The old AttachUser() that sent JOIN without topic or names replies
would leave clients in incomplete/broken state. JoinUser() was doing
two things; depending on passed arguments it was either joining user
on the channel on IRC, or attaching clients (properly). Now JoinUser()
joins the user on IRC, and AttachUser() attaches as expected from the
method names.
- PASS [user[@identifier][/network]:]password
- USER user[@identifier][/network] ...
NOTE: There's a slight ambiguosity with the '@' character, which happens
to be a valid character in usernames, but also acts as a marker for the
identifier. Therefore, '@' is considered as part of the username if it's
followed by non-word characters (as in an email address), otherwise as
a marker for an identifier.
This is only an enabler for #343. The rest can be done with modules:
- managing client ID specific playback buffers
- filtering channels based on the client ID
The reason this should be part of ZNC core is that only global modules
have access to OnUnknownUserRaw(), which is needed to capture USER/PASS.
First of all, the aforementioned modules shouldn't be global. Furthermore,
it would be possible to have only one module that parsed and removed the
client ID so that ZNC core woulnd't choke.
This changes znc to handle multi-receiver messages in the same way it handles multi-channel join or parts. This will ensure that no more than one target is ever given to a module for any given call.
Store query buffers per query the same way it's done for channels.
This allows clients to implement persistent query buffers. Queries
remain open across clients and sessions until a client explicitly
sends a command to clear a (closed) query buffer.
A new config option AutoClearQueryBuffer that default to false
ensures behavioral backwards compatibility, and another config
MaxQueries protects from OOM eg. due to PM attacks.
Currently the connection timeout handling of znc uses three magic numbers, each
of which is at least repeated in two unrelated places. This commits defines the
numbers in CIRCNetwork and makes the other places just use this number.
This also renames PING_TIMEOUT to PING_FREQUENCY because I feel that describes
this constant better.
I am not really happy about the name NO_TRAFFIC_TIMEOUT that is used for the
real timeout, but I couldn't think of a better name. PING_TIMEOUT isn't good
because that sounds like the time between sending a PING and the resulting
timeout.
Signed-off-by: Uli Schlachter <psychon@znc.in>
Previous commit added support of it to Csocket.
When encoding is specified, core will convert incoming messages to UTF-8,
and outgoing messages from UTF-8.
When no encoding is specified, it will do nothing to bytes, like before.
This is to be changed somewhere in future, to have UTF-8 on wire by
default too.
When encoding's name starts with *, incoming messages will be treated as
UTF-8, if it is already correct UTF-8. Otherwise, it's converted.
Fix#151Fix#366
The following people agreed with the change, in alphabetical order:
(people who approved in several ways are listed only once)
By email:
- Adam (from Anope)
- Austin Morton
- Brian Campbell
- Christian Walde
- Daniel Holbert
- Daniel Wallace
- Falk Seidel
- Heiko Hund
- Ingmar Runge
- Jim Hull
- Kyle Fuller
- Lee Aylward
- Martin Martimeo
- Matt Harper
- Michael J Edgar
- Michael Ziegler
- Nick Bebout
- Paul Driver
- Perry Nguyen
- Philippe (cycomate)
- Reuben Morais
- Roland Hieber
- Sebastian Ramacher
- Stefan Rado
- Stéphan Kochen
- Thomas Ward
- Toon Schoenmakers
- Veit Wahlich
- Wulf C. Krueger
By IRC:
- CNU
- Jonas Gorski
- Joshua M. Clulow
- Prozac/SHiZNO
- SilverLeo
- Uli Schlachter
At https://github.com/znc/znc/issues/311 :
- Alexey Sokolov
- Elizabeth Myers
- flakes
- Jens-Andre Koch
- Jyzee
- KindOne/ineedalifetoday
- Lee Williams
- Mantas Mikulėnas
- md-5
- Reed Loden
At the last few pull requests' comments:
- Allan Odgaard
- Jacob Baines
- Lluís Batlle i Rossell
- ravomavain
- protomouse
The following commits' authors didn't respond:
Trivial changes:
- f70f1086fd
- 4ca8b50e45
The changes which are not presented in master anymore:
- 5512ed2ea0
- 960a4498f7
- 0f739de2c0
- 7f53cc810bFix#311Fix#218
The specs are going to change to support leap seconds properly,
so ZNC implementation will need to be updated.
But it would be too big change now, when 1.0-beta is out.
So support for this capability is delayed for 1.2
If user is connected without network, try "default" network first, if it
exists. If not, try "user" network. If it doesn't exist too, just use
the first one.
Also configs converted from old configs have "default" network instead
of "user" one.
Buflines need to know which part of text to wrap with the timestamp. The
second parameter to `AddLine` (and shorthands) is that text, which after
wrapping is added as the `text` parameter to `NamedFormat`.
Timestamps are formatted at the moment buffers are flushed to the
client. The client parameter to `GetLine` provides access to the User
and the new server-time capability.
This fixes a bug where an unauthed IRC connection will cause your
clients nick to be the default nick. But this is before we update
the client to use this nick. If a client tried another nick when
connecting then they are still expecting this nick.
