Re: [hatari-devel] Re: Problems with GEMDOS drive support

["Roger Burrows" <anodyne@xxxxxxxxxxxx>]

On 20 Sep 2018 at 18:25, Thorsten Otto wrote:

> On Donnerstag, 20. September 2018 18:07:48 CEST Roger Burrows wrote:
> > Obviously.  
> 
> Looking at the source it can't find any indication for that.
> 
Mea culpa, you are quite right.

> >So does EmuTOS
> 
> But only if no twisted cable is detected (which hopefully is the case for 
> Hatari) or if the rootsector contains a swapped MSDOS checksum. Was the image
> previously formatted on a PC?
> 
No, as explained earlier, it was done by byteswapping an image that works on 
the ACSI emulation.  And the reason it was byteswapped was that the ACSI image 
did not work as-is on Hatari IDE with Atari TOS or EmuTOS.  But it works when 
byteswapped.

So I guess my conclusion must be that, for IDE emulation, Hatari byteswaps the 
emulated disk itself before presenting it to the operating system.  This looks 
like what happens in ide_data_readw()/ide_data_readl() etc.

So the net is the same: the previous piece of code must be fixed to assume a 
byteswapped IDE image, because Atari TOS & EmuTOS will not work with a 
non-byteswapped image.

Of course, in theory you could change ide_data_readw()/ide_data_readl() etc to 
NOT byteswap the data, then you would not need to change the code that figures 
out the correct letter for the GEMDOS drive.  But that would break existing IDE 
images used in Falcon emulation.

Roger