Comment # 11
on bug 12826
from Guy Harris
(In reply to Peter Wu from comment #8)
> Comment 5 uses an approach that seems quite common for Lua dissectors (I
> think it is also documented in an example Lua dissector): obtain old handle,
> override dissectors, call old dissector and act on it.
>
> I think that https://code.wireshark.org/review/16176 is sufficient for
> correctness (i.e. not crash on missing data), but unfortunately loses the
> possibility to propagate the end-of-stream flag from the TCP layer to HTTP.
>
> In C dissectors, we rely on code review and conventions to avoid illegal
> "data" parameters (though we do have type confusion problems at times).
>
> We cannot rely on the Lua dissector not to pass garbage. Currently it always
> passes a NULL data parameter which is handled gracefully by at least:
> modbus (mbtcp), ethertype, wlan (ieee80211). (Searched for
> call_dissector_with_data and looked at a random sample).
>
> Maybe we should drop this data parameter and use p_add_proto_data:
...which is a mechanism for persistent data attached to packets; I'd prefer not
to use it for data that can be generated by the calling dissector on the fly.
You are receiving this mail because:
- You are watching all bug changes.
