In article <20050816013054.GA7112@proton.internal.fnop.net>,
Rui Paulo  <rpaulo@NetBSD.org> wrote:
>-=-=-=-=-=-
>
>On 2005.08.14 23:10:14 +0000, Mipam wrote:
>| --- Makefile.inc.original	2005-08-14 23:29:23.000000000 +0200
>| +++ Makefile.inc	2005-08-14 23:29:43.000000000 +0200
>| @@ -5,7 +5,7 @@
>|  
>|  WARNS?=	1
>|  WFORMAT?=2
>| -NAMED_USE_PTHREADS?=yes
>| +NAMED_USE_PTHREADS?=no
>|  
>|  .include <bsd.own.mk>
>
>Yes, that's just what we need, I guess.
>Can someone responsible/knowledge for bind comment on this ?

Named works fine for me with pthreads for me on shark, i386, sparc,
and sparc64.  While there might be memory allocation bugs with
pthreads and bind I have not hit them, but my usage is not
representative (I am just using a cacheing name server on most of
these machines).  I am a bit torn about turning pthreads off by
default. On one hand it is better to be safe and release with
pthreads turned off on our release branches, but on the other hand
I'd like to keep bind threaded at head since it is the only (?)
threaded daemon in our source tree and that should help testing
our pthreads code. Importing a new version of bind on NetBSD should
be relatively easy now, so if this bug is already fixed, we can
import the new bind in NetBSD-current.

christos