[mythtv-users] mac build

[Mark Kendall]

On Mon, 1 Feb 2021 at 11:51, James Linder <jam at tigger.ws> wrote:
> > On 1 Feb 2021, at 7:22 pm, John Hoyt <john.hoyt at gmail.com> wrote:
> >
> > Hi Mark, as it stands, the initial bug is fixed - James can now use the cutlist editor (for the most part) and I've resolved my packaging error that caused his metadata grabber issues.
> >
> > Unfortunately, since backporting the MythButtonUIList fix that John P suggested, James is now experiencing new intermittent segfaults with the cutlist editor. I suspect it's another QT5 related issue (have nothing to base that belief on...).  Any suggestions or thoughts on related fixes in master (James is running fixes/31) would be appreciated.
> >
> > It also appears from James' report that macOS is less robust to qt5 related crashes and segfaults whereas Linux (not sure what flavor) does not.
>
> Mostly SuSE Leap 15,2, lubuntu (lxqt) 20.10, xubuntu 20.04 and a sprincle of Debian Buster. Not tried on buster, all rest are fe with suse as be.
>
> > I'm wondering if something could be added to or modified in the code to help macOS handle whatever the issue is and avoid segfaults.
> >
> > I have a development build fired up and can easily patch it / add print statements to help in the bug finding process.
> >
> > If you believe it to be on the qt5 compile side, I can also look into that.  I'm now way too familiar with how qt5 gets compiled in macports.

I would be genuinely surprised if it was a Qt issue - but obviously it
is a possibility.

The only thing I can really suggest is getting a debug build to James
that can hopefully produce a useful backtrace. I forget whether macos
will produce a backtrace without prompting - so may need to run
mythfrontend through lldb and/or gdb.

> Mark before edit on HD resulted in E, <space> one or two moves (<rt> or <left>) and crash.
> Now I've done cutlist on half a dozen videos and have prevoked core dump (sig 11) twice.
> Posted the trace but it just said line 15 /Applications/Mythfrontend.app/Contents/MacOS/mythfrontend
> which is not far from 'something is wrong somewhere'

Yup - 'houston we have a problem'. Not sure there is much point trying
to guess what might be going wrong - short of a full backtrace, maybe
just enable as much as logging as possible:)

mythfrontend -v all --loglevel=debug --noupnp

Regards
Mark