[omniORB] WG: GIOP 1.2 protocol error in input message

[Duncan Grisby]

On Tuesday 11 December, "Michael Sturm" wrote:

> using omniORB 4.0.7 under VxWorks we often face the following
> malfunction in handling GIOP 1.2 input messages:

If possible, update to the latest cut of omniORB on the omni4_1_develop
branch, or failing that, the omni4_0_develop branch. It's much easier to
help if you're using the latest version of the code.

> "From endpoint: giop:tcp:10.49.x.y:z. Detected GIOP 1.2 protocol error
> in input message. giopImpl12.cc:425. Connection is closed."
> 
> This situation seems to occur in alliance with fragmented GIOP messages
> after having updated client side Java Web Start / JRE to 1.6.0_03 (did
> not see this malfunction under JRE 1.4.x).
> 
> Find the traced GIOP messages leading to this in the attachment
> (traceLevel: 30).

The error is that omniORB has received a fragment message that claims to
be the continuation of a request message, but omniORB thinks it has
already seen the complete request. That could be a bug in omniORB, or it
could be that the client is sending invalid data.

Can you run with traceLevel 40 and also traceThreadId 1. That will make
it easier to see what's going on.

Cheers,

Duncan.

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