[mythtv] Qt 5 support (Peter Bennett)

[David Hampton]

On Sun, 2026-08-23 at 21:05 +0800, James Linder via mythtv-dev wrote:
> 
> > On 8/17/26 19:00, David Hampton via mythtv-dev wrote:
> > > On Fri, 2026-08-14 at 15:36 -0400, David Hampton wrote:
> > > > I still regularly compile and test both Qt5 and Qt6, but it
> > > > doubles
> > > > the amount of work.? Two Linux builds, eight android builds
> > > > (four
> > > > architectures times two versions of qt), etc. etc.
> > > Is anyone else running Qt6 based images?
> > > 
> > > David
> > > 
> > > 
> > I have been running QT6 since April. No problems, however I do not
> > use 
> > mythfrontend.
> 
> I’ve changed my mind over the last 2 days:
> 
> I have a headless backend
> 2 linux front ends
> 1 mac frontend
> 
> Watching this thread I decided to try Qt6
> After fighting with cmake for a day Jim A suggested I go from V35 to
> master.
> I was unable to get Qt6 to build on the backend, I eventually got
> master on Qt5 running
> John Hoyt’s mac frontend works, is smooth, but has crashed once.
> I spent a dat with gemini and eventually got Qt6 master built.

I'm sorry to hear you had problems getting qt6 builds to work.  My 
three internal SuSe builders (15.6, 16.1, tumbleweed:20260802) have all
been building qt6 images since May.

> I can, by touch, playback a recording, I hear it but the display is a
> ghost display. ie it picked pixels from the desktop it was started
> on. I can send to another desktop, complete with first desktop part
> image.

I haven't ever seen or heard of this ghost display problem with qt6
builds.  It does sound similar to
https://github.com/MythTV/mythtv/issues/1334, but in that issue the 
screen is never drawn at all.  Do you think its the same problem?  

> On this machine 1080 display on a 4K display. I have tried lots but
> have no display! I am building Qt5 version as we speak

No display at all definitely sounds like
https://github.com/MythTV/mythtv/issues/1334.  

> My software is a little unusual, but not outrageous: openSuSE
> slowroll and leap 16.0 running kde plasma. V35 and Qt5 was (a) easy
> with cmake (b) rock solid.
> 
> (A big slice of my cmake problem stemmed from finding system-build
> files for ffmpeg and cmake refusing (despite lots of coaxing) to
> build ffmpeg)

Are you sure cmake wasn't just detecting that its already built FFmpeg
and was reusing the libraries from last time?  I had added
optimizations to not continuously rebuild libraries that never change,
which using configure/make were blindly rebuilt for every build.  (And
now with the new cmake staging directory patch, if you use the staging
directory [the default] then cmake will also blindly build these
libraries every time as well.)  Problem solved?  :-)

If you look in the output of the MythTV portion of the build you should
see something that looks like this:

-- Checking for module 'libmythavcodec'
--   Found libmythavcodec, version 62.28.100
-- Checking for module 'libmythavdevice'
--   Found libmythavdevice, version 62.3.100
-- Checking for module 'libmythavfilter'
--   Found libmythavfilter, version 11.14.100
-- Checking for module 'libmythavformat'
--   Found libmythavformat, version 62.12.100
-- Checking for module 'libmythavutil'
--   Found libmythavutil, version 60.26.100
-- Checking for module 'libmythswresample'
--   Found libmythswresample, version 6.3.100
-- Checking for module 'libmythswscale'
--   Found libmythswscale, version 9.5.100

This is the cmake build finding its own version of the FFmpeg
libraries.  They are intentionally named differently from the standard
FFmpeg libraries to guarantee that the wrong libraries aren't found.

This bit of cmake code is the same regardless of whether you're
building qt5 or qt6 based images.

David