Re: Unable to compile gnustep-baseadd with apple-apple-apple

[David Ayers]

Adam Fedor wrote:
> On Saturday, December 27, 2003, at 10:37 AM, Stéphane Corthésy wrote:
>
> >  and make wants to compile SSL bundle whereas it shouldn't do it for 
> > apple-apple-apple.
>
> This doesn't happen for me either, as long as I do
>
> make base=no add=yes

Hmmm. actually this patch was supposed to make those the default options 
for OS X.  I don't have it so I can't check what's going wrong.
http://savannah.gnu.org/cgi-bin/viewcvs/gnustep/gnustep/core/base/config.mak.in.diff?r1=1.10&r2=1.11&sortby=date

> > Then here are some patches for sources:
> I fixes these. Thanks.

Thanks, I was hoping to finally get around to looking at them tomorrow.

> > BTW, gsweb uses objc_thread_id() implemented in thr-pthread.m, but 
> > this file is not compiled when we compile only gnustepbase-add.
> thr-pthread is for the gnu runtime on darwin. Perhaps we should just put 
> this function in GSCategoies for MacOSX to use? Or is there (or should 
> we add) an NSThread method that would work better?

I've just grep'ed through gsweb and it is used quite often.  This should 
definiatly either be resloved by using existing NSThread API or 
augmenting it if needed.  It seems to be used for two main purposes:
1. identifing the thread (we can easily replace those with [NSThread 
currentThread]
2. An NSRecursiveLock category that access the GNUstep-base @private 
mutex variable to test the currect thread against the internal state of 
the mutex. (GSWUtil.m)

We really should look into that later usage, as it is non-portable and 
shouldn't be necessary.  I'll put that on my TODO list but can't make a 
commitment on when this can be resolved.

Cheers,
David