UnrealIRCd Bug Tracker - Issues

0004157: config definable oper levels

Tue, 21 May 2013 21:37:13 +0200

Add config defined oper levels/classes along the same concept as charybdis or inspircd.

0004207: RFC: bother with hybrid-style localClient split or not

Tue, 21 May 2013 21:33:26 +0200

do we want to split most of aClient structure out, or continue with the 2.8-style hack?

(this can be sorted post-alpha1, but soft-blocks the memory pooling work)

0004196: TS6-style SID/UIDs

Tue, 21 May 2013 21:12:04 +0200

Implement TS6-style SID/UIDs.

Things for discussion:

1. do we want to keep UIDs secret as required by TS6 (this will never be a conformant implementation of TS6)

2. how will the SID handshake go? PROTOCTL SID= token?

0004210: check for cleanliness with clang-analyzer and coverity

Tue, 21 May 2013 12:06:27 +0200

summary covers it all. i'd like to make a pass with coverity and clang-analyzer to ensure we're not doing anything too messed up.

we've touched 100k lines of code since branching off 3.2 so :)

0004188: Unreal 3.4 alpha1 blockers

Tue, 21 May 2013 12:06:27 +0200

These are my blockers for 3.4 alpha1.

0004209: race-condition (?!) with exit_one_client()

Tue, 21 May 2013 12:05:10 +0200

seems we can corrupt ourselves by constantly connecting and squitting a server.

we should like, take a look at this with purify and stuff.

0004208: Create bounds-checking ircsprintf and use it everywhere

Tue, 21 May 2013 12:02:39 +0200

The patch changes ircsprintf to ircsnprintf, and changes every use to add length checking, and replaces some potentially vulnerable string abuse not directly related to ircsprintf. Replaces several calls to snprintf with ircsnprintf and removes lots of strcat and strcpy.

0004203: convert commonly allocated structures to use mempool code

Mon, 20 May 2013 22:36:45 +0200

aClient and aChannel look like good initial targets. also dbufs.

what else should we do?

0004117: Channel mode - Prevent channel destruction when channel is emptied.

Mon, 20 May 2013 21:54:40 +0200

Currently when everyone leaves a channel, it is destroyed. This causes the ban list/invite list/channel creation date to also be destroyed.

I am suggesting creating a new cmode that is only settable by /oper's to prevent the channel from being destroyed when it's emptied.

This would also make the channel show up in /list with 0 users (pending the channel has the cmode and is not +s).

The only time this would fail is if *ALL* the servers on the network go offline.

0004206: Use C99 booleans everywhere

Mon, 20 May 2013 21:32:05 +0200

Kind of a rainy-day task. Not an alpha1 blocker, but we should have it cleaned up before a final release.

0003946: CS INVITE noisy due to ircd invite notice

Mon, 20 May 2013 21:26:33 +0200

Implemented a CS JOIN that calls CS INVITE (if user has access), so that a JOIN may always succeed (regardless of +i, +k, +b, etc). Unfortunately every time a user with access uses CS JOIN (NS AJOIN uses CS JOIN), an invite notice is sent to the channel. For some channels, this is considered annoying.

Can we add a SVSINVITE that doesn't do this, or make a configurable option to disable the notice?

0004131: modulize client CAP system

Mon, 20 May 2013 19:33:18 +0200

Modulize the new CAP system, so modules can add/remove clicap's... similar to how ISupport, User Modes, and so on are modulized.
This means moving part of m_cap.c to the core.

** OLD STUFF, ALREADY DONE IN 0004104: **
Additionally, when at it, add the possibility to disable SASL.

Also add the possibility to disable CAP altogether.

Since all those happen pre-auth, if a serious (crash)bug has been found it should be possible to quickly disable the feature.

0004191: sha2-512 or sha3-512 (keccak) cloaking module

Mon, 20 May 2013 19:05:09 +0200

we should offer a sha3-512 cloaking implementation (or at least sha2-512).

md5 is becoming a lot easier to crack these days...

0004205: clean compile without -fno-strict-aliasing

Mon, 20 May 2013 19:00:01 +0200

remove -fno-strict-aliasing, fix resulting warnings.

0001873: ELIST

Mon, 20 May 2013 18:40:46 +0200

ELIST=options
Options:
M = can search for a channel matching the given text
N = can search for a channel NOT matching the given text (!)
U = can search based on user count (>, <)
T = can search based on topic setat time (T>, T<)
C = can search based on creation time (C>, C<)

Unreal supports MNU. TC was removed a while ago, I can't see why, the changes says it "fixed a bug" but no mention of what. So perhaps we can look into reimplementing these?

Also, Unreal does something incorrect. For MN, Unreal applies the masks to channel names *AND* topics. Meaning if I do, /list *blah* it checks for #*blah* as well as a channel topic containing "blah." It should not apply to topics. Instead, I propose adding a new char where you can add masks for topic searching. Any suggestion on a char? (note: t can't be used since Bahamut decided to use lowercase for ELIST so it has to be case-insensitive)

0002483: "Numerics" instead of nicknames over links

Mon, 20 May 2013 09:51:56 +0200

Now, before you flame me, I understand this will be a *long* time happening, if at all.
I just wanted to enquire as to whether it will happen at ALL, ie the shortening of nicks to (say,) b64 numerics like in ircu, etc.

0002579: After synch has ended, IRCd will send JOIN instead of token regardless of PROTCTL setting.

Mon, 20 May 2013 09:48:00 +0200

Read topic.

I was working on services, and noted that during synch, the IRCd would use the C token, after synch however the IRCd would start using JOIN instead of C whenever a user parted/rejoined a channel - not a good thing if you're only supporting tokens.

0002691: Some commands don't use token form in server->server

Mon, 20 May 2013 09:47:22 +0200

EOS doesn't

Nickname collision stuff in m_nick doesn't

I'll post as I find more.

0003328: Possible tokenization in the Server 2 Server protocol ( JOIN, NOTICE )

Mon, 20 May 2013 09:46:37 +0200

A copple commands are never (in my experience) send as tokens.
JOIN
NOTICE

there can be a cople more didnt test them all.

0003662: Remove AKILL command

Mon, 20 May 2013 09:40:32 +0200

No point to really leave it, it just takes up space.

IIRC codemastr once said it was there to return something when servers or services issued an AKILL command, but all modern services use the proper TKL commands.

If users would want an AKILL command, they now have the ability to use real command aliases.

0003134: Automatic suffix on loadmodule

Mon, 20 May 2013 09:33:43 +0200

Instead of:
loadmodule "modules/blah.dll";
it should become:
loadmodule "modules/blah";
(it will add .dll for windows, and .so for *NIX, etc)

Additionally, let the make process add a symlink called 'modules' pointing to 'src/modules', so we can make it all loadmodule "modules/etcc" instead of src/modules for half of the *NIX installs (not all..), and modules/ for win32 :P

0004204: Default value for LISTEN_SIZE is empty, breaks build

Mon, 20 May 2013 09:25:04 +0200

When I run ./Config and keep giving the default answers with [Enter], I end up with LISTEN_SIZE being defined to nothing:

include/setup.h:201:#define LISTEN_SIZE

This doesn't compile well:

s_bsd.c:445:32: error: expected expression before ‘)’ token
result = listen(listener->fd, LISTEN_SIZE);
^

Looks like not everyone can reproduce it, but:

05:45 (kaniini) at any rate, LISTEN_SIZE should be set to SOMAXCONN

0002181: WebTV translation of NOTICE -> PRIVMSG doesn't work for server broadcasts

Mon, 20 May 2013 04:36:41 +0200

When enabled, a user with usermode +V normally receives all notices as PRIVMSG. This is not true for a server broadcast notice, as depicted in this debug log:

-> irc.* MODE aquanight +V
<- :aquanight MODE aquanight :+V
-> irc.* NOTICE aquanight :test
<- :aquanight!aquanight@* PRIVMSG aquanight :test
-> irc.* NOTICE $irc.* :test
<- :aquanight!aquanight@* NOTICE $irc.* :test

0002152: NOSPOOF for servers?

Mon, 20 May 2013 04:34:43 +0200

This was recently brought up in the forums here: http://www.phpmemx.net/~unrealir/forums/viewtopic.php?p=6090#6090 [^]

Servers are not checked when NOSPOOF is enabled.

I can confirm this with Anope logs if you want.

The easiest way to deal with this is just send an appropriate PING (with cookie, preferably) prior to sending the server's pass/protoctl/server back. If the PONG is not correctly (exactly) replied before the class::pingfreq hits, ERROR :Closing link (Ping timeout) :P . You could, of course, also go as far as bahamut does, and do PING/PONG exchanges between each stage of the synchronization.

0003726: Custom cloak module

Mon, 20 May 2013 04:06:31 +0200

I've been using the unreal software where for quite sometime, and i've been thinking, Seeing as Unreal is still in development and alot of people are using this that they should have an option to customize there host masking
which presently looks something like
rox-510E1CEB.ip.secureserver.net, of course rox being changeable.

perhaps a more customizeable host mask would appea=l, purely an option to users tho for example CustomeMasking "Customizeable variables"

For example
"1" - Default, Old method of hidden host masks "rox-702039293.ip.ip.ip"
"2" - New Style Perhaps a secure md5 table
"rox-742929dcb631403d7c1c1efad2ca2700.secure
Of course have a var aviaible for the secure version (md5) table of there ip so u can do instead of "1"-"2" you can do
"w/e-u-want-here.%secureip.-w/e-message here"

I think alot of users are wanting modules for this, you should have it built in as an option.