[nanomsg] Re: Assertion failed: src == NN_SOCK_SRC_EP && type == NN_EP_STOPPED (..\..\src\core\sock.c:828) on nn_close
- From: Jack Dunaway <jack@xxxxxxxxxxxxxxxx>
- To: nanomsg@xxxxxxxxxxxxx
- Date: Mon, 6 Oct 2014 16:29:52 -0500
Tom -- I ran into similar failure modes (
https://github.com/nanomsg/nanomsg/issues/269) and have developed a fix
that has so far solved this issue in practice (
https://github.com/wirebirdlabs/ftw-nanomsg/commit/bb5d37c70800aa620391c80a30b07dcc83501e26).
Perhaps, review and even try this fix, and report back findings?
Best regards,
Jack R. Dunaway | Wirebird Labs LLC
On Mon, Oct 6, 2014 at 8:15 AM, Tom Quarendon <tom.quarendon@xxxxxxxxxxxxx>
wrote:
> What might the above assertion failure on attempting to close a socket
> indicate?
>
> It doesn’t happen all of the time, so it’s clearly some issue where I’m
> not closing things down properly.
>
>
>
> I’m working with inproc and an NN_PUSH NN_PULL pair of sockets.
> Unfortunately I can’t reproduce this if I try and reduce my problem down to
> a few lines of code, but my application logic is something like this:
>
> Main thread kicks off background thread.
>
> Background thread creates an NN_PULL socket and calls bind. It then goes
> into a loop reading commands from the socket, which it then executes. On
> receiving a special “shutdown” command, it exits its loop and then calls
> nn_close. It is this nn_close that causes the assertion failure. If I
> remove this close I get no issues.
>
>
>
> Main thread then opens a NN_PUSH socket and calls connect. It then writes
> a list of commands it wants executing. It then closes the socket.
>
> To terminate it then creates another NN_PUSH socket, calls connect, and
> then writes the special shutdown signal, and the calls nn_close().
>
>
>
> The only timing issue I can think of is that the main thread sends the
> shutdown signal, and before it then gets a change to close its socket, the
> background thread processes the message and closes its end. Because it did
> the bind it somehow owns some state that is released and when the main
> thread then does its close, the assertion happens. However I’ve put in some
> Sleeps to try and force this sequence, but it doesn’t make the failure
> happen all of the time.
>
>
>
> At the time that the assertion occurs, src==0 (so this is false) and
> type==33987 (so also false).
>
>
>
> I’m hoping someone might have some ideas as to what the problem is.
>
> Thanks.
>
>
>
>
>
>
>
>
>
Other related posts:
