[kaffe] Problem with kaffe for m68k platform
robilad at kaffe.org
Mon Jan 9 06:32:38 PST 2006
Kiyo Inaba wrote:
> Dalibor wrote:
>>gdb should allow you to set watchpoints on addresses, so that
>>you can see who writes the erraneous value into the structure.
> Based on my experience, (at least for m68k) watchpoint is not easy to
> use. This reduced the performance in 1:100 order, and gdb can not
> detect properly for the memory access. So I gave up to use watchpoint
> on this architecture. I am interested whether it is ok for other
> architecture or not.
Watchpoints work for me on x86, at least, though they are not very fast.
Otoh, they are pretty convenient to detect where some bad value comes
from, if one has enough time to wait it out. :)
>>I guess you are the first one porting to m68k-ecos. I am sorry to hear
>>that you are running into so much trouble. ;(
> The 'ecos' port done by Simby has not been merged, and Coldfire port
> (or actually, it is Dragonball port, if I remember correctly)
> done by Transvirtual has not been merged. So this combination is new
> CPU with new OS, much worse than just trying new OS ;-)
Yeah. That would explain why Daniel had to modify the assembler code
(cas1 instruction, afair).
> BTW, uClinux should be another new OS for Kaffe, based on my understanding,
> am I right? (Of course I know uClinux is very very close to the original
> linux, but again 'slight' difference introduces many fancy thing to debug)
I believe uclinux has now been (at least partially) merged in into the
regular linux kernel, but it does have enough exciting low level
differences to play with, according to documentation on the web.
More information about the kaffe