On Wed, 2010-06-16 at 14:30 +0100, Philip Martin wrote: > Daniel Shahaf <d...@daniel.shahaf.name> writes: > > > Philip Martin wrote on Wed, 16 Jun 2010 at 11:17 -0000: > >> How are you going to check adds? During an add the old value is NULL > >> but must still be checked. > >> > > > > During an add, old_value_p!=NULL and *old_value_p==NULL, and the code > > checks that present_value==NULL as well. > > > > The case old_value_p==NULL means "unspecified" (i.e., just make the > > change regardless of what's there now). > > > > OK, I didn't read the code carefully enough.
I didn't spot that either. Please could you add a second "const" in the function prototype to make it clearer? Otherwise it looks like an output. "const svn_string_t *const *old_p" - Julian > >> I don't think the old interface should be deprecated. > >> > > > > Why? The new interface has all the functionality of the old interface; > > one can do > > > > s/svn_fs_change_rev_prop(*args)/svn_fs_change_rev_prop2(args, > > old_value_p=NULL)/g > > > > with no change in functionality. > > Agreed. >
