Re: [egit-dev] EGit Corruption @ GitHub

["Shawn O. Pearce" <spearce@xxxxxxxxxxx>]

Robin Rosenberg <robin.rosenberg@xxxxxxxxxx> wrote:
> s??ndagen den 31 januari 2010 18.27.53 skrev  Chris Aniszczyk:
> > Anyone have any insights on this?
> >     http://help.github.com/egit-corruption/
> 
> Yes. This is https://bugs.eclipse.org/286653

I'm not so sure.
 
The very first error displayed on that GitHub page is:

  $ git fsck
  broken link from  commit 5b90a930763c442f0fc3d819685083b4eda69f8e
                to  commit e1ea55d308b7808cb982f509c8dfa199ada4677e
  missing commit e1ea55d308b7808cb982f509c8dfa199ada4677e

This implies JGit didn't pack a commit that it should have when
building up the object list to send to the remote peer.

But bug 286653 is about ObjectWalk aborting tree/blob discovery early
because it found an empty tree in the project.  That phase is *after*
JGit has walked through the entire set of commits it needs to pack.


These are two different bugs.  The GitHub page must be talking
about a potential bug in RevWalk, not ObjectWalk.

-- 
Shawn.