changed index parents?

[Robert Collins]

So Aaron's not at fault here....

On Tue, 2007-09-18 at 14:28 +1000, Robert Collins wrote:
> On Tue, 2007-09-18 at 10:40 +1000, Robert Collins wrote:
> > Hi Aaron, I just saw a bunch of changed index parents in bzr.dev - I'm
> > guessing that you had run your improved reconcile on your repository, so
> > merging your reconfigure patch changed bzr.dev.
> > 
> > Could you confirm this, so I don't have to keep looking for a bug in
> > packs :).
> 
> Further to this the new parents are *wrong*.
> 
> details:
> for NEWS, 
> robertc at robertcollins.net-20070720032020-xiftpb5gqeebo861
> has two parents at the revision level:
> parent: pqm at pqm.ubuntu.com-20070720015347-eaeqmggngaemmbde
> parent: pqm at pqm.ubuntu.com-20070705224207-7pslqt12ofh4vnzx
> 
> now, NEWS was altered by my commit, so it got its own revid.
> At the file graph level it can have either one parent (if the merging
> revisions had collapsed) or two (if they hadn't).
> 
> My laptop where the commit was made has: ['robertc at robertcollins.net-20070720005841-xnu6um6vx0n41h0k']
> 
> bzr.dev has: ['robertc at robertcollins.net-20070720005841-xnu6um6vx0n41h0k', 'pqm at pqm.ubuntu.com-20070705224207-7pslqt12ofh4vnzx']
> 
> Now, pqm at pqm.ubuntu.com-20070705224207-7pslqt12ofh4vnzx changed NEWS, as did 'robertc at robertcollins.net-20070720005841-xnu6um6vx0n41h0k'
> 
> but...
> ....nzx is in the ancestry of ...h0k
> 
> And doing heads() of the two correctly returns ...h0k.
> 
> So I'm not sure whats going on here, but as of *now* bzr.dev has incorrect, corrupt per-file graph data.

I was able to pull bzr.dev yesterday morningish so I am suspecting one
of the merges from the following people caused the problem:

Martin
Jelmer
Keir Merle
D.M Watkins
Ian C

Its *possibly* caused by running the reconcile logic Aaron was working
on, or possibly by something else.

If you are on this list of names please do the following in your
repository that bzr's merge was pulled from:

grep qeebo861 .bzr/repository/knits/ed/%4e%45%57%
53-20050323055033-4e00b5db738777ff.kndx

Please paste the output into a mail to me/the list.

An uncorrupted repo will have 3 numbers at the end of the line, and only
one line.

A repo that has had its indices recreated incorrectly will show one line
with 4 numbers (offset, length, parent1, parent2).

A repo that has had the corruption propogate to it by the index
parent-update code will show 2 lines, one correct, and one with the
additional parent added.

So we're looking to find the repository that introduced the glitch, and
what was done to it, so that we can track down the root cause.

-Rob

Here is how to determine if you

-- 
GPG key available at: <http://www.robertcollins.net/keys.txt>.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20070918/92936b71/attachment-0001.pgp