On Nov 22, 2007, at 11:22, Chris Ross wrote:
> On Nov 22, 2007, at 04:26, Michael-John Turner wrote:
>> Another possibility is to build milter-greylist using pth rather than
>> native threads. I had a lot of crashes with my 3.1 system when  
>> using clamav
>> with native threads, but since switching to pth it's been rock  
>> solid, even
>> with fairly high paging. May require some edits to system headers  
>> (the
>> PREFER_NATIVE_PTHREADS=NO and PTHREAD_TYPE=pth pkgsrc build  
>> options are not
>> perfect), but it's more than likely possible.
>
>   I can try this too.  Do you just suggest those two options to a
> pkgsrc "make" of mail/milter-greylist might do much [if not all] of
> the work?

   Okay, trying that makes it fail on the autoconf checks for libmilter,
because the compilation of the test program fails to link looking for
pthread_detach.  I see that /usr/lib/libpthread.a (and presumedly .so)
has:

dumbo% nm /usr/lib/libpthread.a | grep pthread_det
0000000000001680 T pthread_detach

   And pth's pthread.a has:

dumbo% nm /usr/pkg/lib/libpthread.a | grep pthread_det
00000000000023c0 T __pthread_detach

   Any ideas as to why?  This a pth issue?  Adding pkgsrc-users to the
list, in case someone else there knows more about devel/pth.

                      - Chris