> I had this problem too. Reason is, Chris calls copyinstr() in the new
> exec code with a "pretty large" maximum buffer size, which is clearly
> larger than the 64k supported by the original locore.s. I just added some
> test statements to forcibly limit the maximum to 64k, and exec worked 
> fine again.

the solution to this problem does not lie in exec, it lies in the m68k
locore.c files.  there's no such restriction for the i386, and
there's no place documented that says "copy* must be limited to
64k.

I'd say: "one of you should fix the copy functions so they DTRT,
then make them m68k common code."  It's not a very hard fix to make;
just add another dbeq loop and a few instructions to the various
functions.  On the other hand, i ask (looking at the hp300 locore)
"why isn't this stuff done as a long-word-at-a-time copy, when
possible?"  alignment considerations?  (those could be worked around,
and it'd be a fair amount faster...)

> PS: hope to check in current sources today.

cool!



chris

------------------------------------------------------------------------------