Fixing signal records when ktrace'ing under netbsd32 emulation

[Martin Husemann]

I recently had to debug something not working for netbsd32 emulation and
found that signals created bogus (64bit) ktrace records.

I came up with the attached patch which seems to work but is a bit ugly
and intrusive. It may not be fully modular and is missing an indirection
from the original ktrpsig -> ktrpsig(), so it is not checking ktrace_on.
I am not sure we can ever get to this code w/o ktrace_on.

I wonder if we should instead of the emulation redirection check
(#ifdef'd properly) for a 32bit process and then hard-coded test against
netbsd32 emulation. No other emulation would ever want to interfere with
the netbsd-specific kte records - or do I miss something?


Martin
P.S.: sendmsg also creates 64bit records, but that is pretty intrusive to
fix too, leaving that for another day.