Re: MacOSX GNUstep build issues

Adam Fedor wrote:

>
> On Dec 9, 2004, at 2:24 PM, Richard Frith-Macdonald wrote:
>
>>
>> That being said ... I'm not completely sure of all this, it's just 
>> the impression I've gained when trying to use MacOS-X.   Someone else 
>> may have a surer idea of the exact behavior.  Even better, someone 
>> may know of compiler/linker flag combinations which could be used to 
>> make MacOS-X act like a normal unix system.
>>
>>
>
> It works fine for me. I can compile the whole GNUstep system and run 
> apps. Thus gnustep-back is a bundle which links in libX, etc...
>
> I've been messing with it for a long time, so perhaps I've done 
> something that is not reflected in the README.Darwin, but I'm not sure 
> what that could be.  I'll have to look again to see if I'm doing 
> something different.
>
Okay I think I've figured this out. The mac does appear to behave the 
way Richard describes if you do not use the '-undefined' flags to the 
linker. It appears that  the GNUstep make system deals with this: when 
linking dylibs it uses '-undefined warning', when linking bundles it 
uses '-undefined suppress'. The problem is these flags don't seem to 
apply when linking tools or apps. It seems to me that one of these 
should be added to INTERNAL_LDFLAGS one mac systems so that it is used 
for all linking commands. But, I don't really know enough about this. 
For the time being I've added the correct flags to ADDITIONAL_LDFLAGS in 
the appropriate GNUmakefile.preamble, which is much less annoying than 
adding all the libraries.

BTW the problem with gdnc turned out to be my libobjc. I forgot about 
the line in README.darwin that said to:

    ln -sf libobjc.a libobjc.dylib

I had manually linked a dynamic version of libobjc, which made gdnc 
crash. Once I linked libgnustep-base against the linked static version 
everything seemed to work fine.

Thanks for all the help.