Hi Ralf, Eric, Sorry for the late reply.
On Fri, Oct 10, 2008 at 09:35:13PM +0200, Ralf Wildenhues wrote: > [ in case you're not subscribed: ] > <http://thread.gmane.org/gmane.comp.sysutils.autoconf.bugs/6319/focus=6320> I'm not subscribed. > Hi Eric, Thomas, > > * Eric Blake wrote on Fri, Oct 10, 2008 at 09:20:47PM CEST: > > Hmmm. This is potentially buggy, anyway. It is unsafe to call cd > > "$srcdir" if > > $srcdir is a relative path, does not contain /, and CDPATH is set in the > > environment (POSIX requires that it output the destination path as > > discovered > > by the CDPATH lookup). Of course, configure has already tried to unset > > CDPATH, > > but what if the user declared it to be a read-only variable? > > I think we've at one time declared the user deserving the consequences > if she sets CDPATH to be read-only. There is lots more code in > autotools that can break in this case, and not much you can do about it. > > I'm not sure whether letting configure barf over read-only CDPATH > (if it can portably be tested, that is) would be a good idea. > As long as CDPATH contains only paths that don't deal with this code, > users could rightfully(?) complain about that. Maybe a big fat warning > instead? > > Thomas, is read-only CDPATH the issue here BTW? No idea. My .bashrc contains "export CDPATH=/some/path:/or:/another" but I've not marked it readonly. On the other hand, I've just unpacked autoconf-2.63 in /tmp and rerun the testsuite, and this time it worked without any problems. I'll report if I stumble over this or another error. Cheers, Thomas
