Merge branch 'master' of ssh://git-annex.branchable.com

1 files changed, 10 insertions, 0 deletions

diff --git a/doc/bugs/can__39__t_annex_get_from_annex_in_direct_mode/comment_2_f26e0f763f9027d9dfc08cd840ced153._comment b/doc/bugs/can__39__t_annex_get_from_annex_in_direct_mode/comment_2_f26e0f763f9027d9dfc08cd840ced153._comment
new file mode 100644
index 000000000..913f5304d
--- /dev/null
+++ b/doc/bugs/can__39__t_annex_get_from_annex_in_direct_mode/comment_2_f26e0f763f9027d9dfc08cd840ced153._comment
@@ -0,0 +1,10 @@
+[[!comment format=mdwn
+ username="http://joeyh.name/"
+ ip="4.153.254.222"
+ subject="comment 2"
+ date="2013-07-07T17:43:49Z"
+ content="""
+You generally cannot use a git-annex repository that is in direct mode as a remote over http. A remote git-annex does not have sufficient information to safely use a direct mode repository in that way. I don't think I can fix that. The http transport will work with indirect mode repositories (not supported on Windows), and with bare repositories (should work ok on Windows).
+
+I'm perplexed by what you show happening in the comment. It appears that the content of the files has been staged into the git repository as the symlink target on Windows. I have never seen that happen, cannot imagine how git-annex could do that. My best guess is you might have run `git commit -a` after `git annex add`, and on Windows, since it doesn't really have symlinks, that could leave the symlink bit set while staging the full content of the file. You should never run `git commit -a` in a direct mode repository (Windows always uses direct mode). See [[/direct_mode]] for caveats about git commands that are unsafe to run in direct mode.
+"""]]