upgrade: Handle upgrade to v6 when the repository already contains v6 unlocked files...
authorJoey Hess <joeyh@joeyh.name>
Mon, 17 Oct 2016 19:19:47 +0000 (15:19 -0400)
committerJoey Hess <joeyh@joeyh.name>
Mon, 17 Oct 2016 19:19:47 +0000 (15:19 -0400)
commit8e22114735344b96b3cb558c1fee5537783bd7c6
tree654c283789aad0f3482303358f4846244f5c0400
parent148bd0dbfde142b79411967926b1400af5590175
upgrade: Handle upgrade to v6 when the repository already contains v6 unlocked files whose content is already present.

Closes https://github.com/datalad/datalad/issues/1020

The use of runWriter in scanUnlockedFiles broke due to this change;
it failed with blocked indefinitely in mvar, because the database write
handle was taken while linkFromAnnex needed to also write to it (to update
the inode cache). So, switched to using a separate runWriter for each call
to addAssociatedFileFast. A little less efficient, but not greatly; the
writes should all still be cached.
Annex/WorkTree.hs
CHANGELOG
Command/Unlock.hs
doc/bugs/remote_repository_must_be_version_6_as_well_to_have_copied_files_appear_correctly.mdwn