Ben Collins-Sussman <sussman@collab.net> writes:
> On Mar 19, 2005, at 3:23 AM, cmpilato@tigris.org wrote:
> >
> > * subversion/include/svn_fs.h
> > (svn_fs_lock): Update docstring to warn callers that the lock they
> > send in might not be the exact lock data used (because of the path
> > canonicalization stuffs).
>
> Why is this necessary to document? Doesn't libsvn_fs
> auto-canonicalize *every* incoming path already?
>
> Ah... perhaps it's because this is the only API in which the caller
> passes in a structure as input, one that might be reused. Hm.
>
> I wonder if svn_fs_attach_lock() should canonicalize the path within
> the incoming structure...? Wouldn't that be more elegant than making
> the caller re-fetch the svn_lock_t?
svn_fs_attach_lock () can't canonicalize the structure's 'path'.
Philosophically, that's just a not-nice thing to do, but technically,
the caller would have no clue that some item in his structure now has
a piece in it with a potentially different pool lifetime than the rest
of his structure.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Mar 19 16:19:48 2005