BIND 8.2.2-P5 crashing on Solaris 2.7

[Matt Ragan]

Are there any known issues with BIND 8.2.2-P5 and statistics logging?  I
have a pretty simplistic named.conf on a test nameserver that is
receiving a few queries approximately every 10 to 15 minutes or so as I
test out a script.

After the process has been active for exactly an hour, the message:

19-Jan-2000 11:47:21.581 Cleaned cache of 0 RRsets

appears in the log file, and named dies.  A traceback on the core file
shows:

core file = core -- program ``named'' on platform SUNW,Ultra-5_10
SIGBUS: Bus Error
$c
ns_logstats(ffbef76c,a9800,ce800,ffbef758,4e5a0,d4909) + c4
__evDispatch(ffbef7f0,d48fc,114988,d48fc,114a28,cec00) + 1b4
usage(a4c00,a4c00,ffbef7f0,ce800,4,ce000) + a54

If I add:

options {
    statistics-interval 0;
};

then named doesn't die until it is stopped with 'ndc stop', 'ndc
restart', or with a kill, so there is evidently a problem in the
statistics processing.  Has anyone else seen this?

The server on which this is running is a Sparc system running 64-bit
Solaris 2.7.  BIND was compiled with GCC 2.95.2.  I haven't yet tried
recompling it with a different version of GCC or on a different Solaris
version.

--
Matt Ragan                                EMEA IT/Network Support
Tivoli Systems (UK) Ltd.                      +44 (0) 1753 780070
Sefton Park    Bells Hill    Stoke Poges    Bucks   SL2 4HD    UK