Binding tests

[Matthew Johnson]

On Wed, 26 Jul 2006, Daniel P. Berrange wrote:

> Rather than having it make assumptions about the names of sub-dirs,

I was assuming they'd just get symlinked in from some mechanism.

> could we re-arrange things a little - I think it'd be good to have
> the ability to run this kind of test harness independant of the source
> tree. And in particular keep a clean separation between the mechanisms
> for checking out the code from source / building it and actually performing
> the tests. For example, I already have an automated builder for the Perl
> bindings[1] which deals with the issue of checking code out from SCM[2], so
> it'd be handy to just be able to call out to a script to run the test harness
> alone.

I was imagining you would check out the harness and run make
BINDINGS=perl if you just wanted perl tested against perl. (it's what
I'm doing with Java atm).

> One simple approach would be to have the stuff currently in the Makefile
> just be in a standalone shell script. Each binding would provide two
> executables - one implementing the client side of the test suite, the
> other implementing the server side of the test suite. The test harness
> script could accept two parameters - the path to the client executable
> and the path to the server executable, run the two & report on outcome.

Well, there needs to be some standard hooks into make to present an
standard method of building them, why not also of running them. Also,
java doesn't neatly build to an executable (although it can be wrapped
in a script).

> Certainly, when I back from vacation I'll take a stab at writing a Perl
> impl of the test suite.

Cool,

Matt

-- 
Matthew Johnson
http://www.matthew.ath.cx/