Glynn Clements writes:
Craig Lanning wrote:
> Craig Lanning writes:
> > I thought about using __MINGW32__, but I was trying not to restrict
> > the fix to just mingw32. Apparently, I ended up being too general.
> > This looks like an acceptable alternative.
>
> It's not as acceptable as I thought. I updated from cvs yesterday (I
> was out of the office last week) and rebuilt xemacs. It still created
> the link. After a little investigation, I realized that the makefile
> is being constructed by using 'gcc -E'. This means that __CYGWIN32__
> is defined, not __MINGW32__. It needs to be using 'gcc -mno-cygwin
> -E' instead of 'gcc -E'.
To build a non-Cygwin version with Cygwin's gcc, I suspect that you
will need to use something like:
./configure --target=i386-pc-mingw32 ...
or
CC='gcc -mno-cygwin' ./configure ...
or
CC='gcc -b i386-pc-mingw32' ./configure ...
I haven't checked if the above work with Cygwin, but (IMHO) they
probably should.
I can't see any way that configure can automatiacally deduce that you
wish to perform a cross-compilation (which is basically what building
for an i386-pc-mingw32 target on an i386-pc-cygwin32 platform amounts
to).
--
Glynn Clements <glynn(a)sensei.co.uk>
You are right, configure can't automatically deduce the mingw32
platform. That is specified on the command line like so:
./configure i386-pc-mingw32
This is all possible thanks to changes that Andy Piper has made to the
sources and the configure script. I am composing this response in the
XEmacs that I built yesterday. From the splash screen:
XEmacs 21.2 (beta19) "Shinjuku" (mingw32) of Mon Sep 27 1999 on NGDS1
During compilation the -mno-cygwin option is included in the call to
gcc. It's only when the preprocessor is called that the -mno-cygwin
option is omitted.
--
----------------------------------------------------------------------
Craig Lanning Principal Engineer
Logicon, Inc.
a Northrop Grumman company E-Mail: Lanning(a)scra.org
5300 International Blvd. Voice: 843-760-3246
North Charleston, SC 29418 FAX: 843-760-3780
USA