Re: Moab Channels

[Ravi Chamarty <ravi@ittc.ukans.edu>]

Hi Mike,

  Moab does not blow up.

> When you say it "encounters an error", what exactly does it do?  Does
> Moab blow up?  What I would expect would happen is that the received
> packets would be dropped until the parent finally ran and registered
> the demuxkey.  Note that this isn't necessarily the correct behavior,
> just what I would expect in the current implementation.  It may be that
> we need to export the enable function so that an EE can do:


This is exactly what is happening. However, the parent thread does not
seem to progress for some strange reason. For debugging purposes, I
commented out the 'socketdata_intr' thread and observed the channel setup.
It progresses well and the rest of channel operations occur correctly.
Of course, the recvfrom part is now absent.

> 
> 	chan = an_inchan_create(...);
> 	<do some more initialization>
> 	an_inchan_enable(chan);
> 	...
> 
> Would this help in your situation?
>

I am not sure how this would help. Right now, enabling the channel is
being done as a part of 'an_inchan_create'. 
 
> Sorry about all these problems.  As you can tell, most of our testing has
> been using "raw" ethernet channels and we haven't done much with socket
> channels.  I will work on testing these better...
> 

Thanks,
Ravi