[dmd-beta] dmd 1.071 and 2.056 beta

[Brad Roberts]

On Tue, 25 Oct 2011, Don Clugston wrote:

> On 24 October 2011 23:01, Brad Roberts <braddr at puremagic.com> wrote:
> > And then there was one.. just:
> >
> >  http://d.puremagic.com/issues/show_bug.cgi?id=6681
> which is only a regression for D1, not for D2.
> >
> > Almost to the point of a 0 regression release.  Hopefully the start of a
> > very important and wonderful trend.
> 
> I think that the number of regressions introduced in release RRR or
> greater is a far more important metric than the numbers of regressions
> reported since DDD. The crucial value of RRR for D2 is 2.048, since
> 2.047 coincided with the release of TDPL.
> Compared to previous releases, we're doing much better in that respect too:
> 
> 6177  Regression(2.053): ICE backend/cgcs.c: struct with destructor in
> assoc. array or typesafe variadic functions
> http://d.puremagic.com/issues/show_bug.cgi?id=6177
> 
> 6278 Regression(2.054): 'in' contract inheritance doesn't work with safe code
> 
> 6398 Regression(2.054): Property getter/setter confusion
> 
> 6329 (2.054, Windows only) Out of range exceptions not thrown in certain cases
> 5533 Regression(2.051, Linux only): segv: -gc, associative arrays,
> const pointers to self
> 
> There are a couple of Phobos regressions:
> 
> 5876 std.container doesn't work with delegates
> 5718 Can't demangle symbol defined inside unittest block
> 
> one druntime regression
> 6088 Stack trace or some info with stack overflow
> 
> and one packaging regression
> 6406 [2.054] Libraries not working on Ubuntu

Oh, I don't disagree with any of that at all.  I'm happy making progress 
on any front so I pressed on the easiest one to achieve.  This is the best 
release to date (I think, haven't actually done any indepth analysis) in 
terms of regression fixing.  My real goal is to get the number down to 0 
and for it to stay there from releaes to release.  I just don't expect it 
for a good while still.

I firmly believe that we can and must get there sometime soon though.  
Keeping to a hard line of no unfixed regressions is an important release 
management goal.

Later,
Brad