In article <199811202204.OAA04825@lestat.nas.nasa.gov>
thorpej@nas.nasa.gov writes:

[ use d0 to pass syscall # ]

> That can't work... it clobbers the value of d0 (which breaks
> longjmp).

Alas....  I see.

Well, then, we waste a trap every time we change signal interface.
I don't think this is a good way.

How about using d1/a0/a1/stack ?
--
ITOH, Yasufumi <yasufu-i@is.aist-nara.ac.jp>, <itohy@netbsd.org>