Mailing List Archive

Experience with upgrade of Xubuntu 18.04 to 20.04
I recently upgraded one of my Xubuntu 18.04 MythTV v31-fixes machines to
the latest 20.04 Xubuntu release. For the most part, the upgrade was
painless but there were a few issues I had to navigate through afterwards.
The issues were not specific to mythtv but impacted the functionality as a
frontend none-the-less.

The involved frontend uses intel VAAPI graphics and a MCEUSB infrared
receiver via kernel driver and ir-keytable for receiving remote control
input.

*First Issue:*
After the upgrade, I found mythfrontend would immediately segfault when
starting. After troubleshooting, I discovered that this issue was actually
more broadly affecting any opengl application as glxgears would also show a
black screen and eventually segfault. Fortunately, glxgears provided
reference segementation fautl in iris_dri.so in the associated lines in
syslog whereas with mythfrontend oddly showed nothing in syslog or
mythfrontend.log associated with the segfault.

I use a xorg.conf file on this machine to manually specify an edid file and
to also force the preferred mode to 1920x1080p @ 54.94hz for judderless
playback with NTSC television. I found that if any xorg.conf file is
present (even with nothing more than a simple DEVICE section present
specifying intel driver and no other options), then a segfault occurs in
opengl applications. Apparently, the latest Iris (Gallium3D) driver is the
default in Mesa 20.0 and some bug in the Iris driver impacts this
combination. In this case, a workaround to allow use of a xorg.conf with
intel graphics and allow opengl programs to run is to preface the startup
command like so:
$ MESA_LOADER_DRIVER_OVERRIDE=i965 mythfrontend
One can also just remove the xorg.conf file altogether losing any
customizations and then opengl programs will run without issue using the
intel driver.

*Second issue:*
ir-keytable has undergone several changes and now uses a different format
for keymap files. Keymap files residing in /etc/rc_keymaps now have a .toml
extension and the required layout and syntax have changed. Therefore, if
you have done any customization of /etc/rc_maps.cfg or have generated any
custom keymaps in /etc/rc_keymaps, those will have to be updated to work
with the latest required format. I can elaborate more on the specific
changes but would just suggest reading the man files or looking at the
other packaged examples as a start.

My guess is that if using a default installation with little to no
customization, then the upgrade will be seamless. You will only have
add/enable the 20.04 mythbuntu repo post upgrade and update.

I haven't performed the 18.04 --> 20.04 upgrade on my combined
frontend/backend that uses nvidia graphics and lirc for remote control.
When I get around to it, I will share anything noteworthy that I encounter
with that machine as well. My BE/FE has many other duties and I will
probably hold off on that one for a while.

-Tim
Re: Experience with upgrade of Xubuntu 18.04 to 20.04 [ In reply to ]
On Mon, 11 May 2020 10:41:43 -0400, you wrote:

>I recently upgraded one of my Xubuntu 18.04 MythTV v31-fixes machines to
>the latest 20.04 Xubuntu release. For the most part, the upgrade was
>painless but there were a few issues I had to navigate through afterwards.
>The issues were not specific to mythtv but impacted the functionality as a
>frontend none-the-less.
>
>The involved frontend uses intel VAAPI graphics and a MCEUSB infrared
>receiver via kernel driver and ir-keytable for receiving remote control
>input.
>
>*First Issue:*
>After the upgrade, I found mythfrontend would immediately segfault when
>starting. After troubleshooting, I discovered that this issue was actually
>more broadly affecting any opengl application as glxgears would also show a
>black screen and eventually segfault. Fortunately, glxgears provided
>reference segementation fautl in iris_dri.so in the associated lines in
>syslog whereas with mythfrontend oddly showed nothing in syslog or
>mythfrontend.log associated with the segfault.
>
>I use a xorg.conf file on this machine to manually specify an edid file and
>to also force the preferred mode to 1920x1080p @ 54.94hz for judderless
>playback with NTSC television. I found that if any xorg.conf file is
>present (even with nothing more than a simple DEVICE section present
>specifying intel driver and no other options), then a segfault occurs in
>opengl applications. Apparently, the latest Iris (Gallium3D) driver is the
>default in Mesa 20.0 and some bug in the Iris driver impacts this
>combination. In this case, a workaround to allow use of a xorg.conf with
>intel graphics and allow opengl programs to run is to preface the startup
>command like so:
> $ MESA_LOADER_DRIVER_OVERRIDE=i965 mythfrontend
>One can also just remove the xorg.conf file altogether losing any
>customizations and then opengl programs will run without issue using the
>intel driver.
>
>*Second issue:*
>ir-keytable has undergone several changes and now uses a different format
>for keymap files. Keymap files residing in /etc/rc_keymaps now have a .toml
>extension and the required layout and syntax have changed. Therefore, if
>you have done any customization of /etc/rc_maps.cfg or have generated any
>custom keymaps in /etc/rc_keymaps, those will have to be updated to work
>with the latest required format. I can elaborate more on the specific
>changes but would just suggest reading the man files or looking at the
>other packaged examples as a start.
>
>My guess is that if using a default installation with little to no
>customization, then the upgrade will be seamless. You will only have
>add/enable the 20.04 mythbuntu repo post upgrade and update.
>
>I haven't performed the 18.04 --> 20.04 upgrade on my combined
>frontend/backend that uses nvidia graphics and lirc for remote control.
>When I get around to it, I will share anything noteworthy that I encounter
>with that machine as well. My BE/FE has many other duties and I will
>probably hold off on that one for a while.
>
>-Tim

You would have had to force the upgrade, which is not recommended. The
upgrade path from 18.04 LTS to 20.04 LTS will only be available when
20.04.1 is out (June?). There are a lot of web sites out there that
blithely tell you if the do-release-upgrade command does not work,
just add the -d option to make it work. They completely fail to
explain that the upgrade is not working because it is not supposed to
be working yet. Adding -d allows you to upgrade on the basis that
20.04 is a development version for upgrading 18.04 LTS, not a release
version. So anyone doing that should be prepared for problems. That
said, I have not heard of any show stopper problems from anyone who
has forced the upgrade, unlike what happened with 16.04 LTS to 18.04
LTS before the .1 update was out.
_______________________________________________
mythtv-users mailing list
mythtv-users@mythtv.org
http://lists.mythtv.org/mailman/listinfo/mythtv-users
http://wiki.mythtv.org/Mailing_List_etiquette
MythTV Forums: https://forum.mythtv.org
Re: Experience with upgrade of Xubuntu 18.04 to 20.04 [ In reply to ]
On Mon, May 11, 2020 at 11:18 AM Stephen Worthington <
stephen_agent@jsw.gen.nz> wrote:

>
> You would have had to force the upgrade, which is not recommended. The
> upgrade path from 18.04 LTS to 20.04 LTS will only be available when
> 20.04.1 is out (June?). There are a lot of web sites out there that
> blithely tell you if the do-release-upgrade command does not work,
> just add the -d option to make it work. They completely fail to
> explain that the upgrade is not working because it is not supposed to
> be working yet. Adding -d allows you to upgrade on the basis that
> 20.04 is a development version for upgrading 18.04 LTS, not a release
> version. So anyone doing that should be prepared for problems. That
> said, I have not heard of any show stopper problems from anyone who
> has forced the upgrade, unlike what happened with 16.04 LTS to 18.04
> LTS before the .1 update was out.
>


Indeed, I forced the upgrade and only share in case anyone else likes to
live on the edge.

An alternative workaround to specifying the i965 driver is to add the
following to the device section of xorg.conf:

Option "DRI" "3"

The problem appears to be a known issue with intel UHD500 / latest Mesa and
intel driver (
https://gitlab.freedesktop.org/xorg/driver/xf86-video-intel/-/issues/193).

Hopefully will be fixed soon.
Re: Experience with upgrade of Xubuntu 18.04 to 20.04 [ In reply to ]
>
> An alternative workaround to specifying the i965 driver is to add the
> following to the device section of xorg.conf:
>
> Option "DRI" "3"
>
> The problem appears to be a known issue with intel UHD500 / latest Mesa
> and intel driver (
> https://gitlab.freedesktop.org/xorg/driver/xf86-video-intel/-/issues/193
> ).
>
>
If using the default modern kernel modesetting intel driver rather than the
legacy intel xorg driver package (xserver-xorg-video-intel), the issue with
opengl applications is altogether avoided. However, without xorg.conf
options provided by the legacy driver, I had to rework my approach for
ensuring the system always boots to the correct resolution / refresh rate
in XFCE whether receiver/television are on or off and for enabling RGB
Broadcast Full for proper colorspace once x display is available.

I have also discovered another issue on 20.04 with this hardware (Apollo
Lake J3455 - Intel Gen 9 graphics) when using VAAPI deinterlacing. The
installation now defaults to the iHD driver rather than i965 and the iHD
driver does not correctly expose all of the supported VAAPI functionality
as available.

This was causing validation of the VAAPI deinterlacers (motion_adaptive
motion_compensated etc) in mythfrontend to fail according to the
mythfrontend.log. It wasn't obvious in the video without inspecting the log
since myth will fall back to opengl deinterlacing gracefully. The issue can
also be verified by comparing the output of vainfo using iHD vs i965.

To work around, you must force the use of the i965 driver by invoking
mythfrontend as follows:
$ LIBVA_DRIVER_NAME=i965 mythfrontend

Take away is anyone on 20.04 using VAAPI for hardware deinterlacing with
the iHD driver might want to verify whether it is working or not for their
combination of hardware and driver version. Possibly related to the issue
under discussion here (https://github.com/intel/media-driver/issues/804) so
hopefully will get fixed in a future iHD driver release.