Crash in 1.0.2: "TypeError: b2a_hex() argument 1 must be string ..."

Benoit Boissinot benoit.boissinot at ens-lyon.org
Fri Jan 9 22:57:09 UTC 2009


On Fri, Jan 09, 2009 at 05:47:24PM -0500, Greg Ward wrote:
> [me]
> > Even if I wipe the bad .hg/patches/status file, it's still busted:
> > 
> >   $ cat /dev/null > .hg/patches/status
> >   $ hg up -C
> >   abort: unknown revision 'beda36e80868b4f17736f9e7eec0a96a925a77ca'!
> 
> [Benoit Boissinot advises]
> > Yeah the problem is the dirstate, not .hg/patches. Use debugrebuildstate
> > to regenerate it.
> 
> No dice.  Even after upgrading to 1.1.2, I cannot fix my broken
> repository:
> 
>   $ hg parents                # demonstrate brokenness
>   abort: unknown revision 'c86a0eb29c417037bb2a8db5ea672205e2f8d893'!
> 
>   $ hg debugrebuildstate      # try Benoit's suggestion
>   abort: 00changelog.i at c86a0eb29c41: no node!
> 
>   $ hg debugcheckstate        # saw this one in "hg help --debug"
>   abort: unknown revision 'c86a0eb29c417037bb2a8db5ea672205e2f8d893'!
> 
> Interestingly, "hg debugstate" gives meaningful output.  It even reports
> the rename that appears to have given MQ a fit.  But I can't shake that
> pesky error
>   abort: unknown revision 'c86a0eb29c417037bb2a8db5ea672205e2f8d893'!
> 

Can you give the first csets from hg log ? (ie is tip == c86a0eb29c417?)
And can you do debugrebuildstate with --traceback ?

thanks,

Benoit

-- 
:wq



More information about the Mercurial mailing list