Re: [hatari-devel] Release time for 1.8.1 ?

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 25/05/2015 21:45, Eero Tamminen a écrit :
> Hi,
>
> I don't think next release version discussion got concluded.
> I still propose it to be 1.9.0.
>
> Fixing Hatari snapshot handling before release is too
> large change, but IMHO we need to do it before next release.
> It can easily be done in such a way that previous Hatari
> versions reject new versions correctly.

Hi

I also think 1.9.0 can be used, and after that we break/change the 
snapshot format to sthg more modular with backward compatibility in mind 
for Hatari version released after 1.9.

ie : any version > 1.9 should be able to load any memory snapshot made 
after 1.9, using default values when needed or displaying a warning to 
the user about some missing data, or ignoring unknown value.

- Hatari 1.15 should be able to load and use snapshot from 1.10, adding 
default value if needed

- Hatari 1.10 should be able to load and use snapshot from 1.15, adding 
default value if needed and ignoring unknown values



> > Anybody on the list can do testing (only SDL2 item
> > needs specific build), it doesn't need to be Hatari
> > developers.
>
> I think SDL2 window position issue can be left after
> release if Thomas doesn't have time to look into it
> as SDL2 support is new and isn't enabled by default.

See my other mail, I might be wrong but I'm not sure this is an issue 
that can be dealt easily at the application level, looks like a window 
manager problem to me.

> > for MIDI,
> > running different MIDI programs and seeing whether
> > there are regressions/problems.
> >
> > For filename support, it's enough to try Hatari on
> > directory with weird file names, and see whether programs
> > (e.g. replacement file selectors) behave with them as expected.
>
> If somebody on list could look at these,
> it would be appreciated!

I tested MIDI with the main MIDI programs used on Hatari ('Notator', 
'M', 'Realtime') and other people also reported the result is now 
similar to real STF, so I think we don't have regression here, only 
better support.

> Instruction cycles should be now checked to be OKish,
> but I would like data cache stuff to be fixed too.
>
> If it doesn't get fixed, new profiler support for
> data cache would need to be reverted before release
> (memory needed for profiling 256 MB TT-RAM is
> too much to keep useless data members).
>
> If it gets fixed, I'll still need to update Hatari
> manual a bit and remove debug warnings.

But it was not confirmed that data cache counting was wrong. You saw 
more hits/misses that you expected, but Doug didn't confirm those 
instructions were wrong.
Unless we have some proper testing program to compare Hatari with real 
Falcon, and some possible confirmation that sthg is wrongly counted, I 
don't think it's the case.

Nicolas