Wireshark-bugs: [Wireshark-bugs] [Bug 11134] CORBA dissectors show malformed packet errors for v
Comment # 13
on bug 11134
from Guy Harris
(In reply to Evan Huus from comment #12)
> I guess my instinct was that most protocols shouldn't claim
> their payloads, so e.g. TCP shouldn't be passing -1 or tvb_reported_length()
> or anything like that, it should be passing 20 (or more, adjusting for
> options).
I think they should start out saying "I'll claim the whole thing" and, when
finished, say "OK, this is all I needed"; that obviates the need to calculate
the length in advance (assuming it *can* be calculated in advance without first
looking into the packet - which it can't, for IPv4 or TCP, for example).
We might also be able to leverage that into a way to have dissectors just
return a Boolean "I accepted this"/"I rejected this" indication and provide the
"how much I dissected" indication separately (which would avoid problems if
there's a case where the dissectors 1) can reject packets and 2) can also not
dissect any bytes, e.g. if you have protocol A transporting request/response
protocol B, with protocol A responses not having any data in them - I seem to
remember there was a case where that could happen, which is why I didn't
vigorously pursue changing all "old-style" dissectors to "new-style"
dissectors).
You are receiving this mail because:
- You are watching all bug changes.
