Jeremy Allison jra at
Fri Jun 9 00:33:02 GMT 2006

On Fri, Jun 09, 2006 at 10:11:50AM +1000, James Peach wrote:
> So, the proposal is
> 	if (libpthread is linked)
> 		if (we have the RT_SIGNAL_NOTIFY bug)
> 			AC_ERROR with blame message
> 		else
> 			allow, but warn

That seems the best idea.

> I wonder whether allowing libpthread will expose us to more bugs though?

Maybe - libpthread shouldn't be needed for aio on sane
platforms, but on earlier Linux versions without aio support
in the kernel the aio is emulated with pthreads. This is
probably why the two get tied together.

We should be able to survive with pthread linked it, it
shoudn't actively break code, just make the malloc's a
little slower (actually in a good implementation the
malloc should get switched to a locking one as part of the first
thread initialization).


More information about the samba-technical mailing list