On Thu, Jun 25, 2009 at 04:13:12PM +0100, Philip Martin wrote:
> HuiHuang <yellow.flying_at_yahoo.com.cn> writes:
>
> > +/** Given @a *path, find its strictly working copy root path @a **wc_root_path
> > + Allocate the result in @a result_pool.
> > + Use @a scratch_pool for temporary allocations.
> > + The error @c SVN_ERR_WC_NOT_DIRECTORY will be returned if
> > + * @a path is not a versioned directory.*/
> > +svn_error_t *
> > +svn_wc__find_wc_root(const char **wc_root_path,
> > + const char *path,
> > + apr_pool_t *result_pool,
> > + apr_pool_t *scratch_pool);
> > +
>
> I haven't really been following the discussion about why you are
> writing this particular function, but the experience from the pre-1.0
> days is that interfaces based around paths don't work very well. The
> problem is that one needs to read the WC metadata before the paths can
> be manipulated and that data then gets thrown away, this makes the
> function expensive and slow.
>
> In the longer term, assuming WC-NG ends up with a single database in
> the root, it's hard to see why one would want this interface at all.
We've been asking WC-NG folks about how to do this in the
new world. They don't know yet. So for now, we're going for a
wc-1 style approach (paths and adm_access), and it will be changed
later.
Thanks for your comments, you've found quite a lot of things
I've overlooked!
Stefan
Received on 2009-06-25 17:29:34 CEST