FYI: mingw cross

[Ralf Wildenhues]

* Bob Friesenhahn wrote on Sun, Jan 23, 2005 at 06:35:05PM CET:
> On Sun, 23 Jan 2005, Ralf Wildenhues wrote:
> 
> >Hehe, Debian sarge has mingw cross-compile packages.  :)
> >
> >Testing this uncovered a bug: $host_os is tested but not set in libtool.
> >Furthermore, a few rather simple fixes to make cross `make check'ing
> >bark less.
> >
> >OK to apply (HEAD and branch-2-0, patch below against HEAD)?
> 
> Yes, it does not appear to harm anything.

Done (also put the additional declarations into branch-1-5, but not the
tests/ changes).  

> >Do we need a $LN_S_FOR_INSTALL?  I do not actually know whether there
> >are systems which put versioned libs in place as well as the unversioned
> >and also do not have symlinks.
> 
> If such conditions exist, I expect that we will hear about it.  Might 
> was well wait until then.

Fair enough.

> >We might need to choose between .libs and _libs based on $host_os?
> >(Imagine a cross-compile tree which is also samba-exported to win?)
> >This would affect some of the tests as well.
> 
> Modern Windows does appear to properly support creating .libs 
> subdirectories.  I just verified that I could create a .libs 
> directory, create a file in that directory, and access the file via a 
> path.  Samba might be (optionally) programmed to do some sort of 
> remapping for old Windows clients.
> 
> The primary requirement is to satisfy the operating system used to do 
> the build (rather than the target) since otherwise the build may not 
> succeed at all.  This implies that we should respect $host_os when 
> making filesystem type choices.

Alright with me.

Thanks,
Ralf