Jeff King <peff@xxxxxxxx> writes: > Hrm. That only works in the current code because we apply the deletion > in the directory (and then clean up the now-empty directory) first. So I > think you would need to check the paths progressively as you apply them, > since those other parts of the diff "haven't happened yet". Just to make sure that I am not hallucinating, I added this one: diff --git a/t/t4114-apply-typechange.sh b/t/t4114-apply-typechange.sh index ebadbc3..83ddf62 100755 --- a/t/t4114-apply-typechange.sh +++ b/t/t4114-apply-typechange.sh @@ -119,4 +119,12 @@ test_expect_success 'directory becomes symlink' ' test_debug 'cat patch' +test_expect_success 'directory becomes symlink' ' + git checkout -f foo-becomes-a-directory && + printf "%s\n" foo/baz foo >order && + git diff-tree -Oorder -p HEAD foo-symlinked-to-bar >patch && + git apply --index <patch + ' +test_debug 'cat patch' + test_done It is a copy of the original, only forcing the patches in the input in the opposite order. Having said that and also having read your two-phase internal application change, I think that two-phase thing is probably a good way to go (we may even want to ignore "previous_patch()" stuff, as its "was_deleted()" and "tobe_deleted()" are all about "force the application of a later patch to depend on the result of application of an earlier patch"). -- 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