Re: comm_failure for large sequences using giop betweentwo machines

[Michael Meeks <michael ximian com>]

Hi Bowie,

On Wed, 2003-06-18 at 05:32, Bowie Owens wrote:
> I have attached a patch which extends test/everything to reproduce this 
> problem. For me test/everything then fails with an exception for 
> --ORBIIOPIPv4=1 --ORBIIOPUSock=0. Can you reproduce the problem?

	I've committed this; ( disabled, and added an annoying warning to help
it stay in mind ;-).

	So - I think one of the most interesting things is that it should be
even easier than expected to hook this up for you; since we are getting
a reply - we don't in fact have to have much of the buffer read before
we can trivially validate that it is in fact a correct reply to
something we sent.

	So - I guess, we need (somehow) to split out (and re-use) the cnx
checking code in giop_recv_buffer (handle_ent) - particularly that the
incoming reply comes from the same cnx. the outgoing call was on ;-) and
use that to validate / bin requests when we hit the size barrier to see
if we should continue / authenticate that cnx.

	How does that sound ?

	Regards,

		Michael.

-- 
 michael@ximian.com  <><, Pseudo Engineer, itinerant idiot