[omniORB] COMM_FAILURE_UnMarshalArgument - explanation needed

[Duncan Grisby]

On Thursday 29 September, "Wernke zur Borg" wrote:

> Further to my previously reported problem of an endless loop (RE: [omniORB]
> Endless loop in omniORB - need help!) we observe spurious
> COMM_FAILURE_UnMarshalArgument exceptions on both sides of a server-client
> connection. No endless loop is entered in these cases.
> 
> This is the trace:
> 
> omniORB: (5) inputMessage: from giop:tcp:195.74.176.35:57104 12 bytes
> omniORB: (5) 
> 4749 4f50 0102 0005 0000 0000           GIOP........
> omniORB: (5) throw giopStream::CommFailure from
> giopImpl12.cc:1243(0,NO,COMM_FAILURE_UnMarshalArguments)
> omniORB: (5) Server connection refcount = 1
> omniORB: (5) Server connection refcount = 0
> omniORB: (5) Server close connection from giop:tcp:195.74.176.35:57104

This is just the server noticing the client cleanly closing the
connection. It is not an error at all. The 12 byte message is a GIOP
CloseConnection message. The sending side doesn't log that it's sending
the message, because the message is sent from a different bit of code to
all the other kinds of messages. It probably ought to be logged.

> Another occurence of the same exception but with "sendChunk" is this:
> 
> omniORB: (3) sendChunk: to giop:tcp:195.74.176.35:57093 24 bytes
> omniORB: (3) 
> 4749 4f50 0102 0001 0000 000c 0000 033c GIOP...........<
> 0000 0000 0000 0000                     ........
> omniORB: (3) throw giopStream::CommFailure from
> giopStream.cc:825(0,NO,COMM_FAILURE_UnMarshalArguments)

This one is the server noticing that the client has gone away. Again
nothing to worry about (well, you might be unhappy that the client timed
out, but it's not an omniORB problem).

Both of these things are perfectly normal, and nothing to do with the
looping you're seeing.

Cheers,

Duncan.

-- 
 -- Duncan Grisby         --
  -- duncan at grisby.org     --
   -- http://www.grisby.org --