> > That's because the depend routines only scan for assembler files beginning
> > with .s (not .S).  I don't know why arm32's are so named -- maybe it's
> > because we used to use gcc to pre-process them at one point.

> Perhaps we should rename them (I often wondered why they were .S), or expand 
> the build system to treat .S as .s's

Fixing the build system means fixing /sbin/config to handle this.  
Unfortunately, the rules it uses are fairly generic, to either this would 
have to be special cased, or we would end up having the case where .C 
files would be treated the same as .c files (not good, since .C implies 
c++).

Renaming .S to .s would therefore seem to be the obvious choice, but then 
we'd loose CVS history...

R.