[plt-scheme] reporting on planet packages and versions used by a set of collections & other tools

From: Neil Van Dyke (neil at neilvandyke.org)
Date: Mon Sep 1 09:07:16 EDT 2008

Noel Welsh wrote at 09/01/2008 08:32 AM:
> For a given Planet major version all minor versions should be the
> same, modulo bug fixes.  So it is reasonable to test using only the
> latest minor version, as it should not differ in functionality, only
> correctness.
>   

Agreed, *should*.  You know I'm going to play devil's advocate here. :)  
Rarely do we actually prove that given changes are backward-compatible 
wrt a specification, and even more rarely do we prove that all the 
interactions among units rely only on specified behavior.  By "rarely," 
I mean "never," for virtually everyone, including myself.

Just within my own systems, I'd like to know whether my assertions of 
version compatibility are correct.  Testing unit version combinations 
within my systems can inform me of reliance on unspecified behavior or 
of a non-backward-compatible change that was not caught in unit testing 
with the versions I happened to be using.  Ideally, I would be informed 
before I submitted code to PLaneT that falsely claimed to be 
backward-compatible or that had overly liberal "require"'d versions.

A fun project for an undergrad once there's a dense web of "require" 
relationships among PLaneT packages might be to brute-force test all the 
version combinations and see whether any new test failures fall out.  
Maybe not fun, but if it found a new defect, that might suggest they've 
found a useful new tool and/or a tangential benefit of PLaneT-like 
repositories.

-- 
http://www.neilvandyke.org/


Posted on the users mailing list.