[mythtv-theming] Font face naming differences on different platforms (was Re: What fonts are guaranteed to be on a Linux box?)

[Michael T. Dean]

On 08/22/2013 03:11 PM, Yeechang Lee wrote:
> That said, I am not sure if the cause is necessarily OSD/theme-related
> per se, or even with MythTV at all. I use two 0.26 frontends, one
> Linux and one OS X. With the stock/fallback osd_subtitle.xml, both the
> FreeMono and Droid Sans Mono fonts behave badly on the OS X frontend
> (whether Jean-Yves' standalone or the MacPorts one). With FreeMono,
> some punctuation marks are either partially obscured (question mark)
> or don't appear at all (apostrophe, period). With Droid Sans Mono,
> they appear but parts are cut off, as are other letters, notably the
> capital W and capital O.
>
> Similar behavior occurs with other free mono fonts I've tested,
> Liberation, DejaVu, and Ubuntu. Only the proprietary Bitstream Vera
> Sans Mono--that Liberation and DejaVu are clones of--seems to appear
> correctly on OS X, but of course we can't distribute it. Since, as I
> mentioned, this only seems to occur on OS X, it's quite possibly an
> Apple-specific font-rendering issue.

Oops.  This reminds me that k-man on irc noticed that (sometimes? for 
some fonts? always?) the font face name is exposed differently on his 
Mac OS X system than on his GNU/Linux systems:

http://irc.mythtv.org/ircLog/channel/3/2011-08-15

Try putting the .otf, .ttf, or .ttc file(s) for the fonts into your 
theme directory (at least temporarily) and running the frontend on both 
platforms with -v gui,file --loglevel debug and you should see 
information like:

In file '<filename>', found font(s) <face name>'
or
Unable to register font(s) in file '<filename>'

Do you get different face names or errors or anything?  The log messages 
will show exactly what Qt gets from the underlying platform font 
managers.  If they're different, we'd have to fix this with code (and it 
wouldn't be that much code--like I have no excuse for not having 
finished it in 2 years :( ).  I still think the approach he asked for 
(and I described) in the IRC conversation is the best 
approach--basically a fallback face name approach, similar to that used 
by CSS.

I'd also like Stuart Morgan's opinion on the approach before we put it in.

Mike