2004/08/09
IRC [09:21] <Dossy> Morning.
IRC [10:04] *** iharding joined the chat.
IRC [10:08] <iharding> \?
IRC [11:40] <jhavard> whoa, dossy's back.
IRC [12:37] <Dossy> yup.
IRC [12:37] <Dossy> hi ian!
IRC [12:37] <Dossy> yay.
IRC [12:43] *** frodoroot joined the chat.
IRC [12:52] <Dossy> hi nate.
IRC [12:52] *** bartt joined the chat.
IRC [12:52] <frodoroot> hey Dossy
IRC [12:53] <frodoroot> I'm having nsopenssl problems
IRC [12:54] <Dossy> yeah, i see
IRC [12:54] <Dossy> ^^^^^^^^[5] [12:54PM] Dossy (+i) via localhost on #aolserver (+nt) ^^^^^^^^^^^^^^^^^^^^
IRC [12:54] <Dossy> [5] [12:54PM] Dossy (+i) via localhost on #aolserver (+nt) ^^^^^^^^^^^^^^^^^^^^
IRC [12:54] <Dossy> [5] [12:54PM] Dossy (+i) via localhost on #aolserver (+nt) ^^^^^^^^^^^^^^^^^^^^
IRC [12:54] <Dossy> [5] [12:54PM] Dossy (+i) via localhost on #aolserver (+nt) ^^^^^^^^^^^^^^^^^^^^
IRC [12:54] <Dossy> [5] [12:54PM] Dossy (+i) via localhost on #aolserver (+nt) ^^^^^^^^^^^^^^^^^^^^
IRC [12:54] <Dossy> [5] [12:54PM] Dossy (+i) via localhost on #aolserver (+nt) ^^^^^^^^^^^^^^^^^^^^
IRC [12:54] <Dossy> ack
IRC [12:55] <Dossy> crap-ass lag
IRC [12:55] <jhavard> So, where have you been? Vacation?
IRC [12:55] <Dossy> yup.
IRC [12:55] <Dossy> All last week.
IRC [12:55] <frodoroot> all I see are ^^^
IRC [12:55] <frodoroot> where'd you go?
IRC [12:55] <Dossy> Hot Springs, Va.
IRC [12:56] <jhavard> And you didn't get HEAD to a can-handle-more-than-64-hits state before leaving. bastard.
IRC [12:56] <Dossy> I know.
IRC [12:56] <frodoroot> heh
IRC [12:56] <Dossy> Jim's doing some more work on HEAD -
IRC [12:58] <Dossy> there's the one bug I just couldn't find.
IRC [12:58] <Dossy> grr, my connection through the VPN has been completely flaky all day today
IRC [12:59] <jhavard> the "p" stands for "packet loss"?
IRC [12:59] <Dossy> seriously.
IRC [12:59] <Dossy> i thought it was the 802.11b wifi suckage but i'm now on hardwire and it's still crap
IRC [13:23] <Dossy> AURHGHG
IRC [13:23] <Dossy> i can't get at my email. ughmf
AIM [13:50] *** GizmoBeastLives joined the chat.
AIM [13:54] <GizmoBeastLives> I've done some cleanup and enhancement of Eric Lorenzo's tclprof module, you can see it running at http://www.jamierasmussen.com/_tclprof
AIM [13:54] <GizmoBeastLives> I reused a bunch of the stats.tcl code to have a similar style interface with sortable columns, etc.
AIM [13:55] <GizmoBeastLives> The changes I made were pretty extensive and include:
AIM [13:55] <GizmoBeastLives> - Some changes to follow AOLserver conventions. Renamed to nstclprof, source reformatted, fix small AOLserver 4 and WIN32 build issues and compiler warnings, add version number.
AIM [13:55] <GizmoBeastLives> - Use config file section "ns/server/$server/module/nstclprof" instead of "tclprof/parameters"
IRC [13:56] <Dossy> cool
IRC [13:56] <Dossy> oh, it's only URL-levle?
IRC [13:56] <Dossy> er level?
AIM [13:56] <GizmoBeastLives> - API restructured. Now adds just one command, ns_tclprof
IRC [13:56] <Dossy> oh, you can drill down
AIM [13:56] <GizmoBeastLives> You can get a complete trace as well
AIM [13:57] <GizmoBeastLives> I also fixed a number of crash bugs and added some new features
IRC [13:57] <Dossy> all the times are 0.00000
IRC [13:57] <Dossy> does it use clock clicks -milliseconds or whatnot?
AIM [13:57] <GizmoBeastLives> I have a fast machine :-)
AIM [13:58] <GizmoBeastLives> one sec
AIM [13:59] <GizmoBeastLives> It uses Ns_GetTime, I'm not sure how good the resolution of that is
IRC [14:00] <Dossy> oh ugh
IRC [14:00] <Dossy> hmm
IRC [14:00] <Dossy> well times in full seconds 0.00000 is kinda annoying.
IRC [14:00] <Dossy> can we do it all in one unit
IRC [14:00] <Dossy> like msec?
IRC [14:00] <Dossy> an integer rep number of msec or somesuch?
AIM [14:00] <GizmoBeastLives> my site has very little dynamically generated content at the moment, the module is more useful on our development servers, which are not publicly available to show people unfortunately
IRC [14:00] <Dossy> you going ot import nstclprof into CVS?
AIM [14:01] <GizmoBeastLives> I would like to import first Eric's version and tag it and then mine on top
IRC [14:01] <Dossy> where does it store the "Recent requests" data? NSV
IRC [14:01] <Dossy> yeah, sounds good
IRC [14:01] <Dossy> but still import it into nstclprof dir
AIM [14:02] <GizmoBeastLives> The module is C-based, only some of the interface code taken from stats.tcl
IRC [14:02] <Dossy> please god, let fanfare start up right this time
AIM [14:03] <GizmoBeastLives> The C code doesn't use NSVs internally
IRC [14:03] <Dossy> does it register a proc, or is the UI an ADP?
IRC [14:03] <Dossy> interesting. mutex locks around a C structure?
AIM [14:03] <GizmoBeastLives> It isn't really an ADP even though it uses that ext, reg. proc like _stats
IRC [14:03] <Dossy> aww
AIM [14:04] <GizmoBeastLives> Yes to mutex locks around C structure
IRC [14:04] <Dossy> i'd like to get back to having an /NS/Admin-like setup
IRC [14:04] <Dossy> with ADPs rather than registered procs
IRC [14:04] <Dossy> I'm not 100% sure what the real cost/impact of having lots of registered procs are, but it's GOTTA be more than using ADPs :})
AIM [14:04] <GizmoBeastLives> It wouldn't be too bad to separate into two different files
IRC [14:05] <Dossy> the difference is using registered procs defined from tcl modules, it's predictable where the files live
IRC [14:05] <Dossy> whereas with ADPs, you have to tell people where/how to install them
AIM [14:06] <GizmoBeastLives> sure. the stats file also has an independent perms check that we've conditionalized so that you can just use nsperms with it and nstclprof
IRC [14:06] <Dossy> hmm
IRC [14:07] <Dossy> yeah, next thing we need to do is implement nsauth and kill nsperm
AIM [14:07] <GizmoBeastLives> that would be nice
IRC [14:07] <Dossy> with 4.1's new pre-queue stuff, nsauth would just be a registered prequeue callback, I think
IRC [14:07] <Dossy> should be REAL easy
AIM [14:08] <GizmoBeastLives> I'm looking forward to trying 4.1 out. I still haven't had time to look into making it build on Win32
IRC [14:09] <Dossy> it builds on win32 just fine with the build.tcl scrit
IRC [14:09] <Dossy> er script.
AIM [14:09] <GizmoBeastLives> Really? when did you last try that?
IRC [14:10] <Dossy> oh
IRC [14:10] <Dossy> i haven't updated it for the new limits.c and pools.c --
IRC [14:10] <Dossy> need to add those two ... yeah
IRC [14:10] <Dossy> but before the recent commit, it definitely worked
IRC [14:10] <Dossy> if you see in sourceforge, i posted 4.1.0 win32 binaries
AIM [14:11] <GizmoBeastLives> yeah, I meant the most recent checking
AIM [14:11] <GizmoBeastLives> *checkin
IRC [14:11] <Dossy> yeah
IRC [14:11] <Dossy> should be easy, just add those two things to the build.tcl
AIM [14:12] <GizmoBeastLives> I think there may be other problems, but I'll get back to you when I've actually investigated
AIM [14:22] <GizmoBeastLives> ah, some preliminary problems are with driver.c, STATUS_PENDING conflicts with its use in winnt.h, where that is a #define, I think service-related
AIM [14:22] <GizmoBeastLives> also with use of mmap
IRC [14:23] <Dossy> eek
IRC [14:23] <Dossy> use of mmap should be guarded with #ifndef _WIN32
IRC [14:23] <Dossy> but STATUS_PENDING conflicts with winnt.h? uh
IRC [14:23] <Dossy> ugh
IRC [14:23] <Dossy> is that winnt.h our code, or windows standard .h that's getting included?
AIM [14:24] <GizmoBeastLives> that's a windows standard header file, I think it get included by ns.h
IRC [14:25] *** aolserver-scribe joined the chat.
IRC [14:25] * aolserver-scribe is logging
IRC [14:26] <bartt> Yay, freenode allowed the logging bot back in.
IRC [14:29] <iharding> Is the mailing list admin here?
IRC [14:30] <iharding> My messages are silently dropped and I can't seem to signoff to try re-subscribing.
IRC [14:34] <Dossy> iharding: Yes.
IRC [14:34] <Dossy> ian: Welcome to the chat.
IRC [14:35] <Dossy> Ian, you're subscribed as: ianh@TPCHD.ORG Ian Harding
IRC [14:36] <iharding> Hmm. That should be OK. That is the reply-to in my emails, although I receive most of my mail as iharding...
IRC [14:36] <frodoroot> Dossy: any ideas on the nsopenssl thing? is scottg interested in this or is it an aolserver problem?
IRC [14:37] <bartt> frodoroot: Have you tried the latest nsopenssl CVS tagged version instead of HEAD?
IRC [14:38] <frodoroot> yesyesyesyes
IRC [14:38] <jhavard> How does one get a list of tags from cvs?
IRC [14:38] <bartt> http://cvs.sourceforge.net/viewcvs.py/aolserver/nsopenssl/
IRC [14:38] <bartt> See drop down at bottom of the screen
IRC [14:39] <frodoroot> bartt: oh wait instead of head? I just did cvs co nsopenssl
AIM [14:39] <GizmoBeastLives> Are you able to replicate the problem with a particular browser version? I found a bug with old 40-bit browsers that I filed a SF bug report on.
IRC [14:39] <bartt> frodoroot: I've been running v3_0beta21 of openssl w/o problems.
IRC [14:39] <frodoroot> bartt: with aolserver 4.0.7?
IRC [14:40] <bartt> cvs co -r v3_0beta21 nsopenssl
IRC [14:40] <tekbasse> I see the same errors nate and janine see. for aolserver4.0.7, nsopenssl3beta21 on freebsd 4-10 stable
IRC [14:40] <bartt> Yes, w/ aolserver 4.0.7
IRC [14:40] <bartt> But then again, this is on relatively low volume sites.
IRC [14:40] <jhavard> ah, rlog command.
AIM [14:40] <GizmoBeastLives> I ran into problems with the HEAD version going into an infinite loop, there were some messages to the list about this
IRC [14:41] <frodoroot> mine is a high volume site
AIM [14:42] <GizmoBeastLives> I hacked in a patch that seemed to fix the problem, but I don't know why it is happening, and I haven't heard back from Scott about it
IRC [14:43] <bartt> frodoroot: Do you have a recipe to replicate the nsopenssl problems you are seeing?
IRC [14:43] <frodoroot> there are so many variables, hmmmm
IRC [14:43] <frodoroot> I am Nate BTW
IRC [14:44] <tekbasse> I am torben btw =)
AIM [14:44] <GizmoBeastLives> I am Jamie
IRC [14:45] <bartt> Hi Nate, I usually use handles in conversations like these so that people receive alerts when they are being addressed.
IRC [14:46] * bartt is Bart Teeuwisse, see /whois bartt
IRC [14:47] * jhavard is havard.
IRC [14:47] <jhavard> I have only one name.
IRC [14:47] * tekbasse looks for a deeper sense of self. no /whoami
IRC [14:47] <bartt> Anyway natexor_, let me know if there is anything I could do to replicate the problem on one of my sites.
IRC [14:48] <natexor_> hmm I'm going to try to replicate it on another machine in order to find out the steps
IRC [14:49] <Dossy> ian: Reply-To: doesn't count. what matters is the From: header
IRC [14:49] <Dossy> nate: i haven't heard from scottg in ages. i'll try to find some time to look into it ...
IRC [14:50] <Dossy> nate: you just can't decide on a /nick huh?
IRC [14:50] <natexor_> Dossy: nate was taken
IRC [14:50] <Dossy> and frodoroot?
IRC [14:50] <Dossy> heh
IRC [14:50] <frodoroot> oh yeah
IRC [14:51] <frodoroot> Dossy: Janine is having this problem too
IRC [14:52] <iharding> Ah. They decided I am iharding, so they created me a mail alias, and probably broke the FROM header.
IRC [14:54] <iharding> I will send myself an email real quick and see who it's from.
IRC [14:58] <Dossy> nate - yeah, I know
IRC [14:58] <Dossy> iharding: please do.
IRC [14:58] <bartt> Janine is also running CVS HEAD of nsopenssl
IRC [14:58] <Dossy> yeah
IRC [14:59] <Dossy> first i want to get AOLserver 4.1 HEAD working again
IRC [15:00] <iharding> All appears to be well: From: "Ian Harding" <ianh@tpchd.org>
IRC [15:00] <iharding> Must be operator error.
IRC [15:00] <Dossy> send me an email to dossy@panoptic.com
IRC [15:00] <Dossy> let me look at the headers
IRC [15:01] <iharding> OK
IRC [15:01] <Dossy> hopefuly it won't get nuked by my spamfilter ... hee
IRC [15:02] <iharding> I shouldn't have mentioned V * I * A * G * R * A in it, I guess...
IRC [15:15] <Dossy> strange.
IRC [15:15] <Dossy> From: "Ian Harding" <ianh@tpchd.org>
IRC [15:15] <Dossy> LISTSERV should have no problem with you. what's the bounce message you're getting?
IRC [15:15] <iharding> I get nothing. My messages just don't show up.
IRC [15:16] <Dossy> you see any bounces from MAILER-DAEMON on your local mailserver?
IRC [15:17] <Dossy> Do you know who your mail server admin is? Could you ask him to inspect the logs?
IRC [15:17] <iharding> OK, I will see where that leads.
IRC [15:17] <Dossy> I used to see mail from you as iharding@tpchd.org --
IRC [15:17] <Dossy> when you send mail to the listserv address, could your mail server be rewriting the from: to come from that addr instead?
IRC [15:18] <iharding> I dunno. Things are kinda messy here. Maybe I should subscribe from home.
IRC [15:18] <Dossy> not a bad idea.
IRC [15:20] <frodoroot> ok I'm going to try nsopenssl3.0beta21
IRC [15:22] <Dossy> just got a series2 tivo ... playing with the Home Media Option now ...
IRC [15:23] <Dossy> think I could implement nstivo module so you can use aolserver as a tivo beacon server + media server :)
IRC [15:34] *** tekbasse parted the chat.
IRC [15:40] <Dossy> everyone fall asleep?
AIM [15:41] <GizmoBeastLives> Just working. Crazy Mondays :-)
IRC [15:45] <iharding> Anyone using NetBSD?
IRC [15:45] <iharding> Or pkgsrc?
IRC [15:47] <bartt> Nope
IRC [15:49] *** tekbasse joined the chat.
IRC [16:13] * tekbasse fixed his aolserver4.0.7/nsopenssl3b27 prebind problem... by commenting out the outboundssldriver in his homemade config.tcl file =)
IRC [16:29] <frodoroot> what outboundssldriver?
IRC [16:32] <tekbasse> I made a ns/server/${servfer}/module/nsopenssl/ssldriver/outboundssldriver but ssldrivers are for server contexts (inbound) only as I re-read the config.tcl file comments
IRC [16:33] <tekbasse> not client contexts
IRC [16:34] <tekbasse> also, looks like maxconnections is not used. Am reviewing bartt's nsd file for hints: http://www.mail-archive.com/aolserver@listserv.aol.com/msg07092.html
IRC [16:35] <frodoroot> I have a ssldriver/users section
IRC [16:36] <tekbasse> users = server context, should be fine
IRC [16:40] <tekbasse> I wonder if the upper/lowercase letters in the ns_params matters, such as serverlog vs ServerLog. It doesn't seem to (log works), but maybe for not all cases?
IRC [16:43] <bartt> Never matters
IRC [16:47] <tekbasse> thanks, bartt!
IRC [16:56] <tekbasse> Hmm.. does the order of ns_section configurations matter? ie ns_section ns/servers before ns_section ns/server/${server}
IRC [17:02] <bartt> yes, it generally does
IRC [17:08] <tekbasse> aha! also, noting no ns_section ns/server/${server}/tcl section.. in the posted config.tcl.. will try w/o it later. bbl
IRC [17:19] *** iharding parted the chat.
AIM [17:27] <GizmoBeastLives> I have a TCL internals question. I have some code that calls the MD5 tcllib package, which does math with rather large numbers. (It is actually using the tcllib smtp, with uses the tcllib mime, which uses md5)
AIM [17:28] <GizmoBeastLives> Although I can't find anything that has changed, the md5::md5 routine is now throwing "integer value too large to represent" on one of my servers
AIM [17:29] <GizmoBeastLives> So, question one is:
AIM [17:29] <GizmoBeastLives> set AA [expr {-758138321665285842 + 0}]
AIM [17:29] <GizmoBeastLives> set A [expr {6401081433749451061 + 0}]
AIM [17:29] <GizmoBeastLives> incr A $AA errors
AIM [17:29] <GizmoBeastLives> set A [expr {$A + $AA}] works
AIM [17:37] <GizmoBeastLives> Question 2: MD5 uses the incr method and this is the source of the error. However it only causes the error on some servers. On a server that is still working, the incr throws the error when I evaluate it one context, but not when it is called within the context of the MD5 proc (!)
AIM [17:37] <GizmoBeastLives> The only thing I can think of is that something in the MD5 context is causing Tcl to do the string -> integer conversion differently
AIM [17:40] <GizmoBeastLives> Anyone ever seen anything like this? The Tcl version is the same on both servers - 8.4.4 - though the AOLserver version is a bit different. In the case where it has started failing for no apparent reason TCL was built with debugging symbols.
IRC [18:00] *** frodoroot parted the chat.
AIM [18:08] <GizmoBeastLives> OK, the good news is that I figured out what changed and triggered this problem. The bad news is the change was loading nstclprof on the server. nstclprof does a rather unfortunate thing with if / for / while that must be the root cause of this problem. Another thing to look into before the module is ready for production use I'm afraid.
IRC [20:25] *** aolserver-scribe parted the chat.
IRC [20:30] *** bart1 joined the chat.
IRC [20:32] *** bart1 parted the chat.
IRC [20:32] *** bart1 joined the chat.
IRC [20:41] *** bartt parted the chat.
AIM [21:14] *** GizmoBeastLives joined the chat.
AIM [22:13] *** UD Rules All joined the chat.
AIM [22:13] <UD Rules All> hello
AIM [22:14] *** UD Rules All joined the chat.