2014-07-09, 21:38
2014-07-09, 22:45
Discontinue support for a LTS version of Ubuntu? Might as well just say Ubuntu isn't officially supported since the alternative is upgrading the OS every 9 months when all you use the HTPC for is a media center. Or, worse, 3rd party builds that we have no say or control over will become increasingly popular :\
Users are more understanding that we might give them credit for. I think most of them would be okay with a warning when using Precise rather than not being able to install it at all. Like in this case, we could have an Ubuntu-specific settings description for hardware decoding, so the warning is in the right spot.
Users are more understanding that we might give them credit for. I think most of them would be okay with a warning when using Precise rather than not being able to install it at all. Like in this case, we could have an Ubuntu-specific settings description for hardware decoding, so the warning is in the right spot.
2014-07-10, 00:23
LTS releases are supported for five years.
The 9 month cycle is not the LTS so you don't have to update.
The 9 month cycle is not the LTS so you don't have to update.
2014-07-10, 00:26
How Ubuntu LTS releases are supported by cannonical does not really matter - if they don't provide drivers for your new GPU ... xbmc cannot do anything about it.
The OP used 12.04.x on his HSW GPU and the libva-driver-intel are 2 years too old to make any hw decoding going ...
The OP used 12.04.x on his HSW GPU and the libva-driver-intel are 2 years too old to make any hw decoding going ...
2014-07-10, 07:51
(2014-07-09, 22:45)Ned Scott Wrote: Discontinue support for a LTS version of Ubuntu? Might as well just say Ubuntu isn't officially supported since the alternative is upgrading the OS every 9 months when all you use the HTPC for is a media center. Or, worse, 3rd party builds that we have no say or control over will become increasingly popular :\
Users are more understanding that we might give them credit for. I think most of them would be okay with a warning when using Precise rather than not being able to install it at all. Like in this case, we could have an Ubuntu-specific settings description for hardware decoding, so the warning is in the right spot.
What does "support" mean in this case? Providing some code which compiles without an error? No dev uses this old OS and we don't do any testing on it. I consider this as NOT supported.
2014-07-10, 08:01
Also note that other applications like vdr have no major upgrades on 12.04. Following this strategy we should not provide Gotham builds for 12.04.
2014-07-10, 08:06
OK I am not a dev or anything but I vote for continued precise packages. My (nVidia graphics chipset) systems work just damn fine on precise.
2014-07-10, 13:25
Gents, please, don't throw out the baby with the bathwater! The last time I upgraded from one Ubuntu LTS version into another, it went sour, and I had to work for hours on end to recover my data. My Ubuntu 12.04 still has a lifetime of 3 more years!
And to be clear: my original problem was not va-api not working. What I know now is that the non-working vaapi made the system fall back to software, and it worked fine. The real problem was that on the system where va-api appeared to be working, it was pretending to do deinterlacing, though it was not. Eventually I learned that I had to switch off vaapi support, to get a good picture.
Now then, why should I upgrade an old vaapi to something new, which I still need to inactivate, as at present "exactly no(!) intel hardware with xbmc mainline code" supports hardware deinterlacing? If that is impossible, any potential other advantage of vaapi makes no sense to me.
And to be clear: my original problem was not va-api not working. What I know now is that the non-working vaapi made the system fall back to software, and it worked fine. The real problem was that on the system where va-api appeared to be working, it was pretending to do deinterlacing, though it was not. Eventually I learned that I had to switch off vaapi support, to get a good picture.
Now then, why should I upgrade an old vaapi to something new, which I still need to inactivate, as at present "exactly no(!) intel hardware with xbmc mainline code" supports hardware deinterlacing? If that is impossible, any potential other advantage of vaapi makes no sense to me.
2014-07-10, 13:27
Eventually:: hw decoding on hsw uses 10% CPU usage where as multithread decoding of e.g. 1080p60 h264 contents warms up the machine by using a full one and a half core in sum?
2014-07-10, 16:30
"Eventually::" Agreed. And it will be much appreciated. But until then my guess is that several driver versions will come and go.
The issue here was that xbmc was fully aware that deinterlacing was not working (you pointed me to the log file, which says: DEBUG: VAAPI - deinterlace not support on this intel driver version), yet it left me with the impression that it was, by allowing to switch the option "Deinterlace video auto/on/off". But irrespective of my setting, it was always off, because it wasn't working, and wouldn't fallback to software!
xbmc was not giving any feedback - to a normal user - of that fact; instead it left me (and my guests, who made sneery remarks on the video quality; that hurts me too!) with a very poor visual experience. That was unneccesary.
People like me are nowhere near developer status, but will try to help with problems they come across, which then, like here, will needlessly take up my time and your time, when they are put on the wrong track by lack of feedback.
The issue here was that xbmc was fully aware that deinterlacing was not working (you pointed me to the log file, which says: DEBUG: VAAPI - deinterlace not support on this intel driver version), yet it left me with the impression that it was, by allowing to switch the option "Deinterlace video auto/on/off". But irrespective of my setting, it was always off, because it wasn't working, and wouldn't fallback to software!
xbmc was not giving any feedback - to a normal user - of that fact; instead it left me (and my guests, who made sneery remarks on the video quality; that hurts me too!) with a very poor visual experience. That was unneccesary.
People like me are nowhere near developer status, but will try to help with problems they come across, which then, like here, will needlessly take up my time and your time, when they are put on the wrong track by lack of feedback.
2014-07-10, 19:09
https://wiki.ubuntu.com/1204_HWE_EOL <- hehe, even Ubuntu realized it and wants their users to update
2014-07-11, 00:03
(2014-07-10, 07:51)FernetMenta Wrote: What does "support" mean in this case? Providing some code which compiles without an error? No dev uses this old OS and we don't do any testing on it. I consider this as NOT supported.
It's one thing if it is getting in the way of development, I could understand that argument, but this sounds like cutting off the nose to spite the face.
We all want XBMC to not be a pain in the ass, for developers and users, right? But if we try to use a blunt hammer for every situation like this, removing builds or forcing people to patch code to make their own builds, then we run the risk of being like XBMC for Android, which has a ton of PITA 3rd party builds.
Ask yourself, which would be better, telling people "you can use XBMC for Ubuntu 12.04, but hardware video decoding might not work on X or Y hardware." but still having them use builds we control
or
having to deal with some asshole's 3rd party build that breaks a bunch of other things, but the users refuse to use our builds because our builds require them to update the entire OS every 9 months. Or the number of users who refuse to update from old versions of XBMC continues to grow and grow. Look at the OP of this thread, he's not demanding anything, but he just wants to know why it didn't work when he thought it would.
Which is easier for you?
Like I said, as long as users know what the issue is then they're far more likely to be understanding than we might give them credit for.
2014-07-11, 00:11
(2014-07-10, 00:26)fritsch Wrote: How Ubuntu LTS releases are supported by cannonical does not really matter - if they don't provide drivers for your new GPU ... xbmc cannot do anything about it.
The OP used 12.04.x on his HSW GPU and the libva-driver-intel are 2 years too old to make any hw decoding going ...
I think everyone is okay with the fact that the older OS won't work with the newer hardware. The issue is punishing everyone that has a working setup, just because OP didn't see any documentation about what is supported in his situation, and then asked a question about it. We can fix documentation, both in XBMC in settings descriptions and in the manual, without having to drop support for 12.04.
I don't see how this is any different than explaining to users that XBMC for Mac OS X doesn't do HD audio passthrough, or that certain drivers for a motherboard might not work with specific OS's.
*shrug*
2014-07-11, 07:22
(2014-07-11, 00:03)Ned Scott Wrote:(2014-07-10, 07:51)FernetMenta Wrote: What does "support" mean in this case? Providing some code which compiles without an error? No dev uses this old OS and we don't do any testing on it. I consider this as NOT supported.
It's one thing if it is getting in the way of development, I could understand that argument, but this sounds like cutting off the nose to spite the face.
We all want XBMC to not be a pain in the ass, for developers and users, right? But if we try to use a blunt hammer for every situation like this, removing builds or forcing people to patch code to make their own builds, then we run the risk of being like XBMC for Android, which has a ton of PITA 3rd party builds.
Ask yourself, which would be better, telling people "you can use XBMC for Ubuntu 12.04, but hardware video decoding might not work on X or Y hardware." but still having them use builds we control
or
having to deal with some asshole's 3rd party build that breaks a bunch of other things, but the users refuse to use our builds because our builds require them to update the entire OS every 9 months. Or the number of users who refuse to update from old versions of XBMC continues to grow and grow. Look at the OP of this thread, he's not demanding anything, but he just wants to know why it didn't work when he thought it would.
Which is easier for you?
Like I said, as long as users know what the issue is then they're far more likely to be understanding than we might give them credit for.
I am just telling the truth. Fact is that we don't test on Precise nor recommend it. Providing a build is pretending pretending to support it. Users stay with 12.04 because they want "stable" but they get the opposite in regard to XBMC.
