Under NetBSD-1.3_ALPHA/alpha (latest toolchain) it gets a warning
   on the compile and also won't link unless I add -ltermcap as
   the final step. Oh, I had to fix an alpha machine include file
   too- it was propagating inline splraise back to userland just
   because someone include <sys/param.h>.
   
Apparently anything that uses iostreams requires -ltermcap for some
reason, we noticed this even on the first 1.3 snapshot.  The bug about
splraise I reported months ago, and I know someone on the core team said
they'd bring that up with the rest of the netbsd developers (param.h gets
machine/interrupt header which uses K&R C to define splraise, which causes
g++ to choke).  We fixed it here by moving an #ifdef KERNEL one line down,
as I recall.  It is not unusual that no kernel developer uses C++, but it
would be nice ....

   Thoughts?
   
I'm downright scared by source archives that don't have enough info to
build the compilers such that they work.  I'm working with a commercial
adopter of NetBSD, and "snapshots" which cannot be used to bootstrap
themselves (rebuild from source, get same output) are not going to be
trusted enough to go into development.  I'd like to reraise the suggestion
that NetBSD developers put out snapshots that include source and binary
that is consistent, if I may....