> It would appear that the directory `sys/lib/libkern' assumes that the
> compiler has access to `installed' header files (for example
> quad.h:#include <limits.h>).  If cross compiling, this may not be the
> case.

It (and lib/libsa) also assumes:
	(1) that every architecture wants to build the same set of files,
	(2) every architecture can build those files with the same set
		of compiler flags.

All of these assumptions are wrong.


To handle the #include problem, i'd use #ifdef _KERNEL in the libc
files, and then #include the appropriate includes (e.g. libkern.h) for
the _KERNEL case.


for the two problems that i enumerated above:

Throw away the machine-independent makefile garbage (yes, i know i was
instrumental in creating and propagating some of that garbage 8-),
have each arch that needs the files build them someplace convenient,
maybe even per-kernel in the kernel Makefile.  (frankly, i don't
_care_ about the wasted space for a few dozen small .o's...  8-)




chris