Chris G. Demetriou wrote:

> matthew green <mrg@towers-of-dub.eterna.com.au> writes:
> > the reason builds for <sys/endian.h> failed to build lately is that
> > two of the games have helpers that are built with ${HOST_CC} rather
> > than ${CC} and also were changed to include <sys/endian.h> -- but
> > the host compiler won't have a <sys/endian.h> yet [ ... ]
> 
> Of course, if the goal here is to actually support cross-compiling in
> the general sense -- which it should be, in my opinion -- then it's
> silly to assume that the host system will ever include a
> <sys/endian.h> or lots of other NetBSD-specific features.
> 
> Cross-compiling NetBSD-NetBSD is all well and good, but it's not
> nearly so useful as letting NetBSD be cross-compiled from random
> hosts.  _That's_ what we should be shooting for.

So how do we solve the problem of generating a file, in this particular
case of having 64 bit data of a specific endianness?  Even before my
changes to fortune to use <sys/endian.h>, off_t's were assumed to
be the same size on the generating and target systems - I couldn't
cross-compile a fortune.dat file on my Ultrix box the could be read
by a NetBSD box without shoving some real kludges in strfile(8).

Simon.