The downstream connection callback must only be invoked when the event
that SSL requests for the connection to make progress has actually
occured. Otherwise, the downstream callback might do nothing but
re-queue an unrelated event (e.g. in user_net_io_want_write), and the
event loop comes around instantly while making no progress. Track the
SSL-requested events separately and deliver the required downstream
event when they fire.
Sample strace:
epoll_wait(0, {{EPOLLIN, {u32=96, u64=96}}}, 91, 10000) = 1
: net_ssl_callback in state tls_st_need_write calls cb NET_EVENT_WRITE
: User writes data, OpenSSL tries to write data
write(96, <snip>..., 170) = -1 EAGAIN (Resource temporarily unavailable)
: handle_openssl_error requests NET_EVENT_WRITE
epoll_ctl(0, EPOLL_CTL_MOD, 96, {EPOLLOUT, {u32=96, u64=96}}) = 0
: User callback then requests NET_EVENT_READ|NET_EVENT_WRITE
epoll_ctl(0, EPOLL_CTL_MOD, 96, {EPOLLIN|EPOLLOUT, {u32=96, u64=96}}) =
: Data available for *reading*
epoll_wait(0, {{EPOLLIN, {u32=96, u64=96}}}, 91, 10000) = 1
: net_ssl_callback in state tls_st_need_write calls cb NET_EVENT_WRITE
: again...
This could essentially happen due to time drift,
high load, or the process being put in sleep for a while.
The reason is that recurring timers could be added to the same time slot
as the timeslot being handled.
Previously you would have to do something like this:
for (type foo = (type) list_get_first(list); foo; foo = (type) list_get_next(list)
{
/* code */
}
Now, you can instead write this as:
LIST_FOREACH(type, foo, list,
{
/* code */
})
Basically, boilerplate stuff including the casting is gone.
This API provides transparent asynchronous DNS lookups
with both IPv4 and IPv6 support, and in addition will try
to connect to all addresses until one of them work.
This implements the "happy eyeballs" algorithm provided that the client supports IPv6
and that the DNS records provides both IPv6 and IPv4 addresses.
This caused the backends to return an error code, which
in turn ended the mainloop. However, several other things
also might occur in the main loop, such as DNS lookups which
come prior to creating any connections that in turn would be monitored.
Fix bug #185 - Args of !commands lost/damaged.
All string arguments were incorrectly freed after being added to the argument list for a command.
Instead this fix makes sure it is properly copied into a new string, and by doing so this requires
a new API for dealing with hub command arguments in a type safe manner, and also allows for each
argument to be cleaned up properly when the command is no longer needed.
This also fixes issues with parse errors for certain types, and optional arguments (previously it was impossible
to tell the difference for an integer with value 0 or if no integer was given).
All arguments can now be accessed through the new functions
hub_command_arg_reset() and hub_command_arg_next().
These functions are also exposed to plug-ins.
The argument type notations for 'n' has changed to mean nick (string),
and 'u' is used for a user (struct hub_user - must be online).
Jan Vidar Krey
Hector Martin
Tillmann Karras