Re: [Ethereal-dev] Ethereal addition for analysing RTP data

[Guy Harris <guy@xxxxxxxxxx>]

On Thu, Mar 06, 2003 at 04:48:47PM -0700, Greg Morris wrote:
> That did the trick on that one...

OK, I've checked in a change to precede all the error enums with
TAP_RTP_, to avoid name collisions.

> Now I'm almost thru the code... At
> least down to line 1683 when I encounter hopefully the last error...
>  
> tap_rtp.c
> tap_rtp.c(1683) : error C2466: cannot allocate an array of constant
> size 0

Yup, you can't, in any *sane* compiler.

Unfortunately, it appears that, in this case, GCC is not sane.  OK, I
can see value of their hack for allowing zero-length arrays:

	http://gcc.gnu.org/onlinedocs/gcc-3.2.2/gcc/Zero-Length.html#Zero%20Length

as the last member of a structure, if the structure is really a
fixed-length header followed by a variable amount of data.  (However, I
don't *use* that hack, as it's non-portable, and you *can* make it work
with a one-element array - you just have to do a little more work to
compute the total length of a structure with N bytes of data.)

But allowing an array *outside* a structure to be zero length (or, for
that matter, allowing a zero-length array to be the *only* member of a
structure), just seems stoopid to me.

It's definitely a bug in tap_rtp.c, given that the code later assigns a
value to the non-existent first member of that array.

I've also checked in a change to make it a 1-element array, rather than
a 0-element array.

Try removing gtk/tap_rtp.c and doing a "cvs update" after a 10-minute or
so delay, to allow the change to propagate to the anonymous CVS tree -
you should get version 1.4 of gtk/tap_rtp.c (if not, wait some more time
and try again; lather, rinse, repeat).