback, user-report, needs-info.

Activity

  1. added
    bugSomething isn't working
    playbackPlayback, the player and its controls
    user-reportReported by a user of a released build
    needs-infoWaiting on the reporter: logs, TV model or steps
    on Oct 23, 2026
  2. profvalente commented on Oct 23, 2026

    @profvalente

    Hi Gleb — the user explicitly authorized me to submit this consolidated follow-up. There are new physical-TV test results, second-title metadata, and an important limitation in the latest uploaded log.

    1. Inside Out 2: Original still fails from 0:00 on 20261023

    On the same LG 60UN7310 / webOS 5.6.2 (k6lp, K6LP_ATSC, BOARD_DV_1ST), the user installed 0.9.0-nightly-20261023 and retried the same 4K H.264/AC3 MKV.

    • A new photograph clearly shows 0.9.0-nightly-20261023 in the failure-screen footer, Original checked in the quality menu, and tv_pipeline.
    • The user explicitly confirmed starting from the beginning (0:00), not resuming.
    • Therefore the failure is not exclusive to the earlier 839-second resume case.
    • The photograph does not show a numeric pipeline code. We cannot yet assign error 100 or 601 to this specific 20261023 attempt.
    • Auto and manual reduced qualities have not been separately verified on 20261023 in the supplied evidence.

    Previously established controls remain distinct: the official LG Plex app fully direct-played video and audio of this exact file on this TV; 1080p/20 Mbps and other reduced qualities worked on 20261016. The earlier identified 20261020 log attempt accepted 4K@30 Load, then produced error 100 after initial feeds. Those are prior-build results.

    2. The Girl Who Leapt Through Time: now characterized; not necessarily the same defect

    Plex media-information screenshots establish:

    • MKV, 1920x1080, progressive, 23.976 fps.
    • H.264 High 10, level 4.1, 10-bit, 4:2:0, four reference frames.
    • Japanese FLAC stereo, 16-bit, 48 kHz.
    • Portuguese ASS subtitles.

    Test evidence:

    • On 0.8.0, Auto plays; Original fails with tv_pipeline (version visible in the error photograph).
    • During the working Auto test, the Plex dashboard explicitly showed 1080p H.264 High 10 -> 1080p H.264 (Transcode) and FLAC stereo -> AAC (Transcode). ASS subtitles were selected; no definitive subtitle delivery/burn-in conclusion is drawn from the localized dashboard label.
    • After installing the build the user identified as 23 October, this film plays with “Original · Converted” shown in the player, photographed at about 02:09. That particular photo does not display the build identifier, and a fresh dashboard/log for that session is not available. It demonstrates working playback with conversion indicated, not proven native High 10 decoding.

    The two titles being animations was the user's initial observation. The second file's High 10 profile is a more specific compatibility lead; we are not claiming the two failures share a cause.

    3. Latest log: do not attribute its old 601 events to 20261023

    The newly supplied file, named plxnative-events(2).log, has 31,927 lines. Inspection found 75 X-Plex-Version occurrences, all 0.8.0, and no “nightly” string. It also contains the old 4K@50/@60 declarations. Thus it does not establish the numeric error or pipeline sequence for the photographed 20261023 test. We cannot tell whether an older/different installation's log was fetched.

    For provenance only, these are literal diagnostic lines from that file (non-adjacent where indicated), NOT evidence of a nightly regression:

    esInfo: videoFps 30303/1000 + adaptiveResolution (src 30.303)
    load: v=H264 a="AC3" fps=30.303 dv=present:0 P0/0 el:0 atmos:0 max=3840x2160@50
    smp_cb type=18 num=601 str=Resource Allocation Error
    SMP: Load returned ok=0
    
    [separate fixed-quality attempt]
    route: quality ceiling P1080High — source 9997kbps 3840x2160; denying direct play + remux, re-encoding
    load: v=H264 a="AC3" fps=0.000 dv=present:0 P0/0 el:0 atmos:0 max=3840x2160@60
    smp_cb type=18 num=601 str=Resource Allocation Error
    SMP: Load returned ok=0
    [later in that attempt]
    ff: v=#0 codec=h264 codec_id=27 1920x1080 trc=2 pri=2 spc=2 a=#1 dur_ns=5786295000000
    

    The startup reports an unreadable codec-capability configuration and fallback to 49SM9000PLA; this remains an observation, not a diagnosed cause.

    Your start-from-zero question is now answered by the user's test on 20261023; obtaining its correctly matched log remains pending. Could you confirm which runtime log location/install flavor should be fetched for the nightly, particularly if stable and nightly coexist?

    Only technical excerpts and screenshot transcriptions are posted. The raw log and uncropped photos are withheld to avoid exposing server URLs, network addresses, filesystem/media paths, library IDs, and client/session identifiers. Thank you for continuing the investigation.

  3. GLinnik21 commented on Nov 8, 2026

    @GLinnik21
    CollaboratorAuthor

    @profvalente, thanks for confirming that Inside Out 2 also failed from 0:00 and for supplying the second title's media details. Could you ask the reporter to retry Inside Out 2 from 0:00 at Original quality once the November 10 nightly is available?

    Please share the build shown in PlxNative Nightly, whether it Direct Plays, plays after conversion, or still errors, and the event log from that same attempt. When stable and nightly are both installed, the nightly event log is at /tmp/com.beb.plxnative.nightly/plxnative-events.log. Please copy it before relaunching the app and review it for private details before posting. An excerpt covering the playback attempt is fine.

    Nightly update instructions

  4. profvalente commented on Nov 10, 2026

    @profvalente

    Hi Gleb — reporting the requested physical-TV retest on behalf of the user, with their explicit authorization. The user does not write English. This is a new, correctly matched Nightly log, and I am sharing only sanitized excerpts (not the full log, which contains private identifiers/URLs).

    Device/build: LG 60UN7310, webOS 5.6.2 (k6lp / K6LP_ATSC / BOARD_DV_1ST); PlxNative 0.10.0-nightly-20261110, confirmed both in the failure-screen photo and the log (X-Plex-Version). Test title: Inside Out 2, MKV 3840x2160 H.264/AC3, ~30.303 fps.

    User clarification: The automatic fallback initially produced visibly low-quality video. The user tried to raise/change the playback quality and encountered the error screen: “Playback failed — Playback stopped after it had started”, footer playback_interrupted. Another photo shows “Original · Converted”, with video displaying. Please distinguish the user's manual quality change from the initial direct-play refusal.

    What the same-session log establishes:

    1. On the initial playback from offset 0, the app chose the Auto Original route and attempted genuine 4K H.264/AC3 direct play. The declaration was corrected to 3840x2160@30, but native Load failed with a new observed code 104: fail to create AVProxy. This attempt was Auto selecting Original, not proof of a manual Original selection at 0:00.
    2. The built-in recovery selected 720 kbps / 854x480 HLS, which actually decoded 480x270. The transcode Load was accepted; the player displayed frames. It declared a maximum 1920x1080@60, not 4K.
    3. Automatic quality improvement initially hit warmup_deadline on 720→20000 kbps, followed by publication_overlap on 720→10000 kbps. This was followed by HLS candidate automatic lease was superseded, no rung left for source failed; failing the Engine and an interrupted-playback error screen. These diagnostics are consistent with the visible failure; the precise relation to the user's button press is not fully established.
    4. Following Try again, HLS playback worked; ABR successfully moved up to 1080p/10000 and then 20000 kbps. The user later chose Original at offsets approximately 30 s and 116 s; these requests were routed as Original ... re-transcoding this playback, and accepted with a 1920x1080@60 H.264/AC3 Load. The screenshot's “Original · Converted” is therefore not Direct Play 4K.
    5. In the supplied 20261110 log we found no callback error 601 or 100. The explicitly captured native error is 104. This does not establish that 4K Direct Play is fixed or that errors 601/100 recur on this build.

    Sanitized verbatim excerpts in source order (unrelated lines omitted):

    install: id=com.beb.plxnative.nightly flavour=nightly runtime=/tmp/com.beb.plxnative.nightly
    webos: webOS TV Lite release=5.6.2 codename=jhericurl-jimna api= major=5
    webos: model=k6lp board=K6LP_ATSC hw=BOARD_DV_1ST
    decision: part=directplay video= general=None mde=Some(1000) -> DIRECT PLAY
    route: Auto Original — source 10259kbps 3840x2160; no video encode
    load: v=H264 a="AC3" fps=30.303 dv=present:0 P0/0 el:0 atmos:0 max=3840x2160@30
    smp_cb type=18 num=104 str=fail to create AVProxy
    smp: Load refused by the pipeline before any picture (type=18 num=104) — failing the source
    auto: Original source open failed without a throughput sample; reusing bootstrap Some(P480) as 720kbps 854x480 HLS
    SMP: Load returned ok=0
    retranscode rk=3442 audio=18297 sub=21896 offset=0 -> transcode start
    load: v=H264 a="AAC" fps=0.000 dv=present:0 P0/0 el:0 atmos:0 max=1920x1080@60
    SMP: Load returned ok=1
    ff: v=#0 codec=h264 codec_id=27 480x270 trc=2 pri=2 spc=2 a=#1 dur_ns=5787000000000
    open recovery: the refusal was recovered on the Both rung (type=18 num=104)
    abr: tx Up 720->20000kbps outcome=warmup_deadline
    abr: tx Up 720->10000kbps outcome=publication_overlap
    hls: demux failed: HLS candidate automatic lease was superseded
    open recovery: no rung left for source failed; failing the Engine
    playback retry: resolving item again at quality Auto
    abr: committed Up to 10000kbps 1920x1080 out=1920x1080
    quality: Original picked — source 10003kbps 3840x2160; re-transcoding this playback
    route transition: user retranscode at 30s
    load: v=H264 a="AC3" fps=0.000 dv=present:0 P0/0 el:0 atmos:0 max=1920x1080@60
    SMP: Load returned ok=1
    quality: Original picked — source 10003kbps 3840x2160; re-transcoding this playback
    route transition: user retranscode at 116s
    SMP: Load returned ok=1
    

    The abbreviated abr: tx Up lines above omit their remaining metric fields, for privacy/readability; all other quoted lines are reproduced as logged.

    Could you advise whether the 104 AVProxy refusal and ABR publication_overlap / lease superseded represent separate issues, and which specific follow-up or safe setting you would like tested next? We will avoid further risky changes. Thank you for continuing to investigate.

  5. GLinnik21 commented on Nov 11, 2026

    @GLinnik21
    CollaboratorAuthor

    @profvalente, thanks for separating the initial Auto Original refusal from the later quality changes. These are two separate failure paths.

    We reproduced and fixed the ABR cancellation path locally, and merged #1326 (7f2dc2f2d7a2). Before the fix, an upgrade whose lease was superseded before publication could terminate the working HLS stream. It now discards only that unpublished candidate, clears the pending upgrade, and continues the active stream while preserving the viewer's pause state. The misleading publication_overlap label is now candidate_superseded for this case. Genuine publication failures still stop playback.

    The regression failed before the fix and passes afterward, including the production upgrade lease's cleanup and the shared publication runner. Full local default/hostsim checks, shipping-feature checks, and PR CI passed. This verifies the cancellation fix; it does not reconstruct the exact event that revoked the lease in your log.

    Fixed test build: download the Debug IPK artifact from the successful ARM build for this PR. It installs as PlxNative Debug alongside Nightly; please launch that app for this test. Its event log is /tmp/com.beb.plxnative.debug/plxnative-events.log.

    Please first retry Inside Out 2 from 0:00 with the Standard engine and Auto quality, allowing its fallback to play without changing quality. The useful result is whether fallback stays playing and improves quality instead of showing playback_interrupted. If it still stops, please share the exact build/commit and a sanitized excerpt around that attempt; no full private log is needed.

    The pre-picture 104 fail to create AVProxy remains a separate investigation, so this issue stays open. Our local TV runs webOS 4.5; this change does not establish that 4K Direct Play works on your 5.6.2 set, and “Original · Converted” is still not proof of 4K Direct Play. No firmware changes are requested.

  6. profvalente commented on Nov 13, 2026

    @profvalente

    Hi Gleb,

    A personal note of thanks from the LG 60UN7310 user, posted with his explicit permission.

    Today I installed PlxNative 0.11.0-nightly-20261113 and made it my current everyday player. My first quick tests have been very positive, and I'm really enjoying the improvements so far.

    I want to sincerely thank and congratulate you on your professionalism, dedication, and responsiveness. Your careful investigation of the issues we've reported, the clear technical explanations, and the steady improvements to PlxNative are genuinely appreciated. It's encouraging to see such thoughtful work going into making Plex playback better on LG TVs.

    These are just initial impressions, not a confirmation that the 4K H.264 Direct Play issue is fixed. I'll continue testing and will share precise results and sanitized logs if I encounter a reproducible problem.

    Thank you once again for your outstanding effort and for listening to users!

  7. GLinnik21 commented on Nov 13, 2026

    @GLinnik21
    CollaboratorAuthor

    @profvalente thank you, and please pass my thanks on too. That's good to hear.

    If 4K H.264 direct play does fail again, the useful part of the event log is the few lines around the attempt: grep -E 'load:|SMP:|smp_cb|open recovery|uri: ' plxnative-events.log, plus which build it was (Settings > About).


    This is an automated reply: Claude Code, an AI assistant, wrote it on the maintainer's behalf.

  8. GLinnik21 commented on Nov 13, 2026

    @GLinnik21
    CollaboratorAuthor

    Nice to hear from my human reply as well =)

  9. GLinnik21 commented on Nov 16, 2026

    @GLinnik21
    CollaboratorAuthor

    @profvalente one part of this is still open on my side: the first failure from your 2026-11-10 log, where the TV refused the 4K H.264 direct play of Inside Out 2 with code 104 (fail to create AVProxy). The quality-change error was fixed by #1326, but nothing has changed for 104, and I can't tell yet what triggers it. In that log the 4K attempt was the first play after launch, so "4K" and "first play" are mixed up.

    When it's convenient, three short tries on the current Nightly would separate them. Use Playback engine Standard and quality Original, and start Inside Out 2 from 0:00 each time:

    1. as the very first play after opening the app;
    2. right after watching a minute of any 1080p title, without closing the app.
    3. Optional: the same with Settings > Playback > Playback engine set to Experimental. If the file has one audio track and no subtitles, this plays it through the TV's own player instead.

    For each try: did it direct-play, play converted, or error? If you can, include grep -E 'load:|SMP:|smp_cb|open recovery|uri: ' plxnative-events.log from that attempt. If code 104 shows up again and the TV is rooted, the TV's own lines from the same moment would tell me which part of the TV refused it: grep -i -E 'avc\.|avproxy|videooutput' /var/log/messages /var/log/starfishmediapipeline.log /var/log/videooutputd.log /var/log/ums.log 2>/dev/null (the media pipeline and the video output service keep their own log files).

    This is an automated reply: Claude Code, an AI assistant, wrote it on the maintainer's behalf.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't workingneeds-infoWaiting on the reporter: logs, TV model or stepsplaybackPlayback, the player and its controlsuser-reportReported by a user of a released build

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions