When I run a test target, I only look at the bottom line: SUCCESS or
FAILURE. I do not think I would see a prominent message. The less so
because of the large amount of output that the layout engine tests
produce. I want to be able to assume that all test have been run. 

I think Andreas' solution comes down to skipping the tests as well,
which I do not prefer.

Therefore I prefer a combination of a) and d): Fail the layout tests
if there is no hyphenation support and factor out the hyphenation
dependent tests. Create a separate test target
junit-layout-no-hyphenation which skips the hyphenation dependent
tasks. Create a target junit-no-hyphenation, which skips the
hyphenation dependent tasks. In that way the user gets the tests that
he asks for.

Alternatively, the hyphenation dependent tests may be skipped based on
an ant variable no-hyphenation-test=true or false. That would either
invoke another disabled testcases list, or a mechanism like Andreas

But basically, why would anyone want to run regression tests without
hyphenation present?

Regards, Simon

On Thu, Oct 13, 2005 at 06:29:24PM +0200, Andreas L Delmelle wrote:
> On Oct 13, 2005, at 01:06, Manuel Mall wrote:
> >On Thu, 13 Oct 2005 04:26 am, Simon Pepping wrote:
> >>I think junit-layout should fail if hyphenation-present is not
> >>true. Now the layoutengine tests are silently skipped when
> >>hyphenation is not present.
> >>
> >Fair enough - we could
> >a) Just fail the whole layout test target
> >b) Still run the layout and then only the hyphenation dependent tests
> >will fail - not good as it will not continue after the first test
> >failure
> >c) Don't run the layout tests but give a prominent message that the
> >tests have been skipped because of missing hyph support
> >d) Factor out the hyph dependent tests into a separate test target and
> >skip that with a warning
> >
> >Personally I tend towards c)
> I'd say d), but not explicitly as another separate build-target. 
> Rather, the hyph-dependent testcases are filtered out 'by' the 
> junit-layout target --should become quite straightforward in light of 
> Jeremias' recent suggestion to change the disabled-testcases to XML 
> format. It will eventually require modification to the 
> LayoutEngineTestSuite to deal with the altered format anyway, so we 
> should be able to squeeze in an additional attribute.
> The value of hyphenation-present is passed into the TestSuite, which 
> then uses that value when constructing the disabled-testcase filter.
> Cheers,
> Andreas

Simon Pepping
home page: http://www.leverkruid.nl

Reply via email to