[Jack-Devel] Client timeout from jackd1 when calling get_all_connections from set_port_registration_callback

Mike Kazantsev mk.fraggod at gmail.com
Sun Oct 18 12:40:15 CEST 2015


I have a question:

Is it expected that calling "get_all_connections(port)" from a port
registration callback in libjack can cause client being timed-out by

See more detailed test-case description and error output below.

I was creating an app where JACK (used only jackd1) manager code would
connect/disconnect JACK ports as they appear or disappear, with the
assumption that it'd be the only app managing jack connections

Fairly obvious (expected) way to do it seem to be:

* Init client.
* Run "set_port_registration_callback(my_callback)".
* Enumerate all current ports, create all connections between these.

* On every port registration callback:
  * Run "get_all_connections(port)".
  * connect() to ports which should-be - but not yet - connected.
  * disconnect() from ports that should not be connected, but got
    returned from that call.

But when I tried doing that, it seems that running
"get_all_connections(port)" causes (unexpected) error when port with
same name appears for the second time during app lifecycle.

I.e. starting jackd1 as "jackd -d dummy", then starting the client,
then running "alsa_out -d hw:CARD=SB", stopping alsa_out and running it
again, causes the client to time-out.

Error message in client (from libjack, I think):

  cannot read result for request type 10 from server (Connection reset by peer)
  cannot send event response to engine (Broken pipe)
  cannot continue execution of the processing graph (Bad file descriptor)
  jack_client_thread: graph error - exiting from JACK

jackd output when this happens:

  timeout waiting for client test-client to handle a port registered event
  cannot send port registration notification to test-client (Resource temporarily unavailable)

One complication here is that I'm using not a C client, but a python
CFFI wrapper around it (see links below), which seem to translate API
pretty much verbatim from what I see in doxygen docs.
But I can't be sure it's not an issue with the wrapper, of course.

I've attached the python code that 100% reliably reproduces this
behavior on this machine, and on clean a Debian Jessie VM.

Running "get_all_connections(port)" anywhere outside of that callback
seem to resolve (or work around) the issue.

I wasn't able to find any reference to the issue (i.e. something like
"you should not do that!") in the docs or the google, and python module
author can't confirm (see links below) whether it's an API bug/feature
or a problem with python module itself.

So the question is basically whether it's expected behavior of jack API
(and maybe is or should-be documented somewhere), or totally
unexpected, and likely an issue with python module?

Python API wrapper module:

Related issue I've opened with the python API wrapper module:

Python (python2) code to reproduce the issue (also attached to this mail):

Code link above also contains instructions on how to run it to
reproduce the issue.

This code depends on python jack client module linked above, which can
be built/installed with e.g. "pip2 install --user jack-client"
(to ~/.local/, "pip" script is usually packaged in distros as
"python2-pip" [arch] or "python-pip" [debian]).

Would appreciate any suggestions. Thanks!

Mike Kazantsev // fraggod.net
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 213 bytes
Desc: OpenPGP digital signature
URL: <https://lists.linuxaudio.org/archives/jackaudio/attachments/20151018/822912ba/attachment.sig>

More information about the Jackaudio mailing list