Re: Patch: New Test File.getParent

[Wolfgang Baer <WBaer at gmx dot de>]

Hi Mark, David,

David Gilbert wrote:
> Mark Wielaard wrote:
...
> > Why do you have the following check at the start of the test?
> >
> >  
> >
> > >    SimpleTestHarness testharness = null;
> > >    try    {
> > >          testharness = (SimpleTestHarness) harness;
> > >    }    catch (ClassCastException cce)
> > >    {
> > >          harness.fail ("Harness not an instance of SimpleTestHarness");
> > >          return;
> > >    }
> > >   
> >
> >
> > I couldn't find anything in your test that needs anything not in the
> > TestHarness interface.

I have this from the jdk11.java test for java.io.File and its needed or
used to get the actual source directory where mauve is executed as
this is the absolute path of the empty file.

See test:
String srcdirstr = testharness.getSourceDirectory();

> Yes, please write tests to the TestHarness interface only.  This 
> matters, for example, in the Mauve test report that I run:
>
> http://www.object-refinery.com/classpath/mauve/report/
>
> ...because it is generated using gnu.testlet.runner.Mauve, and not 
> SimpleTestHarness.  So if your test class checks for SimpleTestHarness, 
> it will have to be excluded from the Mauve report, which would be a shame.

I assume this is the cause why your java.io.File test results are quite
different than mine. For me only 2 test fail in java.io.File and some
are not executed due to one of the two failures. So only 124 tests are
run instead of 141, whereas for your test (linked on the GNU classpath
homepage) only 44 passed and 18 failed.

> New tests in Mauve are always appreciated, so thanks for your efforts!

No problem. As you both a much more experienced - any clues how I can get this
directory used in my test and the jdk11 test without this trick ?

Regards,
Wolfgang