[V3dot10] making and recording changes

stephane ducasse stephane.ducasse at free.fr
Fri Jan 19 12:33:33 UTC 2007

On 19 janv. 07, at 11:23, goran at krampe.se wrote:

> Hi folks!
> Ok, I have two silly and probably totally way off things I wonder  
> about:
> 1. Is there a major reason why we couldn't use a "good ole" update
> stream in which we can mix:
> 	- Regular old style changesets, though typically phased out in favour
> of:
> 	- Changesets that only load a *specific* MC snapshot (stored  
> somewhere
> where it never goes away)
> Since I haven't looked at the architecture of 3.9 (ScriptLoader etc)
> bear with me.

this is what we have. Now the problem is that if you load something  
with CS (sometimes you even
have to manually edit them), then you end up having dirty packages.
So you save packages so that they are not dirty (or because you  
merged a nice fix another guy did)
but  the packages saved that way may not load again because of wrong  
order that you do not control.
But you can check that only if the code that would break needs to be  
In addition cs may break MC because of reordering of category and we  
lost methods
because of that.
With atomic loading this would be much easier.

> 2. Couldn't someone take a stab at reimplementing ChangeSet in a
> reasonable way? :) Currently they are quite confusing to the newbie  
> (who
> doesn't realize that they do not actually *contain* the code - just
> references to touched methods). It seems to me that they would be much
> more useful if they actually contained their state (when reified in
> memory) and of course handled for example class renames etc better.


> Or perhaps Colin has already done this too in MC2 - there is something
> in there called Slices IIRC.

I should really have a look at work on MC2

> regards, Göran
> _______________________________________________
> V3dot10 mailing list
> V3dot10 at lists.squeakfoundation.org
> http://lists.squeakfoundation.org/mailman/listinfo/v3dot10

More information about the V3dot10 mailing list