Bug fixes, new features, etc..

• fix crash with qsort

• fix for crash/coredump with DNS check
• Darwin/OSX 10.4.x build cleanups
• reduce calls to time()
• new queuer for objects (may not work on all systems)
• numerous cleanups
• some excess noise reduced in logs
• add some checkpointing of icmp packets and timing
• minor dependency builder fixes
• allow sets to work in the root = object declaration

• fix bug where numfailures would get out of sync with some objects
• cleanup some very verbose logging
• rename yylex sysmon_conf_yylex
• cleanup unused variable in icmp check
• cleanup of some "MODE xml" SHOWOBJ output
• changed stale/killed check to only log once/second
• many warnings cleaned up -Wall
• pingv6 check implemented, tested (MacOS 10.3, linux, freebsd)
• fix segfault if no description configured
• change icmp hold queue size calculation
• autodetect max number of checks that can be done at once
• add MODE outagelog to tcp/1345 sysmon service
• add pragma no-cache to html file
• fix for snmp based checks and SIGHUP state
• dns check fix for double free()
• FreeBSD malloc/abort() recursive fix
• more SNMP build cleanup
• queue based off of next_queuetime instead of doing math each time around
• NETSNMP_CALLBACK_OP_RECEIVED_MESSAGE vs RECEIVED_MESSAGE compile issues

• crude 32-bit vs 64-bit counter hack code
• fix net-snmp compile dependency, if you get error about RECEIVED_MESSAGE in snmp.c when compiling, update your net-snmp version, they now use NETSNMP_CALLBACK_OP_RECEIVED_MESSAGE
• fix autodetection of openssl, need to search /usr/local/ssl
• fix autodetection of snmp support (redhat users, you need: net-snmp-devel and openssl-devel installed)
• snmp-rate, snmp-octets support, snmp fixes
• oh wait, need to parse config for pause, resume and stop to get alternate pidfile path
• clean up gethostname calls to reduce cpu
• don't log more than once if not queued in a long time
• make snmp be more graceful when it encounters an error instead of creating a core file

• print out if snmp support was detected/used
• remove spurious logging for objects without a contact
• log hostname being queried when gethostbyname() takes more than 1 sec
• no need to parse config to send pause, resume, or stop
• fix some parser memory suckage and information spillage btw objs
• statuschanged needs to be reset to false after textfile write
• minor code audits w/ rats
• remove unused buffer in clientstruct
• autodetect gethostbyname2
• force numfailures = 1 when doing snmp_type_reboot/sysuptime query

• fix sighup semantics for snmp queries, two queries to same host could cause data mixup when sighup is caught
• fix coredump related to snmp queries
• increase snmp reliabiltiy
• fix parser for snmp-high, snmp-low, snmp-exact to make them consistent
• add localized debugging to snmp subsystem
• fix signedness in parser for snmp-high snmp-low and snmp-exact
• include strings.h in config.h to quiet -Wall
• remove unused variables in srvclient.c/do_client_http
• fix some information leakage btw objects in parser

• bugfix for memory leak when lots of sighups are sent to process

• pop3 fix
• basic ssh support from Patrik Henningsson

• fix some poor parser semantics that would cause objects to improperly inherit previous object values
• add dns query -- type dns; dns-query "example.com";
• fix noheartbeat typecast and code
• minor changes on html page and how we display our version
• add config savestate "/path/to/file";
• catch some signals more to provide a bit more coredump info if necessary
• fix 'make register'
• fix 'CONF'

• update copyright and license information for net-snmp and ssl req
• add SHOWOBJ and MODE xml

• print filename in errors
• start to add xml cli hook
• add -l command line option (no syslog)
• fix initalization of upct variable
• fix problem with queueinterval

• fix issue with spawn
• add per-object queue timing
• add warning if hostname contains a space

• fix for icmp loss/checksum bug

• fix log_orphans() and loadconfig() to work as intended
• integrate Leon patches into cvs
• clean up a few very minor bugs

• add config and code for notify on down/up only or both
• count up-cycles as well as down-cycles
• add formatting codes for up count, last recovery, service state
• add up-count in status pages
• add config for HTML refresh meta-header
• tidy the resulting HTML output (eg consistent case in tags)
• make memory debugging work for strdup() too
• add config for PID file (/etc/ lives in readonly land here)
• convert some random numbers to symbols (e.g. SYSM_OK, TRUE)
• add commandline keywords pause/unpause stop/shutdown/suicide
• document new and existing commandline options
• collapse disparate signalling functions into one
• make signal flag variable names consistent
• collapse no-contact-address code in page_someone()
• canonicalise existing X-Sysmon-unique-ID mail header
• add X-SysMon-Host, X-SysMon-Uptime mail headers
• add config for date (config date "%x %X";) use strftime codes
• add date "format" shorthand ISO="%F %T" and DEC="%d-%b-%Y %T"
• reformat some diagnostics to be much more informative
• update man pages a bit
• update the HTML documentation
• formalise a stray entity in the Changes.html file
• make sysmond tell about all of its options

• add new translation feature %c - count
• fix debugging
• client tty fix
• lots of mini patches from john heasley
• more radius fixes from john heasley
• config converter by steve neighorn and john heasley
• radius fix
• fix some snmp bugs and parser related snmp issues

• some basic snmp support
• fix memory leak in parser
• fix some excess config parsing time
• add some warnings that can help diagnose config file problems

• fix bug whereby pings would not be sent

• fix some warnlog+killafter stuff
• change ipv6 support for non-autodetect
• warn if no flex
• warn if no yywrap found
• fix for saddr, daddr on FreeBSD & Solaris

• fix parser to make it a bit more tolerant of irrelevant whitespace
• disable auto-utilization of ipv6
• fix bug related to icmp rework that caused it to never count icmp replies (in my lab at least)
• fix some icmp.c vs icmpv6/pingv6.c semantics in error messages
• add some of the hooks for customizing number of pings to send and receive
• ability to disable listenport - set it to 0 on cli

• fix for build issue
• fix segfault when configuring custom colors.

• division by zero coredump thanks to Bengt Gördén
• lots of IPv6 voodoo started to add ipv6 ping test.
• fix solaris lex issue
• make ./configure detect IPv6
• fix some getuid/geteuid semantics that would cause pid file to not be written

• change holdqueue managment stuff
• change default maxqueued from 100 -> 75
• some tweaking of the ping test
• add some logging if there are a lot of objects and we can't queue new checks after some time.
• fix bug with pmesg where we would core if one wasn't configured in /etc/sysmon.conf
• make sensitivity a per-host configuration option
• make paused data visible in html file
• warn if more than one configured root

• fix for root = statement to be more tolerant
• warn for multiple configured roots
• fix for uninitalized variable
• fix for percentage uptime to be correct
• fix for improper variable casting
• fix for sysmon check not detecting non-responsive daemons

• prototype some snmp-trap stuff
• log unknown client messages
• fix port number handling of checks that don't use ports (ping)
• fix for embedded \n in parser causing line_number info to be wrong

• fix more parser issues

• fix some problems w/ crashes in invalid config files

• fix parser bug that would cause us to not compile

• fix bug in smtp check that would cause it to be killed
• fix bug in parser with missing quotes
• fix bug when we catch sighup
• fix bug in dnscache code that would cause us to core
• fix some parsing issues that would cause us to core dump
• insure that port number is set to zero
• make us portable to real lex instead of just flex
• fix multi-line imap banner (heas)
• fix casting issues (heas)
• fix comment handling (heas)

• fixed portability
• fix http check
• fix only one object in config file case
• fix some more memory handling, and make MALLOC() not memset memory to zero anymore

• fix some memory leaks and improper accesses
• fix typo in parser.l that would cause properly formatted objects to not be marked as valid.
• Fix segfaults in the parser
• Fix logging bug - Null pointer access in the case of fopen() failing for the file we should be logging to.

• remove a ton of internal timing checks and calls to gettimeofday()
• if we are initally starting up the daemon, use specified logging immediateley. Useful when logging to a file instead of via syslog for debugging info that will impact syslogd performance when we can locally optimize it.
• insist that they specify a radius secret if it's a radius check
• insist that they specify a username and password for checks that require it
• fixed parser issue where we would free data before it was supposed to be.
• added -q command line option to daemon to be quiet when sending sighup from "sysmond reload"
• last message-id not preserved after sighup
• fix for http check core dumping when parser should reject object
• fix for whitespace between object and object name.

• set variable = "value"; now works in 0.90 series
• we now log "orphan" objects (that have no relationship)
• fix comment handling
• fix for a few parser issues

• zero is never greater than zero. this is a bug. we fixed comparison issue.

• config logging file /path/to/file
• fix for stale textfile
• maxqueued was being ignored
• improper port number stuff going on caused sighup to not copy the state over
• if a hostname is not valid, we didn't reject the object
• fix problem with copy_alerts when we catch sighup
• fix bug that would cause crash if no contact configured
• update text output by configure

• releasing to get the bugs out
• fix numerous crashes/bugs in 0.89.99
• generate unique_id
• fix bug if configed_root = null
• make install now saves old version as sysmond.old and sysmon.old
• make install no longer will get Text file busy message

• finished porting all options (Except generic sets) to new parser. Released to public for testing.

• fix bug in adjacency builder (stupid me, trying to be too smart)
• rewrite queuer from 0.89->now to fix dep handling
• remove the pid file when we die

• fix for LinuxPPC under RedHat-6.2

• added http/host-header support. thanks to patrice@mac.com for the patch
• NEW config file format. Must update, no backwards compatability
• rewrote the config file parser
• reworked the internal engine to use a graph

• fix for some excessive logging
• fix for some minor smtp sanity for possible bug
• fix for NULL pointer access in dnscache.c
• compile fix for OpenBSD

• update configuration options in man pages
• fix build issue related to curses

• attempt to fix problem with pid file being wrong
• fix memory leak with icmp ident hash
• fix memory leak in dns cache code
• fix in lib.c where we were possibly corrupting the stack
• fix massive memory leak in page.c where we do not free() strdup()'ed strings
• fix memory leak in config parser
• free dns_cache when sysmond is stopped
• modify internal MALLOC to take a description for memory leak watching
• fix bug in SMTP check. If we read 0 bytes from socket, we would not know, and would clobber a pointer because we were messing up some of the memory space.
• rework some internal FD managment and storage to increase performance
• sysmond pause added. Will pause monitoring until you send a unpause signal.
• fix for pages going out to smartlist mailing lists, and sendmail -f
• sysmond reload added. Parses config file and reloads if it works without a problem.

• added "config authkey"
• fixed divide by zero case in new %r feature
• fix for numfailures
• addition of %V to pmesg
• fix/feature for pmesg to handle \n and \r
• new feature in pmesg %r for reliability of the site
• fix for 'pmesg' handling. would overwrite some of the message depending on order of options.
• added config header header headerval
• fixes for radius check
• misc buffer overflow fixes (increased buffer size that snprintfs dump stuff into.. would discard some data in cases)
• fixes for nntp and http check to reduce chances of internal-killed messages

• fixes for curses so it should actually work properly now (sorta)
• MAX in config.h
• holdqueue fix in icmp.c
• radius.c fix for typo

• fixed bug where sysmond would crash if monitoring radius server and got an error from read()

• added radius
• added new command line option, -i to disable icmp if not running as root
• different rand number seeding
• different randchar calculation
• added gen_rand_ascii(char,len)
• loadconfig speed tweaks
• loadconfig misbalanced braces warnings
• syswatch element counter stores number now for later use
• fix problem with too many icmp checks going on at once
• fixed improper logging with dnscache code
• [PORT] Apple OS/10 Server port. Now builds on this OS

• added -w (warnlog) toggle option. can also do with SIGUSR1
• rework some more dnscache issues
• fix possible overflow issue related to dnscache code
• fix some queueing semantics
• fix for possible dnscache bug in lib.c (possible precision loss)
• fix for possible issue with icmp code
• fix for check of smtp server on NT that returns multi line banner
• logging cleanups
• fix for $replacement stuff with spawn
• fix for -n to cause it to not spawn
• added %p (port number)
• icmp.c: rewritten ICMP packet catcher
• icmp.c: improperly ignoring return value of setsockopt
• patch for broken dnscache code
• do'h missing IMAP check code in lib.c when parsing config file
• added zombie collect function -- should fix zombies with spawn on fail, and some other things.
• added warn_dnscache_lameness()
• we printed WARNING: %s not checked in %d seconds when the parents were down, so we could not have checked them.
• fix for SYSMON_HOST in curses client not being able to be overiden on command line.
• fixed bug in tcp check that would cause checks to return "Internal-Killed" when they should have been timedout.
• Added warning message that will appear in logs and on console if something is not checked at 3*queuetime
• Made update for folks that are checking *really* large networks that weren't being monitored properly. new config option resulted. config maxqueued # to increase the number of checks that can be going on at once.
• updated some of the way stuff is displayed in html file if a site has never been up
• made it so you can configure http check as http:8080 to check web server on port 8080 instead of port 80
• added a warning if you do not have a whitespace before { on a config file line to aide in debugging config files
• fix for a logic issue. we now only show the host that is down instead of all the "children" that depend on that host in the client, and in the html file. if you don't like this, let me know and i'll put in a option for it.
• fix for FreeBSD ncurses so we will compile
• fix for nonexistant ssh check causing core when folks have attempted to configure it

• adjusted hold queue setting format to help with memory use
• added "make register"
• added make register, and e-mail help contact information to configure script
• adjusted the icmp.c packet size to help with possible memory problems
• fixed bug in loadconfig.c where we used wrong print format and caused a segfault/core when there was invalid hostname
• fixed stack clobering feature when we expanded variables set with set varname = "string", and expanded them to somethign *really* much longer than we could handle -- this made debugging complicated as there was a corrupt stack which meant cores were not useful at all
• added reading env var of SYSMON_HOST in client as for which machine to connect to.
• fixed a bunch of debugging
• hacked a fix around possible buffer overflow reading config file
• Added message-id generation code, so mailers w/ threads can view sysmon messages with threads
• fix of missing config file name, and line number when there are unknown hosts.
• fix for zombies if using spawn
• possible fix for http check sucking up lots of cpu
• fix for RLIMIT_OFILE in hpux
• added config replyto user@host
• fix bug in tcp check that would cause core in rare circumstance
• changed icmp.c/IHL calculation
• fix for people whose sizeof(struct timeval) != 8
• minor fixes of prototypes in (v)snprintf
• more patches from jhbrown@ai.mit.edu for sysmon to work again on an OSF/1 box
• fixed another core related to null contact. patch by jhbrown@ai.mit.edu
• fixed sysmon so it wouldn't hold on to socket after its death .. patch by jhbrown@ai.mit.edu
• added logging of improper smtp banner when this occurs
• added sysmon check. Useful for monitoring others sysmonds
• fixed bug with null contact causing crash/core
• added "config queuetime 500" where 500 is in seconds
• fixed bug with icmp check and possibly stopping under a rare circumstance
• fixed bug with long contact strings.
• fixed a bug with config showupalso
• fixed something with config nosubject
• fixed downct = 1, when statechange, and statechange = UP
• added %T to do current time hh:mm:ss
• added config nosubject
• added set sender = "username@host"
• adjusted things so errors would be properly printed and syslogged
• fixed problem with last notified time not being copied when doing sighup, caused dup pages
• set subject = "%d is %w" etc..
• fixed printing of messages in pinger unless they're unknown
• made http a valid check type -- basically an alias to www check
• fixed problem that would cause http check to say killed-internal when it should not have, and should have reported No Response
• fixed problem that would make curses client broken
• made sighup set statuschanged = TRUE
• made default statuschanged = TRUE
• fixed some whitespace we were inserting in the html files that could cause funky formatting
• more movement to print_err for proper logging of errors
• fixed $variable,otherdata such that it works as one would expect
• Updated "CONF"
• Fixed sighup stuff where it was set improperly to TRUE at startup, so it would load teh config file twice

• added set randomvariablename = "random thing" then you can use $randomvariablename throughout your config file
• changed abort -> ABORT
• fixed problem with tcp ports not reporting properly and reporting as killed instead of their real error (solaris)
• fixed problem in umichx500 check
• fixed problem with killed checks eating up fds and not giving them back
• moved lots of stuff to print_err
• created ABORT, so it will log when (and where) it dumps core, need to change all aborts() over
• added print_err
• changed dnscache to use timeval instead of time_t

• fixed possible problem in http check w/ down sites
• colors in html file are now configurable (ex:) set downcolor = "ffffff" set upcolor = "ffffff" set recentcolor = "ffffff"
• removed some unused code
• allowed config file changing of PMESG: set pmesg = "whatever you want"
• added pid file /etc/sysmond.pid -- not written if not root
• cleaned up some inconsistent code
• fixed bug with %H and no reverse dns for an ip address (fix your reverse dns people! ;)
• changed PMESG default
• changed ping test to use non-blocking socket IO
• did some work in getline_tcp to help out
• added %H which does gethostbyaddr in PMESG translations
• limited the number of checks that can be going on at once properly
• added set_defaults
• did some optimization to reduce the number of times we call time()
• did some more variable initalization to take care of possible memory corruption issues
• added %I (ip address) and made it part of PMESG on default
• Spawn on fail now works, will e-mail output of program started to contact instead of outage message.
• Changed a few things w/ the way curses detection was done, and fixes some brokeness of autoconf in relation to it.
• Added negation to test type (ie, instead of [type] there is now an optional minus flag you put at the beginning of test. This says only follow children (in curly braces from that host) if it is down. ex: puck.nether.net -tcp 80 puck.www e-mail
• Fixed a NULL pointer assignment in the config parser with included config files
• Optimized the http check some
• made it so you can build dependencies on non-ping lines in your config file
• did some overhauling of config file parser, still needs major rewrite

• Fixed config option "showupalso"
• Fixed config option "noheartbeat"
• Changed semantics for how to deal with empty config file (thanks to Lars Kellogg-Stedman for finding these three)
• Added in patch for port to Unixware
• Fixed potential core in icmp.c
• Changed handling of empty config file

• Fixed some PMESG semantics, PR#42, PR#43, thanks to jpike for finding these.
• Fixed minor quote handing issue which would cause words to be cut off in some cases. PR#44

• made some parser fixes/upgrades, PR#6
• made more debugging available via debug
• fixed bug that would cause 0.78.4.20 to die
• http check broken in beta versions fixed in 0.78.4.21
• fixed sigHUP PR#2
• updated the java client PR#35
• fixed some talktcp.c issues where connect() would return without an error so quickly, we got confused. PR#34
• fixed timing issues in curses client. PR#15
• added tracking db - http://puck.nether.net/cgi-bin/sysmon.cgi
• change to make writing of text file info less frequent to avoid lots of disk i/o going on when nothing has changed.
• fixed a core dump problem if you kill the program as it's loading the config file
• checks are now done in parallel
• paging stuff merged into "new" format, fixed some problems with autoconf and the page stuff so it will work well.
• added config include PR#20
• fixed segfault in textfile.c PR#32
• added config showupalso PR#12
• added config errorsto user@site option in e-mail page stuff
• added uptime tracker (telnet to your sysmond and type UPTIME)
• added debugging toggle option
• implemented stale check killer - needs more work to be complete.
• reworked queuer (only in the beta one) to eliminate some bugs
• Merged in patches from torbjorn@thor.EU.net for using the gnu configure stuff
• changed some of the paging semantics for unpingable hosts so it will send a description

• Fixed bugs with text file writing so that it doesn't nuke the File when it updates the status file.
• Merged in patches from torbjorn@thor.EU.net for lots of things, including authdns, radius and https. https is not on by default, and needs to be configured by you the user when it is to be used. These patches will be reworked when the multiple checks at once is implemented to speed up network monitoring

• Thanks to Dan at FrontierNet for access to his AIX machine to port to there.
• Web page set up for sysmon at http://puck.nether.net/sysmon/
• Made some changes to sighup handling internally
• Cleanup changes made for handling shutdown
• Memory leak plugged in config loading of bogus dns names

• Thanks to Ray Davis at Hewlett Packard for access to a hpux machine to fix the hpux port and make it build with their C compiler
• bug found in dns caching code which would handle invalid hostnames improperly, causing core dumps

• "config logging none" to not syslog stuff
• config noheartbeat to not send heartbeat packet
• moved most config.h options to config file
• added some memory leak debugging code
• multiple clients allowed at once!!! :)
• fixed a bug where multiple clients at once would cause a double free() and kill the program
• periodic re-page option added, config pageinterval 33 where 33 is in minutes

• integrated some fixes from those great folks at ebone that keep me on my toes with the code. (Steen.Linden@ebone.net)
• removed the passing of client_socket all about to use a structure that will later be expanded to do the multiple clients at once stuff.
• changed sendmail to autodetect its path (erik.lawaetz@ebone.net)
• added config statusfile [text|html] /path/to/file to write out network status in
• added dns caching code to reduce nameserver load
• sighup support complete
• moved all server<-> client interface to srvclient.c
• changed config file options for www content checking to allow specifying real urls and text, not hardcoded stuff
• took out checkfile program, use sysmond -t (test config) instead
• heartbeat packet stuff added, will send your hostname and version on bootup for statistical purposes, never called again no other data passed on. You can #undef this in config.h
• added new config option for syslog to get out of config.h
• fixed single fd leak when loading config file :/
• fixed bug in snapshots of syslog not always logging stuff due to leftover #ifdefs
• added fd leaking watch code so we can tell if we're leaking fd's
• after getting sighup, immediateley check the network don't sleep, don't collect $200
• check for sighup within sleep, and break out if you get one
• added %d to pmesg declaration that prints downtime in DD:HH:MM format

• cleaned up code to eliminate some uneeded global variable nastyness
• framework for sighup support, but not all coded in properly yet.
• almost inital addition of dprintf, but need to hack it a bit more for it to be useful

• fixed some udp bugs
• various cleanups and runtime debugging started to be implemented properly.. (sorta)
• syslog cleanup
• if syslog defined, it logs where client connections come from. -- do we want rfc1413 (ident) support to be added? blocking issues with it.. etc..?

• config numfailures support fixed
• curses client fixed (hopefully)
• fixed a lot of compiler warnings
• shuffled some code around and reworked status variables
• comments updated somewhat

• imap config file options fixed
• pop3 config file options fixed
• overlapping internal variable problem fixed
• bug in non-blocking code that caused possible infinate loops and slow checks to be performed, and possible bad data to be returned
• fix in some of the timer code
• TCP connection timed out sensing code fixed

• several changes were done since the last version, not all were documented. They are listed here:
• imap support finished
• pop3 support finished
• smtp support finished
• redundant code elimiated
• filedescriptor leak found and plugged
• modularized the code a fair amount more in preparation for sending out checks in parallel
• testpop3 and testimap programs added to the suite, to test the individual services. There should be a sysmond -s option soon to test them internally via command line before adding them to the config file
• a few new parts of the data structure added for later expansion to the parallel checking services.

• fixed another little memory leak in the ping() function
• hacked on making all connections to block as little as possible at all in this world without threads
• hacked on icmp.c a lot
• hacked on talktcp.c a lot
• made icmp.c not use signals to check for data anymore
• included "pingtest" utility (basically a ping program)

• reduced blocking to an almost minimum under most circumstances
• added a new error report level for nntp checking, No server Response
• made the "block" program take the argument of the hostname to check blocking on instead of it being hardcoded
• inclusion of JAVA client in distribution

• fixed several blocking issues
• fixed a few more problems with client/server interaction
• fixed nntp checking so it won't mailbomb folks
• added block utility to test the blocking of your server
• several minor fixes to improve security

• more client fixes (of course -- the client sucks)
• fixed some inconsistencies in talking to sockets
• put in news server monitoring stuff. Check sample config file for an example
• added -d flag (no fork) to daemon (useful for debugging)
• added -v flag (prints version) to daemon and exits
• added -p flag (changes listening port number)
• added signal handling to properly exit if we've gotten something ugly
• removed lotsa redundant code

• little fixes in client
• now builds under Solaris x86 properly
• a bit more documentation in the code