Karl Fogel <kfogel@newton.ch.collab.net> writes:
> Philip Martin <philip@codematters.co.uk> writes:
> > The problems I was referring to are that the Subversion libraries
> > internally use both "." and "" to refer to the current directory, the
> > svn_path_xxx functions do not always do the right thing when given a
> > "" or "." argument, and some APR functions don't work with "".
> >
> > Some people have expressed the opinion that path canonicalization
> > should be handled by APR. I'm not sure where that leaves the
> > svn_path_xxx functions, since it doesn't make sense to try to split,
> > join and manipulate paths in Subversion if the canonical form is
> > determined by APR.
>
> It doesn't seem like any APR developer is taking this on, and it seems
> reasonable for Subversion to do what it can in the meantime.
>
> Since APR doesn't always work with "", would our potential issue
> description be something like this?:
>
> Make Subversion libraries internally use "." to refer to the
> current directory, and make the svn_path_xxx functions assume (and
> canonicalize) to "." as the current directory. Tweak the command
> line client and notification system as necessary to handle this.
That's about it. There are other things that could go with this in a
"path library consistency" issue, or perhaps they should be separate
issues
svn_path_split_nts("foo") returns NULL and "foo" and so there
are svn_path_is_empty_nts checks all over the code
svn_path_is_single_path_component("..") treats ".." as special
svn_path_is_child(".", "..") doesn't treat ".." as special
svn_path_is_child() handles NULL arguments
svn_path_compare_paths_nts() doesn't handle NULL arguments
what should svn_path_get_longest_ancestor(".", "foo") return?
--
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Aug 20 23:41:50 2002