### Please describe the problem.
Not able to successfully git-annex sync with a remote due to a git fatal. Caused by masters diverging?

### What steps will reproduce the problem?
git-annex sync, or, letting the assistant try.

### What version of git-annex are you using? On what operating system?
git-annex version: 5.20131221+b1 on my laptop
git-annex version: 5.20131224-g6ca5271 on the remote server

### Please provide any additional information below.

Output of a manual git-annex sync in the directory:

[[!format sh """
greg@x200s:~/Documents$ git-annex sync
commit  (Recording state in git...)
Copyright Office/Orphan Works/ARROW/170409_ARROW_Leaflet.pdf: unmerged (783afced6bc43138373fda43edfda0c33be36525)
Copyright Office/Orphan Works/ARROW/ARROWproject_results1.pdf: unmerged (b536e5f3d93e7905e05510f26db1f743e9eae16e)
Copyright Office/Orphan Works/ARROW/ARROWproject_results1.ppt: unmerged (5543049b8940cc5702d37aff18b03c67d9c8374d)
Copyright Office/Orphan Works/ARROW/ARROWstandardPresent2010.pdf: unmerged (54d751bc98cb5da29d3d568856b74675e842072e)
Copyright Office/Orphan Works/ARROW/ARROWstandardPresent2010.ppt: unmerged (efe0e94b51eccb9a6a0c352f4a210bd5a6105050)
Copyright Office/Orphan Works/ARROW/ARROWtrifoldMAR2011.pdf: unmerged (b52ff16178e29261fe00a518c23610a3b0826482)
Copyright Office/Orphan Works/Documentation/20110531/Documentation.doc.odt: unmerged (1348d5f42f7e34706407f7936f4fb0438e4b8ffa)
Copyright Office/Orphan Works/Documentation/AAPpublishers.pdf: unmerged (3f448a03d31a38adb095e3031e4ee13771d22d70)
Copyright Office/Orphan Works/Documentation/Documentation.doc: unmerged (265fdff7787f560e3ba20789a12e15ffb165ec7f)
Copyright Office/Orphan Works/Documentation/Documentation.pdf: unmerged (7a9ff92663ed42b42b9baaefaf4721499d18d82d)
...
fatal: git-write-tree: error building trees
git-annex: failed to read sha from git write-tree
"""]]

See also:

1. the [partial daemon log](http://paste.debian.net/73176/) from the assistant running in that directory on the laptop and 
2. the output of [git fsck](http://paste.debian.net/73175/) on the remote.

git-annex repair on the laptop and the server:
[[!format sh """
greg@x200s:~/Documents$ git-annex repair
Running git fsck ...
No problems found.
ok
"""]]


### How I ended up fixing it:
[[!format sh """
greg@x200s:~/Documents$ killall git-annex
greg@x200s:~/Documents$ git-annex indirect
blah...............
indirect  ok
ok
greg@x200s:~/Documents$ git status
On branch master
Your branch and 'rose/master' have diverged,
and have 294 and 1 different commit each, respectively.
  (use "git pull" to merge the remote branch into yours)

Untracked files:
  (use "git add <file>..." to include in what will be committed)

	.gitrefs/

nothing added to commit but untracked files present (use "git add" to track)
greg@x200s:~/Documents$ git pull
Merge made by the 'recursive' strategy.
 Copyright Office/Orphan Works/staging/reporting/process_report.txt.2 | 1 +
 Copyright Office/Orphan Works/staging/reporting/with-title.xls       | 1 +
 Copyright Office/Orphan Works/staging/with-title.xls                 | 1 +
 Copyright Office/Orphan Works/worker_emails.txt                      | 1 +
 git.fsck.log                                                         | 1 +
 5 files changed, 5 insertions(+)
 create mode 120000 Copyright Office/Orphan Works/staging/reporting/process_report.txt.2
 create mode 120000 Copyright Office/Orphan Works/staging/reporting/with-title.xls
 create mode 120000 Copyright Office/Orphan Works/staging/with-title.xls
 create mode 120000 Copyright Office/Orphan Works/worker_emails.txt
 create mode 120000 git.fsck.log
greg@x200s:~/Documents$ git-annex sync
commit  ok
pull rose 

Already up-to-date.
ok
push rose 
Counting objects: 1658, done.
Delta compression using up to 2 threads.
Compressing objects: 100% (904/904), done.
Writing objects: 100% (1604/1604), 138.97 KiB | 0 bytes/s, done.
Total 1604 (delta 892), reused 1298 (delta 688)
To greg@rose.makesad.us:/home/greg/Documents/
   f1d206e..e836b9b  master -> synced/master
ok
greg@x200s:~/Documents$
"""]]

I restarted the assistant and the daemon.log looks good.

After sync'ing on the server, it appears that this has been the case for quite some time (based off of what symlinks were created).

Lastly: Joey, this is probably what caused that weird behavior in the webapp where it showed the bad transfer each day after the fsck at noon. I never diagnosed that more but I bet I won't see it tomorrow.

[[!tag moreinfo]]