Re: CVS commit: src/lib

["James K. Lowden" <jklowden%schemamania.org@localhost>]

On Fri, 22 Mar 2013 16:11:26 +0000
Emmanuel Dreyfus <manu%netbsd.org@localhost> wrote:

> On Sat, Mar 23, 2013 at 12:56:01AM +0900, Joerg Sonnenberger wrote:
> > > I was lucky, it worked reliabily on netbsd-5. It failed reliabily
> > > on netbsd-6. I call that a regression. The patch makes it working 
> > > reliabily again, even for the unlucky.
> > 
> > Repeating it doesn't make it any more true.
> 
> How is it false? Do you have a case where current code fails to
> support a DSO with -lpthread while it does not use threads?

What Joerg is telling you is that to call it a "regression" is
logically incorrect.  The 5.0 behavior was nondeterministic and
sometimes wrong; the 6.0 behavior is deterministic and always right.
To detect and act on an error is a *correction*, not a regression, even
if you were negatively affected.   

Would applying Joerg's patch not solve the problem you're addressing?  

--jkl