On Fri, 2006-05-26 at 10:55 +0100, James Youngman wrote: > On 5/24/06, Bruno Haible <[EMAIL PROTECTED]> wrote: > > > Then the argument about the glibc behaviour and the *BSD libc behaviour is > > moot: if there's a single platform where it crashes, and if the standard > > says > > that NULL are invalid arguments, gnulib's implementation can crash as well. > > I would go further and say that it _should_ crash. Otherwise, bugs > just hang around the place and don't get fixed.
As Bruno noted, the reporter last year was trying to get debug output. He was bound to be getting null values. I am doing the same thing; testing for nulls is important, but sometimes I need to "see" them. When I get a bug report about something not working, like this: http://www.dbmail.org/mantis/view.php?id=352 It took me < 15 minutes to track down a way upstream function producing null pointers, entirely without a debugger. Crashing on null %s args puts gnulib at odds with all other open source lib's -- glibc, *BSD libc's, so on. People use gnulib to get the gnuish behavior on other platforms. Being safe about nulls is a gnuish behavior that many people appreciate. Thanks, Aaron
