Re: [Monotone-devel] [ANNOUNCE] monotone 0.26pre3 released

[Nathaniel Smith]

On Thu, Mar 30, 2006 at 09:48:52PM +0400, Petr Ovtchenkov wrote:
> > mtn --db=venge.mtn db init
> > mtn --db=venge.mtn pull venge.net "net.venge.monotone*" --exclude 
> > net.venge.monotone.contrib.webhost --dump venge.dump.2
[...]
> monotone: fatal: std::logic_error: roster.cc:1422: invariant 
> 'I(right_uncommon_ancestors.find(right_marking.birth_revision) != 
> right_uncommon_ancestors.end())' violated

I ran this command and the pull worked fine; I don't get this error.
The dump files look very strange, too -- they don't contain any
information on the actual source of the problem.  I went to go add
more instrumentation, so you could re-run it and get more data, but...
the instrumentation is already there -- the only caller of the
function that contains the failing invariant, marks more than a dozen
pieces of data as being interesting in the case of a crash, but
somehow when it crashes it says "work set contains 0 items".

I'm almost wondering about some sort of strange issue with compiler or
libraries?  Though that's strange, since you're running on linux,
which is very well tested.  We might even need to take extreme
measures to reproduce this, like asking you to compile with
--enable-static-boost and debugging, and seeing if the resulting
binary still causes a problem on our system, etc....

Can anyone else reproduce this problem?

-- Nathaniel

-- 
Damn the Solar System.  Bad light; planets too distant; pestered with
comets; feeble contrivance; could make a better one myself.
  -- Lord Jeffrey