[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: svn commit: r10133 - in trunk/subversion: include libsvn_subr tests/clients/cmdline tests/libsvn_subr

From: Josh Pieper <jjp_at_pobox.com>
Date: 2004-07-04 22:44:40 CEST

Ben Reser wrote:
> On Sun, Jul 04, 2004 at 01:26:01PM -0400, Josh Pieper wrote:
> > file:/// URLs would have to be special cased then to leave the ".."
> > path elements in since a portion is directly translated to a
> > filesystem path.
>
> True.
>
> > That seems like an unnecessary restriction, especially for file:///
> > URLs.
>
> I wasn't too fond of it. But obviously we have to either make /../
> behave the way people expect or not at all. I prefer making them work.

Well, I think we have to allow it for filesystem paths, which would be
any relative path and the root portion of the file:/// URL. Since we
don't know what portion of the file:/// URL is the root at
canonicalization time, I don't think it feasible to treat the root
differently from the rest of the path.

> > Are there other cases where this problem can arise besides a
> > repos->repos copy? If not, then maybe we could just not collapse the
> > ".." anywhere and give better diagnostics there.
>
> Not sure. I haven't done extensive testing looking for problems...

Therefore, I think this above suggestion is the best alternative. We
would treat all URLs and relative paths the same with regard to
"/../", namely that they would not be collapsed. Then our job is just
to make "svn copy" and "svn move" return sensible errors when those
elements are present in the wrong place.

-Josh

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sun Jul 4 22:45:55 2004

This is an archived mail posted to the Subversion Dev mailing list.