Re: GNUstep on Cygwin

[Adam Fedor]

On Sunday, December 7, 2003, at 10:06 AM, Fred Kiefer wrote:

> Fred Kiefer wrote:
> > I am now struggleing to get back compile as it insistes to compile 
> > xlib. The problem seems to be the X installation that comes with 
> > Cygwin. I will have to force the winlib evironment. It looks a bit 
> > strange to me that the system automatically detects that it has to 
> > build a window server, so it uses win32, but tries o combine this 
> > with xlib. But I think this can be resolved with the correct 
> > configure switches, which I  have to look up.
>
> To keep you informed of my progress. Starting configure with the 
> correct parameters did the trick:
> configure --enable-server=win32 --enable-graphics=winlib
>
> As Cygwin has also the ablility to compile GNUstep for the X window 
> system, it is fine, that you have to specify which one you want. 
> (Although I would prefer win32 as the default and the mixed state I 
> had yesterday when only setting one switch was horrible)

We could probably just check target_os for cygwin or mingw. That might 
be a better check. Although I know someone was using cygwin and the X11 
backend (of course that doesn't work either, since Cygwin says it 
implements the SHM extension but doesn't actually implement it (which 
confuses configure).

> Now I have problems with the path conversion. GNUstep make did set my 
> GNUstep system directory to c:/GNUstep/System. And now applications 
> are looking for the backend bundle in a directory below this.

Did you try --with-prefix=/usr/GNUstep

since Cygwin should work with that? Just wondering...


> This path is not recognized as being an absolute one and so the 
> loading of the backend bundle fails. I could of course change the 
> setting in the GNUstep.sh, but as this is now a valid Cygwin path, I 
> wanted to change the file handling in NSFileManager to convert this 
> path correctly. Currently we only do conversions for the file names on 
> MinGW, so I tried to switch the condition to #ifdef __WIN32__, but for 
> unknown reasons this condition did not hold on CYGWIN. Can anybody 
> explain this to me? From the gcc specs I can tell, that this define 
> should be set.


Try this script to see what is defined:

#!/bin/sh
for def in `gcc -E -v - </dev/null 2>&1 | grep [-]$`; do
  case $def in -D*) echo $def;; esac
done