[ARGUS] Re: bug?
Peter Van Epp
vanepp at sfu.ca
Fri Oct 15 11:26:33 EDT 2004
Thats the ticket! Thanks. Carter: it looks like adding -mcpu=v9 to
CFLAGS when Solaris9 is found would be a good start. I haven't (in my extensive
test of the same data file 4 times :-)) seen the seg faults Andrew refers to
though. In any case here are the same numbers as for the FreeBSD boxes and the
Mac on what I think is a 4 processor Sunfire 280 to local disk (rather that
the Netapp the provides home directory service here). A fair bit faster, but
I expect a lot more expensive too. I'm reasonably sure this machine/kernel is 64
bits (because we had trouble with identd and had to upgrade GCC to get it to
work correctly here) so this may be a short term answer to the 64 bit question
on Solaris for those that were working on it (modulo the seg faults).
As I get time I'll run other files here and see if I can cause a
bus error and see if we can figure out what it is.
time ./ra -r /tmp/com_argus.2004.10.10.22.00.00.0.gz -nn > /dev/null
81.0u 18.0s 1:30 109% 0+0k 0+0io 0pf+0w
27: vanepp_fraser% time ./ra -r /tmp/com_argus.2004.10.10.22.00.00.0.gz -nn > /tmp/t
85.0u 24.0s 1:41 107% 0+0k 0+0io 0pf+0w
28: vanepp_fraser% time ./ra -r /tmp/com_argus.2004.10.10.22.00.00.0.gz -nn host 142.58.200.82 >/dev/null
26.0u 12.0s 0:30 126% 0+0k 0+0io 0pf+0w
29: vanepp_fraser% time ./ra -r /tmp/com_argus.2004.10.10.22.00.00.0.gz -nn host 142.58.200.82 >/tmp/t
27.0u 12.0s 0:30 128% 0+0k 0+0io 0pf+0w
Peter Van Epp / Operations and Technical Support
Simon Fraser University, Burnaby, B.C. Canada
On Fri, Oct 15, 2004 at 04:58:04PM +1000, Andrew Pollock wrote:
> On Thu, Oct 14, 2004 at 07:52:13PM -0700, Peter Van Epp wrote:
> > To answer my own question (post coffee this time :-)), no this isn't
> > a bug. Attempting to build argus_bpf without the pcap librarys won't succeed
> > and not succeeding isn't a bug :-). Unfortunatly building the right client
> > files looks to run in to 64 bit problems (even with -m32 specified to gcc)
> > ra core dumps on the data file presumably because of time_t being 64 bits
> > (although I didn't run gdb to see), so I guess a test on Solaris wasn't to be
> > unless a bit more poking will fix it (which seems unlikely).
> >
>
> You be needing CFLAGS = -g -mcpu=v9
> to get it to compile on Solaris. You may still experience random bus errors
> though.
>
> regards
>
> Andrew
More information about the argus
mailing list