[plt-dev] Re: [plt-translators] Translations for test engine

From: Eli Barzilay (eli at barzilay.org)
Date: Thu Apr 23 13:29:38 EDT 2009

On Apr 23, Robby Findler wrote:
> On Thu, Apr 23, 2009 at 12:16 PM, Eli Barzilay <eli at barzilay.org> wrote:
> > On Apr 23, Robby Findler wrote:
> >> On Thu, Apr 23, 2009 at 1:46 AM, Michael Sperber
> >> <sperber at deinprogramm.de> wrote:
> >> > If so, is there documentation or anything that describes what
> >> > dependencies are OK and which one's aren't?
> >>
> >> I believe Eli maintains this information. Generally the way it
> >> works is someone adds require somewhere that breaks things and
> >> Eli lets them know when a script fails somewhere. I don't know if
> >> the precise list is available on the web, but I'm sure Eli would
> >> be more than happy to make it be so if it isn't.
> >
> > * This is in iplt/build/distribution-specs.
> >
> > * Yes, there is a dependency now, this is bad -- but making it
> >   worse is not a good idea.  (You can see this by the fact that
> >   "test-engine" appears in dr-extras instead of plt-extras)
> 
> I do not believe Mike has actually made it worse in any interesting
> way. It will require a few more Emacs macros to undo his changes,
> but that can only happen when the larger problem is fixed.

No, it's not worse in an interesting way -- just worse.  Each such
change makes disconnecting DrScheme from such components so that
distribution is sane more a dream than a possible reality.


> Put another way: we should not stand in the way of Mike for this
> reason. It was not his doing that put us in this bind so we should
> not punish him!

It *is* making things worse.


> > * There is another issue here which is that the teaching languages are
> >  intended to be used by themselves, are used in a handin server
> >  situation, etc.  Adding a preference-dependent option to this kind
> >  of thing is bad.
> 
> The word "adding" is misused here, iiuc.

No -- the current use of string-constant in the test-engine collection
is *only* for gui strings, so any behavior that would be seen through
a handin server or through a distributed program would be the same.


> Whatever thing needs to be done so that check-expect works in the
> handin server still needs to be done and what Mike has done does not
> changed that.

Of course it needs to be done still -- but the current problem with it
is unrelated to the current issue.  The current issue is potentially
going to add a new problem.

-- 
          ((lambda (x) (x x)) (lambda (x) (x x)))          Eli Barzilay:
                  http://www.barzilay.org/                 Maze is Life!


Posted on the dev mailing list.