On Thu, Nov 10, 2016 at 12:41:39AM +0900, Norbert Preining wrote: > > But that'd be just bandaid IMO and a real fix would either revert the > > relevant Unicode::Collate change if it's a bug, or adapt biber to work > > If one could find out what the relevant change is. It is rather unfortunate > that the two version of the same things don't produce the same :-(
% perl -MUnicode::Collate -le 'print $Unicode::Collate::VERSION, " ", Unicode::Collate->new->viewSortKey("Foo")' gives this output with the versions available on jessie and sid: 1.04 [1680 176D 176D | 0020 0020 0020 | 0008 0002 0002 | FFFF FFFF FFFF |] 1.07 [1680 176D 176D | 0020 0020 0020 | 0008 0002 0002 | FFFF FFFF FFFF |] 1.14 [19A9 1AA1 1AA1 | 0020 0020 0020 | 0008 0002 0002 | FFFF FFFF FFFF |] 1.17 [1C60 1D58 1D58 | 0020 0020 0020 | 0008 0002 0002 | FFFF FFFF FFFF |] So the U::C sort key generation has changed (again), and the biber test suite hardcodes md5sums of the old sort keys. Given https://github.com/plk/biber/commit/9bebd3ef0cb57b5ca4f15885d0cbfecc4939a835 I suspect biber just needs to adapt again. -- Niko Tyni nt...@debian.org