| Commit message (Collapse) | Author | Age |
|---|
| |
|
|
|
|
|
|
|
|
|
| |
This adds nn_device and nng_device. There were some internal changes
required to fix shutdown / close issues. Note that we shut down the
sockets when exiting from device -- this is required to make both threads
see the failure and bail, since we are not using a single event loop.
I also noticed that the bus protocol had a bug where it would send
messages back to the originator. This was specifically tested for in
the compat_device test, and we have fixed it.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
The CMSG handling was completely borked. This is fixed now, and
we stash the SP header size (ugh) in the CMSG contents to match what
nanomsg does. We now pass the cmsg validation test.
We also fixed handling of certain endpoint-related options, so that
endpoints can get options from the socket at initialization time.
This required a minor change to the transport API for endpoints.
Finally, we fixed a critical fault in the REP handling of RAW sockets,
which caused them to always return NNG_ESTATE in all cases. It should
now honor the actual socket option.
|
| |
|
|
|
|
|
|
| |
I implemented the reqrep compatibility test, which uncovered a few
semantic issues I had in the REQ/REP protocol, which I've fixed.
There are still missing things. and at least one portion of the req/rep
test suite cannot be enabled until I add tuning of the reconnect timeout,
which is currently way too long (1 sec) for the test suite to work.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
This does a few things. First it closes some preexisting leaks.
Second it tightens the overall close logic so that we automatically
discard idhash resources (while keeping numeric values for next id
etc. around) when the last socket is closed. This then eliminates
the need for applications to ever explicitly terminate resources.
It turns out platform-specific resources established at nni_init()
time might still be leaked, but it's also the case that we now no
longer dynamically allocate anything at platform initialization time.
(This presumes that the platform doesn't do so under the hood when
creating critical sections or mutexes for example.)
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
This compiles correctly, but doesn't actually deliver events yet.
As part of this, I've made most of the initializables in nng
safe to tear-down if uninitialized (or set to zero e.g. via calloc).
This makes it loads easier to write the teardown on error code, since
I can deinit everything, without worrying about which things have been
initialized and which have not.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
As part of this, we've added a way to unblock callers in a message
queue with an error, even without a signal channel. This was necessary
to interrupt blockers upon survey timeout. They will get NNG_ETIMEDOUT,
but afterwards callers get NNG_ESTATE.
|
| |
|
|
|
|
| |
Platforms must seed the pRNGs by offering an nni_plat_seed_prng()
routine. Implementations for POSIX using various options (including
the /dev/urandom device) are supplied.
|
| | |
|
| |
|
|
|
|
|
| |
This adds the surveyor protocol, and updates the respondent somewhat.
I've switched to using generic names for per-pipe and per-socket protocol
data. Hopefully this will make 'cut-n-paste' from other protocol
implementations easier.
|
| |
|
|
|
| |
This should eliminate all need for protocols to do their own
thread management tasks.
|
| | |
|
| |
|
|
|
|
|
|
|
| |
In an attempt to simplify the protocol implementation, and hopefully
track down a close related race, we've made it so that most protocols
need not worry about locks, and can access the socket lock if they do
need a lock. They also let the socket manage their workers, for the
most part. (The req protocol is special, since it needs a top level
work distributor, *and* a resender.)
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
PUSH attempts to do a round-robin based distribution. However, I
noticed that there is a bug in REQ, because REQ sockets will continue
to pull down work until the first one no longer has room. This can
in theory lead to scheduliung imbalances when the load is very light.
(Under heavy load, the backpressure dominates.)
Also, I note that mangos suffers the same problem. It does not
make any attempt to deliver work equally, basically each pipe winds
up pulling messages until its own buffers are full. This is bad.
We can borrow the logic here for both REQ and mangos.
None of this is tested yet.
|
| | |
|
| |
|
|
|
| |
This fixes several issues, and brings PUB/SUB to operational
correctness. Included is test code to verify that.
|
| |
|
|
|
|
|
| |
The use of a single function to get both size and length actually
turned out to be awkward to use; better to have separate functions
to get each. While here, disable some of the initialization/fork
checks, because it turns out they aren't needed.
|
| | |
|
| |
|
|
|
|
| |
On retry we were pushing back to the queue. The problem with this is that
we could wind up pushing back many copies of the message if no reader was
present. The new code ensures at most one retry is outstanding.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
| |
This also adds checks in the protocols to verify that pipe peers
are of the proper protocol.
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
This uncovered a few problems - inproc was not moving the headers
to the body on transmit, and the message chunk allocator had a serious
bug leading to memory corruption. I've also added a message dumper,
which turns out to be incredibly useful during debugging.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
This may also address a race in closing down pipes. Now pipes are always
registered with the socket. They also always have both a sender and receiver
thread. If the protocol doesn't need one or the other, the stock thread just
exits early.
|
| | |
|
| | |
|
| | |
|
