Wireshark · Ethereal-dev: Re: [Ethereal-dev] Exception in pdml

[Wesley Shields <wshields@xxxxxxxxx>]

Guy Harris wrote:

> No, ideally, an invalid value shouldn't *exist*.  A field_info structure 
> shouldn't *have* a negative value.

Agreed.  I have a jpeg that when viewed through a browser the resulting 
capture will throw the exception when outputting to pdml.  I can provide 
that jpeg, or the capture, if necessary (it's just a generic green ball 
used as a bullet in webpages).

> There was a bug in the Mobile IPv6 code that caused that to happen; I 
> checked in a fix a couple of days after that message was sent.
>
> I have a change that will cause Ethereal to abort if a dissector 
> attempts to create a field with a negative length or set the length to a 
> negative value, so that bugs of that sort will be caught.  I'm testing 
> it now with a collection of captures I have, in order to see whether any 
> dissectors have bugs of that sort; once the testing is done, and any 
> dissectors it finds bugs in are fixed, I'll check that in.

This idea is much better than my patch of checking for negative values 
when printing the pdml.  If a dissector is setting a negative value what 
should be an appropriate fix?  I'd like to know so that the patch I will 
be writing for this particular dissector will match what you are doing 
with the rest.  Is there any problem with setting the length to 0?

-- WXS