This means that all the handshaking while accepting incoming
connection is done exclusively in I/O threads, thus it won't
overload the application thread's mailbox.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
So far ZMQ_RECONNECT_IVL delay was used only when TCP connect
failed. Now it is used even if connect succeeds and the peer
closes the connection afterwards.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
- ctx constructor was calling mailbox_t constructor implicitly
- moved WSAStartup and WSACleanup to be outside constructor/destructor
Signed-off-by: Pieter Hintjens <ph@imatix.com>
Reaper thread destroys the socket asynchronously.
zmq_term() can be interrupted by a signal (EINTR).
zmq_socket() will return ETERM after zmq_term() was called.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
The internal log socket was subtracted from the number of
available sockets. So, if max_sockets was set to 100,
you could create only 99 sockets. Fixed.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
The patch allows for running 0MQ on Solaris and HP-UX
even though ulimit for max number of file descriptors
is set to unlimited.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
The problem was that RLIMIT_NOFILE can be set to RLIM_INIFINITY
(and that appears to be default on AIX) which caused 0MQ to fail.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
The meat of the patch was contributed by Douglas Creager.
Martin Sustrik implemented storing peer options in inproc
endpoint repository.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
For very high-speed message systems, the memory used for recovery can get to
be very large. The corrent limitation on that reduction is the ZMQ_RECOVERY_IVL
of 1 sec. I added in an additional option ZMQ_RECOVERY_IVL_MSEC, which is the
Recovery Interval in milliseconds. If used, this will override the previous
one, and allow you to set a sub-second recovery interval. If not set, the
default behavior is to use ZMQ_RECOVERY_IVL.
Signed-off-by: Bob Beaty <rbeaty@peak6.com>
So far these were declared as C++ static functions which
was incorrect and caused warnings with SunStudio.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
On systems using GCC 4.0 or newer which support symbol visibility in shared
libraries, use -fvisibility=hidden and only export explict API functions
defined in zmq.cpp. We do not enable -fvisibility on MinGW since this uses a
separate mechanism (__declspec).
Signed-off-by: Martin Lucina <mato@kotelna.sk>
HP's version of STL doesn't allow for initialisation of basic_string
(blob_t) using NULL pointer, while SGI's implementation is OK with
that. Fixed.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
- RAND_bytes function resides in crypto library
- pollfd on AIX used 'reqevents' instead of events and 'retnevents'
instead of 'revents'
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
Added block boolean var to second process_commands() invocation for blocking sockets
instead of always using true. This prevents the process_commands() call from hanging
when a message is received with an empty queue after the call to xrecv() but
prior to the initial call to process_commands() invoked when ++ticks == inbound_poll_rate.
Signed-off-by: Marc Rossi <mrossi19@gmail.com>
For historical reasons queue to transfer commands between
threads was called 'signaler'. Given that it was used to
pass commands rather than signals it was renamed to 'mailbox',
see Erlang mailboxes.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
Threads were so far identified by integers called 'slots'.
This patch renames them to more comprehensible 'tid's (thread IDs).
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
If the socketpair used by signaler_t fills up, this can lead to deadlock.
This patch provides partial resolution by attempting to resize SO_SNDBUF on
the writer side, and if that fails we shall at least assert rather than
hang.
I've also refactored the signaler_t code to make the platform-dependent
parts clearer and have tested both the MSG_DONTWAIT and standard POSIX path
in recv.
The Win32 implementation currently does not implement resizing as I'm not
convinced that it's safe, but it will also assert like other platforms if
signaler_t::send() cannot succeed.
The OpenVMS implementation has been carried forward but is untested.
Signed-off-by: Martin Lucina <mato@kotelna.sk>
devpoll_t used this constant to determine how many events to
retrieve from the poller in one go. The implementation was
changed not to depend on this constant.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
With SunStudio's implementation of STL basic_string constructor
doesn't accept NULL as a parameter even though size of string
is set to zero. Fixed.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
1. ZMQ_LINGER option can be set/get
2. options are part of own_t base class rather than being declared
separately by individual objects
3. Linger option is propagated with "term" command so that the
newest value of it is used rather than the stored old one.
4. Session sets the linger timer if needed and terminates
as soon as it expires.
5. Corresponding documentation updated.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
The comment about memory ownership in uuid_t::to_string
was misleading as on most platforms the buffer is a field
of the object;
The string allocated in UuidToString() wasn't properly freed
with RpcStringFree()
Signed-off-by: Taras Shpot <taras.shpot@eleks.com>
A minor workaround to resolve compilation error with sunstudio compiler,
which does not yet support member templates for std::multimap.
Signed-off-by: Martin Pales <m.pales@gmail.com>
zmq_msg_close now empties the message on zmq_msg_close, thus not
leaving random data in the structure, that may be mistaken for
a valid message.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
Session base class now handles the engine events exclusively. It notifies
derived session types using dedicated "attached" and "detached" events.
Couple of bugs was fixed along the way.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
When a session is being closed down its inbound and outbound
pipe pointers are set to null. If (re) connection happens at
that time, session may try to reinistantiate the pipes which
is wrong. This patch allows session to attach pipes only once
in its lifetime.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
