Wednesday, May 30, 2007


If you want to have more informations about my GSOC project, please visit my blog at:

Sunday, May 27, 2007

Working on Monticello 2

After some discussions with my mentor Stéphane Ducasse, I have decided to improve Monticello 2 (MC2) and make it usable to every squeaker as part of the Google Summer of Code.

I have started by opening a repository on SqueakSource on which I will be able to put my own work. It currently contains the last available versions of MC2 packages.

Then, I sent an email to the Squeak-dev mailing list requesting information about MC2: its design, its state... Colin Putney and Avi Bryant kindly answered, giving details.

I also found old posts on Colin's and Avi's blogs talking about MC2 design.

Finally, I've received a mail from Keith Hodges giving information about the current status of his work on Monticello and what I can do to improve MC2.

The next task I would like to achieve is opening a wiki page on the Squeak Swiki containing all information I can grab about MC2. This pages will also be used as a place for me to write new documentation.

Any feedback will be appreciated.

Monday, May 21, 2007

Seven Days to GSoC start: looking forward and looking back

Next monday will mark the start of the Google Summer of Code 2007: from that day the students will officially start working on their projects (and getting paid, too). But since there's nothing forbidding them to start working earlier, I've just sent our five students an email asking whether they've done anything during this interim period and, in case, what they've done.

While we wait for them to reply, I'd like to look back for a moment to the selection process. Using an agile/XP term, we could call this a little retrospective.

So, what went well with our selection process? First of all, the fact that we had so many student proposal, and most of them came from students sincerely interested in working on/with Squeak, and not just in it for the money. The fact that we had to coordinate with many other GSoC projects in order to select our 5 students is another proff of the level of the students who applied for a Squeak project.
Another thing that I'd like to remark is the level of involvement of our mentors. They all took their job seriously, throughly examining and evaluating all the proposals. Kudos to them!

What could be improved? Personally, I felt that there was a little "creeping" in the mentor and proposal lists. We started with ~10 mentors and 12 proposals on our list, we ended up with 14 mentors and 18 proposals. Freezing the two lists a little earlier can make coordination a little easier.
I'd also like to have more people on the Squeak IRC channel during the whole proposal submission period. This way people coming into #squeak to ask for info will have more opportunities for a correct answer.

What could we experiment with? In the past weeks I was toying with the idea of a Squeak bounties system. If we'd manage to put this or a similar system in place during the autumn, we could also use it for gathering proposals for next year's Summer of Code.
Also, for next year's Summer of Code it would be worth considering positioning Squeak as an umbrella project, managing proposals involving the main Squeak system, Seaside, Squeakland, Sophie, Croquet etc.

These are some of the points I wanted to make. Do you have any comments or any more points to add to the discussion? if so, then the comment form is waiting you!

Monday, May 14, 2007

How to load the NewCompiler

Hi everyone,

I juste want to add a litle tip about how to load the NewCompiler so here is the basic tree dependency:

So in oder to load the NewCompiler you have to fulfill the dependency. You can find them all in Squeaksource.

The next post I will show you how to write bytecode easily.
Let comments if you find it interesting or if you want any tips about the NewCompiler.

Thursday, May 3, 2007

NewCompiler ToDo

This last 3 days I have met Marcus at Bern and we discuss about things to do and the way of doing them.

Then here the work to be done:
- The more important things: make the decompiler work so we can propose it to the 3.10/3.11 release team.
- Have 2 parser so we can work on one without breaking it all
- Remove the reference to the RBParser and RBToken
- Clean the interface of the RBNode
- Integrate the continuations from Seaside
- Make debug it work in the debugger
- Profile the compiler to optimize it.
- Create some visitor to have the AST annotations and remove the token
- Work on the VM and new bytecodes/primitives
- ...

Up to now I have look at VMMaker and implement some bytecodes the improve the performance of the method with closures by 6. Those bytecode are experimental and we should propose some more efficient.

We have also change the organization of the NewCompiler. So now we have 3 packages:

- AST: Package for the RBNode and RBToken (which should be removed)
- NewParser: Package containing the parser so you can use it for your purpose without loading the complete compiler. We make this package mostly to make the RefactoringEngine using the SqueakParser so we can remove the RBParser.
- NewCompiler: The closure compiler