Re: [Discuss-gnuradio] Flow control with message ports

[Michael Wentz]

Nathan, 

I don't have any other queue in my block - it's just the one associated with the message port. The following starts printing once I send too many messages (before the block downstream has started processing them):

WARN ASYNCHRONOUS MESSAGE BUFFER OVERFLOWING; DROPPING MESSAGE

Looking at tpb_thread_body.cc, it seems that maybe I'm in a special case given I don't have a handler for the block receiving the messages (it decides when to check the queue on its own). In that case it looks like the queue is limited to 100 messages?

-Michael

On Wed, Nov 9, 2016 at 6:34 PM, West, Nathan <address@hidden> wrote:

I'm not sure I understand. There was once a proof of concept flowgraph called pmt_smasher that would effectively keep publishing messages and the queue grows without bounds which was generally considered a low-priority issue (having no back pressure/flow control on message ports).

You're describing different behavior than I understand the message ports to have. Is the queue that's overflowing some custom queue in your block that you dump new messages on to? If so just make that queue grow as more messages come in.

Nathan

On Tue, Nov 8, 2016 at 7:27 PM, Michael Wentz <address@hidden> wrote:

Hi,

I've made a block in Python that has one message port out and no other ports. What the block does is simple: read from a file, parse data into a dict, convert to a PMT, and publish as a message. The port is connected to a sync_block that is acting on these messages when it deems fit. My desired behavior is for the publisher to fill up the queue as fast as possible and block if the queue is full (waiting for room to open up). What I've observed is that the queue will instead overflow and messages will be dropped. Is there any way to have a blocking call to message_port_pub()? 

Looking through the code I do see a method in basic_block to get the number of messages in the queue, which I could use to decide to publish a message or not - but this isn't brought out in the SWIG interface. Is there a reason why? If not, I was thinking about re-defining the SWIG interface for basic_block in my OOT with additional methods, but was wondering if that would create conflicts/weird issues.

Any other ideas for how to do this would be appreciated. I realize I could parse the file in my sync_block, but that's my last resort here.

-Michael

_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio