Re: [Monotone-devel] Re: Invariant violated when converting a CVS repository to Monotone using Tailor

[Nathaniel Smith]

On Tue, Sep 12, 2006 at 07:47:22PM +0200, Koen Kooi wrote:
> Nathaniel Smith schreef:
> > On Tue, Sep 12, 2006 at 11:54:31AM -0300, Juan Jose Comellas wrote:
> >> I have been trying to convert a big CVS repository to Monotone using 
> >> Tailor, 
> >> and I got this error in the middle of the process:
> >>
> >> mtn: fatal: St11logic_error: database.cc:845: invariant 'I(cancel_size < 
> >> pending_writes_size)' violated
> > 
> > Wow, lots of people are hitting this.  The bug is already fixed in
> > mainline, if you want to get that; otherwise, you can patch up 0.29
> > simply by finding line 845 of database.cc, and turning the "<" into a
> > "<=".  I.e., it currently says
> >   I(cancel_size < pending_writes_size);
> > but it should say
> >   I(cancel_size <= pending_writes_size);
> > 
> > Sorry about that,
> 
> It also seems 0.28 is a bit less prone to this invariant, so you could try 
> that if you
> have a fairly large repository where 'regenerate rosters' will take some 
> hours (like with
> the OE.mtn from openembedded.org).
> Apart from that mainline rocks with it überfast 'commit' and 'update' :)

0.28 does not have the code that this bug is in, thus it does not have
this bug :-).  For the same reason, though, it eats a lot of memory on
commits, which might be an issue for you running tailor.

Current mainline is definitely much more friendly to tailor than 0.29,
though, because of the "überfast" commits that Koen mentions --
see:
  
http://thread.gmane.org/gmane.comp.version-control.monotone.devel/8093/focus=8093
and the reply:
  
http://thread.gmane.org/gmane.comp.version-control.monotone.devel/8093/focus=8133

-- Nathaniel

-- 
So let us espouse a less contested notion of truth and falsehood, even
if it is philosophically debatable (if we listen to philosophers, we
must debate everything, and there would be no end to the discussion).
  -- Serendipities, Umberto Eco