Failure during remote upgrade

John Arbash Meinel john at
Wed Sep 30 15:38:04 BST 2009

Hash: SHA1


>>>>> Is it safe to run the upgrade again?
>>>> You'll need to rename backup.bzr to .bzr, but otherwise, yes.
>>> I guess I need to do that through sftp, right?
>> Nope, this failed.
> It failed because you're not renaming backup.bzr to .bzr.

I'm pretty sure he can't 'rename backup.bzr to .bzr' because the '.bzr'
directory still exists.

> Codehosting only permits a few top-level filenames: ".bzr", "backup.bzr"
> and ".bzr.backup" (obsolete).  This is because we want to host Bazaar
> branches and nothing else.
> By renaming backup.bzr to backup.bzr.failed, you are trying to create a
> forbidden filename.  The same is true when you try to rename .bzr to
> .bzr.failed.
>> LOSA time?
> No, not needed.
>> However, it's odd that there's a .bzr directory and a backup.bzr
>> directory.  I don't know how to crawl out of this hole. :(
> There is a backup directory in case something went wrong with the
> conversion.  It is retained even after the conversion is complete, as a
> paranoid measure,  in case the conversion was somehow faulty.
> Aaron

I would think you could rm -rf backup.bzr and then try the upgrade
again, rather than renaming backup.bzr => .bzr. However, the original
failure was (I think):
Read from remote host Network is unreachable

So I'm not sure the specific state of the remote object. IIRC we might
rename .bzr/repository .bzr/repository.backup, but otherwise things
would be in a stable state.

I can't really check from here, because that is on the hosted side, and
I don't have write access to the mailman branches. But I can try to help
you either here or on IRC.

Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla -


More information about the bazaar mailing list