Age | Commit message (Collapse) | Author |
|
Create one IOHook instance for each hooked socket which contains all the
hook specific data and read/write/close functions, removing the need for
the "issl_session" array in SSL modules.
Register instances of the IOHookProvider class in the core and use them to
create specialized IOHook instances (OnConnect/OnAccept).
Remove the OnHookIO hook, add a dynamic reference to ListenSocket that
points to the hook provider (if any) to use for incoming connections on
that socket.
For outgoing connections modules still have to find the IOHookProvider
they want to use themselves but instead of calling AddIOHook(hookprov),
now they have to call IOHookProvider::OnConnect() after the connection
has been established.
|
|
SSL profiles are now used instead of fixed SSL settings for everything
SSL, making it possible to use completely different settings for each
listener and outgoing connection.
Outgoing connections are broken until the next commit.
|
|
|
|
both openssl and gnutls
|
|
|
|
Also, fix a bug where eval() caused compile errors.
|
|
automatically
|
|
|
|
and on rehash
This eliminates the need for calling OnRehash() in init()
|
|
|
|
- Add CXX11_OVERRIDE to *Regex::Matches and *RegexFactory::Create.
- Fix documentation comment on regex_string.
- Fix various code duplication/layout issues.
|
|
|
|
- Deduplicate getcompilerflags, getdependancies, getlinkerflags.
- Remove $NoPedantic (add -Wno-pedantic to $CompileFlags instead).
- Remove --enable-freebsd-ports-openssl and all related code (this
will be replaced with --no-pkg-config=[name] in the future).
- Remove some unused build system properties.
- Remove support for caching third party include and library paths
(can cause unexpected problems when they change).
|
|
|
|
GnuTLS versions
Use the (old) LIBGNUTLS_VERSION_* defines as a fallback when GNUTLS_VERSION_* is unavailable
Fixes #595 reported by @rhylan
|
|
Also, make all paths in the config relative to their associated
directory. This reverts a change in 2.0 which turned out to be a
terrible idea, especially for system-wide installations.
|
|
|
|
|
|
The majority of modules were logging with their module name as the
log type. There was a few places which were logging to a non-name
type but, with the exception of CONFIG, those messages are so
uncommon that it doesn't make sense to use a seperate type for
them.
|
|
recognized as trusted
|
|
SSL library
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Quotes from the documentation:
"You must allocate the to buffer to be at least length*2+1 bytes long. (In the worst case, each character may need to be encoded as using two bytes, and you need room for the terminating null byte.)"
"The return value is the length of the encoded string, not including the terminating null character."
http://dev.mysql.com/doc/refman/5.6/en/mysql-real-escape-string.html
|
|
- Modules which use this class will now have to catch a
CoreException when opening files if they wish to ignore
the failed loading of a file.
- m_randquote has been cleaned up massively and the RANDQUOTE
command has been removed as it was pretty much useless.
|
|
config
|
|
|
|
|
|
|
|
- Remove virtual keyword from a ton of methods which don't need it.
- Add override keyword to a ton of methods which do need it.
|
|
* 'const char*' to 'const std::string&'.
* snprintf to std::string concatenation.
* Replace duplicated OneOfMatches with InspIRCd::MatchMask.
|
|
This allows us to send a server notice to a user without worrying
about whether they are registered or not.
If a user receives a server notice and they are not registered
then the nickname field will contain an asterisk instead of their
nick name.
|
|
|
|
|
|
|
|
This greatly decreases the load time because the DH parameters no longer have to be (re)generated each time the module is loaded
|
|
Timer::Tick() now has a bool return value: if false is returned the timer is deleted using operator delete, otherwise, if it's a repeating timer then it's rescheduled (readded)
Timers are removed from the TimerManager automatically at destruction
Timers are now stored in a multimap instead of a sorted vector
|
|
WSAEWOULDBLOCK
|
|
- Use #pragma once instead of include guards.
- Move header files in src/modules to include/modules.
- Fixed various spacing issues.
|
|
|
|
|
|
|