[patch] remove darwin from libgomp's target.

[Shantonu Sen]

But what directory would it go into? Lots of OSes might benefit from  
this implementation, since sem_init is an optional API. Putting it in  
a darwin/ directory and then have AIX or GNU/Hurd use it would not  
make much sense at all.

The condition variable-based implementation and sem_init-based  
implementation both use "posix" APIs, so posix/ seems a reasonable  
location that describes the class of OS targets that would use it.

Shantonu


On Feb 2, 2006, at 2:08 PM, Richard Henderson wrote:

> On Mon, Jan 30, 2006 at 11:06:45PM +0100, Andreas Tobler wrote:
>> Shantonu Sen wrote:
>>> Here's what I have so far, with the new files as attachments. In
>>> particular, I haven't verified that the cancellation point  
>>> semantics are
>>> preserved with respect to the POSIX semaphore version. I also  
>>> dumped the
>>> files in libgomp/config/darwin. A longer term fix would probably  
>>> involve
>>> more autoconf checks, and having this version alongside the POSIX
>>> semaphore version in libgomp/config/posix, and using the appropriate
>>> version, since they both use POSIX features.
>>>
>>> I guess I should get a copyright assignment process going...
>>
>> Shantonu,
>>
>> I've taken your work and tried to integrate it into the current
>> infrastructure.
>
> Actually, I think Shantonu did a better job.  I'm not as in
> favour of the HAVE_BROKEN_POSIX_SEMAPHORES define as I am of
> using the directory search mechanism.  It would be different,
> perhaps, if you were autoconfing and detecting a failure in
> the posix semaphores.
>
> I would like that a comment be added why Darwin can't use the
> regular posix implementation.
>
>
> r~