On 11. juni 2010, at 21.44, Junio C Hamano wrote: > Johannes Sixt <j.sixt@xxxxxxxxxxxxx> writes: > >> I know what your point is. It is still inappropriate to call >> normalize_file() on data that comes from the repository. It is not the >> task of a merge procedure to blindly normalize data. > > It is not "blindly", but "running normalization one _extra time_, as the > repository data is supposed to be canonical already", which is utterly > wrong. I agree that double normalization is evil, but the repository data isn't necessarily canonical if the configuration has changed since the data was added. How do you feel about Finn Arne's idea of first convert_to_work_tree()-ing the data, then convert_to_git()ing it back? That gets rid of the double normalization at the cost of some performance and memory usage (especially with CRLF output enabled). I'm going to do some benchmarks to go along with my next stab at this. - Eyvind -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html