[Libreoffice] concept for c++ based subsequenttests

[Michael Meeks]

On Wed, 2011-11-30 at 23:16 +0100, Bjoern Michaelsen wrote:
> On Wed, Nov 30, 2011 at 04:20:28PM -0500, Kohei Yoshida wrote:
> > Let me cast my vote for the use of C++ too.  Markus has already outlined
> > the benefit of using C++ for debugging point of view. 
> 
> Oh, I have no opposition against writing C++ tests. When have the
> infrastructure for that in the build system and it certainly has advantages
> once the test is there. However, as Michael pointed out a dynamic language has
> clearly some advantages too when aiming for getting the test up quickly.

	Personally I don't care what language the test is written in ;-) but I
do have these requirements for the ideal unit test:

	* easy to debug
		+ is a single process
		+ gives a complete stack-trace of unit test + core code
		+ allows interactive inspection of variables & call
		  frames up and down that stack
	* runs quickly - in a handful of seconds
	* 100% reliable (in the absence of a real bug)

	How we achieve that is of course up to us. To me, as of today, all of
the above says C++; perhaps gdb will show us nice python stacktraces
instead of random python internals in the near future, and allow
inspection of variables: if so, I'm well up for it - assuming we have a
cppunit framework that bootstraps it inside the same process (not that
hard I think).

	HTH,

		Michael.

-- 
michael.meeks at suse.com  <><, Pseudo Engineer, itinerant idiot