Re: Beaming regression 2.15.39 compared to 2.14.2

[Graham Percival]

On Thu, May 24, 2012 at 01:19:29PM +0200, address@hidden wrote:
> On 24 mai 2012, at 12:04, Urs Liska wrote:
> > A regression is something that doesn't work in a later version and that has 
> > _deliberately_ worked in a previous version. I.e. something that has once 
> > been fixed to work in that specific way.
> > If it just was correct and isn't anymore, it isn't considered a regression 
> > but just a newly introduced bug.
>
> Still a regression. Any change in behavior that is not fully accounted for in 
> the change log and that you feel leads to worse behavior than a previous 
> version is a regression.

I believe that Urs has the correct definition, although I don't
have a reference handy.  It should be in the CG, or at least the
GOP decision from last summer, though.

If we try to account for accidental changes of
accidentally-working stuff, 2.16 will never be out (until/unless
we change the release policy during GOP 2, which I will be
proposing).

- Graham