Re: Short warning on current Cygwin

[Adam Fedor]

On Aug 27, 2004, at 7:33 PM, Fred Kiefer wrote:

> The other problem is even worse. The way we are generating the list of 
> exported symbols is no longer working. nm wont return any class names 
> anymore, so classes cannot be exported, which breaks shared libraries 
> for GNUstep. This may require a switch to a different export mechanism 
> for your symbols. As far as I remember there already where a few 
> patches for this. What I don't know is, if they are specific to 
> certain GCC versions or could be used a s a general replacement.

S.J. Chun sent this patch (among others) that perhaps dealt with this. 
The problem is that it does not work on my system (perhaps because it 
is older?):

diff -urN core/base/Source/Makefile.postamble 
../Build/10.Base/base/Source/Makef
ile.postamble
--- core/base/Source/Makefile.postamble Sat May 15 01:49:58 2004
+++ ../Build/10.Base/base/Source/Makefile.postamble     Sat May 15 
02:42:50 2004
@@ -167,7 +167,7 @@
        rm -f $@
        rm -f _tmp.def
        cat win32-def.top > $@
-       nm $^ | grep '^........ [T] _' | sed 's/[^_]*_//' > _tmp.def
+       nm $^ | grep '^........ [R] _' | sed 's/[^_]*_//' > _tmp.def
        cat _tmp.def | grep "_class_name_" >> $@
        rm -rf _tmp.def



So the question is, is there a way to make this work in all cases 
(maybe do both?) or is there some test we can do to pick which one to 
use?

P.S.  I'd like this code to eventually go in gnustep-make so it would 
work with every library.