ghost revs/inconsistent parents/missing revs in bzr check

[Christophe TROESTLER]

On Tue, 7 Apr 2009 17:25:47 +0200, Tom Cato Amundsen wrote:
> 
> It can get worse.

I just tried on some repositories of mine and it seems not to work
well when using repositories branched from SVN [*].  E.g.

$ bzr check
Checking working tree at 'file:///home/trch/software/OCaml/syntax-ext/pa-do/trun
k/'.
Checking repository at 'file:///home/trch/software/OCaml/syntax-ext/pa-do/'.
bzr: ERROR: parent_id {4 at 8e2808ce-4b4b-0410-b4ba-b33ea86c27bb::trunk} not in inv
entory

or

$ bzr.dev check
Checking working tree at 'file:///home/trch/software/OCaml/fftw3/'.
Checking repository at 'file:///home/trch/software/OCaml/fftw3/'.
bzr: ERROR: bzrlib.errors.SHA1KnitCorrupt: Knit <bzrlib.knit._VFContentMapGenera
tor object at 0x1e8aed0> corrupt: sha-1 of reconstructed text does not match exp
ected sha-1. key ('christophe.troestler at umh.ac.be-20080529074833-7x5b8xt01jv4rrr
a',) expected sha 8fd1763db7261af897f286c705a5484557276329 actual sha 8d1bab21bd
3469636a92a607624591b50a873d8a

Traceback (most recent call last):
...

My day to day interaction with these repositories works fine but these
errors are a bit worrisome.  Is there a way to fix them?  I can send
the repositories or perform some checks if needed.

Regards, 
ChriS

---
[*] A fresh checkout from SVN checks fine but I would prefer to avoid
that as I have some branches that I have not made public.  If I branch
from the bzr repository it also checks fine and I only have some

      ghost revisions
      revisions missing parents in ancestry
      inconsistent parents