Solution: removed definitions and uses in subclasses

Solution: add underscore prefix

Solution: add underscore prefix

Solution: pull up common code, introduce new create_engine function in base class

Solution: pull up

Solution: pull up tcp_connecter_t::rm_handle and use in all subclasses

Problem: process_plug, get_new_reconnect_ivl and add_reconnect_timer duplicated across subclasses of stream_connector_base_t

Solution: pull up to stream_connector_base_t

Solution: pull up to stream_connecter_base_t

Solution: extract common base class stream_connecter_base_t

Solution: replace by checks against valid value of _handle

Solution: replace by checks against valid value of _handle

Solution: rename to _reconnect_timer_started

Solution: rename to _reconnect_timer_started

Solution: add underscore prefix to data members

Solution: add underscore prefix to data members

Solution: declare _endpoint const

Solution: extract initialization code into get_peer_address function and declare _peer_address const

Solution: assign result to address

Fixes #3376

Solution: remove them

Problem: one more test fails on GNU/Hurd

Solution: mark test_rebind_ipc as XFAIL on Hurd as it does not
implement getsockname on IPC and thus it's impossible to use
wildcard IPC binds.
Document that ZMQ_LAST_ENDPOINT does not work on Hurd with IPC.

Problem: have to check ABI compatibility manually

Solution: automate it on Travis

Problem: race condition breaks automated OBS release

Solution: instead of using a temporary release branch that gets
deleted after an arbitrary timer, which never matches the time OBS
actually takes to do a full git clone, keep the branch permanently
in place and simply move it forward to the latest tag on each
release.

Problems: test failures on PPC and GNU/hurd

Solution: mark it as XFAIL like the other tests that use ipc://*

Solution: prefer CXX11 atomics if they are available to compiler
intrinsics.
test_hwm_pubsub fails 50% of the times on PPC64 and RISC-V with an
apparent memory corruption of messages sent by the application thread
to the I/O thread when using compiler intrinsics.
Switching to CXX11 atomics makes the test reliable again. The
standard API should be preferred anyway, if available.

Solution: import macro from autoconf-archive that works better
than AC_CHECK_FUNCS

Solution: detect cacheline size for aligment purposes at build time
instead of hard-coding it, so that PPC and S390 can align to a value
greater than the 64 bytes default.
Uses libc getconf program, and falls back to the previous value of 64
if not found.

Problem: ZMQ_SOCKET_LIMIT and ZMQ_THREAD_PRIORITY have the same value

Solution: remove documents and tests for ZMQ_THREAD_PRIORITY getter. It
never worked and can never work as it has the same value as a get-only
option ZMQ_SOCKET_LIMIT. It cannot be changed without breaking ABI.
Note that the setter works fine as ZMQ_SOCKET_LIMIT is get-only.

Problem: GCC warnings about string overflows in tests

Solution: mostly false positives, but fix them

Problem: gssapi pkg-config check in configure.ac does not work

Solution: correctly enable the definition in platform.hpp so that the
gssapi support is actually built in if requested and available.

Problem: NEWS for 4.3.1 does not mention CVE number

Solution: add it now that it's been assigned

Problem: test_rebind_ipc still fails

Solution: create manually the random IPC file to reuse