Re: [Simulavr-devel] Patches - again

[Joerg Wunsch]

As Onno Kortmann wrote:

> > But if we do not want reviews, ok, then everybody should check
> > in. No problem. I have my local copy :-)

> Well, I do NOT know the policy here for patches and/or check-ins. Is
> "compilable" successful enough, should a test suite finish
> successfully, should it pass additional review, what is it?

> My personal preference is 'compilable & test suite is ok', as this
> is 'development quality' code, isn't it?

I'm afraid once you (or whoever) committed it to CVS, nobody is ever
going to touch it again.  So my very personal opinion: it's OK to
initially commit "development quality" code (basic requirement: it
must compile for everybody), *but* only as long as you continue
working onto it until it's got "good quality" (or even "excellent
quality").  Otherwise, over time, the quality of the final product
will severely suffer.  Seriously, Onno, some of your ideas sound
really cool, but then, some of your comments make me a little scary as
well. ;-)  (Like the one explaining that there's still a memory leak
within.)  For sure, I wouldn't want to have that within a final
release myself.

As Klaus' response above sounded a little annoyed, and even partially
like he might completely abandon this project (at least in public), I
continued some private discussion with him.  That way, we could handle
it in German, which was simply easier for the job.  After a couple of
emails, his initial annoyance was basically gone, and he assured me
that he primarily wants to step back a little, to allow others to
eventually step in.  He feels all the ideas and patches suggested do
require quite a bit of polishing, but he doesn't have the time to
perform that kind of polishing within only a few days, so others have
to take care of it.  Of course, one of his (understandable) hopes is
that he'll recognize the project later on so it's not going to become
a *completely* different thing.

During this discussion, Klaus mentioned me *his* objectives for the
project.  I felt this being quite important, as it explains a bit the
history of the project, as well as the way he maintained it.  As Klaus
currently cannot read/write emails (for different reasons unrelated to
simulavrxx), I asked him for permission to translate that part of his
email into English, and repost it to the list:

Klaus' really *main* objective is simulator speed.  Achieving a good
simulation speed counts above all to him; he rather avoided things
like "politically correct OO methods" if they adversely affect the
simulation speed.  (As one of the things that are still quite
suboptimal in that respect, he mentioned the LCD simulation.  He
committed it to CVS anyway, but considers it inferior in terms of
speed.  Any takers?)

His second objective exactness of simulation, including exactness of
the timings.

The third on his list was stability, and finally, code that is easy
enough to understand it again after a couple of years.


He also mentioned his #1 item on the feature list, which he sees as a
requirement before calling the version 1.0: completeness of the
simulation features of an ATmega128.

-- 
cheers, J"org               .-.-.   --... ...--   -.. .  DL8DTL

http://www.sax.de/~joerg/                        NIC: JW11-RIPE
Never trust an operating system you don't have sources for. ;-)