On Sep 24, 2007, at 2:20 AM, Martin Husemann wrote:

> On Sun, Sep 23, 2007 at 05:09:44PM +0200, Julio M. Merino Vidal wrote:
>> for features that are plain C.  (I'm specially afraid of this because
>> of the sudden interest on pcc, which does not support C++.)
>
> Declare it as a goal but drop it's priority and don't waste time on
> it right now?

Right; it should not be high priority now given that changing the  
tests in the future should be easy as long as they don't use C++.   
But I think it's good to keep this in mind for future releases.

> Unless Anders wants to pick ATF *now* for pcc regression tests (not
> sure how to priorize in that case though)

Jeremy was considering to use ATF for pcc (among other tools, such as  
dejagnu), but I'm not sure if he'll end up using it.  However, pcc  
tests could (and should) be built with the host compiler for now,  
which supposedly is GCC and supports C++.  In other words, I don't  
see why you'd need to build the tests with pcc itself to test pcc.   
As I see it, the tests will invoke pcc and/or interface with its  
internals directly, to check the results of the calls afterwards.   
And the former can be done from sh, for example.

-- 
Julio M. Merino Vidal <jmmv84@gmail.com>