[V3dot10] Re: Monticello Mysteries

Keith Hodges keith_hodges at yahoo.co.uk
Thu May 10 15:59:05 UTC 2007


 cc to squeak-dev
>> bertf wrote:
> The pre/post-load/unload scripts actually could be changesets, yes. 
> How would you imagine "managing" these?

> Thats why I am pondering the idea.
>
>
> Keith
My conclusion so far is that it may not make sense to manage ChangeSets 
in monticello for general use, but I think it makes sense for some 
scenarios.

For example: 1.

Later monticello versions do not load into earlier squeak versions.

One solution is to make all of the code in the later version work in the 
earlier versions. Sometimes this is not easy and results in hack filled 
conditional code such that: "If this method is available in use it if 
not do something else".

A second solution is to add 'extensions' or 'overrides' * to fix the 
base image for earlier versions. This can work well, but if the method 
has been included in the base image  then thes 'extensions' or 
'overrides' are unnecessary, and would pollute the later image.  Also 
the range of images may not have those classes present, and so this 
produces loading errors. (Magma has this problem)

So my pondering has led me to think that if I could publish a base-image 
patching changeset, and manage that in monticello (i.e. have revisions 
of that patch in a repository)  Then  I can provide a specific patch for 
a specific target image.

Having a patch in monticello might then allow monticello configurations, 
or a dependency mechanism to pick up the needed patch automatically.

best regards

Keith




More information about the Squeak-dev mailing list