BTW this is just a request for comments, not a finished patch. (mergeinfo_tests.py 2 fails and I have not yet investigated.)
- Julian I (Julian Foad) wrote: > Hi Paul. > > I think we can tighten the validation of svn_merge_range_t to exclude > change number "r0" (RANGE->start == -1) as in the following patch. > > My reasoning is that a change numbered "r0" is not a valid concept in > any Subversion system because the state (tree-snapshot) numbered r0 is > by definition the beginning. (It also happens to be empty by > definition, but that's not so relevant.) We can say the same in a > different way: change "r0" would mean the change from "r(-1)" to "r0", > and "r(-1)" is not a valid concept. > > Makes sense? > > [[[ > Tighten merge-range validation to not allow "change number r0" aka "revision > range -1:Y". > > * subversion/libsvn_subr/mergeinfo.c > (IS_VALID_FORWARD_RANGE): New macro. > (get_type_of_intersection): Use IS_VALID_FORWARD_RANGE() for tighter > validation of arguments than before: it previously accepted "change 0". > (range_intersect, range_contains): Validate arguments. Add doc strings. > > * subversion/tests/libsvn_subr/mergeinfo-test.c > (randomly_fill_rev_array, rev_array_to_rangelist): Expand doc strings. > (test_rangelist_remove_randomly, test_rangelist_intersect_randomly): Don't > ever include change number r0 in a merge range. > --This line, and those below, will be ignored-- [...]
