| Age | Commit message (Collapse) | Author |
|---|
|
Storing commands in OS socket buffers caused whole lot of
problems when free space in the buffer ran out. This patch
stores commands in ypipes instead and uses socketpair just
to signal the other thread, ie. at most one byte is stored
in the socketpair at any single instant.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
|
|
This patch addresses serveral issues:
1. It gathers message related functionality scattered over whole
codebase into a single class.
2. It makes zmq_msg_t an opaque datatype. Internals of the class
don't pollute zmq.h header file.
3. zmq_msg_t size decreases from 48 to 32 bytes. That saves ~33%
of memory in scenarios with large amount of small messages.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
|
|
On-disk storage should be implemented in devices rather than
in 0MQ core. 0MQ is a networking library and there's no point
in storing network buffers on disk.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
|
|
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>
|
|
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>
|
|
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
|
|
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
|
|
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
|
|
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
|
|
Add signaler_sndbuf_size option to config.hpp which allows the user to
increase the size of the send buffer used by the signalling socketpair.
Implement random backoff for reconnection attempts using a primitive
pseudo-random generation to prevent reconnection storms.
Add wait_before_connect option to config.hpp to allow the user to enable
random delay even on initial connect. Default is false for low latency.
Signed-off-by: Martin Lucina <mato@kotelna.sk>
|
|
|
|
|
|
|
|
|
|
|
|
Sockets may now be migrated between OS threads; sockets may not be used by
more than one thread at any time. To migrate a socket to another thread the
caller must ensure that a full memory barrier is called before using the
socket from the target thread.
The new zmq_close() semantics implement the behaviour discussed at:
http://lists.zeromq.org/pipermail/zeromq-dev/2010-July/004244.html
Specifically, zmq_close() is now deterministic and while it still returns
immediately, it does not discard any data that may still be queued for
sending. Further, zmq_term() will now block until all outstanding data has
been sent.
TODO: Many bugs have been introduced, needs testing. Further, SO_LINGER or
an equivalent mechanism (possibly a configurable timeout to zmq_term())
needs to be implemented.
|
|
|
|
parameter in zmq_init is unused and obsolete
|
|
|
|
|
|
configure.in is now the master source for the package version number,
this propagates to src/platform.hpp (for zmq_version) and doc/Makefile.am
(for documentation generation) automagically.
|
|
|
|
|
|
|
|
|
|
|
|
|
