summaryrefslogtreecommitdiff
path: root/doc/bugs/Crash_trying_to_sync_with_a_repo_over_ssh.mdwn
blob: 38f54d2b6140579454a29b3cbfc68ee3e8b5a1fd (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
30
31
32
33
34
35
36
37
38
39
40
41
42
43
What steps will reproduce the problem?

I create a new annex, added in a bunch of files.

I cloned this annex to another machine, where I already had those files, so I copied them into a directory named "foo", did "git annex add foo", and then did "git rm -r foo", and git commit'd my clone of the annex.

Then I try to "git annex sync" with the remote.

What is the expected output? What do you see instead?

I don't know, I've never used git-annex before.  This is what I get each time:

    Hermes ~/Products/tmp/Movies (master) $ ga sync
    git-annex-shell: Prelude.(!!): index too large

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

It's the 'master' as of yesterday: c504f4025fec49e62601fbd4a3cd8f1270c7d221

I'm on OS X 10.8.2, using GHC 7.6.1.  The annex in question has 38G in a few hundred files.

Please provide any additional information below.

I'm willing to help track this down!

> I've got it, October 9th's release 
> included commit bc649a35bacbecef93e378b1497f6a05b30bf452, which included a
> change to a `segment` function. It was supposed to be a
> rewrite in terms of a more general version, but it introduced a bug
> in what it returned in an edge case and this in turn led git-annex-shell's
> parameter parser to fail in a code path that was never reachable before.
> 
> It'd fail both when a new repo was running `git-annex-shell configlist`,
> and in `git-annex-shell commit`, although this latter crash was less
> noticible and I'm sure you saw the former.
> 
> Fixed the reversion; fixed insufficient guards around the partial code
> (which I cannot see a way to entirely eliminate sadly; look at
> GitAnnexShell.hs's `partitionParams` and weep or let me know if you have 
> any smart ideas..); added a regression test to check the non-obvious
> behavior of segment with an empty segment. I'll be releasing a new
> version with this fix as soon as I have bandwidth, ie tomorrow.
> [[done]] --[[Joey]]