Wireshark-bugs: [Wireshark-bugs] [Bug 9510] qtshark Packet list Protocol and Info columns are sp
Date: Mon, 09 Dec 2013 18:48:35 +0000

Comment # 4 on bug 9510 from
(In reply to comment #3)
> Hi,
> 
> (In reply to comment #2)
> > Column details are NOT set inside if (tree) or PINFO_FD_VISITED().
> 
> Can you share source of your dissector and some sample pcap file?

Yes - we plan to submit it to Wireshark. However, it consists of 8 separate
dissectors, 41 files and over 38,000 lines of code (not including the GUI
stuff). So short of submitting it now and going through the acceptance process,
I'm not sure how helpful it would be.

> 
> > Note that the behavior under the GTK version of Wireshark is correct - only
> > the Qt version exhibits the behavior.
> 
> AFAIK the only difference is that GTK+ version is caching column info, and
> Qt version is not.

That's interesting - though I'm not quite sure what the implications are.

So what is the proper way to set column info? Inside a "if PINFO_FD_VISITED()"
or "if (tree)" block? Or every time the dissector is called?

Interestingly, another one of our protocols is built on UDP multicast. In this
case, the protocol column under Qt is always UDP, but under GTK+ is LBT-RM
(which is correct). But the info column is correct in both cases. But this one
doesn't exhibit the same odd behavior as TCP (resizing the protocol column in
this case does NOT display the correct protocol). And for another protocol
(also built on UDP multicast), both the protocol and info columns are correct
with both GTK+ and Qt. AFAIK the columns are set the same basic way - but it's
possible I'm doing something subtly different.

I can't help but wonder if bug 9511 is somehow related. Just a gut feeling at
this point.

I can package things up and send them to you if you think that would be
helpful. Or if you have some ideas about things to check, I could investigate
those.


You are receiving this mail because:
  • You are watching all bug changes.