I'm unhappy about twalk_r

[DJ Delorie]

Florian Weimer <fweimer@redhat.com> writes:
> Exposing this information using the VISIT argument seems wrong to me.

I agree.  It would be more useful if the binary-ness of the tree were
*more* exposed, so the user could take advatage of the walk-order, but
we don't expose it.

> I think this is much more useful: It avoids pointless repeated calls to
> ACTION, and it allows premature termination of the iteration.

twalk() is defined by POSIX.  If we have a *_r version of it, it should
conform to the POSIX definition as closely as we can, despite these
inefficiencies.

Perhaps we could add a twalk_find_r() function that does something
between twalk and tfind?  Or twalk_sorted_r()?  Or something else not
called twalk_r()?

Of course, that assumes that we have a need for a walk-no-find function
that isn't already solved by either twalk or tfind.