UnrealIRCd Bug Tracker - Issues

0006630: CAP LS * reply is sent to non-302 clients

Sun, 21 Jun 2026 20:03:26 +0200

The "CAP LS *" syntax, which allows sending more than one line of capabilities, is an extension of the capability negotiation version 302: https://ircv3.net/specs/extensions/capability-negotiation#cap-ls-version-features

But with recent additions to the list of capabilities (coincidentally "draft/multiline", which is unrelated to multi-line CAP LS), Unreal's response to "CAP LS" that are not "CAP LS 302" no longer fits on a single line, causing Unreal to violate the spec:

The solution to this is to omit some capabilities to clients that don't support version 302. I guess Unreal (or its modules) could hardcode a set of "old" capabilities, as clients not supporting version 302 are unlikely to support recent and complex specs like multiline, chathistory, labeled-response, or metadata.

0006629: IP matching for security groups not working with /0 or ::/0

Mon, 15 Jun 2026 20:19:18 +0200

this doesn't work:

---
security-group ipv4-users {
ip 0.0.0.0/0;
}

security-group ipv6-users {
ip ::/0;
}
---

does work:

---
security-group ipv4-users {
ip 0.0.0.0/1;
ip 128.0.0.0/1;
}

security-group ipv6-users {
ip ::/1;
ip 8000::/1;
}
---

doesn't seem to like /0

0006611: RPC_CALL_ERROR - Every refresh i have a error to logs.

Sat, 13 Jun 2026 16:15:41 +0200

Hello all,

i have a error with panel, when i click to Servers or any button i see a error to logs.

timestamp:"2026-02-07T10:43:20.404Z"
level:"info"
subsystem:"rpc"
event_id:"RPC_CALL_ERROR"
log_source:"chati.shprehu.net"
msg:"[rpc] Client RPC:adminpanel: RPC call server.module_list"
client:Object
name:"RPC:adminpanel"
id:"2424A7VUA"
hostname:"127.0.0.1"
ip:"127.0.0.1"
details:"RPC:adminpanel@127.0.0.1"
server_port:8600
client_port:42972
connected_since:"2026-02-07T10:43:19.000Z"
idle_since:"2026-02-07T10:43:19.000Z"
method:"server.module_list"

0006628: multiline-concat should not concatenate server-side

Sat, 13 Jun 2026 10:53:41 +0200

When a multiline-aware client sends this:

@label=xyz BATCH +123 draft/multiline #test
@batch=123 PRIVMSG #test :
@batch=123 PRIVMSG #test :how \x0304is
@batch=123;draft/multiline-concat PRIVMSG #test :everyone?
BATCH -123

other multiline-aware clients receive this:

@label=xyz;msgid=iD3UGnJUYeJK5rxS9uFrcE;time=2026-06-12T19:20:07.558Z :alice!~username@Clk-4D552FD5 BATCH +DeAmXVtTyMA8pZ9MEgkzpZ draft/multiline #test
@batch=DeAmXVtTyMA8pZ9MEgkzpZ;label=xyz;time=2026-06-12T19:20:07.558Z :alice!~username@Clk-4D552FD5 PRIVMSG #test :
@batch=DeAmXVtTyMA8pZ9MEgkzpZ;label=xyz;time=2026-06-12T19:20:07.558Z :alice!~username@Clk-4D552FD5 PRIVMSG #test :how \x0304is
@draft/multiline-concat;batch=DeAmXVtTyMA8pZ9MEgkzpZ;label=xyz;time=2026-06-12T19:20:07.558Z :alice!~username@Clk-4D552FD5 PRIVMSG #test :everyone?

and multiline-unaware clients receive this:

@msgid=iD3UGnJUYeJK5rxS9uFrcE;time=2026-06-12T19:20:07.558Z :alice!~username@Clk-4D552FD5 PRIVMSG #test :how \x0304is everyone?

(where they display both "is" and "everyone" in red)

this differs from Ergo where they would receive this:

@msgid=iD3UGnJUYeJK5rxS9uFrcE;time=2026-06-12T19:20:07.558Z :alice!~username@Clk-4D552FD5 PRIVMSG #test :how \x0304is
@msgid=iD3UGnJUYeJK5rxS9uFrcE;time=2026-06-12T19:20:07.558Z :alice!~username@Clk-4D552FD5 PRIVMSG #test :everyone?

(where they display only "is" in red)

Additionally, when a message is long, for example:

@batch=123 PRIVMSG #test :################################################################################################################################################################################################################################################################################################################################################################################################################################################################how \x0304is
@batch=123;draft/multiline-concat PRIVMSG #test :everyone?

then the fallback is simply truncated:

@msgid=4APKGfukNuD9d1E6xvQcX1;time=2026-06-12T19:25:32.766Z :alice!~username@Clk-4D552FD5 PRIVMSG #test :################################################################################################################################################################################################################################################################################################################################################################################################################################################################how \x0304is everyon

(missing "e" at the end)

The multiline spec was designed specifically so that messages can be relayed as-is to clients that don't support multiline, and look as they would if sent be a multiline-unaware client.

0006477: Count "flood-blocked" messages and make counter available in security group

Fri, 12 Jun 2026 17:44:55 +0200

Dear Unrealircd Team

Sorry if this is the incorrect place to make a feature request

My request is about the flooding control. What I propose is there is a session-only score that tells the IRCd how many (flood-blocked) has happened because of something that user has tried to do, that can be used as part of the criteria for a security group

I currently have a script that reads server notices but if this was available at the config level it could be part of a useful measure building on the existing features like max-concurrent-conversations

0006304: Gline hit stats

Mon, 08 Jun 2026 13:45:23 +0200

I was just curious if there's a way to see last hit on a gline. This way I could delete it if it didn't hit for a month for example. If it hits 3 times a day, I'd think to leave it there instead of removing and adding it again and again.

0006144: Make zline, gline, kline, shun create an ID.

Sun, 07 Jun 2026 17:55:23 +0200

Will be nice if this can be done directly from the IRCD and not only with operserv.
/gline *@etc - whatever reason -> this part be automatic ID: XZ9YCRP1E8.

Will make my life easier.

thanks.

0006579: usermodes/privdeaf

Fri, 05 Jun 2026 18:14:40 +0200

It seems logical that usermodes/privdeaf should ignore TAGMSG otherwise users receive the "User does not accept private messages" notice multiple times before they've even sent the message that should trigger it.

0006610: Move "make pem" to its own "./unrealircd" subcommand

Fri, 05 Jun 2026 17:39:53 +0200

Currently if we want to generate a new certificate/key, we nee to run the "make pem" command.

While I expected the "make install" command to replace the older certificate/key, it doesn't, which is by design to avoid some rare issues, like replacing certs on test instances, debugging, etc.

After discussing it on IRC, it was decided that it should be better to move it to its own "./unrealircd" subcommand and get rid of the "make pem", while being able to generate them from "./Config" on initial setup.

The command can probably be called "./unrealircd makecert" and throw a warning to the user, like: "This command will replace your existing server certificate and key. Do you wish to proceed? [Y|N]"

0006626: Documentation (connthrottle.set)

Tue, 26 May 2026 18:47:52 +0200

In the documentation it says that the argument enabled should be a string "on" or "off"
in my case it is more True or False.

here is the error message when i try to pass a string in the parameter enabled :

RPCResult(jsonrpc='2.0', method=None, error=RPCErrorModel(code=-32602, message="Missing parameter: 'enabled'"), result=None, id=123)

0006625: antimixedutf8 ignore except directive

Fri, 15 May 2026 15:39:27 +0200

While tuning a working antimixedutf8 setup, using the documented syntax for except yields: "set::antimixedutf8::except with no value"
The except block was verbatim from the documentation, ie.:
except {
security-group known-users;
}

0006617: There seems to be an issue with the updated GeoIP feature on 6.2.4-rc1 on Windows

Fri, 17 Apr 2026 06:04:12 +0200

I've been playing with the new release (6.2.4-rc1) and there seems to be an issue with the updated GeoIP feature affecting the Windows version of the software.

I've tried 6.2.4-rc1 on two new virtual machines (so there would be no left overs of a previous installation of unrealircd); one running Debian 13 and the other Windows 10 LTSC.

On Linux, the code compiles fine and after editing unrealircd.conf, I get the irc server running without any issues.

On Windows, I always get a fatal error:
[error] conf/modules.default.conf:318: loadmodule classic: failed to load: Cannot open module file: No such file or directory

In modules.default.conf, the line 318 reads:
loadmodule "$GEOIP_ENGINE";

To make sure it wasn't a user error on my part, I've copied both unrealircd.conf and modules.default.conf from my Linux install to my Windows virtual machine but I still get the same error.

If I manually edit modules.default.conf and change the line 318 to
loadmodule "geoip_classic";
the server starts fine. Obviously, commenting out the line also allows Unrealircd to run.

0006619: I think the operclass permissions "channel:override:secureonly " doesn't work

Tue, 07 Apr 2026 18:15:18 +0200

It seems I'm unable to join channels +z when I'm not connected through SSL.

Here is the relevant part of my unrealircd.conf

oper test {
class opers;
mask {
*@192.168.*;
*@127.0.0.1;
*@localhost;
}
password **REDACTED**;
operclass netadmin-with-override;
}

From my understanding of the doc available at https://www.unrealircd.org/docs/Operclass_permissions#channel
I should be able to join a channel +z even if i'm not connected via a secure connection.

yet...

/mode #test
#test +ntzPZ
#test created on Mon Apr 06 23:47:28 2026

/join #test
#test unable to join channel (not using secure connection)

even if I modify unrealircd.conf file to explicitly gave myself channel:override:secureonly, it still doesn't work

operclass oper-test {
parent netadmin-with-override;
permissions {
channel {
override {
secureonly;
}
}
}
}

oper test {
class opers;
mask {
*@192.168.*;
*@127.0.0.1;
*@localhost;
}
password **REDACTED**;
operclass oper-test;
}

/sajoin myself #test also doesn't work.

I had this issue with 6.2.1 on Windows but the bug is still present with 6.2.4-rc1 on Debian.

0006616: Error

Sat, 04 Apr 2026 06:59:44 +0200

ERROR

0006615: libsodium does not compile on arm64 with gcc -- replace the .tar.gz

Sat, 04 Apr 2026 06:56:33 +0200

Compile fails (doesn't matter if 'unrealircd upgrade' or manual compile:

Making all in src
make[1]: Entering directory '/usr/src/unrealircd-6.2.3/extras/libsodium-1.0.21/src'
Making all in libsodium
make[2]: Entering directory '/usr/src/unrealircd-6.2.3/extras/libsodium-1.0.21/src/libsodium'
Making all in include
make[3]: Entering directory '/usr/src/unrealircd-6.2.3/extras/libsodium-1.0.21/src/libsodium/include'
make[3]: Nothing to be done for 'all'.
make[3]: Leaving directory '/usr/src/unrealircd-6.2.3/extras/libsodium-1.0.21/src/libsodium/include'
make[3]: Entering directory '/usr/src/unrealircd-6.2.3/extras/libsodium-1.0.21/src/libsodium'
CC crypto_aead/aegis128l/libsodium_la-aegis128l_soft.lo
during RTL pass: subreg1
In file included from crypto_aead/aegis128l/aegis128l_soft.c:52:
crypto_aead/aegis128l/aegis128l_common.h: In function 'decrypt_detached':
crypto_aead/aegis128l/aegis128l_common.h:252:1: internal compiler error: Bus error
252 | }
| ^
0x1df34bb diagnostic_context::diagnostic_impl(rich_location*, diagnostic_metadata const*, diagnostic_option_id, char const*, std::__va_list*, diagnostic_t)
???:0
0x1e04dd7 internal_error(char const*, ...)
???:0
gcc: internal compiler error: Segmentation fault signal terminated program cc1
Please submit a full bug report, with preprocessed source (by using -freport-bug).
See <https://github.com/archlinuxarm/PKGBUILDs/issues> for instructions.
make[3]: *** [Makefile:2768: crypto_aead/aegis128l/libsodium_la-aegis128l_soft.lo] Error 1
make[3]: Leaving directory '/usr/src/unrealircd-6.2.3/extras/libsodium-1.0.21/src/libsodium'
make[2]: *** [Makefile:3715: all-recursive] Error 1
make[2]: Leaving directory '/usr/src/unrealircd-6.2.3/extras/libsodium-1.0.21/src/libsodium'
make[1]: *** [Makefile:413: all-recursive] Error 1
make[1]: Leaving directory '/usr/src/unrealircd-6.2.3/extras/libsodium-1.0.21/src'
make: *** [Makefile:528: all-recursive] Error 1

0006612: Implement extbans support in the SILENCE list

Sat, 14 Feb 2026 20:05:16 +0100

The feature request is to extend the SILENCE list to supporting extbans as well, in the same fashion as the channel b/E/I lists. Currently the SILENCE list only supports the format nick!user@host.

It could be convenient for the user to be able to use extbans in the SILENCE list also, such as: ~account, ~asn, ~country, ~realname, ~security-group, ~certfp and ~text. Maybe ~time also in case it is possible to manage temporary items in the SILENCE list.
The other extbans might not be relevant for the SILENCE list.

This would also opens the possibility of remotely managing the SILENCE list through the services and the use of SVSSILENCE.

Once implemented the usage could be similar (and a bit different due to the SILENCE syntax) to the channels: SILENCE +~realname:qwerty, SILENCE -~account:some-account

SILENCE list could look like:
-> server. SILENCE
<- :server. 271 Nick *!user4@host4
<- :server. 271 Nick nick3!*@*
<- :server. 271 Nick nick2!user2@host2
<- :server. 271 Nick ~asn:65536
<- :server. 271 Nick ~security-group:some-insecure-group
<- :server. 271 Nick ~text:a_spammy_spam
<- :server. 272 Nick :End of Silence List

0005849: WHOISASN like on Inspircd - Very useful

Fri, 30 Jan 2026 16:27:46 +0100

Would it be possible to add ASN to whois in the future?
It would be very useful to me

On Inspircd it looks like this:

if (asn)
whois.SendLine(RPL_WHOISASN, asn, "is connecting from AS" + ConvToStr(asn));
else
whois.SendLine(RPL_WHOISASN, "*", "is connecting from an unknown autonomous system");

https://github.com/inspircd/inspircd-contrib/blob/master/3.0/m_asn.cpp

Where is it possible to build in module like showwebirc.c with whoisasn ?

0005278: Add extban to ban users banned in another channel

Fri, 30 Jan 2026 16:00:24 +0100

We currently have the extban to ban users that are in a channel, with "+b ~c:#channelname".

Would be nice if we could have an extban to ban users banned in another channel (so we can keep some kind of banlist sync across channels registered by the same person.

I was thinking in something like "+b ~b:#channelname" or "+b ~C:#channelname".

The latter goes in hand with the extban "~c" as they're kinda related.

Thoughts?

0006607: Documentation update

Fri, 30 Jan 2026 07:53:21 +0100

Hello team,

in the documentation of the connthrottle.set, you've said that we should use action = on in the parameter.
while it works well when you provide : enabled : true or false

0006608: connthrottle.set not working

Fri, 30 Jan 2026 07:41:17 +0100

Hello,
when you set the connthrottle to false and you call the status endpoint are we expecting to have the enabled field to false ?

0006578: Failed unrealircd module install exits with code 0

Fri, 23 Jan 2026 13:16:52 +0100

The exit code of "./unrealircd module install" is 0 (Success) even if the command has failed

0006602: Invalid channel user limits are normalized to 1

Fri, 23 Jan 2026 08:46:32 +0100

When sending "MODE #chan l :abc" or "MODE #chan l :-1", Unreal converts it to "MODE #chan l :1": https://github.com/unrealircd/unrealircd/blob/1c461db46de2b0d7d0c5b0b1f4e71439170f87e0/src/modules/chanmodes/limit.c#L191-L199

This is inconsistent with other IRCds: Charybdis/Solanum, Ergo, Hybrid/Plexus4, irc2, ircu2/Nefarious ignore the MODE command entirely, while InspIRCd and ngIRCd send ERR_INVALIDMODEPARAM then ignore it too.

0006606: Regression: services kill users who use post-registration SASL

Fri, 23 Jan 2026 07:48:55 +0100

1. User "bar" connects (without authenticating), gets RPL_WELCOME, MOTD, etc.
2. User "bar" SASLs, gets back "My.Little.Server 903 bar :SASL authentication successful"
3. User "bar" gets RPL_WELCOME and MOTD **again**
4. User "bar" gets killed: :My.Little.Services KILL bar :My.Little.Services (Ghost detected via nick collision (new))

This happens with both Atheme and Anope, since https://github.com/unrealircd/unrealircd/commit/0cf0c0faa27ccfc4308834cb98ae92dcc0c31b64

0006591: OOM kill because of low memory during installation

Mon, 19 Jan 2026 10:33:38 +0100

Upgrade fails at this stage:

CC src/libpcre2_8_la-pcre2_match.lo
gcc: internal compiler error: Killed (program cc1)
Please submit a full bug report,
with preprocessed source if appropriate.
See <file:///usr/share/doc/gcc-6/README.Bugs> for instructions.
Makefile:2721: recipe for target 'src/libpcre2_8_la-pcre2_match.lo' failed
make[1]: *** [src/libpcre2_8_la-pcre2_match.lo] Error 1
make[1]: Leaving directory '/home/test/unrealircd-6/extras/pcre2-10.45'
Makefile:1581: recipe for target 'all' failed
make: *** [all] Error 2

dmesg tells me it's because of low memory:

and I have this error in dmesg: Out of memory in UB 1905: OOM killed process 32385 (cc1) score 0 vm:173224kB, rss:128360kB, swap:0kB

0006604: ~msgbypass:censor does not work on ~text:censor bans

Thu, 15 Jan 2026 08:21:43 +0100

A ~msgbypass:censor exemption does not work on ~text:censor bans, as it may be expected.