jit.net.send/recv stack overflow when connection interrupted


    Aug 16 2006 | 9:45 am
    dear jitter,
    i experience problems in a setup where i have a tiBook connected to a
    g5; both hava a patch with the jitter network objects (as 'appatched'
    below), with corresponding port settings.
    mainly, i send control messages from the tiBook to the G5, and
    matrices in the opposite direction (low res, low framerate).
    in normal use, this works fine.
    but the problem appears in 'unnormal' use:
    for this performance, i need to disconnect my tiBook from the
    network, walk to a different room, and reconnect.
    even that works fine - unless i forget to switch of the matrix-
    sending from the G5.
    if the G5 keeps sending matrices while the tiBook is disconnected,
    the G5 will produce a stack overflow @ the moment that the tiBook is
    reconnected.
    when i resume processing, the G5 will work normally again, but the
    connection to the tiBook won't work anymore.
    when i then restart max on the G5, it will crash at the moment of
    shutting down.
    what would help me is a way to make sure that the connection exists
    (to interrupt the streaming when it is gone); this is not correctly
    reported from the net-objects, it seems.
    any other tips to make this 'fool safe'?
    thanks,
    jan

    • Aug 16 2006 | 5:41 pm
      by the way, this happens with max 4.6 / jit 1.6.
      an 'anormality' of the jit.send/recv objects might be connected to
      this problem:
      -open the jit.net.send / recv helppatches, start transmitting on the
      local loopback.
      -sending the 'connected' message to jit.net.send should report '1',
      which it does, but transmission is interrupted by this, whil a
      subsequent 'connected' still reports a 1.
      -selecting 'any IP' on jit.net.recv brings streaming online again,
      but this procedure can't be repeated often it seems
      i guess the unsent matrices are buffered, until there is a buffer
      overflow. can't that be prohibited one way or the other?
      thanks,
      jan
      On 16.08.2006, at 11:45, Jan Klug wrote:
      > dear jitter,
      >
      > i experience problems in a setup where i have a tiBook connected to
      > a g5; both hava a patch with the jitter network objects (as
      > 'appatched' below), with corresponding port settings.
      > mainly, i send control messages from the tiBook to the G5, and
      > matrices in the opposite direction (low res, low framerate).
      > in normal use, this works fine.
      >
      > but the problem appears in 'unnormal' use:
      > for this performance, i need to disconnect my tiBook from the
      > network, walk to a different room, and reconnect.
      > even that works fine - unless i forget to switch of the matrix-
      > sending from the G5.
      > if the G5 keeps sending matrices while the tiBook is disconnected,
      > the G5 will produce a stack overflow @ the moment that the tiBook
      > is reconnected.
      >