[darcs-users] darcs, lost in space

[Vigor van Ort]

On Fri, 18 Feb 2005 07:51:57 -0500, David Roundy
<droundy at abridgegame.org> wrote:
> On Wed, Feb 16, 2005 at 03:57:46PM +0100, Vigor van Ort wrote:
> > I had created the repo on my local system, and then yesterday I copied
> > it over to a server and used 'darcs get' to make a local copy through
> > ssh.  Since then I made one or two small patches and successfully did
> > a couple of 'darcs push' from my local copy.  So, I thought everything
> > was working fine, and I have been learning my way around darcs
> > command.
> >
> > Until today.  I just made another patch, and when I do 'darcs push',
> > it goes away and never comes back.  The darcs process has racked up 15
> > minutes of CPU time as of this writing on the remote server, and is
> > going strong.  The patch is less then 3k and touches only a couple
> > files.   I tried aborting the push and doing 'darcs check' and 'darcs
> > optimize' on the server, but both return immediately saying there is
> > nothing to do.
> >
> > Now, I don't want to shit on your program, but I have to wonder what
> > could possibly take so long with such a small patch when there is only
> > 1 developer and no conflicts!?
> 
> My guess would be that for some reason you do have conflicts (usually
> behavior like this is caused by large-scale conflicts, which are
> surprisingly easy to create).  (Or--less likely--as someone I think
> suggested, it could be a failed ssh connection.)
> 
> Did you use unrecord or amend-record at any point?

No.

> 
> Did you try the suggestion to do the push locally and see if it still
> hangs?

I just now created a second local branch by doing a get from the
server.  Then I pushed from the local development branch to the second
local branch and it returned success within a second or two.

> 
> Can we take a look at the two repositories to see what may be causing this?
> 
> What is the output of darcs whatsnew on the server? 

Interesting.  Even though all the files look like they are there,
whatsnew says that they are removed? ("R ...." output).  The second
local copy does not show this.

I used md5sum to find out what files are different, and the
differences are small:
> diff limp-md5.localcopy.sorted limp-md5.server.sorted
33a34
> 6b5da2c73bce18aa5abd43074f73c0c2  ./_darcs/patches/pending
55,56d55
< c6f738b6d9b43a13ae6970969f610d83  ./_darcs/prefs/defaultrepo
< c6f738b6d9b43a13ae6970969f610d83  ./_darcs/prefs/repos
84a84
> ec7b9f8be50bb0ebf367f35370ba435b  ./_darcs/prefs/author

So the server has a patches/pending file, which the local client does
not..   Even more interesting - the "pending" file shows lots of
deletions of whole files (and rmfile operations).  I wonder where that
came from - like I said, all the files seem to be accounted for!

I removed that pending file from the tree and then doing a push from
my local system took just a few seconds.  Thanks, that was helpful.

So, this was a large-scale conflict, which is why the server appeared
to be hung?  Is there a way this could be made more friendly?  Because
I expected 0 conflicts, I had no idea about this list of pending
deletes.  Is there a way, sort of like the --dry-run, to find out how
many conflicts are involved, to help figure out what's going on when
darcs eats up CPU time?

thanks!
V.