Re: Octave 2.9.12 available for ftp

[Paul Kienzle]

On Jun 15, 2007, at 3:44 AM, Thomas Weber wrote:

> Am Freitag, 15. Juni 2007 06:22:10 schrieb John W. Eaton:
> > On 14-Jun-2007, Paul Kienzle wrote:
> > | So try:
> > |
> > |    %!xtest assert()
> > |    %!xtest fail()
> > |
> > | and it should work.
> >
> > OK, sorry for not realizing this earlier.  I did this for the two
> > tests in set/ismember.m that are known to fail and also changed the
> > test/fntests.m script to report the total number of expected failures
> > encountered in the summary.
>
> How about telling the users that they might want to look at these 
> failures? As
> a sort of integrated TODO list? This obviously means we only use xtest 
> for
> bugs, not for stuff that fails because some library is not installed, 
> etc.

What about the tests for random number generators which fail 
occasionally?

If they are never looked at because they are expected failures then we 
should just drop them.  If we drop them, then such a problems such as 
the byte swap problem on PPC wouldn't show up.  Maybe just have one 
grossly overpowered test for rand and randn and leave it at that?

What should we do with the rest of the tests?  Would it be worthwhile 
to move them to a 'random_test_battery' function which checks the 
expect mean, variance, skewness and kurtosis for all the generators?  
This would have to be run by the user rather than the test framework.

        - Paul