2012/10/2 Vincent Rivière <vincent.riviere@freesbee.fr>:
On 02/10/2012 19:27, Helmut Karlowski wrote:
Why not use prg-flags? There are enough bits available and it's
independent from MiNT.
Yes, we can put that CPU flag in the standard 28-byte header.
We can also put it into the extra 228-byte header of the a.out-mintprg
executables, but that one is only present in executables produced by
gcc/gas.
Anyway, that new flag will have to be tested by anything implementing
Pexec(): EmuTOS, FireTOS, FreeMiNT...
This is not really a MiNT question, but that kind of reflexion is
usually discussed with MiNT people.
Maybe I simplify this matter too much, but since the CF is fairly
incompatible with "real" 68k CPUs (at least when compared to the user
mode compatibility of e.g. the 030 vs. the 000), why not just use a
new "magic" value in the 28-byte header (change the initial branch
instruction to something different for CF-only binaries). That way,
incompatible binaries will be rejected at the Pexec() level.