On Mar 1, 2012, at 9:03 PM, Mariano Martinez Peck wrote:
> I have also experimented the problem of Monticello being unusable
The sentence should stop here...
We really need to think hard about Monticello... just look at 1.4 from
https://ci.lille.inria.fr/pharo/job/Pharo%201.4/lastSuccessfulBuild/artifact/Pharo-1.4.zip
Image size: 17.7MB
Hmm... that's large. Pharo is growing out of leaps and bounds! We add crap all
the time!
No, actually not.
1. Monticello ancestry.
MCVersionInfo allInstances do: [ :each | each instVarNamed: 'ancestors'
put: nil ].
doit, save. --> 12.7 MB.
So MC wastes 5MB for history. Now history is important. But can we actually use
the history?
The Changes and Source are together now 53MB... because we need the history.
But, isn't
that what MC should do? Why can't it do it? Even though it wastes 5MB
constantly?
Continuing...
2) Scriptloader. Big piece of crap. Delete the class. Save ---> 10.3 MB
What it does is to define which exact package versions made up a pharo version.
So we know what
14345 was *exactly*. And it records the comment of what changed.
Again: Shoudn't a "software versioning and a revision control system" do
*exactly* that? No?
3) "ScriptLoader new cleanUpForProduction" does 1 and 2 and removed the tests,
does some cleanup
--> 8.3 MB And that is with all the fonts! Fully usable. All tools. (but
not SUnit). 8.3 MB....
So I really wonder if we don't need to do something different....
Marcus
--
Marcus Denker -- http://marcusdenker.de
