on ia64 architecture libunwind comes with gcc. Unfortunately
libunwind is not directly usable as-is and fails at link time:
```
ia64-unknown-linux-gnu-g++ -o perf/.libs/local_lat perf/local_lat.o src/.libs/libzmq.so -lsodium -lrt -lpthread -ldl
src/.libs/libzmq.so: undefined reference to `_ULia64_step'
```
The change adds --{enable,disable}-libunwind flag to control
automatic dependency. The default is unchanged: use if available.
Signed-off-by: Sergei Trofimovich <slyfox@gentoo.org>
Solution:
* Document the new behaviour when generating 'ZMQ_POLLOUT' events
for ZMQ_ROUTER sockets with 'ZMQ_ROUTER_MANDATORY' set to `1`
* Add clarifications for 'ZMQ_ROUTER' socket when
'ZMQ_ROUTER_MANDATORY' is set to `1`
Before this commit, xhas_out() was returning true regardless. This
was correct before the ZMQ_ROUTER_MANDATORY flag as introduced.
However, ZMQ_POLLOUT.
With this commit, _if_ ZMQ_ROUTER_MANDATORY is set, xhas_out() will
return false if ALL peer's outgoing pipes are full.
There is an outstanding high-level design question:
If ZMQ_ROUTER_MANDATORY is set, and zmq_poll() waits for ZMQ_POLLOUT
events, zmq_poll() will immediately wake up if only 1 pipe has
room to send, regardless of the peer, creating a busy loop of
zmq_poll() wake-up, zmq_send() (EAGAIN). There is no way for
the application to selectively wait for ZMQ_POLLOUT for specific
peer(s), which seems somehow necessary in ZMQ_ROUTER_MANDATORY.
This discussion will be addressed in a separate issue.
Signed-off-by: Marc Sune <marc@voltanet.io>
Signed-off-by: Fredi Raspall <fredi@voltanet.io>
Solution: if a CURVE server is using zmq_connect, the same session
will be used for any client "reconnect" (actual binds). This is
acceptable, so do not assert if zap_pipe already exists during the
handshake, but simply reuse it.
Fixes#2608
Problem: When using print_backtrace() on Linux with libunwind, printout of stack traces from multiple threads are interleaved. Solution: added static mutex to serialize printing of stack traces.
Solution: remove the pipe from the real trie when a peer disconnects.
Also add a unit test that exercises the behaviour by reconnecting
a different socket and sending a message that matches.
Fixes#2601 and introduced by #2042
Solution: we can use 'CMAKE_SYSTEM_VERSION' instead of
'${CMAKE_SYSTEM_VERSION}' for the 'if' clauses.
CMake fails to evaluate condition when CMAKE_SYSTEM_VERSION is
empty, which can happen with a default installation of Mingw-w64
in Linux.
Solution: use lower case ws2_32.lib to enable cross compilation
from platforms with case-sensitive filesystems.
When cross compiling the tests with Mingw-w64, CMake cannot
locate Ws2_32.lib
Solution: revert change that made ZAP mandatory.
The "Stonehouse" pattern, where CURVE is used only for encryption and
without authentication, is a valid use case so we should still
support it.
Also restore CURVE testing in the test_heartbeat.
Fixes#2594
Solution: if inproc://zeromq.zap.01 exists, which means ZAP is
enabled, abort immediately if it cannot be used (eg: out of memory)
or it is configured incorrectly (eg: wrong socket type).
Otherwise authentication failures will simply be ignored and
unauthorised peers will be allowed to slip in.
problem: for zmq radio/dish pattern, if the radio process restarts, the dish will not resend subscriptions to radio. And the result is that the dish will never receive any more messages.
solution: in session_base_t::reconnect (), take ZMQ_DISH into consideration when invoking hiccup method.
In function session_base_t::reconnect, if we set immediate to 1 and set linger, we will get into first block of reconnect function, and set pipe to NULL, but we forget to cancel timer of linger. Once timer tiggered, we will get coredump. Solution: cancel timer in the end of set pipe to NULL
Some #define switches cause the body of entire files to be omitted. This
causes a linker warning on Visual Studio 2017, for example
warning LNK4221: This object file does not define any previously
undefined public symbols, so it will not be used by any link
operation that consumes this library
Since this is warning us about something that shouldn't be
earth-shattering news, we add a linker flag to suppress this warning on
MSVC builds.
Solution: don't set thread name on Android
Setting a thread name on Android may fail with "permission
denied" error and abort the process due to failed assertion.
Tested on Android 5 and 6 (two phones).
Strangely enough, it only happens on signed APKs and is fine
in debug. Using JeroMQ is not an option as we need TCP keepalive
settings and authentication which JeroMQ doesn't support.
Solution: roll back the pipe if writing messages other than the
first fails in router::xsend. Roll it back also when the pipe is
terminating.
Also add test case that reproduces the memory leak when ran with
valgrind.
Fixes#2567
Solution: roll back the pipe if writing messages other than the
first fails in router::xsend.
Also add test case that reproduces the memory leak when ran with
valgrind.
Fixes#2567
Solution: use either a wildcard IPC, or where the codepath needs to
be tested a file named after the test, so that it is unique and there
is no clash on the filesystem, allowing parallel test runs.
Solution: use ZMQ_LAST_ENDPOINT in most places. This alllows running
tests in paralle, and on over-booked shared machines where many of
the ports would be already in use.
Keep 3 tests with an hardcoded port, as there are some code paths that
require it (eg: connect before bind), but list those ports in
tests/testutil.hpp as macros so that they do not overlap and still
allow parallel runs.
These changes were inspired by a patch uploaded to Ubuntu by the
package maintainer, Steve Langasek <steve.langasek@ubuntu.com>.
Thank you Steve!
Solution: add new tokens to .travis.yml and change ci_deploy.sh
script to use Github APIs to create a temporary branch at the tag,
and the OBS APIs to trigger a source service run in the stable and
draft release projects:
network:messaging:zeromq:release-stable
network:messaging:zeromq:release-draft
The branch hack is unfortunately necessary as it is not possible to
modify OBS sources with the token APIs, and it is also not possible
to automatically fetch the latest tag in the service files.
The temporary branch is immediately deleted.
Solution: a new option has been added to the tar_scm service, so use
it to set the version to last_tag+git<last commit date> which is more
useful.
Unfortunately it's not possible to set it to the current version as
set in the header files, as it's not possible to parse files, only
commit ids, dates and git tags. But it's a step forward.
Solution: do not define ZMQ_HAVE_POLLER in src/zmq_drafts.h otherwise
src/zmq.cpp will implement zmq_poll using the new poller classes.
Same for ZMQ_HAVE_TIMERS, even though it has no internal effect, but
to be safe against future development.
Problem: there is no test coverage for GSSAPI.
Solution: add a test structured like the CURVE test.
The test is not built if libzmq is not configured with
--with-libgssapi_krb5. It will report SKIPPED status
if the required environment is missing (see below).
Environment: KRB5_KTNAME and KRB5_CLIENT_KTNAME
environment variables must point to a keytab file
containing creds for a host-based test principal
(see comment at top of source for details).
Kerberos must be configured and a KDC containing the
test principal must be running, otherwise the test
will fail/hang.
N.B. For now, the test must use the same principal for
both client and server roles because it seems impossible
to set them to different principals when they are
threads in the same process. Once one principal is
cached in credential cache, attempts to acquire creds
for a different "desired name" seem to be ignored and
the cached principal is used instead.
Problem: GSSAPI DRAFT code was made conditional on
ZMQ_BUILD_DRAFT_API, but zmq_draft.h duplicates the DRAFT
symbols definitions from zmq.h so this is unnecessary.
Solution: drop the extra ifdefs
Problem: GSSAPI NAMETYPE socket option numbers were modified
+1000 when moved to DRAFT section, but should use the definitive
values while in DRAFT to minimize disruption later.
Solution: renumber the socket options
Problem: GSSAPI NAMETYPE options were not documented in man
pages for zmq_getsockopt() and zmq_setsockopt().
Solution: add new options to these manual pages.
Problem: The new GSSAPI NAMESPACE options should have been
added to the DRAFT section of the API so they can be changed
until stabilized.
Solution:
- Move defines to the DRAFT section of zmq.h
- Duplicate them in zmq_draft.h, as is the local custom
- Compile only if defined (ZMQ_BUILD_DRAFT_API)
- Refactor internals slightly to avoid #ifdef hell
Problem: principals are looked up unconditionally
with the GSS_C_NT_HOSTBASED_SERVICE name type.
Solution: Add two new socket options to set the name type
for ZMQ_GSSAPI_PRINCIPAL and ZMQ_GSSAPI_SERVICE_PRINCIPAL:
ZMQ_GSSAPI_PRINCIPAL_NAMETYPE
ZMQ_GSSAPI_SERVICE_PRINCIPAL_NAMETYPE
They take an integer argument which must be one of
ZMQ_GSSAPI_NT_HOSTBASED (0) - default
ZMQ_GSSAPI_NT_USER_NAME (1)
ZMQ_GSSAPI_NT_KRB5_PRINCIPAL (2)
These correspond to GSSAPI name types of:
GSS_C_NT_HOSTBASED_SERVICE
GSS_C_NT_USER_NAME
GSS_KRB5_NT_PRINCIPAL_NAME
Fixes#2542
