[Coco] HDBDOS - RETRIEVE FIXED!!

[Tormod Volden]

On Mon, Jan 19, 2015 at 4:26 AM, Robert Gault wrote:
> Chad H wrote:
>>
>> Yes I know right!!   I played around with that section of code ALL DAY
>> yesterday!! You have to adjust this section every time you change the length
>> of the code before it...
>>
>> * HARD DISK DRIVER
>>
>>                 org       MAGICDG+$193D       It all starts here!
>>
>> * Indirect Jump Table ( jsr [$MMMM] )
>>
>>                 fdb       DISKIO              universal hard disk input /
>> output
>>                 fdb       SETUP               Setup command packet
>>                 fdb       BEEP                Make a beep sound
>>                 fdb       DSKCON2             DSKCON Re-entry
>>
>> HDBHI          fcb       $00                 HDB-DOS Offset hi-byte
>> HDBLO          fdb       $0000               HDB-DOS Offset lo-word
>>
>> PORT           fdb       DATAADDR            Interface base address
>> CCODE          fcb       TDELAY              IDE: startup delay / SCSI:
>> wakeup delay
>>
>> That "MAGICDG" Offset, if not right, made it freeze instantly.     If you
>> use a hex editor on the original HDBDOS.ROM file for your computer you will
>> see at the offset they have there this long string of ff's with a data
>> marker at that address. (ff ff ff 99 da ad da b3 ...."   It's that first
>> 'da' that you have to find the offset of and change it beside the MAGICDG.
>
>
> I know what is happening but not why. The problem almost certainly is with
> lwtools.

I think the reason is how lwasm implements "org", which is documented
in the lwtools documentation. For non-raw output formats it will
change the loading address, but naturally not for raw formats, which
is used here for assembling HDB-DOS. And it will not pad the binary
either... So fill is required, unless the HDB-DOS assembly workflow is
changed to using non-raw formats.

Tormod


> All the intermediate org statements in the program are being ignored. If you
> replace
>   org MAGICDG+$1930
> with
>   fill  $39,MAGICDG+$1930
> then the HDBDOS jump table and offset will always be at $D930.
>
> Given this problem, all the org sections probably should be changed to equ.