This matches the behaviour of zmq_setsockopt(ZMQ_SUBSCRIBE, ...), which also
silently drops subscription messages if the SNDHWM is reached.
Signed-off-by: Staffan Gimåker <staffan@spotify.com>
This allows us to actually report an error to the caller on resolve
failure, rather than asserting later on in the io thread.
Signed-off-by: Staffan Gimåker <staffan@spotify.com>
We have to use an incomplete type in the interface.
The definition of iovec is only in the implementation.
This appears to following existing practice in 0MQ.
It seems a bit flakey that zmq.h is not included in zmq.cpp,
which is normal practice to ensure the implementation matches
the specified interface. YMMV. I follow 0MQ style.
Also move iovec definition from implementation to interface.
Not clear to me at present if an incomplete type should
be used to avoid gratuitously including <sys/uio.h> in
the interface. The interface can't be used with this include.
We use a distinct context initialisation function to specify
all sockets derived therefrom will be thread safe.
However the inheritance is done exclusively in the C interface.
This is not really correct, but it is chosen to minimise
interference with the existing C++ code, including any
construct or other calls within the C++ code base.
Semantically the C++ code should be unchanged,
physically some data structures and extra methods are
provided by they're only used from the C binding.
1. Reorganise C API socket functions to eliminate bad practice
of public functions calling other public functions. This should
be done for msg's too but hasn't been in this patch.
2. Reorganise code in C API socket functions so that the
socket is cast on one line, the C++ function called on
the next with the result retained, then the result is returned.
This makes the code much simpler to read and also allows
pre- and post- call hooks to be inserted easily.
3. Insert pre- and post- call hooks which set and release
a mutex iff the thread_safe flag is on.
4. Add the thread_safe_flag to base_socket_t initialised to
false to preserve existing semantics. Add an accessor for
the flag, add a mutex, and add lock and unlock functions.
Note: as yet no code to actually set the flag.
With the introduction of subscription forwarding, the first message sent
on a PUB socket using a unidirectional transport (e.g. PGM) is always
lost due to the "subscribe to all" being done asynchronously.
This patch fixes the problem and also refactors the code to have a single
point where the "subscribe to all" is performed.
Signed-off-by: Martin Lucina <martin@lucina.net>
Add the '-Ae' flag and check for gethrtime() on HPUX
Check if CLOCK_MONOTONIC defined before using it - if not, use
gethrtime() if it's available, otherwise fall back to the old
behavior.
Signed-off-by: AJ Lewis <aj.lewis@quantum.com>
This patch fixes the problem described in LIBZMQ-205. The assertion itself
is probably caused by previously queued POLLIN events arriving after POLLIN
has been disabled on the socket.
The following additional bugs have been fixed as part of debugging this
problem:
- pgm_receiver_t does not flush messages written to the session in all
cases which can lead to a stalled reader. Add calls to session->flush ()
in the appropriate places.
- ensure to restart polling when a pending message is flushed in
activate_in ().
Signed-off-by: Martin Lucina <martin@lucina.net>
Opening any PGM socket gives this assertion. The problem is in
pgm_sender_t::plug() which is incorrectly testing the return value from
session::write().
Signed-off-by: Martin Lucina <martin@lucina.net>
When (X)SUB socket is being shut down there may be pending
outbound subscription commands in the pipes. In such case
we want to close the socket immediately instead of waiting
for the commands to be sent.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
When polling on XREP socket in incoming message part was prefetched,
but not the identity of sender. The problem is fixed by this patch.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
These include configurations for both Win32 and x64 platforms. All project
settings have been normalised in property sheets (the ".props" files under
builds/msvc/properties) to simplify maintenance. Build artefacts are all
generated in platform-specific subfolders of bin, lib and obj directories.
Also enables the use of precompiled headers with MSVC10.
This significantly reduces the time required to compile libzmq with Visual
Studio on Windows. It should have no impact on other platforms.
Signed-off-by: Stuart Webster <sw_webster@hotmail.com>
failed: s == retired_fd (ipc_connecter.cpp:174), as reported in LIBZMQ-294.
This patch fixes the bug, and also an identical problem in tcp_connecter
which has not hit people since TCP connect() usually completes via the
asynchronous code path (poll, out_event).
Signed-off-by: Martin Lucina <martin@lucina.net>
Older versions of gcc have problems with in-line forward declarations
when there's a naming conflict with a global symbol.
Signed-off-by: AJ Lewis <aj.lewis@quantum.com>
Expand the original patch to all such forward declarations.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
The new function allows to retrieve options (flags)
from zmq_msg_t.
Signed-off-by: Chuck Remes <cremes@mac.com>
Renamed from zmq_msg_flags to zmq_getmsgopt
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
When there were both '0' and '255' subnodes in (mtrie)
the removal of the node resulted in an infinite loop.
Fixed.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
Previous patches have missed the case when the identity should
be sent from an inproc endpoint. Fixed.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
However, the "durable socket" behaviour wasn't re-added.
Identities are used solely for routing in REQ/REP pattern.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
When zmq::req_t::xrecv detects that a response has no request ID
label, or the ID is the wrong size, it would return an EAGAIN, but
would not discard the remainder of the message. This could allow the
remainder of the message to incorrectly "leak" into a future response,
if it is crafted to look like a reply with a valid response ID.
Discard all remaining message blocks, if the ID is invalid in any way.
This patch changes the Win32 version to call RaiseException instead of abort
(which eventually calls TerminateProcess). This allows crash dumps to be sent
correctly instead of the process disappearing.
Signed-off-by: Paul Betts <paul@paulbetts.org>
