Services is a package of services for IRC networks. Read the documentation for more information.
The latest version can always be found at the official Services distribution site.
Services has been reported to run when compiled under the Cygwin [sources.redhat.com] environment for Windows, but has not been tested extensively and is not currently supported by the author. If you need a stable environment, please use a supported environment (such as Linux or FreeBSD).
Services was at one point reported to be usable under OS/2; however, it has not been tested recently and is not supported. The author has also heard reports of Services being used with moderate success on AmigaOS.
Only if you like getting your messages ignored and/or publicly flamed.
Because, as Red Hat themselves admit (and just about anyone involved in Linux development can tell you), some releases of GCC 2.96 did have serious bugs in them, and I have personally verified that at least some releases of 2.96 fail to compile Services correctly. Since Red Hat, in one of their many stupid decisions, chose not to change the version string in the fixed releases, I have no way to tell whether a given release of GCC 2.96 is a "fixed" one or not. Rather than have to wade through assembly listings for every case of a bad GCC release, I've chosen to disallow the use of GCC 2.96 entirely. The previous version of GCC, 2.95.3, is known to work correctly with Services, as are more recent versions (3.2 in particular); either downgrade or upgrade before compiling Services, or install from a binary distribution instead of the source code.
If you absolutely must compile Services using GCC 2.96, you can override the configure check with the -force-gcc-2.96 option. However, the author will not provide support for any problems which occur if you override this check.
Incidentally, Services is believed to not have any of the programming bugs Red Hat discusses on the page referenced above. The one exception is pointer aliasing optimization (the "int i; *(float *)&i = 2.0;" example), which is a bug in the C standard, and is disabled when compiling Services anyway.
GCC versions earlier than 3.4 have bugs which cause the __builtin_apply() and __builtin_return() pseudofunctions to be miscompiled (see GCC Bugzilla, bugs 8028 and 11151 [gcc.gnu.org], for details). Working around this problem requires the use of assembly language, so workarounds are platform-specific; Services currently has workarounds for these bugs on the Intel x86, SPARC, and PowerPC platforms, but on other systems, you will need to use GCC 3.4 or later, or the configure script will report an error and refuse to continue.
Many systems, including Linux/glibc and Solaris, have subtle bugs in their strtok() implementations which cause Services to perform improperly, or even crash in some cases. The configure script checks for these bugs, and reports strtok() as "not found" if one or more bugs are present. (In fairness, the definition of strtok() is vague in some respects, and developers of these OS's may have interpreted it differently than the way it is used in Services.)
Your make program isn't compatible with the Makefile for Services. The Makefile was designed to work with GNU make, version 3.79 or later, and may not work with older versions or other systems' "make" programs. In particular, the make programs bundled with SunOS/Solaris and FreeBSD have been reported not to work; you will need to use GNU make on these systems. See the system requirements section for details.
You forgot to run the configure script first. See the installation directions.
Services puts itself in the background when it starts, so you get your shell prompt right back. Meanwhile, Services will continue setting up, then connect to the IRC server specified in ircservices.conf (or on the command line). If it doesn't connect, you probably have the wrong IRC protocol module loaded in ircservices.conf.
Also make sure that you are actually running one of the supported servers. There are many, many different variations on the basic IRC protocol out there, and I have neither the time nor the desire to add support for all of them. See section 2-1 for information on which servers are known to work and not to work with Services. If your server is not listed in the "interoperates" list as a supported server, chances are it will not work. As always, you can check the log file (ircservices.log by default) for error messages.
Tough luck, unless you want to write support for it yourself, or you can supply a complete RFC-style description of the protocol to work from.
You need to configure your IRC server to accept Services as an IRC server. For most IRC servers (those based on the irc2.x distribution), that involves adding two lines like the following to your ircd.conf file:
C:127.0.0.1:password:services.whatever.net::99See the example ircd.conf file provided with most IRC server distributions for the meaning of each field.
N:127.0.0.1:password:services.whatever.net::99
Note particularly that "no N:line" errors can mean any of several things:
This is typically caused by including a port number in the C:line for Services, which tells your server to try to autoconnect to it (depending on the class (Y:line) settings). This is not what you want, because Services will connect to the server itself, but does not listen for servers to connect to it. The solution is to remove the port number from the C:line.
Of course not. RTFM, and see the previous answer.
You forgot to run "make install" after compiling Services.
This can occur if Services was interrupted (e.g., by a reboot) or crashed while writing the databases. Remove the file given in the warning message (this is the file specified by the ircservices.conf LockFilename directive, by default ".lock" under the Services data directory), then use the OperServ UPDATE command to save the databases; alternatively, the FORCE option can be used with the OperServ UPDATE command to perform both steps at once.
Make certain that (1) the disk on which the databases are stored is not full, and (2) that Services has permission to create files in the data directory (on Unix, this means that the user-ID that Services runs as must have write and execute permission on the data directory, as well as execute permission on all parent directories of the data directory).
If you have a really large network (tens of thousands of simultaneous users), Services in its default configuration may not be able to keep up with all the network traffic. Here are some possible ways to speed Services up:
Services has been successfully used on networks as large as 22,000 simultaneous users with 260,000 registered nicks.
Make sure you've selected the correct IRC protocol (IRC server) module in ircservices.conf; see question B.3 for details.
See if you can reproduce this by doing a certain sequence of actions. If so, please report it in detail (see question Z.3 below). If not, you're probably out of luck; if you like, you can report it anyway, but chances are it won't get fixed if it can't be reproduced. If you do have such a problem, you may find the cron utility useful for dealing with it; see the section regarding the ircservices-chk in section 2-5.
If you're running the DALnet, Unreal, or Undernet IRC server, make sure every server on your network has a U:line for Services in ircd.conf; for example:
U:services.whatever.net:*:*Services will report this in a WALLOPS or GLOBOPS message the first time it discovers it cannot change modes.
If the ircd.conf file was created, edited, or copied from a Windows computer, then it may be corrupt (Windows inserts ^M characters at the end of every line), which will cause the ircd to not recognize the U:line. Use a text editor (like vi or Emacs) to remove the ^M characters, or re-create the file from scratch.
This is normal, and happens whenever ChanServ kicks a user from a channel in which they were the first to enter (for example, a user without privileges entering a channel with the RESTRICTED setting active, or a user on the channel's autokick list). In this case, ChanServ joins the channel for a short period of time to prevent the user from joining again immediately after they've been kicked. If you are using the Bahamut or Unreal IRC servers, this will result in a message like the above; it can be safely ignored.
You can't. RFC 1459 (the IRC protocol definition) requires that all automated clients send all replies using NOTICE rather than PRIVMSG, and Services follows that requirement. If you want to change how Services replies appear in your IRC client, change your client's settings.
Yes; enable the MergeChannelModes directive in ircservices.conf. (Note that this has minor security implications--be sure to read the comments in the example configuration file.)
These messages can appear when Services disconnects a user from the network or removes (kicks) a user from a channel, and are the result of time lag inherent in all networked systems. For example, if an autokilled user connects to the network and immediately sends a /join command to join a channel, the user's server may receive the /join before it receives the disconnection message from Services. Services will then receive the /join for what it thinks is a user who no longer exists, and reports that it received a message for a non-existent nickname.
These messages do not indicate a bug or other problem, and can be safely ignored.
This is caused by a bug in the "mIRC" IRC client for Windows computers; if you try to send a very long message using the /operserv (or /nickserv, etc.) command, such as a long global notice, then mIRC will insert a colon in the middle of it. There is nothing Services can do to fix this (since Services can't tell whether the colon was intentionally put there or not). If you experience this problem, try upgrading mIRC to the latest version or using another IRC client. Alternatively, as a workaround you can use /msg OperServ instead of /operserv.
See also this thread [trout.snt.utwente.nl] on the mIRC forums.
This is due to a bug in some versions of GCC running on SPARC systems. Services contains a temporary workaround for this problem, but it is very compiler- and environment-specific, and it may not work for you. The bug has been filed in the GCC bug database as bug 11151 [gcc.gnu.org], and has been fixed as of GCC 3.4.
Make sure you have Guest* (or whatever prefix you specified in ircservices.conf) listed in a Q:line in the ircd.conf file for all of your servers. For example:
Q::Reserved for Services:Guest*This prevents people from using Guest### nicks and colliding people who get their nicks changed. Note that some IRC servers will generate messages whenever someone gets their nick changed to Guest### by Services and a configuration line like the above is used; you may want to modify your IRC client's configuration so it doesn't display the messages (or change the IRC server itself so it doesn't generate them).
Alternatively, you can set an SQline on the same mask using the OperServ SQLINE ADD command (in the operserv/sline module), and Services will take care of making sure all servers have Q:lines for the nick.
Another user may have set the same E-mail address on their nickname. Use the LISTEMAIL command to see what nicknames have the problem address associated with them.
Because allowing users to send mail (using SENDPASS) without first confirming that the address is actually valid can allow malicious users to send spam via Services. Consider the following example:
This would allow the attacker to flood the victim's mailbox with potentially dozens of messages a second. Even if an IRC operator noticed the unusual activity quickly and autokilled the attacker, the victim could still be left with hundreds or thousands of junk messages. It is to avoid this potential situation that the SENDPASS module has been designed not to function unless the mail-authentication module is loaded.
You don't. Use a bot instead. (See question Z.6.5 for detailed discussion of why Services does not have this functionality.)
Normally, this is because the successor had too many channels registered; in this case, you will see an entry in the log file like the following:
[date] Successor (SuccessorNick) of channel #somechannel owns too many channels, deleting channel #somechannelIf you don't get a message like this or you can verify that the successor wasn't running into the channel limit, please report it using the bug-reporting procedure below (question Z.3).
This has nothing to do with Services, and is a feature of the IRC server itself which occurs when Services sets the topic on a channel. The "(nickname)" is not actually included in the topic, and the second and later users to join the channel will not see it. There is no way to disable it for the first user, unless you modify the IRC server itself.
No. Just tell your users to not use the ACCESS command. (You can also disable the ACCESS command for the whole network by removing or commenting out the LoadModule chanserv/access-levels line in ircservices.conf, but this will prevent even users who understand the ACCESS command from using it.)
Such functionality is unneeded, because ChanServ requires users to register their nicknames before allowing them to register channels. As a result, as long as you require users to include E-mail addresses with nicknames, you can just look up the address for the channel founder's nickname if a problem arises with a particular channel.
Set the RESTRICTED option for the channel, as described in the help for SET MLOCK, or use a bot to keep the channel open.
Use the ChanServ LEVELS command to disable the AUTOPROTECT level; doing so will also disable automatic setting of the "channel owner" mode for the channel founder. Setting the AUTOPROTECT level to anything other than "disabled" will likewise re-enable the "channel owner" mode.
This is a side-effect of security measures which prevent non-autoop users from getting a +v in before Services deops them. Either do the +v before the -o or use the ChanServ VOICE command.
In recent versions of Unreal, it is possible to set bans on a user's IP address. However, Unreal 3.2 servers do not send client IP address information to Services, so it is impossible to tell which IP-address bans match the user; thus the UNBAN command will fail to remove them. Unreal 3.2.1 and later do send IP addresses to Services, so if you upgrade all of your servers the problem will go away.
The STATUS command is intended primarily for use by bots, to allow them to take advantage of access level information stored in Services. For this reason, the responses to the STATUS command, including error messages, need to have a fixed format that can be understood by a computer.
These messages and mode changes are harmless, and are caused by the CSSetChannelTime configuration option (see also section 3-2-1). This functionality is implemented by sending a message to the network that Services has the "proper" timestamp (TS) and modes for the channel. Unfortunately, many IRC servers blindly cancel their own modes, including the channel ops of the user who initially joined the channel, before noticing that Services is not actually changing any modes; this resultes in the server sending out a "-o" mode change immediately followed by a "+o". Some servers also seem to think that the timestamp change is a bug (or hack attempt) and send out a warning to IRC operators about it.
All of these messages are harmless and can be safely ignored, but if they bother you, you may want to consider disabling the CSSetChannelTime option.
In all irc2.x-derived IRC servers (and possibly others), every server on the network must have an H:line for Services in the ircd.conf file, which looks something like:
H:*::services.example.net
Did you define yourself as the Services super-user? You need to insert your nickname in the ServicesRoot directive in the operserv/main section of modules.conf.
You can't. However, you can allow Services administrators to obtain Services super-user privilege by setting a password with the OperServ SET SUPASS command; Services admins can then use the SU command to obtain Services super-user privilege.
Services does not kill users when an autokill is added; it only kills them when they next connect. This is designed behavior, intended to reduce the possibility of a mistyped autokill causing the wrong users to get killed. (Suppose you typed "AKILL ADD *@*" and then accidentally hit Enter before finishing the host mask . . .)
Services doesn't count its own pseudo-clients (NickServ, ChanServ, etc.) in its user count, so Services will normally report a number somewhat lower than /lusers. This number will vary depending on which pseudo-clients are loaded, and will also change as nickname enforcers are introduced and removed.
If you have a very large and/or busy network, there may be an additional offset, either positive or negative, caused by lag (1) between users signing on/off and Services recognizing them, or (2) between Services killing a user and the servers receiving the kill. On a network with under 500 simultaneous clients, this difference should typically not be more than one at any time, but it can increase quickly as the network grows in size. (If the difference is abnormally large, see question C.2 for ways to speed Services up.)
Finally, you may have encountered a bug in Services. If the difference in counts increases over time without a corresponding increase in the number of users online, then this is the most likely explanation. Make sure you're using the most recent version of Services, then see question Z.3 for how to report the problem.
You need to be opered (i.e. user mode +o from using /oper; this is different from channel operator mode) to access OperServ.
To quote from the help for the RAW command:
This command has a very limited range of uses, and can wreak havoc on a network or cause Services to crash if used improperly. DO NOT USE THIS COMMAND unless you are absolutely certain you know what you are doing!In particular, Services does not process any messages sent with the RAW command, so if you (for example) use KILL to kill a user, Services will think the user is still online, and if they connect to the network again Services may well crash, or at least fail to recognize the new user.
To summarize: DO NOT USE the RAW command under any circumstances.
This is designed behavior. Use the OperServ AKILL command instead of the /gline command.
If you have EnableExclude defined in the operserv/akill section of modules.conf, but your IRC server does not support autokill exclusions, then Services will not add AKILLs or G:lines, because doing so would prevent autokill exclusions from working properly. See section 3-4-3 for details.
Services never kills users when a new autokill is added; the ImmediatelySendAutokill configuration directive only causes Services to send the autokill itself (that is, the user/host mask to prohibit new connections from) to the IRC servers on your network. This is a safety feature intended to limit the damage caused by a mistyped autokill.
Note that some IRC servers will themselves kill users matching a newly-added autokill; this is unrelated to Services.
If you would like to supply a new language module for Services, take a look at lang/en_us.l, which is the language module for English, as well as the comments at the top of lang/langcomp.c, which describe the language file format. If/when you have completed a module for your language, E-mail it to the author for inclusion in Services. However, by sending the author of IRC Services a language file for inclusion in IRC Services, you agree to waive all copyright and other rights to that file and the text it contains, and you agree and state that no one else (such as your employer) has any claims of copyright or any other rights to said file and text. (You will receive credit, of course, but the copyright remains the author's. If you are not sure whether your employer might have rights to your translation, consult with your employer.) Also, it would be very helpful if you would be willing to continue updating the module as changes are made to the base English file.
Some language files are not complete—in other words, they don't have a translation of every message Services uses, but only some of them. In this case, the missing messages will be displayed in English (or in the default language set in defs.h, if that language file has the appropriate message). You can either wait for the primary translator to complete the translation, or do the translation yourself and send the author the messages translated into your language.
First, check the Services home page and make sure you have the most recent version of Services; if not, upgrade to the most recent version before reporting your bug, because it may have already been fixed. Also check the setup of your IRC network and Services; for example, make certain you have the proper IRC protocol module loaded. The Services log file (ircservices.log by default) may contain information that will help you locate the problem.
If you already have the most recent version or upgrading does not fix the problem, contact the mailing list and report it. Make certain you include as many details as possible, even things which seem obvious to you--because they may not be obvious to other people. (I have seen bug reports like "I have a problem with expiring nicks and channels"; this is useless, because it doesn't give any details on what the "problem" is, such as when it happens, what the last-used times of the nicks and channels were, or anything else that could be used to find the alleged bug.) Include at least the following information:
If possible, try and reproduce the problem on a clean install (i.e. empty databases), starting from registration of nicknames and channels; this will help pinpoint where the problem is.
You may be asked for a "debug log"; this means a log created by starting Services with the -debug command-line option, which writes a lot of debugging information to the log file (such as data received from and sent to the uplink server). Note that the logfile can grow in size quickly when debugging is active, so make sure you have enough disk space available. It may be convenient to use the -log option as well, to write the log to a different file than the default logfile; for example:
ircservices -debug -log=debug.log
If you do not get a response to your message to the mailing list in a reasonable amount of time, or for some reason you cannot subscribe to the mailing list (you must subscribe in order to send mail to the list), then you can try contacting the author directly.
A "backtrace" is a list of what functions have been called in order to reach a certain point in a program. For example, the function main() may have called a function foo(), which in turn called a function bar(), so the backtrace at a point inside function bar() would look like:
Backtraces are most useful in determining where (and often why) a program crashed, since they indicate both the exact point at which the program crashed and the sequence of actions that led to the crash; for this reason, they are a standard feature of most debuggers. The instructions below are for the GNU debugger, GDB, but should be mostly applicable to other debuggers as well.
The most common way to obtain a backtrace is from a core file, a file created when a program crashes that contains information about the program's state at the time of the crash. However, Services normally avoids the creation of core files in order to provide a somewhat cleaner response to such emergency conditions. In order to force Services to create core files, it is necessary to pass the -dumpcore option to the configure script and recompile Services. Additionally, since compiler optimization can cause the crash location to be reported incorrectly, and since the default compilation options do not include debugging information helpful in interpreting a backtrace, it is preferable to turn optimization off and debugging on by using the configure option "-cflags -g". If you have already run the configure script once, the following command line can be used to enable core dumps:
./configure -dumpcore -cflags -g -defaults
Once you have reconfigured Services, compile, install, and start it as usual. When a crash occurs, a core file will be created, usually named "core" or "ircservices.core" and usually stored in the Services data directory (though both of these are operating system dependent). This file can be used to generate a backtrace for the program at the time of the crash.
To generate the backtrace, load the program and core file into the debugger with the command line:
gdb executable-file core-fileFor example, if your Services executable file is located in /usr/local/sbin, your data directory is /usr/local/lib/ircservices, and the core file is named core, you would use:
gdb /usr/local/sbin/ircservices /usr/local/lib/ircservices/coreAfter the executable and core file have been loaded, you will be presented with a "(gdb)" prompt. Type "bt" (short for "backtrace") and press Enter at this prompt, and the backtrace will be printed.
If Services still does not produce a core file after following the above instructions, it is possible to obtain a backtrace from the program as it is running. To do this, recompile Services as above, but instead of starting it the ordinary way, first load it into GDB:
gdb executable-fileand then run Services in the foreground, rather than in the background as usual:
run -debug -noforkWhen the program crashes, GDB will print a message like "Program received signal SIGSEGV, Segmentation fault", and you will be presented with the "(gdb)" prompt. If you type "bt" at this prompt, you will be able to get a backtrace as above.
Nothing is wrong, except your expectations. IRC Services is a completely different program from the Services program used on DALnet; they are similar in concept only.
New ideas are always welcome; however, not all ideas will necessarily be included in Services. Some suggestions are simply impossible or not feasible to implement; others, even if technically possible, would complicate and/or slow down Services too much for too little benefit. In general, new features should have wide applicability (be useful for many people) and be as simple as possible to use, and they should not duplicate features provided just as well or better by another program (such as an Eggdrop bot or the like). If you have written code to implement your suggestion, feel free to send it along with (but not instead of!) a description of the suggestion itself.
If your suggestion is not accepted, you are of course welcome to implement it and distribute it yourself, either as a patch to Services or as a completely separate program.
As described in Z.5 and Z.6 above, this runs counter to my development plans, and I frankly don't believe such a feature is needed. (Just because you can get a hundred people to agree with you doesn't in and of itself make your beliefs correct.) I expect that if there is truly enough need for a feature like this, someone will write a module to implement it.
You may have reported it after it was already fixed or added. In general, the Changes file lists only the first person to submit the bug report or feature request, except when that person is the author or a lot of people all report the same problem or request, in which case no individual credit is given. Mistakes do happen, though; if you really think you should be given credit for something, or if you find yourself being given credit for something that wasn't yours, please inform the author.
The short answer is, no.
The long answer is, while I accept suggestions and patches from others, I have tough standards on what code I actually allow into Services itself; in fact, even when I get a patch, I never use it as is but rewrite it so that it fits in with the rest of the Services code properly. This is the main way I'm able to keep Services as stable and coherent as it is: I don't let anyone else touch the actual code base. (There are also licensing-related issues with allowing multiple people to contribute code; see the copyright page.)
Every now and then I have someone tell me, "But it's open source! You have to let other people help too!" This is, however, a misunderstanding of what open source is about. Open source is not about having lots of people work on a single project; that's the "bazaar model" of software development, popularized by Eric S. Raymond's essay "The Cathedral and the Bazaar" [www.catb.org] (and I don't entirely agree with his arguments, but that's another story). Open source is only about giving users of software the freedom to modify that software as they see fit, by giving them a copy of the source code for the software; the difference is that those users may not necessarily be all working together on a single project, and may even be in "competition" with one another, so to speak, each releasing their own version of a program. While it may seem that this wastes effort, the advantage is that each of those people has their own ideas for what the software should do; each program evolves in its own way, letting people experiment with more features, more ways of doing things than if they were all working on a single program. This is, in fact, the case with Services as well; I have personally seen around a dozen derivatives of my Services program, each with its own particular changes, and I have taken several ideas from those other programs as well.
So in summary, while you're welcome to contribute ideas (or patches), and you're also welcome to take the Services source code and start your own version of it, I have no plans to allow anyone else direct access to the Services development code base or to delegate out various functions or modules to other people.