Next steps for MCP?

Daniel Vainsencher danielv at netvision.net.il
Mon May 26 13:18:41 UTC 2003


Now posted. What's changed - 
* PackageGraph generation for all of Squeak is down to 34 seconds (from
830 - all on my machine, of course).
* The packages name is now updated to MudPie.
* the method allShortestPathsFromPackage:to: is added, which is pretty
useful for finding flaws, as shown in the example below (btw,
explore-it)
* Includes some basic tests that show how the Kernel and Morphic
framework are completely tied into practically everything else. These
tests are stil coarse, especially the kernel tests, so that they will
only pass when everything is clean. Finer grain tests I'm still thinking
about.

Daniel

Stephane Ducasse <ducasse at iam.unibe.ch> wrote:
> please publish the newest version
> 
> Stef
> On Sunday, May 25, 2003, at 03:05 PM, Daniel Vainsencher wrote:
> 
> > Stephane Ducasse <ducasse at iam.unibe.ch> wrote:
> >> For example when working on KCP, I noticed that Dictionary was
> >> referring to MorphGraphicalMenuDictionary.
> >> While it should be in the other directions. So I fixed it if I 
> >> remember
> >> correctly.
> > Do you have a test for it? if it's not tested, it's not done ;-)
> >
> >> It would be excellent to have a bunch of Spaghetti tracer scripts
> >> identifying different flaws in Squeak.
> >> We could then evaluate them to see if we progress. Alex plays with > 
> >> SpagTr but I did not had the time to discuss
> >> with him.
> > I'm building tests of exactly that form (self assert: (mma
> > areInDisjointComponents: {Dictionary. MorphGraphicalMenuDictionary})).
> >
> > If a test is red, you do
> > mma allShortestPathsFromPackage: (mma packageDefiningClass: Dictionary)
> > to: (mma packageDefiningClass: MorphGraphicalMenuDictionary) to get
> > hints on why.
> >
> > BTW, this is using a slightly more updated version than is on SM, I can
> > update tommorrow if someone is interested.
> >
> > Anyway, the tests I'm making will cover everything I can understand how
> > to cover in an intelligent manner in Squeak.
> >
> >> I created a page on the KCP for all kinds of stuff we could fix such 
> >> as
> >> the deepHorror from Ctuluh: Utilities.
> >> Please add ideas there too.
> > Yes. Algorithm for separating Morph from the Squeak Kernel
> > 1. Run
> > mma _ MultiModuleAnalyzer onCategoriesWithPrefix: ''.
> > mma allShortestPathsFromPackage: (mma packageDefiningClass: Object) to:
> > (mma packageDefiningClass: Morph).
> > 2. Refactor every bad link it shows.
> > 3. repeat until fix point. ;-)
> >
> > Obviously it'll take a long time for that test to pass, so we'll need
> > more specific ones to show progress, which is what I'm thinking about
> > now.
> >
> > Daniel
> >



More information about the Squeak-dev mailing list