please take a look at this repo:

http://common-lisp.net/~alendvai/wui-corrupted-darcs-repo.tgz

i haven't done anything unusual, maybe some C-c's. the problematic patch has
been recorded by an 1-2 month old darcs.net darcs2 checkout.

the server side version is available at:

http://common-lisp.net/project/cl-dwim/darcs/wui/

some transcripts:

ati@ed101:~/workspace/wui$ darcs repair
darcs: failed to read patch:
Sat Oct 18 15:35:03 CEST 2008  norbert.madarasz@gmail.com
  * am chart examples
/home/ati/workspace/wui/_darcs/patches/20081018133503-520bc-6ce7944d5d32f5f903e93274d13546e4233eb9d0.gz:
openBinaryFile: does not exist (No such file or directory)
ati@ed101:~/workspace/wui$ 






ati@ed101:~/workspace/wui$ darcs unrec
Fri Oct 24 12:25:09 CEST 2008  attila.lendvai@gmail.com
  * housekeeping
Shall I unrecord this patch? (1/453)  [ynWvpxdaqjk], or ? for help: y
Thu Oct 23 18:05:18 CEST 2008  attila.lendvai@gmail.com
  * renamed to layer-context-capturing-component-mixin
Shall I unrecord this patch? (2/453)  [ynWvpxdaqjk], or ? for help: d
darcs: failed to read patch in get_extra:
Sat Oct 18 15:35:03 CEST 2008  norbert.madarasz@gmail.com
  * am chart examples
Sat Oct 18 15:35:03 CEST 2008  norbert.madarasz@gmail.com
  * am chart examples
/home/ati/workspace/wui/_darcs/patches/20081018133503-520bc-6ce7944d5d32f5f903e93274d13546e4233eb9d0.gz:
openBinaryFile: does not exist (No such file or directory)
Perhaps this is a 'partial' repository?
ati@ed101:~/workspace/wui$ 





ati@ed101:~/workspace/wui$ darcs pull
Pulling from "alendvai@common-lisp.net:/project/cl-dwim/darcs/wui"...
No remote changes to pull in!

Thanks for the tarball!

So here are some steps I've taken to probe this:

# generate a list of patch hashes
# there might be an easier way (maybe we should have a darcs show patch-hashes)
# and note that this is a bad bad way in theory to deal with XML
darcs changes --xml | grep "hash=" | sed -e "s/.* hash='//" | sed -e "s/'>//" > patch-hashes

# list all the patches we cannot find
for i in `cat patch-hashes`; do if [ ! -e _darcs/patches/$i ] ; then
echo "NOT FOUND: $i"; fi; done 
# NOT FOUND: 20081018133503-520bc-6ce7944d5d32f5f903e93274d13546e4233eb9d0.gz

I've also diffed the darcs changes output from the tarball and the
official repository to find out if there is a potential recovery
solution.  Unfortunately, the patches are in a different order, which
means that simply copying the patch file by hand will not do.  On the
other hand, Attila, it should be possible to reorder the good repository
so that you can indeed just copy the patch file in question.  I can help
you with that if you need.

Thorkil: where shall we go from here?

On Fri, Oct 24, 2008 at 11:05:58 -0000, Attila Lendvai wrote:
> /home/ati/workspace/wui/_darcs/patches/20081018133503-520bc-6ce7944d5d32f5f903e93274d13546e4233eb9d0.gz:
> openBinaryFile: does not exist (No such file or directory)
> Perhaps this is a 'partial' repository?

You mentioned a Ctrl-C... darcs people: maybe there is some chance the
Ctrl-C interrupted fetching a patch, but failed to make darcs quit?

thanks for the offer, but i've already pulled the pending patches into a fresh
checkout and i've copied the changed files, too. so i'm all good, just thought
that this tgz may help tracking down an ugly bug...

good luck!

So let's go over what we have:
- an old-fashioned repo with a single missing patch file
- possible control-C during the darcs get of the repo

Unless I'm mistaken there is no other corruption than that one file being
missing (and confusing the heck out of Darcs)

So it sounds like what we need somebody to do is to study more closely what
happens when you control-C during old-fashioned fetches.  Perhaps one way to do
it is to set up a sample repo with some particularly large patch files so that
we can reproduce the condition of interrupting in the middle of getting one of
these files and see what happens.