--> benr has joined #net-snmp
--> hampa has joined #net-snmp
[2002/09/27 05:27:51] #net-snmp <hampa> where should I put the sharedobject produced by mib2c ?
--> TrogL has joined #net-snmp
--> wes has joined #net-snmp
[2002/09/27 06:29:37] #net-snmp <TrogL> morning. might as well warn you now, I'm going to be working on the 100-disk problem
[2002/09/27 06:30:04] #net-snmp <wes> :-)
[2002/09/27 06:30:52] #net-snmp <TrogL> swap -s returns bytes allocated, reserved, used and available - is it worth reporting all four?
[2002/09/27 06:31:01] #net-snmp <TrogL> is it worth reporting the output of swap -l?
[2002/09/27 06:33:51] #net-snmp <TrogL> Dave thinks I should start reporting disks starting at #10, I'm trying to figure out what he thinks belongs in 0-9
[2002/09/27 06:34:22] #net-snmp <TrogL> if we're doing individual swap partitions, i'm right back where I started 'cause some of my machines have 12 swap partitions
[2002/09/27 06:35:30] #net-snmp <TrogL> personally, I think the RFC is badly written - I'd rather have RAM and swap in a separate data strucgture
[2002/09/27 06:37:55] #net-snmp <wes> well, the reason being they tried to do things generically.
[2002/09/27 06:38:11] #net-snmp <wes> with respect to swap -s, reporting them all may not be possible?
[2002/09/27 06:38:41] #net-snmp <wes> As far as numbering goes, I think you shouldn't try to use a fixed numbering sequence like was done in the past (we've already proven how bad that can be)
[2002/09/27 06:39:25] #net-snmp <wes> It makes the code a little bit harder to write, but if you just start numbering and go up as you count things it makes it much easier since you'll never run into situations like yours.
[2002/09/27 06:39:32] #net-snmp <TrogL> agreed on fixed numbering - swap "used" is a computed field
[2002/09/27 06:39:38] #net-snmp <TrogL> yep
[2002/09/27 07:02:15] #net-snmp <TrogL> ACTION installs current CVS to make sure nothing broke in last two weeks
[2002/09/27 07:11:17] #net-snmp <TrogL> heh - agent/mibgroup/host/host.h "
[2002/09/27 07:11:38] #net-snmp <TrogL> "the disk device discovery code...should never have seen thelight of day"
--> mike has joined #net-snmp
--> jbpn has joined #net-snmp
--> rstory has joined #net-snmp
--> wes has joined #net-snmp
--> wdebruij has joined #net-snmp
[2002/09/27 09:35:02] #net-snmp <wdebruij> hi
[2002/09/27 09:35:27] #net-snmp <rstory> hi
[2002/09/27 09:36:02] #net-snmp <wdebruij> does anyone know if it's true that only udp transport is opened at startup
[2002/09/27 09:36:07] #net-snmp <wdebruij> it seems that way
[2002/09/27 09:36:39] #net-snmp <rstory> yes, UDP is the default transport.
[2002/09/27 09:37:09] #net-snmp <wdebruij> ok, cause I'm tring to write an extra transport domain
[2002/09/27 09:37:26] #net-snmp <wdebruij> and I started with simply rewrting the tcp file
[2002/09/27 09:37:28] #net-snmp <jbpn> cool -- what for?
[2002/09/27 09:37:49] #net-snmp <wdebruij> for raw sockets (running below the TCP/UDP stack)
[2002/09/27 09:37:51] #net-snmp <jbpn> (I wrote the majority of the transport-related code)
[2002/09/27 09:38:14] #net-snmp <jbpn> Mmmm... interesting.
[2002/09/27 09:38:35] #net-snmp <rstory> aha.. I'll leave ya with the transport expert.. Mr. jbpn! ;-)
[2002/09/27 09:38:43] #net-snmp <wdebruij> thanks ;)
[2002/09/27 09:38:50] #net-snmp <jbpn> ACTION blushes
[2002/09/27 09:39:15] #net-snmp <wdebruij> the code is written for a specific active network listener, but can - if I get it to work - be used for anything really
[2002/09/27 09:39:30] #net-snmp <wdebruij> the problem is, how do get the tcp listener to start running
[2002/09/27 09:39:50] #net-snmp <jbpn> Ah, it's done semi-automagically.
[2002/09/27 09:39:59] #net-snmp <wdebruij> right?
[2002/09/27 09:40:34] #net-snmp <wdebruij> from the tdomain_transport function perhaps?
[2002/09/27 09:40:41] #net-snmp <wdebruij> in snap_transport.c
[2002/09/27 09:41:05] #net-snmp <jbpn> First, you need to get the transport to register itself -- this *ought* to be done via its constructor.
[2002/09/27 09:41:30] #net-snmp <jbpn> But C doesn't have constructors, so it's done by calling a function from netsnmp_tdomain_init() in snmp_transport.c
[2002/09/27 09:41:43] #net-snmp <wdebruij> sorry, that was what I meant
[2002/09/27 09:42:17] #net-snmp <wdebruij> I added some code to snmp_tcp_ctor, but I seems that this function is never called, although it should be called from netsnmp_transport_init
[2002/09/27 09:42:20] #net-snmp <jbpn> This sets up the transport domain specifier list so that "tcp:blah" gets routed to the tcp transport for turning into an address
[2002/09/27 09:42:43] #net-snmp <jbpn> But -- you need to configure it --with-transports=TCP
[2002/09/27 09:43:01] #net-snmp <jbpn> Otherwise SNMP_TRANSPORT_TCP_DOMAIN isn't defined
[2002/09/27 09:43:37] #net-snmp <wdebruij> I - ahem - removed the #ifdef just to be sure that wasn't the problem
[2002/09/27 09:43:46] #net-snmp <wdebruij> and configured with transport TCP
[2002/09/27 09:43:49] #net-snmp <jbpn> ah
[2002/09/27 09:43:51] #net-snmp <jbpn> hmm
[2002/09/27 09:44:00] #net-snmp <wdebruij> are there testprograms to see if TCP is running
[2002/09/27 09:44:03] #net-snmp <wdebruij> such as snmpget
[2002/09/27 09:45:39] #net-snmp <jbpn> snmpget tcp:localhost sysUpTime.0
[2002/09/27 09:45:49] #net-snmp <jbpn> sorry -- I have to go now
[2002/09/27 09:46:04] #net-snmp <wdebruij> ok, thanks
[2002/09/27 09:46:07] #net-snmp <wdebruij> bye
[2002/09/27 09:46:14] #net-snmp <jbpn> good luck !
[2002/09/27 09:46:57] #net-snmp <wdebruij> thanks :)
--> benr has joined #net-snmp
--> ChanServ has joined #net-snmp
--> rstory has joined #net-snmp
--> wes has joined #net-snmp
--> mike has joined #net-snmp
--> openproje has joined #net-snmp
[2002/09/27 16:32:19] #net-snmp <tcaine> anyone know if the SNMP.pm module is being supported anylonger?
[2002/09/27 16:32:29] #net-snmp <tcaine> s/supported/developed/
[2002/09/27 16:33:20] #net-snmp <tcaine> seems to be having major problems building under solaris 7.
[2002/09/27 16:34:09] #net-snmp <tcaine> Old versions of the net-snmp and ucd-snmp code don't seem to work any more either.
--> wes has joined #net-snmp
[2002/09/27 16:40:38] #net-snmp <tcaine> does the SNMP.pm code work any longer?
[2002/09/27 16:41:08] #net-snmp <tcaine> I can't get it to build under a few different platforms, solaris 7 or redhat something.
[2002/09/27 16:41:44] #net-snmp <wes> it should be fixed in 5.0.4 I think.
[2002/09/27 16:42:46] #net-snmp <tcaine> it gets a little further but doesn't actually pass tests.
[2002/09/27 16:43:02] #net-snmp <tcaine> or so it seems.
[2002/09/27 16:47:07] #net-snmp <wes> is this 5.0.4.pre2 or the cVS tree?
[2002/09/27 16:49:22] #net-snmp <tcaine> 5.0.4.pre2
[2002/09/27 16:49:35] #net-snmp <tcaine> should I grab the stuff from the CVS tree?
[2002/09/27 16:49:58] #net-snmp <tcaine> it looks like the OID tests are failing for perl in 5.0.4.pre2
[2002/09/27 16:50:40] #net-snmp <wes> it might need a make install first
[2002/09/27 16:50:51] #net-snmp <wes> If you don't have the mibs installed, it might be problematic.
[2002/09/27 16:51:17] #net-snmp <tcaine> I've got em, but I'll give that a go.
[2002/09/27 16:51:31] #net-snmp <wes> cvs wouldn't hurt.
[2002/09/27 16:51:37] #net-snmp <wes> http://www.net-snmp.org/cvs/
[2002/09/27 16:51:40] #net-snmp <wes> for instructions.
[2002/09/27 16:51:47] #net-snmp <wes> make sure you configure with --enable-shared.
[2002/09/27 16:53:42] #net-snmp <tcaine> ah, that would fix the crt error complaining about not being able to find main.
[2002/09/27 16:54:16] #net-snmp <tcaine> thanks for the tips.
[2002/09/27 16:56:50] #net-snmp <tcaine> should --enable-shared be default?
[2002/09/27 16:57:35] #net-snmp <wes> it is for non-cvs
[2002/09/27 16:57:58] #net-snmp <wes> the problem is that when it was disabled for cvs (it doubles compile time and makes debugging more of a pain) perl didn't require it.
[2002/09/27 17:00:06] #net-snmp <tcaine> ok, that make sense.
--> ua has joined #net-snmp
[2002/09/27 17:02:46] #net-snmp <ua> Re tcaine, hi wes
[2002/09/27 17:02:55] #net-snmp <wes> howdy.
[2002/09/27 17:03:14] #net-snmp <ua> Good to speak with you, in realtime. ;-)
[2002/09/27 17:03:32] #net-snmp <wes> :-)
[2002/09/27 17:03:50] #net-snmp <wes> Do you have a name I should recognize? (ua isn't very informative ;-)
[2002/09/27 17:03:56] #net-snmp <ua> ua is short for Joshua.
[2002/09/27 17:04:01] #net-snmp <ua> I'm
[2002/09/27 17:04:07] #net-snmp <wes> Ah.
[2002/09/27 17:04:20] #net-snmp <ua> I wrote Net::Telnet::Cisco which you can read about at NetTelnetCisco.SourceForge.net
[2002/09/27 17:04:41] #net-snmp <ua> and I'm the current maintainer of SNMP::Multi.
[2002/09/27 17:04:51] #net-snmp <wes> ahhhhhh
[2002/09/27 17:04:56] #net-snmp <tcaine> which is a subclass of SNMP.pm
[2002/09/27 17:05:26] #net-snmp <wes> yep.
[2002/09/27 17:05:38] #net-snmp <ua> So, you know, we're a *little* familiar with SNMP.pm. ;-)
[2002/09/27 17:05:50] #net-snmp <ua> We have a vested interest in seeing it succeed.
[2002/09/27 17:06:28] #net-snmp <tcaine> you can say that again.
[2002/09/27 17:06:29] #net-snmp <ua> wes: are you the project lead, the Perl Guy, or...?
[2002/09/27 17:06:43] #net-snmp <wes> sort of both.
[2002/09/27 17:07:09] #net-snmp <wes> I'm the originator of the project and I happen to dabble in whatever I need, which means pretty much everything.
[2002/09/27 17:07:33] #net-snmp <wes> SNMP.pm was originally written by Joe Marzot but he's been absent for a long time so I'm generally th only person maintaining it now.
[2002/09/27 17:08:18] #net-snmp <ua> Well, I'm just about done with the Test::More overhaul. I've got ASN, OID, agent, agent/default_store, default_store all converted. That also includes patches to the 2 'gen' files btw.
[2002/09/27 17:08:34] #net-snmp <wes> Ah, did you see my note back about that?
[2002/09/27 17:08:44] #net-snmp <ua> I'm not sure if I can adequatly say just how much Test::More kicks ass.
[2002/09/27 17:08:47] #net-snmp <ua> Yeah, I replied too. ;-)
[2002/09/27 17:11:12] #net-snmp <ua> Schwern wrote it. He has his slides for the lecture "Writing a Test Library" up. They explain how Test, Test::Hardness, Test::Simple, and Test::More came about.
[2002/09/27 17:11:14] #net-snmp <ua> http://magnonel.guild.net/~schwern/talks/Writing_A_Test_Library/full_slides/
[2002/09/27 17:11:31] #net-snmp <ua> I'm sure with your knowlege you could breeze through it in no time at all.
[2002/09/27 17:12:19] #net-snmp <ua> s/Hardness/Harness/
[2002/09/27 17:15:52] #net-snmp <wes> ua: just read your response.
[2002/09/27 17:15:57] #net-snmp <ua> *nod*
[2002/09/27 17:16:15] #net-snmp <wes> Yes, i'm sure it's perl-independent but I don't think people should be required to have it
[2002/09/27 17:16:25] #net-snmp <wes> it certainly looks very useful, however.
[2002/09/27 17:16:32] #net-snmp <ua> It's the Good Stuff.
[2002/09/27 17:16:36] #net-snmp <ua> It's the Light Side of the Force.
[2002/09/27 17:16:40] #net-snmp <wes> there *are* ways around it so that we can support both methods.
[2002/09/27 17:16:56] #net-snmp <ua> Are there?
[2002/09/27 17:17:01] #net-snmp <wes> most likely.
[2002/09/27 17:17:06] #net-snmp <ua> Would that not require forking the tests?
[2002/09/27 17:17:31] #net-snmp <wes> I need to see the patch you're creating, so submit it to the coders list so I can take a look at it.
[2002/09/27 17:17:43] #net-snmp <wes> what i'm thinking is this:
[2002/09/27 17:18:17] #net-snmp <wes> take a look at the top of local/tkmib or local/mib2c and note how it determines if you have some of the required perl modules availabel (it searches for Tk, SNMP, ...)
[2002/09/27 17:18:24] #net-snmp <wes> you can do different things based on those results.
[2002/09/27 17:18:56] #net-snmp <wes> if you don't have the new test suite, then you can fall back to doing simple ok()s or whatever but if you have the newer one then you can use the extra functionality.
[2002/09/27 17:19:50] #net-snmp <ua> I'll have the patch done in about an hour.
[2002/09/27 17:20:06] #net-snmp <wes> ok, I won't be able to look at it till monday I'm sure though.
[2002/09/27 17:20:07] #net-snmp <ua> You want a tarball or a makepatch dist?
[2002/09/27 17:20:08] #net-snmp <wes> (just to warn you)
[2002/09/27 17:20:13] #net-snmp <ua> s'ok
[2002/09/27 17:20:23] #net-snmp <ua> In that case, I may grab a beer now. ;-)
[2002/09/27 17:20:32] #net-snmp <wes> cvs diff would be the best way to go, assuming you're using a cvs checkout (which you had better be if you're sending patches)
[2002/09/27 17:20:45] #net-snmp <ua> Oh, I have access to the CVS server?
[2002/09/27 17:20:55] #net-snmp <wes> everyone has read-access
[2002/09/27 17:21:01] #net-snmp <wes> http://www.net-snmp.org/cvs/
[2002/09/27 17:21:18] #net-snmp <wes> you can check it out and then modify things and use "cvs diff" to generate a patch
[2002/09/27 17:21:25] #net-snmp <wes> that's the prefered method.
[2002/09/27 17:21:31] #net-snmp <ua> cool, you got it.
[2002/09/27 17:21:44] #net-snmp <wes> Anyway, a patch is always prefered.
[2002/09/27 17:22:07] #net-snmp <wes> a tarball will overwrite changes made between the point you got a copy of the code and other changes were made.
[2002/09/27 17:22:16] #net-snmp <wes> (ie, i've checked in a few things in the last few days to the perl tree)
[2002/09/27 17:22:35] #net-snmp <ua> right. that's why I was leery of the .tar. Also: the .tar won't delete any test.pl files.
[2002/09/27 17:22:37] #net-snmp <wes> ie, always send patches based on the code you originally got
[2002/09/27 17:23:17] #net-snmp <ua> ok, I'm going to get back to it. TTYL
[2002/09/27 17:23:25] #net-snmp <wes> thanks.
[2002/09/27 17:23:28] #net-snmp <ua> np
[2002/09/27 17:23:35] #net-snmp <wes> you may wish to send a patch before you finish so I can see what it's doing.
[2002/09/27 17:23:49] #net-snmp <ua> Fair enough.
[2002/09/27 17:23:55] #net-snmp <wes> as I said, i'm really not wishing to have people install that perl module if there is a way around it.
[2002/09/27 17:24:16] #net-snmp <ua> It's possible to build Test::More and yet not install it.
[2002/09/27 17:24:31] #net-snmp <wes> true
[2002/09/27 17:24:36] #net-snmp <ua> viz. 'make distcheck`
[2002/09/27 17:24:41] #net-snmp <wes> I don't really like redistribution though, but...
[2002/09/27 17:24:46] #net-snmp <ua> and then play games with @INC or -I
[2002/09/27 17:25:15] #net-snmp <ua> me either, but if we don't trust users to know how to install CPAN modules, it's probably the simplest solution to guarantee successful builds.
[2002/09/27 17:26:04] #net-snmp <wes> it has nothing to do with not knowing how.
[2002/09/27 17:26:04] #net-snmp <ua> anyway, we can talk about that later.
[2002/09/27 17:26:13] #net-snmp <wes> It's purely about reducing dependecies.
[2002/09/27 17:26:33] #net-snmp <wes> right now the SNMP perl module is not dependent on any external modules outside of the core perl set.
[2002/09/27 17:26:39] #net-snmp <ua> Hmm. When you put it like that.... testing really isn't a dep. Users don't *have* to type 'make test'
[2002/09/27 17:26:43] #net-snmp <wes> In an ideal world, I'd prefer to keep it that way.
[2002/09/27 17:27:08] #net-snmp <wes> actually, the CPAN module does it for you by default (but then it also would auto-install the prerequisite)
[2002/09/27 17:27:38] #net-snmp <ua> what about distributing the perl code only on CPAN?
[2002/09/27 17:27:43] #net-snmp <wes> I'll understand more when I see the patch, of course.
[2002/09/27 17:27:58] #net-snmp <wes> we went that route and it really messed people up.
[2002/09/27 17:28:08] #net-snmp <ua> I've seen one or two CPAN packages that download and compile C .src first...
[2002/09/27 17:28:11] #net-snmp <wes> the module is way too-tied to the current version of the net-snmp package.
[2002/09/27 17:28:18] #net-snmp <ua> hahaha yeah, probably
[2002/09/27 17:28:48] #net-snmp <wes> Joe and I agreed a long time ago that distributing it with the package would reduce the error reports we saw, and it did (greatly!)
[2002/09/27 17:29:56] #net-snmp <ua> got net-snmp co'd from cvs...
[2002/09/27 17:30:03] #net-snmp <wes> :-)
[2002/09/27 17:30:12] #net-snmp <wes> make sure you don't just copy the files over.
[2002/09/27 17:30:35] #net-snmp <wes> apply a patch to it generated from your current tree (or else the patch you send us will erase things I've done lately)
[2002/09/27 17:31:00] #net-snmp <ua> I'm not quite sure I understand.
[2002/09/27 17:32:26] #net-snmp <wes> for now, keep working off your current code base and use that to generate a patch, not CVS
[2002/09/27 17:32:54] #net-snmp <ua> ok, here's the part I don't understand:
[2002/09/27 17:33:05] #net-snmp <wes> what I don't want you to do is copy SNMP.pm or Makefile.PL from your current tree into the CVS checkout you just did and then use "cvs diff" to get a patch, as you're likely to remove changes I've made recently to CVS
[2002/09/27 17:33:20] #net-snmp <ua> ok, I'm with you there.
[2002/09/27 17:34:04] #net-snmp <ua> I know how to get diffs of individual files, what I don't know is how to get diffs of directories, adding and deleting files.
[2002/09/27 17:34:28] #net-snmp <ua> <-- ign'rnt
[2002/09/27 17:34:34] #net-snmp <wes> diff -r I think handles this.
[2002/09/27 17:34:49] #net-snmp <wes> ACTION always uses cvs which does the right combination of diff -r or whatever...
[2002/09/27 17:34:57] #net-snmp <ua> looks like it does
[2002/09/27 17:35:02] #net-snmp <ua> cool, you got it.
[2002/09/27 17:35:49] #net-snmp <ua> ACTION & # hacking
[2002/09/27 17:43:28] #net-snmp <ua> diff is en route.
--> df has joined #net-snmp
--> wes has joined #net-snmp

Powered by: