RABRET Laurent RD-MAPS-ISS wrote:
Another ugly patch to fix the same problem (I forgot to submit it last
week).
Checked in - except for the file type part; that's a "should never
happen" case, i.e. it's a bug in Ethereal if that ever happens, so we
want a crash so it gets reported, and it really "shouldn't happen", i.e.
Wiretap should never return a file type for which it has no name.
Did you see a case where that happened? If so, the fix is to give the
file type a name.
BTW, why are there several different taps displaying what appears to be
the same pieces of summary information? There's stuff in
gsm_map_summary.c, mtp3_summary.c, and summary_dlg.c.
I was wondering if it would have been possible to use the kazlib
exception library... I know systems have different behaviours for
exception generation. For example, some don't throw exceptions for
invalid access memory and some do.
All UN*Xes generate a signal for invalid memory accesses (SIGSEGV on
most UN*Xes; SIGBUS, for some reason, on BSDs). It's not an "exception"
in the sense of, say, a C++ exception.
I think Windows NT (NT 4.0, W2K, WXP, WServer2K3) generate an exception
for invalid memory accesses; I don't know what Windows OT (95, 98, Me) does.
Do you know if the KazLib provides a common behaviour for all systems?
The part of Kazlib in Ethereal doesn't do anything about UN*X signals or
Windows exceptions - it implements its own exception mechanism atop
setjmp() and longjmp().
