> I dunno, but I hope one programming error doesn't mean every pointer
> dereference should be checked so a nicely formatted error can be
> displayed before crashing.

That's not necessarily the action that would be taken, though.  There
are cases where a NULL pointer might be acceptable (this is already the
case in FreeUrl, for example--but it only validates request->url and
request->urlv, rather than validating request itself).  In those cases
it's worth checking.

In fact, the patch I submitted for this problem doesn't just display a
nicely formatted error before crashing; it attempts to repair the pointer
in a sane way.  Sometimes that's a possibility as well.

> This isn't any better than a segv, IMO.

IMO, an error in the server log is much easier to track down in the code
and troubleshoot than a SEGV.

- John

Reply via email to