Source code and test suites

[Jeroen Dekkers]

At Thu, 19 Jan 2006 12:57:20 +0100,
Florian Weimer wrote:
> 
> * Jeroen Dekkers:
> 
> >> For sufficiently complex software, a suite of regression tests is
> >> extremely helpful for making modifications.  However, in some areas,
> >> there is a strong interest in proprietary test suites (in the GCC
> >> context, for example).  Forced publication of test suites would unduly
> >> extend the scope of software copyright.  That's why I'm not sure if
> >> anything should be done about this.
> >
> > Because you don't distribute (or propagate in GPLv3 terms) anything
> > when running a test suite on a program, I don't see how the GPL could
> > enforce anything about test suites.
> 
> The same argument could be applied to the encryption keys and build
> scripts, but providing them is explicitly required.

That's because you can't exercise your rights (modifying the software)
without them. Build scripts are also clearly part of the
software. Test suites don't have to be that, you might not even know
whether it exists.

To give an example, if I create my own GNU/Linux distribution and want
to have it POSIX-compliant, I can download and run a POSIX test
suite. But being forced to publish that test suite doesn't sound right
to me.

Jeroen Dekkers