> So, given gcc's conventions about invoking a preprocessor on
> assembly-code, I think the right long-term fix is to rename all those
> files from <foo>.s to <foo.>.S
> 
> For now I'm making symlinks...

the way they came out of berkeley, they were .s's, and i believe there
were special rules to run cpp, etc., on them.

However, if you use .S's, you solve that problem.  The "NetBSD Way" is
to have them named .S.

I've run into cases where this is a problem.  For instance, the OSF/1
compiler sees .S files as output produced by the 'ucode' object file
splitter, i.e. a temporary file, and it deletes the hell out of them!
Looking at ULTRIX man pages, it looks like the same thing is true of
the ULTRIX compiler, so you'd better be using gcc, else it's not going
to work right.



chris

------------------------------------------------------------------------------