Re: [gnugo-devel] bug in find_neighbor_dragons()

[Inge Wallin]

Gunnar wrote:
> The first one looks rather accidental and the last two are related to
> semeai reading mistakes being uncovered. In my opinion there is no
> question that we should fix the bug even though the regression delta
> is disappointing, but I'll give you a chance to protest before
> applying the patch.

I found it unbelieveable that you even considered if it would be a good
idea *not* to fix a bug even though it creates a few FAILs.

I have seen this thinking in a lot of cases, many of them being my
patches.  I once started a long mail about this, but I never sent it
because I didn't think I had managed to formulate my thoughts in a
neutral manner.

So, this time I will just say what I think without bothering about
formulation.

*Of course* we should fix bugs.

*Of course* we should enhance the code and if the patch uncovers a few
bugs, then so be it.  Bugs that are known can be fixed.  Bugs that are
covered up will fester and cause lots of extra work.

Now, I understand that some times there are "bugs" or perhaps rather
simple solutions to complex problems that have to be used since the full
solution is simply not feasible, or perhaps that the infrastructure in the
rest of the program is not ready yet.

But what I see is a tendency to hold on to these oversimplifications and
to stick to old solutions even though the time to improve them has clearly
come.

So, in short, I think that development version should be allowed to be
weaker, at least for a short while.  When the time for release becomes
close, then we should postpone these types of patches.

Are we close to a release now?

        -Inge