This also fixes a bug in tcp_connecter and tcp_listener, which
generated the event not when they failed to close the socket but
when the succeed to close it.
The patch is meant to make the code easier to understand.
The 'wait' attribute is replaced by 'delayed_start'
and 'timer_started' attributes. The former is constant and
is initialized in the constructor. The latter is a flag
reflecting whether a timer has been started and changes during
the lifetime of the object.
Once the object has been terminated, it is unsafe for this object
to refer to its parent.
The bug was responsible for occasional
test_shutdown_stress failures.
The TCP keepalive tuning code has been moved into the newly added
files; this also allows future TCP-specific code to be added into
these files, without bloating the IP level code and establishes a
known file structure for other IP-based transports.
Remember: this is a no-op change, hence no API or functionality
was changed as part of this commit.
shadowing a real EAGAIN return value from the OS. This caused later
assertions of "Invalid argument" in stream_engine.cpp when it attempted to
use a socket which was not connected.
I also add EINTR to mean EINPROGRESS, as per the POSIX and FreeBSD
documentation which specifies that a connect() call interrupted due to a
signal will complete asynchronously.
Signed-off-by: Martin Lucina <martin@lucina.net>
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>
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>
This is a preliminary patch allowing for socket-type-specific
functionality in the I/O thread. For example, message format
can be checked asynchronously and misbehaved connections dropped
straight away.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
When exec is executed to start a different process image old
0MQ file descriptors could stay open, thus blocking TCP ports
and alike. This patch should solve the problem.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
The engine was not used exclusively for TCP connections.
Rather it was used to handle any socket with SOCK_STREAM
semantics. The class was renamed to reflect its true function.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
Till now the random fraction of the reconnect interval was
computed based on process ID. This patch uses pseudo-random
generated (seeded by exact time of when the process was
started) to compute the interval.
Signed-off-by: Martin Sustrik <sustrik@250bpm.com>
zmq_engine and tcp_socket merged into tcp_engine
zmq_connecter and tcp_connecter merged into tcp_connecter
zmq_listener and tcp_listener merged into tcp_listener
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>
