Re: [Monotone-devel] Monotone bug, maybe workaround

[Justin Patrin]

On 7/19/05, Leif Jakob <address@hidden> wrote:
> On Tue, Jul 19, 2005 at 09:52:20AM -0700 Justin Patrin wrote:
> 
> > Date: Tue, Jul 19, 2005 at 09:52:20AM -0700
> > From: Justin Patrin <address@hidden>
> > To: To address@hidden
> > Subject: [Monotone-devel] Monotone bug
> 
> Hi Justin,
> 
> > monotone: fatal: std::logic_error: revision.cc:185: invariant
> > 'I(*changesets.find(old_id)->second == *old_to_child_changes_p)'
> > violated
> 
> Guess I've bounced into this bug multiple times as well.... But I
> could't provide a database because of closed source... maybe your
> database would help?

If I can get it back into that state I'll strip my key and send it on.
It's >50MB, though.

Actually what I did was to go back through the merge and not eall of
the manual merges. Then I made all of the manual merge changes to my
checked out copy, checked in those changes, and re-ran the merge. The
merge worked this time. As I said, I'll see if I can put it back into
the broken state and see if the below works.

> 
> If you look at the output of your "monotone merge" you will see a parent
> revision that was the root of the revision tree (use monotone-viz) and
> that merge crashes because of plenty of adds/removes (dont't know why)...
> 
> But I've a workaround:
> 
> run "monotone heads" and see the two (hex) revisions, then run
> "monotone lca $rev1 $rev2" and get the common acestor, now run
> "monotone explicit_merge $rev1 $rev2 $ancestor $branch" and
> it magically should work...
> 
> The "monotone merge" uses some "other" ancestor-algoritm, but that
> seems to degrade sometimes...
> 
> To be on the safe side make a backup of the database.
> 
> Have fun
> 
> Leif
> 


-- 
Justin Patrin