What steps will reproduce the problem?
What is the expected output? What do you see instead?

I can reproduce it locally, but don't know what's causing it. The file content is the same, I checked with md5sum and sha512sum. But fsck still thinks the content is different. Are there other factors I could check which fsck looks at? I'm using SHA512E backend.

What version of git-annex are you using? On what operating system?

git-annex version: 3.20120807 Ubuntu 12.04 updated on Aug 20th annex was installed via cabal on Aug 20th, all other packages are from ubuntu.

> What is the error message from fsck? --[[Joey]] 

This is the output:

> reinject ....Moon.avi (checksum...) 
> Bad file content; moved to /mnt/.../.git/annex/bad/SHA512E-s94402560--ead9db1f34739014a216239d9624bce74d92fe723de06505f9b94cb4c063142ba42b04546f11d3d33869b736e40ded2ff779cb32b26aa10482f09407df0f3c8d.Moon.avi
failed
> (Recording state in git...)
> git-annex: reinject: 1 failed

The original file also has sha512 ead9db1f34739014a216239d9624bce74d92fe723de06505f9b94cb4c063142ba42b04546f11d3d33869b736e40ded2ff779cb32b26aa10482f09407df0f3c8d

>> And what sha512 does the file in .git/annex/bad have **now**? (fsck
>> preserves the original filename; this says nothing about what the
>> current checksum is, if the file has been corrupted). --[[Joey]]

The same, as it's the file I was trying to inject:

ead9db1f34739014a216239d9624bce74d92fe723de06505f9b94cb4c063142ba42b04546f11d3d33869b736e40ded2ff779cb32b26aa10482f09407df0f3c8d  .git/annex/bad/SHA512E-s94402560--ead9db1f34739014a216239d9624bce74d92fe723de06505f9b94cb4c063142ba42b04546f11d3d33869b736e40ded2ff779cb32b26aa10482f09407df0f3c8d.Moon.avi

That's what puzzles me, it is the same file, but for some weird reason git annex thinks it's not.

> Ok, reproduced and fixed the bug. The "E" backends recently got support
> for 2 levels of filename extensions, but were not made to drop them both
> when fscking. [[done]] (I'll release a fixed version probably tomorrow;
> fix is in git now.) --[[Joey]]