On Jan 26, 2004, at 1:25 PM, Sean Schneyer (TX/EUS) wrote:
I recently tried to upgrade from 0.9.7 to 0.10.0 and found that the
MMSE protocol support is not working as expected. I then started
working my way backwards and found that it works as expected in
0.9.16. Attached is a trace that demonstrates what I am talking about.
Perhaps it was attached, but the attachment appears not to have arrived
here.
In 0.9.16 packet 149 is properly identified as a MMS m-send-req, but
in 0.10.0 it is simply identified as an HTTP Continuation.
When I did a diff between the packet-mmse.c from the two versions,
there were some changes made to the code. Unfortunately, I'm not
familiar enough with the ethereal codebase to know if they are
significant or not:
Those changes probably aren't significant.
Changes to the HTTP dissector, however, might be.
Try turning on the "Desegment all HTTP headers spanning multiple TCP
segments" and "Trust the 'Content-length:' header and desegment HTTP
bodies spanning multiple TCP segments" options for HTTP, and the "Allow
subdissector to desegment TCP streams" option for TCP, and see if that
fixes the problem. If not, and if any of the packets are reported as
having bad TCP checksums, try also turning *off* the "Check the
validity of the TCP checksum when possible" option for TCP.
Also, while we are on the subject of the MMSE protocol, is there
anyway to get Ethereal to reassemble all of the packets in the stream?
Right now it only decodes the first packet and then declares
[Unreassembled Packet].
Again, that's an HTTP issue - the option changes listed above should
help.
