Josef Moellers wrote:
> For some time now, we are having problems with dying snmpd under some
> (not yet fully understood) circumstances.
>
> What I can see from coredumps, traces and tcpdump output suggests that
> the problem is that
> 1. an agentx subagent is non-responsive (it first registers
I'll be publishing 5.4.rc1 mid-day tomorrow if all goes well. I got
caught off guard a bit today with other things that were more
important so I didn't get it out today as planned.
One issue that has obviously become a problem is the deadline for
which patches can be submitted for upcoming relea
Thomas Anders wrote:
> Successfully tested on AIX 5.3 and IRIX 6.5.
It also just passed another quick build test on Solaris 10, RHEL 4, HP-UX 11.23
and AIX 5.2, done for safety.
+Thomas
--
Thomas Anders (thomas.anders at blue-cable.de)
---
-Coders,
a number of systems (including OSX, AIX, IRIX and Tru64) have known problems
when using nlist/klookup (/dev/kmem access) for a number of reasons. A minimal
stop-gap fix is to turn off nlist/klookup usage completely, most conveniently
done by "configure --without-kmem-usage" (which prevent
The only machines affected are those reporting their statistics from picld.
Of those, the only ones having the bug are the 240-class machines including
netra-240. I don't have one to test, but I've sent the code to the poster
on -users having the original problem.
I did run the code on a V880. I
Bruce Shaw wrote:
> I just found out on -users that fan sensors aren't being reported on
> 240-class Sun servers.
>
> Picld is reporting the fans differently than what I'm used to. The
> infrastructure is already in place for fixing this - it's just a different
> call to an existing module known
I just found out on -users that fan sensors aren't being reported on
240-class Sun servers.
Picld is reporting the fans differently than what I'm used to. The
infrastructure is already in place for fixing this - it's just a different
call to an existing module known to work.
This is a call for v
On 30/10/06, Robert Story <[EMAIL PROTECTED]> wrote:
> Give that there are many *Real variables, including memAvailReal, I would
> have to guess that memTotalFree would include real and swap. Thus I'm
> leaning towards a +1 as well...
Having sat down and looked properly at both the code and the MI
Wes Hardaker wrote:
> You (Thomas) have at least +3... Please apply asap.
Done.
+Thomas
--
Thomas Anders (thomas.anders at blue-cable.de)
-
Using Tomcat but need to do more? Need to support web services, security?
Get st
> "RS" == Robert Story <[EMAIL PROTECTED]> writes:
RS> Something is better than nothing, imho.
RS> +1
You (Thomas) have at least +3... Please apply asap.
--
Wes Hardaker
Sparta, Inc.
-
Using Tomcat but need to do more
Thomas Anders wrote:
> Wes Hardaker wrote:
>> Is there someone with a broken sed machine that could verify this
>> fixes the problem? (it should).
>
> I've started a build test on a number of Solaris machines, both with and
> without a problematic sed. If it works on all, +1 from my side.
Build
+1 Net-snmp definitely needs these stats on irix
Regards,
--
Herve Boulouis
-
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your jo
On Mon, 30 Oct 2006 07:13:51 -0800 chris wrote:
CJ> On Leopard and on Intel-based Tiger systems, --enable-mini-agent is
CJ> required because there is no /dev/kmem. (Configure says --without-
CJ> kmem-usage can only be used with --enable-mini-agent.)
Hmmm... I suggest, as a test, that you try wi
> "DS" == Dave Shield <[EMAIL PROTECTED]> writes:
DS> If 'mk' is a directory, then shouldn't this be either "rm -rf mk" or
DS> "rmdir mk" ?
Yep. Good catch.
--
Wes Hardaker
Sparta, Inc.
-
Using Tomcat but need to do mo
On Sat, 28 Oct 2006 14:54:12 +0200 Thomas wrote:
TA> The attached patch fixes the linking behavior *by default*. It's much
easier to
TA> understand this way (IMHO) and it'd be pretty straightforward to make it
depend
TA> on --enable-as-needed once the approach is agreed upon. It passed last
nigh
On Sat, 28 Oct 2006 01:56:04 +0200 Thomas wrote:
TA> The patch will only affect IRIX by nature, cannot cause regression over
5.3.x
TA> (memory values are reported as all zeros for IRIX by 5.3.x) and has been
tested
TA> successfully on my IRIX 6.5.x test machine. It now reports:
Something is bett
On Sat, 28 Oct 2006 10:14:32 +0100 Dave wrote:
DS> I don't mind switching to use VIRTMEM instead of PHYSMEM - but we need to
DS> be aware that this will change behaviour for some other systems.
DS>
DS> I vote -1 - if only as a holding mechanism.
DS>
DS>
DS> Note that this situation can be trac
On 30/10/06, Wes Hardaker <[EMAIL PROTECTED]> wrote:
> Anyway, enclosed is the patch. It's a more major change than I'd
> ideally like for a last minute change, but it was something we agreed
> upon and thus I should at least offer it. I'm voting +1, but barely.
Wes Hardaker wrote:
> Is there someone with a broken sed machine that could verify this
> fixes the problem? (it should).
I've started a build test on a number of Solaris machines, both with and
without a problematic sed. If it works on all, +1 from my side.
+Thomas
--
Thomas Anders (thomas.a
I promised to handle this at the admin meeting. Little did I know how
much time it was going to eat up from my life as it was worse to
implement than I thought (though the patch size doesn't reflect this;
eval of eval of evals just sucks to figure out in you head for sh).
Anyway, enclosed is the
On 26-Oct-2006, at 4:27 PM, Wes Hardaker wrote:The patch below fixes the problem stated in patch #1578339 but in adifferent way (IE, the way we've been doing it in the past). But nowrather than creating a zillion .so files in CVS, the makefilegenerates them and cleans them up... A bit more effici
> "WH" == Wes Hardaker <[EMAIL PROTECTED]> writes:
WH> I'm willing to consider applying an update to the feature for 5.4.1,
WH> yes. Let's discuss that post 5.43
Not sure how that 3 slipped in there, but you got the message.
--
Wes Hardaker
Sparta, Inc.
On 25-Oct-2006, at 2:56 PM, Robert Story wrote:On Wed, 25 Oct 2006 16:57:45 -0400 chris wrote: CJ> On 20-Oct-2006, at 5:04 PM, Thomas Anders wrote: CJ> CJ> > Robert Story wrote: CJ> >> Darwin 8 & 9 don't include the freeBSD or generic system header, CJ> >> and thus end CJ> >> up with no INP_NEXT
> "TA" == Thomas Anders <[EMAIL PROTECTED]> writes:
TA> If there isn't a +3 consensus to fix this now, can it be done at least
TA> for 5.4.1? I'd rather not wait until 5.5 to fix --enable-as-needed
TA> (still default to off, mind you).
I'm willing to consider applying an update to the feature
On 30/10/06, Thomas Anders <[EMAIL PROTECTED]> wrote:
> If there isn't a +3 consensus to fix this now, can it be done at least
> for 5.4.1?
I'd be a lot more sympathetic with that as a possibility, certainly.
Dave
-
Using To
Dave Shield wrote:
> On 29/10/06, Wes Hardaker <[EMAIL PROTECTED]> wrote:
>> TA> - enable the proposed linking changes if --enable-as-needed is
>> TA> given (because it's the Right Thing and more convenient) ?
>>
>> I definitely think that's the right way to go for a release or so
>> (though I'm le
On 29/10/06, Wes Hardaker <[EMAIL PROTECTED]> wrote:
> TA> - enable the proposed linking changes if --enable-as-needed is
> TA> given (because it's the Right Thing and more convenient) ?
>
> I definitely think that's the right way to go for a release or so
> (though I'm leaning more toward 5.5 bein
For some time now, we are having problems with dying snmpd under some
(not yet fully understood) circumstances.
What I can see from coredumps, traces and tcpdump output suggests that
the problem is that
1. an agentx subagent is non-responsive (it first registers its OIDs,
then tries to contact
Thomas Anders wrote:
> Following up the IRC admin meeting, I'm hereby proposing a patch that makes
> the
> "as needed" support (fulfill library dependencies at library link time rather
> than application link time) more complete. It adds libwrap, libnm (HP-UX),
> libperfstat (AIX) and libadm (Sola
29 matches
Mail list logo
