Re: [hatari-devel] Regression: Lethal Xcess fullscreen

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 28/05/2014 18:53, Matthias Arndt a écrit :
> Am 28.05.2014 12:01, schrieb Nicolas Pomarède:
> > Hi,
> >
> > Lethal Xcess has a calibration routine at start to compute values to
> > remove top border and this calibration is rather short.
> > Result is rather random, it works often in STF mode and works
> > sometimes in STE mode.
> But it worked in the past....

So ? Many things were also broken in the past and are now fixed. And 
many programs worked in the past because emulation was broken. The fact 
that a program worked before doesn't mean emulation was good, it 
sometimes means that emulation was not good enough to trigger the 
correct result.

> > I'm planning to add a quick fix for this, but the root of the problem
> > is much more complicated at the MFP level (will wait for after 1.8)
>
> I prefer a quick fix. A playable game (and it was as playable as on my
> STFM and on the STE in older Hatari versions) is more worth than that
> last cycle accurate emulation which breaks the game.

Let's not mix things, please, what "cycle accurate emulation" are you 
talking about ? I don't add things by pleasure, but because they were 
measured on real HW and are required, this can have temporary side effects.

Lethal Xcess uses a self calibration routine to remove top border, which 
is rather broken in my opinion (many games/demos never used such 
routines and they managed to work correctly).

This calibration routine requires a combination of MFP and video 
emulation. In previous version before 1.7 (note that it was already not 
always working in Hatari 1.7), the calibration worked because the 
emulation inaccuracy in mfp and video "compensated" each other, so it 
was mainly luck.

In 1.7, many video timings were fixed, they were required, they fixed 
many overscan and there's no reason to revert those changes as they are 
correct.

But now, calibration routine has a correct video emulation and a not 
precise enough MFP emulation : this results in the breakage. Only 
solution will be to fix MFP with even more accuracy.

In the meantime, I added a fix for this (but the calibration sometimes 
randomly fails, I think it's the same on real hardware)


Nicolas