Harvesting Process is not working

Peter van Rooijen squeak at vanrooijen.com
Mon Nov 3 21:42:35 UTC 2003


From: "ducasse" <ducasse at iam.unibe.ch>
> Use the BFAV and comment the fix and enh you want.

Stephane,

Thanks for this advice and the other tips you and others gave in other
posts.

I've probably gained enough familiarity with the BFAV now to be able to
contribute effectively. I was afreaid to add clutter by appending posts
without sufficient understanding.

> >> Your change is on my stack I read it but I was so confused by the
> >> discussion that I could not get an idea myself.
> >
> > How might I help with that? I had figured that my fix, containing only
> > minor
> > changes to three methods, with comments, was easy to assess. It seems
> > I was
> > wrong.
>
> Not it was not wrong. Your fix hit a political wall this is something
> else :)

I see what you are saying. I had the impression the "political" element was
more to do with the user interface aspects of warning about duplicate names
in nested shared variable namespaces. I don't have any strong opinions about
warnings when one is programming interactively.

About the nestedness of class variables, pool variables, global variables, I
think it was generally accepted that you can't really avoid that sometimes
something gets shadowed. Then if it happens, better to make sure the code
keeps working correctly. You made this argument too.

> Now we have to really assess how to proceed and as I said I'm still
> confused so I cannot help you in te battle now.

Perhaps emphasizing that this is not about how the browsers should deal with
interactive warnings (which is an issue for the ClassBuilder, not for the
variable declaration mechanism or the name lookup logic), would help, as
well as adding tests that demonstrate that the bug exists and that the fix
works.

> > It would not be difficult to add another change set to the submission
> > with
> > tests that fail without the fix and pass when the fix is installed.
> > Then all
> > that is needed is assessment of correctness of the assertions and the
> > observation that there is no regression on the other image tests.
> >
> > Does that make sense? Would that help? Should I do that?
>
> Please this would help to identify the problem.

Okay, will do.

Thanks again and best regards,

Peter




More information about the Squeak-dev mailing list