fixed problem where two consoles with same call stopped the cluster

[spider.git] / Changes

10Jul99=======================================================================
1. added an explicit disconnect from my side for already connected calls. This
appears to have fixed a problem where if two consoles connect at the same time
with the same (sysop?) callsign the cluster would stop.
2. added some debugging in the Message handler to allow more state to be seen.
05Jul99=======================================================================
1. added stat/msg and stat/cluster commands to allow me to poke about inside
the msg and cluster node tables.
04Jul99=======================================================================
1. removed silly 'new message has arrived' to connected nodes if message is
2. added a ! command to the console.pl program; works like the bash shell.
addressed to the node call.
01Jul99=======================================================================
1. Altered priv to 5 for connect.pl to match disconnect for nodes
29Jun99=======================================================================
1. fixed merge command that I broke.
2. added some randomness in waittimes for aborted outgoing messages 
3. Message Queuing now only occurs every 5 minutes (configurable if anyone 
wants to write the command for doing it) nad on receipt of a PC16. This is
a further attempt at sorting the crossing of messages!
4. a PC33 will cause another attempt at queuing. If you are on a roll, keep
going!
28Jun99=======================================================================
1. console.pl now works with history and editting.
2. added scrolling to stuff sent by the cluster.
3. added colours to things.
27Jun99=======================================================================
1. added WWV filtering
2. added timeouts to for forwarding Messages, also added a waiting time for
failed (stalled) outgoing messages.
3. Incoming messages will now have priority over outgoing messages to the
same node.
4. Added 'catchup' command which will 'catchup' messages to date for a node. 
This means that when you start forwarding to a node, it doesn't get all the
messages queued up that are probably old.
5. added 'uncatchup' which does the opposite of the above.
6. fixed kill full and PC49 handling so that it actually works.
7. Fixed local aliases so they add to the front of the standard ones. This
means you only need to add your specials or override the system ones you need
to.
21Jun99=======================================================================
1. changed regex for cluster->client msgs so that strings like |---| are no
longer ignored.
20Jun99=======================================================================
1. prevented a merge to myself (causes a loop)
18Jun99=======================================================================
1. fixed a problem with DXLogPrint repeating log entries.
2. altered the permissions on disconnnect so that priv 5 users can disconnect
clusters, but still require priv 8 to disconnect users.
3. made sure that an incoming PC18 clears out all the cluster info for
that channel before re-initialising.
4. add init command.
5. disconnect users before nodess on shutdown
15Jun99=======================================================================
1. added grepdbg program (so you can search your debug files and get times
and dates of incidents).
2. Moved DXVars.pm to DXVars.pm.issue to force new users to cp it to their
local directory.
3. added $allowzero to DXProt to allow thru PC19s with zero protocol versions.
14Jun99=======================================================================
1. changed IO::File in help and apropos to IO::File
2. added a check for isa('DXNode') for PC50s
08Jun99=======================================================================
1. Started work on the curses based operator console.
03Jun99=======================================================================
1. cluster seems to have a memory leak, put DESTROY functions in where 
appropriate.
2. try to make sure that PC21 commands are not issued inappropriately and
also reformat PC19 for onward broadcast so that nodes coming in on loops are
dropped from those broadcasts.
3. make sure PC16,17,19,21 doen't affect locally connected nodes.
4. Changed all instances of FileHandle to IO::File.
01Jun99=======================================================================
1. removed a output of an unwanted pc21 for isolated nodes
31May99=======================================================================
1. Added checking for PC16,17,19 and 21 to make sure that they come in from
the channels that they are supposed to.
2. Pass the RCMDing node's priv level to the command (rather than the 
connecting node's privilege level).
3. Pass thru merge requests for other nodes.
4. Removed a warning from client.pl
5. routed external DB commands and drop them (for now) locally
6. routed external mail commands (ie those addressed elsewhere)
7. routed external PC26/27 lines 
25May99=======================================================================
1. try to make the lastin value correct even for local users
2. got rid of the stupid bug that failed to print out the offline message
but put a "*main::STDOUTmain" for every line of that message instead.
3. Ported to 5.00503 (well actually just installed it and it just runs). 
4. Added who command
23May99=======================================================================
1. Got rid of some of the useless error messages on global destruction when
quiting a program.
2. Stuck -w back on client.pl
23May99=======================================================================
1. tried to change connection to raw mode for cluster connections
2. sh/c and sh/u are now sorted in alphabetical order
3. Limited the number of queued dx spots when composing messages to 20 (the 
oldest one is lost for every one added above 20).
4. Added generalised shell globbing everywhere I think it is useful, including
sh/dx, dir, sh/files etc
5. Made dir more compatible with ak1a and added 'to' and 'from' call searching
it should now do everything that ak1a does except counts.
6. added similar functionality to kill command so that you can kill ranges, to
and from callsigns etc.
22May99=======================================================================
1. added check for -1 from Date::Parse and return undef for out of range dates
2. added show/files and type commands
21May99=======================================================================
1. made set/nodx work again.
2. made dx stuff queue nicely again when sending messages.
18May99=======================================================================
1. Added announce dup checking.
2. Added system announce filtering.
08May99=======================================================================
1. Added a bit more checking to the rcmd thing to make more sure that this
rcmd isn't coming from an imposter. Not possible to prevent everything 
ofcourse, just make it a bit harder.
2. altered set/lockout and set/privilege so that they can't executed remotely.
3. Added some code to reject rubbish (with invalid binary characters) protocol
or user lines as input.
4. prevented PC command being executed remotely :-)
05May99=======================================================================
1. change the connection overlap from an abort to a disconnect.
02May99=======================================================================
1. Added the mechanism for doing user spot filtering for spots in the cluster 
(no user control just yet - but sysops can do it).
2. Fixed problem with nodes sending out the node lists of isolated nodes.
30Apr99=======================================================================
1. Try to make it even more AK1A friendly by putting the node list out in the
'correct' order.
2. added code so that local users and commands go to all connected nodes 
INCLUDING the isolated ones.
26Apr99=======================================================================
1. added set/user command as a hack, please use this with care, there is
very little checking.
2. added export_opernam.pl which is a basic reader of AK1A opernam.dat files
to stdout
25Apr99=======================================================================
1. Fixed problem with filtered spots not coming out.
2. Added the possibility of filtering on channel callsign
20Apr99=======================================================================
1. altered order of undefing and closing of user file in an attempt to make 
corruptions/missing users less likely.
2. made messages that are addressed to the node call appear to the alias call
3. check dates much more rigourously.
4. ignore SIGTERM and SIGINT during shutdown of the cluster
15Mar99=======================================================================
1. added $actiondata to filter line to allow per action data such as no of hops
2. fixed a silly problem in talk for non-existant callsigns
3. Added sysop command
02Mar99========================================================================
1. Changed DXUser so that it uses a homemade import/export hash routine 
2. DXUser now uses a DB_BTREE file
3. Added export_users.pl, import_users.pl and convert_users.pl
25Feb99========================================================================
1. added NL language stuff.
2. added set/language
21Feb99========================================================================
1. Allowed multiple 'on <freq>' for show/dx
2. Made sure the 'on 20m/ssb' thing worked (also 'on hf/cw').
3. first cut of the Filtering code, no user commands yet but the file
format is defined and manually added filters should work for spots
17Feb99========================================================================
1. added export_user.pl to export user files (for interest and safety)
2. changed DXUser::init to allow O_RDONLY access which may limit the number
of coredumps G0RDI seems to get.
15Feb99========================================================================
1. Added msg forwarding code which uses $main::root/msg/forward.pl.
14Feb99========================================================================
1. Changed A & K in WWV to be the other way around (I am told by G3IOR that
this will mean that we won't have pole to pole aurora nor all the power lines
overloading everywhere as well as a result of having values of K > 10).
2. allow the '~' character in message bodies.
3. Moved $main::data/badmsg.pl to $main::root/msg/badmsg.pl.
11Feb99========================================================================
1. Fixed a problem with isolated nodes' configurations being sent on sending 
local configs to new connections.
08Feb99========================================================================
1. added a $SIG{ALRM} in the client to actually catch the alarm when it goes 
off which means that the netrom/ax25 call programs terminate properly (and not
loop as fast as their little legs can paddle, soaking up CPU time).
2. Implemented read receipts as an especial request from G4PDQ.
3. Fiddled with DXUser a bit to see whether I can stop it core dumping on new
users in PC16s on G0RDI's machine.
4. Added E4 (Palestine) to Prefix data.
30Jan99========================================================================
1. Some of the dates we get can cause crashes, tried to make it more robust (oh
and Y2K compliant)
2. PC16 seem to come in with missing callsigns from somewhere
3. added $main::data/baddx.pl which prevents callsigns that are in the list being
a) forwarded b) stored and c) (except for the originator, if local) being
displayed locally.
4. added $main::data/badmsg.pl which deletes any messages whose TO address is
in this list (this is largely for european sensibilities). 
18Jan99========================================================================
1. added present(),presentish() and disconnect() to DXCron so that you can see 
(easily) if a station is on the cluster anywhere and also disconnect them
locally. 
2. added rcmd() to DXCron so you can send an rcmd to someone else from crontab.
3. <embarrassment>Fixed create_sysop.pl so that longitude is spelt correctly
there as well</embarrassment> made the update work properly. NOTE create_sysop
will only do what you expect if all cluster.pl and client.pl programs are 
stopped.
4. DXCron wasn't reading in reliably on startup, this is now (hopefully) fixed.
the -M semantics are rather strange!
5. The -w switch is now standard on cluster.pl
6. Added last_connect() which gives the last connect time of a user (or now
if connected).
7. Added present_on(call, node) and presentish_on(ditto) which returns true if 
the call is connected to the node.
8. incoming messages for users will now send a 'new message' message (instead
of merely generating the message)[who wrote this rubbish?].
9. Hooked $SIG{__WARN__} and $SIG{__DIE__} so we store these nasty error 
messages.
10. Print the no of users on restricted protocol links if we have had PC50
11. Had a go at making the sh/cl stats more accurate.
12. PC11 was trying to call Local::spot1 instead of Local::spot, no warning 
previously.
13. Added a simple lockfile to prevent more than one cluster running at a time
17Jan99========================================================================
1. fixed some permission problems on DXLog.
2. There is a circumstance in DXMsg which caused the cluster to stop on an 
null ref pointer which I can't get to the bottom of but have bandaided out 
for now.
3. Started filter and Database code.
4. Upper cased all the TO callsigns for sending to more than one station in
messages.
07Jan99========================================================================
1. Fixed Geomag after emacs reformatted it wrong so sh/wwv works again.
2. Added more range checking in WWV spots (they really are a heap of ...)
3. Changed create_sysop.pl so that you can update as well as create the user 
file
4. Remove spotter's SSID from WWV spots
06Jan99========================================================================
1. Do some range checking for spots and WWV in the future (got a WWV for Oct 
2034 whhich caused a bit of confusion!)
2. Make WWV spots broadcast them to the users! (as opposed to merely storing 
them)(thank you G0RDI).
3. Allow users to do show/announce (thank you JE1SGH).
4. Delay broadcasts to users if they are not in a 'prompt' state (means you can
add messages and see what you are doing on a busy system)
5. Made set/unset dx,ann,wx,talk,wwv do what is expected
6. added set/sys_location and set/set_qra to set the cluster lat/long and qra
7. New messages will now be announced on logon (if there are any)
03Jan99========================================================================
1. Upped the version no !!!!
2. made the DXProtocol routines much less sensitive to '~' characters (JE1SGH)
3. added '^' -> ':' conversions on ann, talk and dx commands
4. You can now send messages containing blank lines.
5. Fixed merge command problem reported by G0RDI
28Dec98========================================================================
1. Crossed fingers, got rid of the instabilities caused by execing programs
at vast speed (reduced it significantly anyway).
2. Personals will now move if homenode is known.
3. Added Local.pm for doing local specials
27Dec98========================================================================
1. Various detail changes to remove some more warning with -w on
2. Added DXCron handling - you can do crontabs now.
3. Added show/program command so you can see where it is loading your .pm files
from.
4. Added pc26/27 replies to locally connected cluster's merge (pc25) requests
5. Added spotters DXCC and original cluster to Spot data files.
23Dec98========================================================================
1. Reindented various things
2. Added missing $main::mycall on the end of outgoing PC11s (!)
3. Check that PC16/17 node calls really are nodes (AFAIAC).
22Dec98========================================================================
1. Added brackets round callsign if not here for prompt
2. Added Iain's set/password routine
3. Fixed sh/dx problems (the xemacs indent region messed Spot.pm up!)
4. Issued 1.15
21Dec98============= late! ====================================================
1. fixed problem with missing DXDebug in DXProt.
2. Fixed DXDebug so that it actually works as advertised with and without 
trailing \n. 
3. Added deduping of WWV spots as well (at for date,time,sfi,k and i) dups
4. Replaced the 0 in "02-Dec-1998" with ' ' so it reads " 2-Dec-1998", it seems
hard to credit it but some 'programs' out there that connect to clusters have
problems with the leading '0'!
5. In the same vain, included a strictly AK1A compatible sh/heading, apparently
this is necessary for the same reason as 4.
6. Started contrib tree stored the old show/heading in contrib/g0rdi/show.
7. Because I now correctly dedupe spots and wwv (there's a hostage to fortune..)
I have added a merge command. 
21Dec98========================================================================
1. Added "issue" to the client program for 'login' connections
2. Added more docs for client program.
3. Fixed problem introduced in 1.12 where a connect locally with a callsign of
someone connected remotely would stop the cluster.
4. issue release 1.13
20Dec98========================================================================
1. Removed all the warnings I get with perl -w (at least for just starting the
cluster and running a few commands).
2. Added per node hop control.
3. Added some docs on how to use it and isolation
4. Made talk command more intelligent in that if the user isn't seen and the
user's last node is visible it tries the talk anyway.
19Dec98========================================================================
1. Fixed problems with sh/rcmd (talk/ann/log) with a callsign as argument and
also made what G0RDI wanted work as well!
2. Added isolate flag to allow isolation of different networks at a gateway.
3. Changed make Makefile.PL to perl Makefile.PL in instructions
4. Copied latest instructions to html directory
5. Had another attempt at making clients disconnect without hanging everytime.
6. Changed msg queuing semantics so that the system routes on exact callsign.
7. Changed the protocol version so that it increments with the DXSpider version.
14Dec98========================================================================
1. Made the telnet thing work a bit better. It still will not work reliably to
a real telnetd on port 23.
2. Allowed network logins on client by specifying login instead of call.
3. made msg handling more robust (PC30 with unknown streams cause PC42), queueing
is only done on channels that are in state 'normal'.
4. Added pc command which takes a callsign and some text and sends it without
mods to the callsign, useful for sending manual PC protocol to unstick things.
Also for sending anonymous messages to online users.
5. Stopped duplicate messages being stored (it receives them and then bins them)
6. Implemented PC49 delete/full from outside (kill full on the inside)
7. Implemented the client command in connect scripts so that you can have 
different scripts for the same callsign.
8. Added sh/wwv command
13Dec98========================================================================
1. Fixed VS6 lat/long in prefix_data and wpxloc.raw
2. Sorted out last in times for remote users
3. editted and substitued G0RDI's version of the install instructions over
mine in spider/html
4. fixed telnet connections
5. added connect instructions to html tree.
6. Added some more help
7. Upissued version
12Dec98========================================================================
1. Incorporated sh/st, (un)set/lockout, forward/opername from Iain G0RDI
2. Added group handling with (un)set/group, show/group. This allows arbitrary
groups to be formed (for ann/<group> for example) and is also used to 'filter' 
spots, announces, wwvs etc into groups of cluster nodes (useful for creating 
disjoint cluster networks)
3. **** CHANGE mylongtitude to mylongitude in DXVars.pm ****
4. Altered QRA locator routines so they work correctly!
5. Fixed all commands that had the wrong mylat(itude) and mylong(itude) names
in.
07Dec98========================================================================
1. Added set/name, set/qra, set/location, set/qth, set/homenode, set/bbs, 
set/email
2. changed the Msg socket close to a shutdown (hopefully more reliable), this 
should stop clients hanging on exit.
3. Added Alias support for help
4. Added lltos and stoll routines to DXBearing
5. Added auto prompting for name, qth, location and homenode if any of these
aren't known.
06Dec98========================================================================
1. Fixed DXBearing::is_qra so that it correctly detects full QRA locators
2. Added sh/qra for doing locator distances and bearings
3. Added some stuff to Notes.txt on hints for command writers
4. Changed help system to use just one file per language
5. Started adding more help
6. Fixed 24Mhz band limits
7. Added apropos command and lots of new help text
05Dec98========================================================================
1. Added sh/heading and QRA locator calculation stuff
04Dec98========================================================================
1. Started Changes file
2. Added load/cmd_cache.pl to clear cmd cache shortcuts (use this if cluster.pl
get confused about where it is getting its commands from eg if you have
removed a file from local_cmd 'cos you want to go back to standard one)
3. Fixed bull handling at least to the extent that it stores them and
doesn't mess up the normal directory handling.
4. Fixed the unset/debug command so that it actually does the job!
5. Removed the middle line 'written in perl for unix' from show/version.pl