aboutsummaryrefslogtreecommitdiff
path: root/doc/bugs/git-annex_confuses_Git_with_nested_submodules/comment_3_e1bc8eb7f6ce0d6f2d2f2b0ea6f20862._comment
blob: 49e883c7ca0a65e886af4e7f7020ae8827ad7daa (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
[[!comment format=mdwn
 username="joey"
 subject="""comment 3"""
 date="2016-03-02T16:48:24Z"
 content="""
[git bug report](http://news.gmane.org/find-root.php?message_id=20160301204218.GA4083%40kitenet.net)

So far, the git devs admit this is a problem, but don't seem too keen on fixing
it, even though it breaks backwards compatability with repositories git
submodule add created (circa 2012).

It might be that git-annex init could work around git's bugginess by,
instead of making submodule/.git a symlink to ../.git/modules/dir, making
submodule/.git be the git directory, and converting ../.git/modules/dir
to a symlink. In very limited testing, that setup seems to work. 

I don't know if all the submodule stuff would work, perhaps it would break moving
submodules etc. And, since git likes to chdir around (not the best idea),
if it expected to be able to chdir from .git/modules to dir and chdir .. to
get back, changing that to a symlink would defeat it.

BTW, I found another way, unrelated to git-annex or symlinks at all,
that git submodule add's broken path handling makes it fall over with
nested submodules.
<http://news.gmane.org/find-root.php?message_id=20160302165240.GA17654%40kitenet.net>.

(It's almost like myrepos was a better idea than this submodule stuff, or
something...)
""]]