[Boost-users] Re: MultiIndexSet invariant test failure

From: Joaquin M Lopez Munoz (joaquin_at_[hidden])
Date: 2005-04-14 14:16:33

Manges, Ross G <Ross.G.Manges <at> boeing.com> writes:

> 
> Joaquin M Lopez Munoz wrote:
> > You've got to implement semantics equivalent to that of
> > the < relationship between numbers:
> Thanks for the info; I now believe I'm doing this correctly.
> 
> > * Now, in the piece of your code where the insert asserts,
> > add a precheck of the invariant:
> > 
> >   m.check_invariant_(); // point A
> >   m.insert(...);        // point B
> > 
> > * Run your test till the assert fires. Is it in point
> > A or B? If A, then most likely you're somehow modifying
> > the elements in some other thread. Get the idea?
> This is a great idea! I added the precheck as you suggested, and it
> appears that I am experiencing the assert in 'point B'.  The invariant
> is OK until the insertion breaks it.  So can I infer that somehow what
> I'm inserting is invalid?  It's a shared_ptr.

Two (main) possibilities: the shared_ptr is invalid or
my code has a bug. Could you test the pointer for validity
before making the insertion?

  m.check_invariant_();
  // check ptr before insertion
  m.insert(ptr);       

If still the error is detected inside insert(), we can try to
produce some sort of poorman dump of the container: define
BOOST_MULTI_INDEX_INVARIANT_ASSERT to throw an exception on
failure and wrap the call to insert() as follows:

try{
  m.insert(...);
}
catch(/* exception thrown by the assert */){
  // dump the contents of m
}

Can you do that? Notice something weird in the dump?
Can you post the dump?

> I recently sent in
> another post regarding problems I'm having with shared_ptrs that might
> be related (see post w/subject: "shared_ptr use_count in multi threaded
> environment").
> 

Maybe. But it could be a bug in Boost.MultiIndex, also.

Joaquín M López Muñoz
Telefónica, Investigación y Desarrollo