Do we need an SDL3 port?

[GitMensch]

While that isn't out yet, at least the header folder changed, so the SDL2 port can likely not be taken "as is".

For more details see the repo at https://github.com/libsdl-org/SDL_ttf

[Bill-Gray]

I've seen occasional references to SDL3 in recent months, but haven't investigated.

If the changes are as minor as changing header folders, we should get by with a -DUSE_SDL3 switch rather than making another port with essentially identical code. (There is already a lot of overlap between SDL1 and SDL2... I've thoughts of moving bits and bobs to the common directory. VT and the framebuffer/DRM ports have lots of overlap, too, and you'll see that I was able to avoid code duplication for the most part.)

[GitMensch]

There is already a lot of overlap between SDL1 and SDL2... I've thoughts of moving bits and bobs to the common directory.

I totally agree that this would be a good idea, should we move that out to a separate issue?

[Bill-Gray]

I'm only reluctant to bring it up as a separate issue because it's likely to at least be some time before it leaves the 'thinking about it' stage. I would do it to make the code easier to maintain; I can't think of any way in which it would add functionality or fix anything.

The most likely motivator for it would be a fix to SDLn (or extension to SDL3) where we say to ourselves : this bit of code needs adjustment in two places; let's pull it into the common directory and fix it once. So I'm not in a great rush to do it.

On a similar level, we have code to read in PSF and some other bitmapped fonts in both the DOSVGA and Linux framebuffer/DRM ports. Ideally, those would be pulled out and put into common to form a bitmapped font subsystem, and that subsystem would then replace the current bizarre font format for SDL1 and SDL2 "narrow" (8-bit character) builds, where the "font" is stored as a Windows .BMP file. That would do a lot more to simplify the code, and would actually add some capabilities in all four platforms (and mean we would have one set of bitmapped font routines instead of three; the DOSVGA and framebuffer/DRM ports have different ways of handling PSF, which in turn bear no real relationship at all to the SDLn .BMP-formatted fonts.)

But again... while it's been in my mind for a while now, basically since I wrote the framebuffer/DRM port, I've obviously not been rushing to actually do it.