Builds in first updates to address issues reported.
Note: I'm testing with slightly different files - I've found and improved some issues that are similar to what was failing in reports, but there could still be issues I didn't see. Please test and report what was fixed and what wasn't.
I did notice a freezing issue on stopping some videos with latest build.
This wasn't specifically 3d (I was hitting it with youtube) but was a regression from previous build.
This was a false alarm. Found out my MySQL server machine had a full disk and all SQL requests were failing. That caused the hang on stop (when it updates elapsed time/played count). Deleted some files and all is good for me now. (But I'll look at any logs for clues from others).
Your branch (popcornmix/xbmc, fix-stereoscopic-3d-gbm-upstream) was last pushed 22 Aug and I cannot find the parsing in it, so I have no idea whether any of the below duplicates what you already wrote - please just say if it does.
I've pushed latest version. Three commits related to sub depth parsing:
MPLSParser: read the stereoscopic stream number table
BlurayOffsetMetadata: read the plane offsets of a 3D dependent view
VideoPlayer: put Blu-ray 3D subtitles at the depth the disc asks for
I did notice a freezing issue on stopping some videos with latest build.
This wasn't specifically 3d (I was hitting it with youtube) but was a regression from previous build.
It is possible that could affect some Blu-ray menu stuff (which tends to include a sequence of short videos)
I'll try to find out if I broke it, or it was a bump to latest master (which has been getting a lot of commits due to beta 2 cut-off).
popcornmix - answering your #72 (a sample with varying subtitle depth), since that looks like where this stalled.
I did get a usable sample with depth from NTM-3D (and after working out the parsing, found a couple of others - Minions was one).
Some of them were quite subtle - without adjacent samples as significantly different depths it can be hard to tell if it's working.
I'm also interested in dual bitmap stereoscopic subs, which appears to be a thing, but I didn't seem to have any.
(it seems more common to have a single bitmap with a plane offset).
Tried to try the subtitle depth parsing in the new build but had some issues. Most of the ISOs I tried played the intro in 3D then black screen. If I tried to skip it would result in instant froozen image or black screen. Audio always kept playing. If I skipped for example 10 minutes into the movie, stopped it and the resumed it would play and activate 3D but the image was 2D.
A few of the ISOs i tried however seemed to work and most if not all the ones that had the issues above was Disney/Pixar movies.
I will be able to get logs and even a video of the problem tomorrow if that helps
Yes, debug enabled log file would be useful. Also is this a regression compared to previous build, or did that do it too?
I run netflix on these builds just fine. Your failure is not due to Kodi 22 (it affects everyone when viewing newer videos)
I've had to manually patch the youtube add with this PR. I don't know why that hasn't been merged when it's reported as fixing it for so many.
Updated builds in first post. Hopeful improvements:
Subtitle depth parsing
Switch to correct 3d mode with 3840x2160 HSBS hevc
Enable Blu-Ray menu in more places (e.g. context-menu/Choose playlist/Show Blu-ray menu)
Awesome, thanks for the tip! I hadn't set up a whitelist. I have one now and it's working
By the way, will we have to create a new image and set everything up from scratch when you release a new version? Or will it appear in the LibreELEC update process?
If you download the latest version (from first post) and put into /storage/.update and reboot it will update in place. No reinstall needed.
Some feedback from playing around: I tested a 3D file in 4K FSBS format (3840 x 2160), which is used for VR. Anyway: The Zidoo Z9X Pro appears to recognise, downsample and send it as 3D Framepacking, which is quite impressive. On the Pi 5, it is just sent as a standard SBS image which has to be manually switched to 3D using the projector's remote control. I guess this results in more resolution loss than with the Zidoo processing. However, to be fair, with the highest 4K FSBS resolution we sometimes see (7680 x 2160) the RTD1619BPD based Zidoo completely refuses to play.
That should be supported as frame-packed. Post a debug log so I can see what's going on.
edit: do you have a whitelist set up? I think if you do (and it includes [email protected]/1920x1080@24) then you'd get the FP output. I'll look into making that happen without a whitelist.
Fully agree - that's why I haven't tried it any sooner. If I turn "HDMI LINK" off, the glitches come back immediately. Turn it on again and they disappear for good.
For interest, what happens with "HDMI LINK" on, but CEC settings disabled on Pi.
That should narrow down if its purely enabling "HDMI LINK", or if it's something it interprets from the CEC data that makes it choose to do the glitches.
Also does the projector have any sort of firmware update mechanism? It could be a bug they've fixed.
But, with last LE 3D build and CE22, even if using Files view, as soon as I choose ISO, no chooser, rather plays main title directly:
Root causeHis library holds the movie's file path as the playlist, not the ISO. CVideoDatabase builds every list item with
Codeitem->SetDynPath(std::move(movie.m_strFileNameAndPath)); // VideoDatabase.cpp:8117so the item already carries bluray://…/3-D Rarities (2015).3D.iso//BDMV/PLAYLIST/01001.mpls before he presses Enter. The new log shows exactly that:
Code15:49:04.593 CInputManager::HandleKey: enter pressed, window 10025, action is Select 15:49:04.593 OnPlayMedia bluray://udf:///storage/Videos1/Movies3D/3-D Rarities (2015).3D/ 3-D Rarities (2015).3D.iso//BDMV/PLAYLIST/01001.mpls 15:49:04.596 CPlayerSelectionRule::GetPlayers: considering rule: discimageNote discimage is only considered, never matched — because the path is a bluray:// playlist, not an ISO. Compare the Spider-Man log, where the same rule reported matches rule: discimage.
From there, GetMenuDecisions gives NO_ACTION under Automatic menu selection: isBluray is false (dynpath isn't a disc image), isBlurayPath is true, and with BD_PLAYBACK_AUTO neither forceSelectionAtStart nor useDiscMenuSetting nor mainTitle is set, so every branch falls through. It plays 01001.mpls directly. AUTO is documented as "Only use simple menu if playlist not known" — the playlist is known, so it doesn't ask. Same for both volumes.
Why "play from Files" changed nothing: the dynpath comes from the database row, not from how he navigated to it. Window 10025 serves both views, and a Files item matched to a library entry gets the same dynpath. My suggestion was based on the assumption that Files view would give a raw ISO path — that was wrong, and I should have checked SetDynPath before offering it as a quick test.
Why IMAX / Best of 3D still work: those live under /storage/Videos2/Various/, and the earlier log shows them opening as raw .iso paths — not library items with a resolved playlist. So isBluray is true for them and the chooser runs.
How the path got stored
VideoInfoScanner.cpp:2140 — at scan time, when a playlist is known for a disc image, the scanner rewrites the item's path to the playlist:
Codeif (playlist > -1 && (IsBlurayDiscImage(path) || IsBDFile(path) || …)) { path = URIUtils::GetBlurayPlaylistPath(path, playlist); pItem->SetDynPath(path); }So any disc scanned into the library with a resolved playlist loses the chooser under AUTO from then on. That's the behaviour change he noticed, and it's a consequence of the playlist-resolution work rather than of the menu code.
What he can do today
- Set disc playback to Simple menu — forceSelectionAtStart && isBlurayPath && atStart then returns SHOW_SIMPLE_MENU, so the chooser comes back for library items too.
- Or Show Blu-ray menu, which he's already found works.
The decision for you
This is a design question, not a defect: under AUTO, a library item with a known playlist currently has no route to the disc menu at all — not via a chooser, and (as he notes) not from the OSD either. Options as I see them:
- Let AUTO return SHOW_SIMPLE_MENU for isBlurayPath at start when HasMenuSupport() — restores his chooser, at the cost of a dialog on every library play of a menu-capable disc.
- Leave AUTO alone and keep the OSD "DVD Menu" action available whenever the disc supports menus — closer to what he expected, and no extra dialog.
- Do nothing: AUTO behaving as documented, and he uses Simple menu.
I'd lean to (2) — it addresses the actual complaint without adding a prompt to every play — but it's your call, and I'll implement whichever you pick.
So the important point is whether it's been scanned to library. library items line direct to feature, not menu.
I'll probably go with (2) so you get back the option of playing the menu.
But wait wait wait!!! I Think I found it: Just went through all the projector settings again and switched "HDMI LINK" ON. And boom! Not one single glitch since then (the movie has been playing for 10 minutes now - fingers crossed...). Why this?
I'm afraid this still doesn't make any sense. "HDMI LINK" is CEC. No obvious reason why it would affect the glitches.
On Claude's suggestion above: "Worth telling him he can confirm it in seconds: play the same film from Files rather than the library and the chooser comes back, because the path is then the ISO."
I did try this, and it still goes directly to the main title, no chooser. Also rebooted Pi, and same behavior. But, with the other ISO's, the IMAX one and 'Best of 3D' compilation, I get chooser.
Can you provide a debug log when playing from files? Claude would appreciate it.
popcornmix, for my edification, is the MVC decoder in your MVC solution ffmpeg and s/w based, and not h/w and Broadcom SoC-based? Asking because w/RPi3, believe it was h/w based.
Also, I personally don't use menus for 99% of my BD/UHD rips. Only have a couple of full ISO's for those that have a lot of playlists and/or separate titles, where having menus do help vs. getting in the way of getting quickly/directly to the main feature.
And this may be an apples vs. oranges comparison, but have been helping test 3D/2D ISO's w/menus with @avdvplus' latest R10 CE test builds, and they're working perfectly. Took him numerous attempts, but he finally did it. If helpful, I can ask him to share any dev insights here that might help w/your LE test builds.
Just tested the same movie (A Turtle's Tale: Sammy's Adventures) in vobs clip above w/Pi5. Played well on my end w/XGIMI Titan Noir Max. Watched about 30 mins.
Software/ffmpeg based decoder. Which I have attempted to upstream.
That route was chosen as Pi5 doesn't contain the hardware MVC (or H.264) decoder and its CPU is more than fast enough.
Sure, any "lessons learned" with menus would be useful.
vobs' issue is hard to understand. Nothing makes much sense. Seems to be a fault of the projector, but it presumably only applies to some configuration. vobs Do you have any picture processing options on projector? (e.g. motion smoothing, de-noising etc? I'd be tempted to turn everything off). If you can configure the input as a PC rather than a BluRay that often disables processing to reduce latency.
Useful test. In settings/video can you change "PRIME Render Method" to "EGL"
The performance will be worse, so not the fix, but an interesting data point if that fixes the glitches.
When you say you've seen the glitch when paused, I assume you mean you've captured a static instance of the glitch (rather than it continues to glitch while paused)? With paused glitch on screen, can you enable/disable 3d mode on projector and report?
"ps: ChatGPT told me that a "HDFury Communicator Dr. HDMI 4K" may solve this issue - should give it a try?" - No, it won't fix this.
A debug log file will confirm if stereoscopicpreferframepacking setting has been read correctly.
If you say you can watch a SBS video without glitches when 3D mode is disabled on projector, but it starts glitching when you press the 3D button on projector (which the Pi will have no knowledge of), it's hard to see how this is the fault of the Pi.
I would bet it's not a MVC decode error, since:
1. the same MVC files do play perfectly on the Zidoos I have (Z9X Pro and Z20 Pro)
I'm talking about the ffmpeg based MVC decoder (which is a new thing). The Zidoo is using a completely different (likely hardware) decoder.
I don't think it's possible for it to be in the actual hdmi output path (it has no knowledge of a frame from 10 frames in the future).
It's going wrong either in the decoder, or somewhere in the render path of kodi. But that wouldn't be affected by the type of display/projector used.
If you add this advancedsettings.xml file:
<advancedsettings>
<video>
<stereoscopicpreferframepacking>false</stereoscopicpreferframepacking>
</video>
</advancedsettings>
it will disable framepacking and output as half-sbs (or half-tab). Is that any different?
May be useful to disable 3d mode and look at the two halves - do both have a consistent glitch or just one eye?