summaryrefslogtreecommitdiff
path: root/doc/bugs/uninit_does_not_work_in_old_repos.mdwn
diff options
context:
space:
mode:
Diffstat (limited to 'doc/bugs/uninit_does_not_work_in_old_repos.mdwn')
-rw-r--r--doc/bugs/uninit_does_not_work_in_old_repos.mdwn9
1 files changed, 9 insertions, 0 deletions
diff --git a/doc/bugs/uninit_does_not_work_in_old_repos.mdwn b/doc/bugs/uninit_does_not_work_in_old_repos.mdwn
index a6f785863..1e1b1fec7 100644
--- a/doc/bugs/uninit_does_not_work_in_old_repos.mdwn
+++ b/doc/bugs/uninit_does_not_work_in_old_repos.mdwn
@@ -4,3 +4,12 @@ As uninit does not need to actually write out any data, just remove it, it shoul
git-annex: Repository version 2 is not supported. Upgrade this repository: git-annex upgrade
If the repo happens to be broken, this essentially locks in data.
+
+> No, because you can always check out the version of git-annex you need
+> for that repository.
+>
+> uninit, as implemented, runs unannex on every file and then does some
+> cleanup. The cleanup does not need to write state, but the unannex does.
+> And it depends on the object directory layout, which has changed between
+> versions. So supporting old versions in this code would complicate it
+> quite a lot. I don't want to go there. --[[Joey]]