diff options
Diffstat (limited to 'doc/bugs/Crash_trying_to_sync_with_a_repo_over_ssh.mdwn')
-rw-r--r-- | doc/bugs/Crash_trying_to_sync_with_a_repo_over_ssh.mdwn | 19 |
1 files changed, 19 insertions, 0 deletions
diff --git a/doc/bugs/Crash_trying_to_sync_with_a_repo_over_ssh.mdwn b/doc/bugs/Crash_trying_to_sync_with_a_repo_over_ssh.mdwn index 1df26fc81..38f54d2b6 100644 --- a/doc/bugs/Crash_trying_to_sync_with_a_repo_over_ssh.mdwn +++ b/doc/bugs/Crash_trying_to_sync_with_a_repo_over_ssh.mdwn @@ -22,3 +22,22 @@ I'm on OS X 10.8.2, using GHC 7.6.1. The annex in question has 38G in a few hun 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]] |