Allocate the pipe capture data buffer upfront.

We were allocating it every time we called cap_pipe_dispatch() (or,
prior to I0256daae8478f1100fdde96a16a404465ec200b3, in
capture_loop_dispatch()) and freeing it before the routine in question
returned.

However, we were treating that buffer as if it persisted from call to
call, which worked *only* if freeing and re-allocating the buffer meant
that we'd get back the same buffer with its previous contents intact.

That is *not* guaranteed to work.

Instead, allocate the buffer when we open the capture pipe, and free it
when we close the capture pipe.

Change-Id: Ic785b1f47b71b55aba426db3b1e868186c265263
Reviewed-on: https://code.wireshark.org/review/21948
Reviewed-by: Guy Harris <guy@alum.mit.edu>

0 files changed, 0 insertions, 0 deletions