Newbie investigating refactoring browser / conceptual typingquestion.

Ned Konz ned at bike-nomad.com
Sat Nov 1 19:25:05 UTC 2003


On Saturday 01 November 2003 9:00 am, G.J.Murphy wrote:
> Ned,
>
> > I think that had you installed the SqueakMap 1.09 package things probably
> > would have worked better, since it would have loaded Monticello, which
> > has a
> > (broken) version of this method. This would have also upgraded your
> > Package
> > Loader as a side effect.
>
> Aha! OK, done it and thanks.
>
> > > FYI, I ran the system SUnit tests before and after the fix:-
> > >
> > > BEFORE:
> > >
> > > 351 run, 230 passed, 4 failed, 117 errors.
> > >
> > > AFTER:
> > >
> > > 351 run, 278 passed, 5 failed, 68 errors.
> >
> > That's interesting. I find it surprising that any of those were even
> > testing
> > Text>>withSqueakLineEndings.
> >
> > Were these the RefactoringBrowser tests that you refer to?
>
> They're the ones that I get from the base 3.6 image plus the two installed
> packages. I ran all the unit tests that were available, which included
> various tests from the Monticello and the Refactoring Browser packages as
> far as I could see. Remember, you're dealing with a novice here!
>

That's odd.
I took a stock 3.6-full image, loaded the new SqueakMap 1.09, loaded the 
Refactoring Browser (make sure to get the one called "Refactoring Browser and 
related tools", and stuck in my fix. I got:

450 run, 447 passed, 2 failed, 1 error

And the 3 that had problems were in Monticello.

> The original problem was manifest in RBParser>>parseMethod, invoked from
> RefactoringBrowser>>parseNode, which I presume is invoked indirectly from
> umpteen higher-level unit tests. My understanding was that a parse tree was
> being built from the complete method text in the lower pane - this was
> going to be searched through to find the node covering as much of the
> selection interval as possible to extract a relevant expression.

I think it might make sense to pull the #withSqueakLineEndings out of the RB 
altogether. It's trying to solve a problem that is more of a package 
management issue; it should be done when packages are filed in.

Plus (as Markus has pointed out to me) this kills string literals that may 
happen to contain LF characters...

-- 
Ned Konz
http://bike-nomad.com
GPG key ID: BEEA7EFE




More information about the Squeak-dev mailing list