Re: [Discuss-gnuradio] stream_to_vector input data size and lock/unlock

[Eric Blossom]

On Tue, Mar 31, 2009 at 04:51:45PM +1300, Kieran Brownlees wrote:
> Hello all,
> 
> I have some interesting behaviour using a stream_to_vector with a large
> number of inputs items (12800).
> 
> The standard method of top_block.connect(), top_block.start() works fine but
> when I try and perform the same operation after a top block has already been
> started I get a sched error.
> 
> Sample test case:
> 
> import time
> from gnuradio import gr
> 
> class top_block (gr.top_block):
>     def __init__(self):
>         gr.top_block.__init__(self)
>         s2v = gr.stream_to_vector(gr.sizeof_gr_complex, 12800)
>         ns = gr.null_source(8)
>         self.connect(ns, gr.null_sink(8))
>         self.start()
>         self.lock()
>         self.connect(ns, s2v, gr.null_sink(102400))
>         self.unlock()
> 
> if __name__ == '__main__':
>     app = top_block()
>     print "End"
> 
> If the above python script is run as it is I get this on my 3.1.3
> installation:
> 
> address@hidden:~/test$ python s2vtest.py
> 
> sched: <gr_block stream_to_vector (1)> is requesting more input data
>   than we can provide.
>   ninput_items_required = 12800
>   max_possible_items_available = 4095
>   If this is a filter, consider reducing the number of taps.
> End
> address@hidden:~/test$
> 
> If however you move the self.start() in the above script below the second
> self.connect and remove the lock() / unlock() then it runs fine (and works
> perfectly). I was just wondering if anyone can explain this behaviour to me?
> 
> Interestingly on my 3.2 install the max_possible_items_available is 8191.
> 
> Thank you,
> Kieran

It looks like a bug in the buffer allocator / buffer reuse code that
is executed when reconfiguring a flow graph.  I opened ticket:380 on this.

Eric