24 lines
1.6 KiB
Markdown
24 lines
1.6 KiB
Markdown
I recently started experimenting with git annex, adding files that I've had
|
||
floating across several computers to repositories. During the testing I had
|
||
a few occasions where I wrecked a repository somehow, and decided to wipe it
|
||
and start anew (at this point there was no important files in them so I thought
|
||
this is the easiest way). Well, as it turns out this interacts badly with location
|
||
tracking, since now `git annex whereis` shows files residing in all those destroyed
|
||
repositories, all having same names as some existing repositories. This makes it hard
|
||
to follow whether a repo actually has a file, or was the file only seen in some dead
|
||
repo with the same name.
|
||
|
||
I planned on cleaning this up by looking up the UUIDs of the now stable, existing
|
||
repos and untrusting all the dead copies (they should effectively disappear from
|
||
git annex´s output then, right?), but I didn't find an easy way to look up the UUID
|
||
of the current repository (maybe this could be included in `git annex status`?)
|
||
I also noticed that untrust cannot remove the trust based on the UUID -- if I try
|
||
it I simply get "there is no git remote named "11908472-...", so I guess untrust
|
||
works with git remote names, which I find a bit confusing, since trust.log logs the
|
||
trust levels based on the UUID. I could just write into trust.log manually, but I'm
|
||
unsure how the changes would get propagated.
|
||
|
||
What should I do? As a related wishlist item I would ask for some additional mechanisms
|
||
for purging known-dead repositories from the location tracking database. And the ability
|
||
to look up the UUID of the current repo, and to use the UUID to specify repositories when
|
||
applicable (untrust and describe maybe).
|