(I apologize for incorrectly posting this earlier on the bro-dev list
which seems to be intended only for the Bro-Tracker bot to post to.)
Hi all:
I am seeing errors like this every time the connection summary script runs:
python:/usr/local/bro/lib/broctl/_SubnetTree.so: undefined symbol '__inet_addr'
python:/usr/local/bro/lib/broctl/_SubnetTree.so: undefined symbol '__inet_aton'
Traceback (most recent call last):
File "/usr/local/bro/bin/trace-summary", line 19, in <module>
import SubnetTree
File "/usr/local/bro/lib/broctl/SubnetTree.py", line 7, in <module>
import _SubnetTree
ImportError: Cannot load specified object
0.20 real 0.00 user 0.00 sys
Incidentally, I get the same kind of linker errors at bro startup that
were seen on OpenBSD with bro a couple of years ago:
# /usr/local/bro/bin/bro --version
/usr/local/bro/bin/bro:/usr/lib/libc.so.60.1:
/usr/local/lib/libbind.so.2.0 : WARNING: symbol(__p_class_syms) size
mismatch, relink your program
/usr/local/bro/bin/bro:/usr/lib/libc.so.60.1: /usr/local/bro/bin/bro :
WARNING: symbol(_res) size mismatch, relink your program
/usr/local/bro/bin/bro:/usr/lib/libc.so.60.1:
/usr/local/lib/libbind.so.2.0 : WARNING: symbol(__p_type_syms) size
mismatch, relink your program
/usr/local/bro/bin/bro version 2.0-debug
# uname -a
OpenBSD <hostname> 5.0 GENERIC#43 i386
# pkg_info | grep libbind
libbind-9.4.2 BIND 8 compatible stub resolver library
Is this a case of some kind of incompatibility between the libc and
libbind resolver functions?
Thanks,
(I apologize for incorrectly posting this earlier on the bro-dev list
which seems to be intended only for the Bro-Tracker bot to post to.)
Heh, it's actually not. That list is for development discussion but we use the tracker heavily enough that it can certainly get overwhelmed with tracker messages if you aren't filtering those to a separate mail box. 
Incidentally, I get the same kind of linker errors at bro startup that
were seen on OpenBSD with bro a couple of years ago:
We're going to begin doing continuous integration (automated building and testing after every commit) of the Bro source code soon and the plan currently is to also do these build/test cycles on a range of operating systems. I think that OpenBSD is going to be one of the ones we test so hopefully our OpenBSD support will be regularly tested (which it currently isn't) so we can prevent regressions from popping up.
Is this a case of some kind of incompatibility between the libc and
libbind resolver functions?
I have no clue, Jon Siwek may know the answer though.
.Seth
Incidentally, I get the same kind of linker errors at bro startup that
were seen on OpenBSD with bro a couple of years ago:
hopefully our OpenBSD support will be regularly tested (which it currently isn't) so we can prevent regressions from popping up.
I saw some discussion in the mailing list archive about those linker warnings, too, and I think it maybe not be a regression, but the original warnings were never resolved.
Is this a case of some kind of incompatibility between the libc and
libbind resolver functions?
That's my guess is that something is out of sync between those two -- compiling and running a simple test program that links against libbind will yield the same warnings, so it might be an issue that the OpenBSD ports mailing list or libbind package maintainer would better help.
+Jon