Re: [MinGW-dvlpr] =?iso-8859-1?q?Problems_with_corss_compiler=09w32ap?= =?iso-8859-1?q?i/lib/Makefile=2Ein?=

[Keith M. <kei...@us...>]

On Monday 11 September 2006 1:19 pm, Earnie Boyd wrote:
> > I just tried to compile w32api for MinGW and found that it's now
> > throwing all the header files in include/w32api.  It's coming from
> > lib/Makefile.in:
>
> That means that Corrina broke the magic.  Only a cygwin target should
> put the headers in /include/w32api and /include/mingw.  Adding
> cygwin-developers to the distribution.
>
> > ifneq (,$(findstring cygwin,$(target_alias)))
> > inst_includedir:=$(tooldir)/include/w32api
> > inst_libdir:=$(tooldir)/lib/w32api
> > else
> > ifneq (,$with_cross_host)

Well, this is wrong in any case; what is `$w'?  No matter, the appended 
`ith_cross_host' can never be null, so this is always true anyway. But 
even allowing that this is a typo, the whole $(with_cross_host) thing is 
dubious; in configure.in I see an AC_SUBST for `with_cross_host', but 
nothing to ever assign a value to it, so it will *always* propagate from 
configure to `Makefile' as a null string.  What is its intended purpose?

> > inst_includedir:=$(tooldir)/include/w32api
> > inst_libdir:=$(tooldir)/lib
> > else
> > inst_includedir:=$(includedir)
> > inst_libdir:=$(libdir)
> > endif
> > endif

As I noted yesterday, this logic really should be handled at the 
configure level; putting it thus, in the Makefile, is just plain ugly, 
and unnecessary.  I stand by my offer to clean it up, but it will not be 
a quick fix.

Regards,
Keith.