Re: svn commit: rev 2821 - trunk/subversion/libsvn_wc trunk/subversion/clients/cmdline trunk/subversion/tests/clients/cmdline

Karl Fogel <kfogel@newton.ch.collab.net> writes:

> Justin Erenkrantz <jerenkrantz@apache.org> writes:
> > The problem I saw when I looked at this before is that even though
> > canonical paths often entered libsvn_wc, the canonicalization was
> > lost at various stages *internally* in libsvn_wc.
>
> That's okay, as long as libsvn_wc knows what it's doing.
>
> That what I was essentially asking Philip in my original mail: are
> they noncanonical coming in (bad), or decanonicalized by choice
> internally (fine, as long as the relevant code is adjusted correctly).

The last problem I saw was when running 'svn ci .'

svn_path_condense_targets calls svn_path_get_absolute passing "." and
gets back a path with a trailing slash, something like "/home/pm/wc/".
Now if the command is 'svn ci /home/pm/wc/', or 'svn ci /home/pm/wc',
then svn_path_get_absolute returns "/home/pm/wc" without a slash.

So should a working copy directory path include a trailing slash?

The locking code uses the path as a hash key. So if the trailing
slash is optional the code needs to ensure that setting and getting
will work both with and without the trailing slash.

-- 
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org