On Tue, Sep 04, 2012 at 11:19:09AM +0200, Jakub Zawadzki wrote:
> On Mon, Sep 03, 2012 at 06:00:06PM -0700, Guy Harris wrote:
> >
> > On Sep 3, 2012, at 4:35 PM, Jakub Zawadzki wrote:
> >
> > > I plan to replace:
> > > typedef void (*dissector_t)(tvbuff_t *, packet_info *, proto_tree *);
> > > typedef int (*new_dissector_t)(tvbuff_t *, packet_info *, proto_tree *);
> > >
> > > with:
> > > typedef int (*real_dissector_t)(tvbuff_t *, packet_info *, proto_tree *, void *data);
> >
> > "Replace"? That means changing *all* dissectors to have the new function signature (not that doing so is the wrong thing to do; it's just going to be a lot of work, even if it's work with a shell script or something such as that).
> >
> > Did you mean "I plan to add to ... this", i.e. keep dissector_t and new_dissector_t around, and add real_dissector_t and have new routines with which "new new style" dissectors can register themselves, with older dissectors not being passed the additional argument when called through call_dissector(), etc.?
>
> I want to have one API, so yes, replace. I know it's monkey job ;-)
>
> > Don't forget to change heur_dissector_t - or add real_heur_dissector_t - and the calls to invoke a heuristic dissector.
>
> Ok, good point, what about another try[1] of merging heur_dissector_t into real_dissector_t,
> heuristic dissectors which haven't accept tvb just return -1.
>
> [1] http://www.wireshark.org/lists/wireshark-dev/201207/msg00197.html
FYI, I added data to new_dissector_t and heur_dissector_t, without real_ prefix.
(commit r44860). I hope to commit changes for call_dissector_only, ... this week.
I'm still ready to merge all *_dissector_t into one API.
Btw. I was thinking it'd be great to have some magic number on begin of
dissector data, instead of 'void *data' we could use some 'struct dissector_data',
defined like:
struct dissector_data {
uint32_t magic;
};
struct dissector_xml_data {
struct dissector_data top;
xml_frame_t *frame;
};
