Anyone else experiencing problems with Widevine (buffering, stuttering)? I noticed the first problems somewhere at the end of last year. Not sure which version exactly. But I found that reverting to 4.10.2252.0 solved the problem. Anything after that is a no go. I was hoping that maybe this was CE related and that the problem might disappear with 19.4, but no.
I’m running CE on a a Tanix TX3 S905X3 4GB and on a TX3 S905W. Both devices have this problem.
I saw this problem mentioned on the OMS forum as well and also the same solution: reverting back to 4.10.2252.0.
It happened to me too in a K1 pro. The solution was to revert to an older version, maybe the one you mentioned, and refusing to upgrade since then.
Same here.
Of course :-). I posted this because I didn’t see it mentioned on this forum yet. Is this problem under investigation already?
Vasco
16 March 2022 09:28
5
You can see here it’s not CoreELEC only related, it’s ARM related:
opened 02:04AM - 14 Mar 22 UTC
Hi there
This version of the CDM has pretty bad issues based on user experien… ces and my own testing. About once a week since mid January users are either opening new issues on the inputstream.adaptive repo or making new forum posts. Unfortunately the cause of their problem isn't very obvious so users are having a hard time getting help.
Is it possible to either:
a) provide a way to be able to retrieve an older image containing 4.10.2252.5
b) prevent upgrading from 4.10.2252.5 to 4.10.2391.0 (or newer versions until we can confirm they function correctly)
c) both of the above
To give a little more detail, printing to log before and after this statement https://github.com/xbmc/inputstream.adaptive/blob/ecccdae020c5e80ff66200664e54bbee990e6474/wvdecrypter/wvdecrypter.cpp#L1269 takes ~5ms on my RPi3b with 4.10.2252.5, and closer to 50-70ms with 4.10.2391.0 . This alone renders streams unwatchable due to the inability of the decoder to provide enough data to keep Kodi's buffer full.
Additionally while trying to implement support for CBC decryption in IA the newer CDM version would crash during routine function calls (CDM timer) for seemingly no reason, the only difference in conditions being the difference decryption scheme and multiple CDM 'sessions'. All is fine in 4.10.2252.5
I can only guess that somehow the TLS relocation stuff/googles compiler somehow affects this or they've released a buggy product. FWIW the libwidevinecdm0 package in Raspberry Pi OS is still shipping 4.10.2252.5
We’re following the issue but there’s nothing we can do. Nevertheless, it’s working perfectly with 4.10.2252.5 so not a real problem atm.
Have you tried one of the 2 tests posted by horstle over at Github?
Nope, didn’t know about those. I looked up Horstle’s git, but can’t find what you’re referring to. Have you got an URL?
Nevermind, found them on the page mentioned by Vasco.
system
15 June 2022 16:24
8
This topic was automatically closed 91 days after the last reply. New replies are no longer allowed.