Sisyphus schrieb:
----- Original Message -----
From: "Jan Dubois"
This happens because the test exits with a non-zero exit code.
Unfortunately a lot of the libwin32 regression tests are in a
pretty sorry state. A few of them are not even proper test but just
sample programs.
I'll try to fix this properly, but a quick workaround for you would be
to add
exit unless Win32::IsAdminUser();
after the 4th test of Win32API/t/test.t.
Yes, I've amended the script so it doesn't fail, and then the tests continue
until one of the netresource tests fail. (When I fix that the tests run to
completion.) But what I find (mildly) annoying is *not* that the tests fail,
but that a failure prevents the remaining test suites from being run. I
thought there would be a way of easily fixing that, but when I asked (on the
MakeMaker list) how that behaviour might be averted, Schwern replied:
"You can try overriding test_via_harness() so it runs all the tests in one
shot <insert hand waving here>.
Or you can disolve the bundle."
Dissolving the bundle is a workaround, rather than a solution - and I could
well understand if you didn't want to do that (though you would have to
unbundle the tests only). The simplest workaround is probably to provide
tests that don't fail :-)
Some tests will also fail on non-english Windows and on
a non-existant Excel.
Patches greatfully accepted :)
Time to change to use Test and Test::More and either make those stupid
tests non-critical, or really override test_via_harness()
My preference would be to see this undesirable behaviour in relation to
bundles fixed - but that's not really an *on-topic* issue for this group.
--
Reini Urban
http://phpwiki.org/
