Cygwin cross compiler

[Stephen Weeks]

> I tried to think of a way to do the 2.2 vs. 2.4 thing all in #ifdef's, but
> there isn't a way without changing the code.  If you change the code so that
> the mem_unit is gotten by what looks like a function call, then you can
> define the function for 2.2 compiles to always return 1.

How would this work?

> Alternatively, it would be easy to just include the 2.4
> 	/usr/include/sys/sysinfo.h
> (so it would definitely have a mem_unit field) and then have the code test
> if it is 0 and, in that case, use 1.
> This isn't great because of future changes, but is probably the best way to go.

I would prefer the function call approach.  Can you show what you had
in mind?  Is there some #define'd variable that's reliably different
on the two kernels?  That would seem sensible.

> As to Matthew's stuff, should I grab
> 	mlton-20020314-1.i386.rpm
> 	mlton-20020314-1.src.rpm
> 	mlton-20020314.tgz
> and should I replace
> 	mlton-20020312-1.i386.rpm
> 	mlton-20020312-1.src.rpm

I would think that you should also grab

	mlton-20020314-1.i386.tgz 

and please do not replace, just add.  Thanks.

> Speaking of build problems, why doesn't the source rpm (and tar file)
> reflect the world one directory higher, including an empty include and bin
> directory?  Then you wouldn't have to manually make them and things would be
> more automatic.

I think that was a problem with the old world, but went away a few
weeks ago, with the cross-compile changes.  Now everything builds in
the 'build' subdirectory of the src dir.  Let me know if you see
otherwise.