[plt-scheme] Recent PLaneT problems and their solutions

Dear Jacob,

Thanks. However, I've still a feature request left.

  Is it possible to provide some language construct to release
  the difference between 'lib' and 'planet'?

--Hans


Jacob Matthews schreef:
> There have been a number of recent issues that have floated over the 
> mailing list surrounding PLaneT, PLT Scheme's automated 
> package-distribution system. I've been working on them, and here are 
> the solutions:
>
> - The "Windows bug" that seemed to cause compilation errors in some 
> packages. This is actually a subtle bug in the SchemeUnit PLaneT 
> package. The problem is that one of SchemeUnit's macros introduces a 
> so-called "3D value" into the compiled code, which is fine for 
> compiled code that you just intend to _run_, but not fine for compiled 
> code that you actually want to write out to disk. In fact certain 
> SchemeUnit files do not compile at all under any platform. We are 
> looking into ways of signalling these kinds of errors earlier to make 
> debugging them easier. Note that due to a recent change in PLT Scheme 
> regarding path-marshalling, the problem will not arise in recent SVN 
> versions of PLT Scheme (but will arise in v301).
>
> - A PLT Redex compilation problem. The PLT Redex package contained a 
> packaging error that referenced a "reduction-semantics" collection 
> that does not exist in release versions of PLT Scheme, but does exist 
> in most developers' systems making the problem harder to see. This has 
> been fixed in a new version of the redex.plt package, already released 
> to PLaneT.
>
> - The planet command-line tool's --remove flag. It was broken in v300; 
> it is fixed in SVN and v301.
>
> - The compile-subcollections info.ss field. In v300, this field had a 
> different layout for PLaneT packages as for collections, creating an 
> unnecessary potential error for package developers. This is fixed in 
> SVN and v301; both layouts are now accepted for PLaneT packages.
>
> - PLaneT server stability problems. I still cannot figure out why the 
> server sometimes stops responding to requests, but while I'm working 
> on it I've instituted an automatic recovery system so that no manual 
> intervention will be required to correct the problem if it happens again.
>
> ;; --
>
> Thanks to the mailing list all for your help with these problems.
>
> -jacob
>
> _________________________________________________
>  For list-related administrative tasks:
>  http://list.cs.brown.edu/mailman/listinfo/plt-scheme