summaryrefslogtreecommitdiff
path: root/doc
diff options
context:
space:
mode:
authorGravatar Joey Hess <joeyh@joeyh.name>2015-06-09 15:59:03 -0400
committerGravatar Joey Hess <joeyh@joeyh.name>2015-06-09 15:59:03 -0400
commit72983d333193a2236bbe0b8d23235933058e938b (patch)
tree6b22a381c7add604afb7e32fcf5f43a9345653dc /doc
parent2206e698051db2f435b249a4988a6550be09a007 (diff)
devblog
Diffstat (limited to 'doc')
-rw-r--r--doc/devblog/day_292__dead_keys.mdwn21
1 files changed, 21 insertions, 0 deletions
diff --git a/doc/devblog/day_292__dead_keys.mdwn b/doc/devblog/day_292__dead_keys.mdwn
new file mode 100644
index 000000000..3ad66edb5
--- /dev/null
+++ b/doc/devblog/day_292__dead_keys.mdwn
@@ -0,0 +1,21 @@
+Ever since `git annex fsck --all` was added, people
+[have](http://bugs.debian.org/753888)
+[[complained|bugs/fsck_reports_unsolvable_problem]] that
+there's no way to stop it complaining about keys whose content is gone for
+good. Well, there is now: `git annex dead --key` can be used when you know
+that a key is no longer available and want fsck to stop complaining about
+it.
+
+Running fsck on a directory will intentionally still complain about files
+in the directory with missing contents, even if the keys have been marked
+dead.
+
+The crucial part was finding a good way to store the information; luckily
+location log files are parsed in a way that lets it be added there without
+breaking backwards compatability. A bonus is that adding a key's content
+back to the annex will automatically bring it back from the dead.
+
+I'm pondering making `git annex drop --force` automatically mark a key as
+dead when the last copy is dropped, but I don't know if it's too DWIM or
+worth the complication. Another approach would be to let fsck mark keys as
+dead, but that would certianly need an extra flag.