Greg Stein wrote: > Woah. This is even crazier than what you fixed. *nobody* asks > for the end revision. In fact, the code would have crashed since > END_REVISION was always NULL.
No, it wasn't that broken, it was only a bit of an ugly API. Both the END_URL and END_REVISION output pointers were (and still are) allowed to be null if the corresponding END input revision is unspecified. > Thus, ->kind must have never been > unspecified. But no matter, as the param is useless. Also note > that only one callsite asks for the start, and only one asks for > the URL. > Seems you could simplify further... Am doing. - Julian > Julian Foad wrote: > > Simplify calling svn_client__repos_locations() by returning > > svn_revnum_t instead of svn_opt_revision_t and by allowing > > NULL for unwanted revision number outputs.
