The gssapi has some helper functions gssalloc_malloc()/gssalloc_free()
which on windows doesn't call malloc()/free(). Instead these are
wrappers around HeapAlloc() and HeapFree(). To complicate matters
gssapi doesn't export these helper functions, so you're left using
the allocation method of your choice.
See Here:
89683d1f13/src/lib/gssapi/generic/gssapi_alloc.h
The zmq gssapi implementation is calling malloc and then calling
gss_release_buffer() to free the memory. gss_release_buffer uses
gssalloc_free() to free this buffer which on windows calls HeapFree()
instead of free(). This causes an access violation on windows.
Linux provides accept4(2) which will return a socket with FD_CLOEXEC set
when called with the SOCK_CLOEXEC flag. So call this when available and
fall back to fcntl(..., FD_CLOEXEC) if not.
getifaddrs() can fail transiently with ECONNREFUSED on Linux.
This has been observed with Linux 3.10 when multiple processes
call zmq::tcp_address_t::resolve_nic_name() simultaneously.
Before asserting in this case, make 10 attempts, with exponential
backoff, given by (1 msec * 2^i), where i is the attempt number.
Fixes#2051
On Windows, the written message does not seem to be guaranteed to be
written to stderr, in particular when stderr is redirected to a file. I
suppose this is because RaiseException terminates the process in a way
that does not give the CRT a chance to flush stdio buffers (or if it
does, there might be a problem when more than one CRT instance is linked
into the program and they overwrite each other's exception handler). Either
way, just make sure the assertion message ends up written to stderr to
ease diagnostics.
Solution: Provide poll() for Windows as well. This is a build option that
defaults to off as the resulting binary will only run on Windows Vista or
newer.
This is not tested with alternative Winsock service providers like VMCI,
but the documentation for WSAPoll does not mention limitations.
On my local machine, throughput improves by ~10 % (20 simultaneous
remote_thr workes to one local_thr, 10 byte messages), while latency
improves by ~30 % (measured with remote/local_lat).
Solution:
- Add checks for **poller_p_ to ensure that we do not segfault when either it
or the value within it are NULL
- Add tests for the above and increase error state coverage
Solution:
Mark them with LIBZMQ_UNUSED macro as per convention; although in future the
appropriate pthread code should be updated to support thread scheduling
priorities (for Mac OS X, et. al.)
The TIPC protocol bindings in ZeroMQ defaults to a lookup domain
of 1.0.0 to prevent 'closest first' search, and instead always
do round robin if several sockets in the network or node have
the same name published. In retrospect, this might have been a
bad idea because it won't work on standalone configurations.
We solve this by allowing an optional domain suffix to be provided
in the address, and 0.0.0 should be used in that case, or if the
TIPC address range in the cluster configuration is defined to some
other value. Domain suffixes are only relevant for connecting
addresses.
Signed-off-by: Erik Hugne <erik.hugne@gmail.com>
