I think I found the bug causing the reported R4000SC/R4400SC crashes.
In r4kcache.h some of the functions for handling the second level caches
were using the wrong cacheops.
I'm rewriting the thing; the file was producing a huge object file and
not able to handle primary instruction and data caches with different
linesize. This was a problem with the Vr4300 and some Magnum 4000 and
the Olivetti M700 where reconfiguring the linesize of the primary cache
What I won't fix for now is the handling of split instruction and data
second level caches a la R4000SC. I don't know of any system using
something like this.