Re: [Monotone-devel] change_set.cc:140: invariant'I(deltas.find(p ath) ==deltas.end()) ' violated

[eli . carter]

On Fri, 2005-04-08 at 16:35 -0500, Emile Snyder wrote:
> On Fri, 2005-04-08 at 13:51, address@hidden wrote:
> 
> > > notices that the same file is going to 2 different targets in the 
> > "same 
> > > commit" and barfs.
> 
> > You might look at cvsps.  It does the same basic task - synthesize 
> > changesets from CVS history.  I know it handles the case you
> describe. 
> > 
> 
> Thanks for the pointer.  I grabbed cvsps 2.0rc1 and it doesn't barf,
> but 
> it does collapse the same changes.  ie. it creates patchsets which
> have 
> both  
> loginfo:1.11->1.12 
> loginfo:1.10->1.11

Please pass the pepper; my words need some seasoning.
*sigh* I thought it wouldn't do that...

> I would be happy to just get monotone to stop barfing ;)  But it does 
> seem like monotone could be taught to deduce that this must have been 
> two separate commits...

Another bit of (perl) code to maybe take a look at is splitdiff from
patchutils.  Its purpose is to take a patch file and split it up so that
a file isn't modified more than once.

I haven't looked at the monotone or cvsps source; but if it builds up a
patchset, it should have a check for time being within bounds, and the
log message to match, and that should be able to do a duplicates check.

Eli