Re: [hatari-devel] Shalll we remove deprecated features like SDL 1.2 now?

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 01/01/2021 à 10:07, Thomas Huth a écrit :
>   Hi,
>
> A happy new year 2021 to everybody! Let's hope that it gets better
> than 2020 (with regards to Covid-19) - though from a Hatari perspective
> 2020 has been pretty good already, with two new releases and lots of
> improvements.
>
> Now that the 2.3 release has been done, I'd like to start a discussion
> about removing some features that have been marked as deprecated. I
> think the most "important" one is currently SDL 1.2, since this version
> of the library is not maintained since 2013 anymore, so most (all?)
> developers use 2.0 these days, and the countless #ifdefs in the code
> can be a real pain if you have to make changes in these areas.
>
> Is anybody still actively using SDL 1.2, without the possibility to
> upgrade to SDL 2.0 ? If so, on which system? Why is SDL 2.0 not an
> option there yet?

Hi

as these items were discussed, I'm still OK for removing both pieces of 
code.
Maybe the lack of SDL2 on some systems (Raspberry or derivatives ?) 
could be a problem, but at one point we can't have hatari evolve and 
take less time to maintain with 2 SDL versions.

For people really not having SDL2, they can use hatari 2.3.1 which 
should really emulate 99% of STF/STE programs. When it comes to 
TT/Falcon, small systems like this are certainly not powerful enough at 
the moment, so it seems normal that more powerful systems (linux 
distros, windows, macos) already have a SDL2 version.



> Another item that I'd like to see removed is the old UAE core. It
> also causes double maintenance in case you have to do changes in a
> related area (e.g. when I did the rewrite of the GEMDOS HD Pexec code
> during the 2.3 development cycle). I can still see that there is a
> benefit of keeping the old UAE core around, e.g. it's a little bit
> faster than the WinUAE core (but of it's of course also less accurate),
> but is anybody really still using it, so that it's worth the effort of
> keeping it in the source tree?

I'm also for removing the old CPU core ; as for the small speed gain 
with old cpu core, I know one or 2 places in the new cpu core that can 
be optimised for the ST case when using cycle accurate mode (because 
they run cycles one by one for the more complex Amiga's DMA, which is 
not needed for ST).
Same as above, people with not fast enough pc can still use hatari 2.3.1 
in the future and old cpu core, but the old cpu core is really adding 
lot of code to hatari and also preventing to add some more advanced 
winuae's cpu core feature (because there's no equivalent with the old 
cpu core)

> For the other deprecated things listed in doc/release-notes.txt I think
> most people won't be affected much anyway, e.g. all modern distros ship
> with python 3, so nobody should be affected if we remove support for
> python 2 (which is EOL anyway) in the future.


Nicolas