On Fri, Dec 18, 2009 at 06:50:59PM +1030, David Kettler wrote:
> > Nice idea, but let me play devil's advocate for a second, to make sure we
> > are clear-headed about this.  Just to go over our terms, our smallest
> > primitive is the "test", and tests are grouped in "suites".  What we are
> > really getting at here is adding a new type---a group of
> suites---which on
> > the surface sounds self-contradictory.  Is there a parallel for this in
> > other testing frameworks, or did we fail in some basic assumption about
> > how testing should work, that we find ourselves needing to reinvent unit
> > testing?
> 
> I don't have much experience with other widely used test frameworks,
> but this does seem reasonable.  The primitive "test" tests some
> individual aspect of the program; the test "suite" aggregates the test
> results for a unit of the program.  We then want to be able to
> aggregate the results of the unit test suites for the whole program.
> Note that the way I've extended walnut_totals is (at least partially)
> recursive, so we won't need to keep adding layers.
> 
> Here's a couple of examples that have something similar to this extra
> layer.
> 
> The git test suite is structured rather like the conkeror tests in
> that groups of tests are written in files that can either be run
> individually, or all the tests can be run, with an aggregation of the
> results.
> 
> http://git.kernel.org/gitweb.cgi?p=git/git.git;a=blob;f=t/README;hb=master
> http://git.kernel.org/gitweb.cgi?p=git/git.git;a=blob;f=t/aggregate-results.sh;hb=master
> 
> 
> In cunit one writes tests in a suite and register the suites with a
> test registry.
> 
> http://cunit.sourceforge.net/doc/introduction.html#usage
> 
> regards, David.


Hi David,

As I think about this, I don't think that program-termination is really
the right time to output a testing summary.  It only appears to be the
right time because the typical way in which we run tests is with the -l
switch from the command line in a batch invocation of a fresh instance of
conkeror, but it is clearly not the right time if conkeror is not
immediately exited after testing.  Instead, I would propose a command-line
switch called -walnut-summary (or -walnut-summarize?), to give the user
true control over the time of summary generation, without introducing new
assumptions about the manner in which tests will be loaded.  What say you?

-- 
John Foerch
_______________________________________________
Conkeror mailing list
[email protected]
https://www.mozdev.org/mailman/listinfo/conkeror

Reply via email to