On Tue, 2 Nov 2004 14:16:48 +1100, ronnie sahlberg
<ronniesahlberg@xxxxxxxxx> wrote:
> error_string = register_tap_listener("frame", &the_tapinfo_struct,
> NULL,
> (void*)frame_reset, (void*)frame_packet,
> (void*)frame_draw);
>
> In the code above you specify NULL as the filter and thus tap will
> give you an unfiltered view of the packets.
> Replace this with "foo" if you only want packets matching "foo".
>
> Can you please create a patch for the readme files and examples if
> this behaviour was unclear so that others are not bitten by this doc
> bug.
README.tapping clearly states this already (my apologies), I used NULL
because I wanted to synchronise the packets that would be tapped and
displayed, and because *fstring (pointer to a filter string) isn't
know at the time the tap gets registered (and is likely to change
during a simulation process).
>
> On Tue, 2 Nov 2004 14:14:26 +1100, ronnie sahlberg
> <ronniesahlberg@xxxxxxxxx> wrote:
> > The tap system is by design decoupled from the display filters
> > applied to the main window.
Maybe I am abusing the tapsystem in a wrong way, but I really don't
know another way for handling changes in display filters, realtime
capturing....
I have tested that tapsystem gets called at the appropriate time (when
the display filter changes and also during realtime capturing), but
because the dfilter flag is set after the frame_packet() is called the
data delivered to the output are the frames corresponding to the
previous display filter.
Question changes to: is there still a way to (at runtime) put a filter
string to the tap? (so tap only packets with the current display
filter.)
> > If you want to push a filtered subset of all packets to the TAP, then
> > you have to specify the appropriate filter string to the call you use
> > when you register the tap listener.
> >
> > display filters on the main window do not affect which packets the tap
> > listener receives, the tap listener itself controls this.
> >
I haven't sit still and I came up with a solution for the non-realtime
case, using a boolean to determine if its the first time tap gets
called (if true retap). This kind of hackery doesn't work with the
option "update list of packets in realtime". (reason: I think and the
one I try to solve is that frame_draw already gets called after all
know packets are tapped). Suggestions to solve this?
With regards,
Koen
ps: here are my advantages:
typedef struct _frame_tapinfo {
gboolean first_time; /* first time we need to retap ! (reason:
dfilter flag is set after the first time tap gets called */
} frame_tapinfo_t;
void
frame_draw(frame_tapinfo_t *tapinfo _U_)
{
printf("frame_draw(frame_tapinfo_t *tapinfo _U_)\n");
if (tapinfo->first_time){
tapinfo->first_time = FALSE;
retap_packets(&cfile);
tapinfo->first_time = TRUE;
}
}
int
frame_packet(frame_tapinfo_t *tapinfo _U_, packet_info *pinfo,
epan_dissect_t *edt _U_, char *data _U_)
{
if (!tapinfo->first_time && pinfo->fd->flags.passed_dfilter){
printf("Frame number: %d\n",pinfo->fd->num);
// printf("frame_packet(frame_tapinfo_t *tapinfo _U_, packet_info
*pinfo _U_, epan_dissect_t *edt _U_, char *data _U_)\n");
}
return 0;
}
--
Programming today is a race between software engineers striving to
build bigger and better idiot-proof programs, and the Universe trying
to produce bigger and better idiots. So far, the Universe is winning.
--Rich Cook--
