Michael van der Gulik wrote:
> Hi all.
>
> I made a Squeak package for the 2011 Google AI challenge
> (http://aichallenge.org), attached to this email. It works (but is
> untested by fire) but I'm giving up work on it.
>
> Basically, the AI challenge does things using shell and python
> scripts. Bots are compiled and then interact with the AI challenge
> through stdin and stdout. This much works for me.
>
> However, I was hoping that I could get the contest organisers to
> install Squeak in /usr/local/squeak, and then to run a user's
> submission, the VM would fire up the image in /usr/local/...
> somewhere, file in MyBot.st, run it and quit without saving anything
> such that no access to the file system is necessary (which is
> obviously a requirement for running code submitted by the general
> public). This doesn't work; Squeak fails to file in code if it cannot
> write to its .changes file.
>
> I can't be bothered to work around this issue. If anybody wants to
> grab my code and get the AI contest to support Smalltalk as a language
> then be my guest.
>
> Michael.
>
>
>
> --- Posted on the Google AI challenge forum in "Language support"...
> Hi.
>
> I've written a starter package for Smalltalk, attacked to this post.
> However, I'm giving up work on it.
>
> This is based on Squeak Smalltalk 4.2 (the current version available
> from http://www.squeak.org/). It works okay with the python tools
> except for a trivial bug: food needs to be removed from the map. I
> fixed this bug last night but then lost the changes somehow.
>
> However, I tried setting up a "server" version of Squeak for inclusion
> into the competition's VMs, and Squeak requires write access to a
> "changes" file (which, for the competition, would be in
> /usr/local/squeak/Contents/Resources/) in order to read in code,
> compile it and run. To me, this is the last straw; I've dodged too
> many issues trying to get this to work and I'm giving up.
>
> If anybody else wants to fix it up and deploy it, be my guest.
>
> Michael.
>
>   
Does Squeak also need to read from valid stuff from the .changes file? 
If not, then perhaps the Google admins could link the .changes to /dev/null.

just wild speculation....
cheers, Ben