I think Richard's patch is fine, i.e., specify a locale to the test,
what we want to test is toPattern()'s logic, not the locale data or
something, and the specified locale is OK to test the logic. And I also
think maybe one locale is not enough, so we may want to include some
more exceptional locale, say, en_UK in this case, by which, we can try
to follow RI as possible. Loose the test by removing the assertion here
is dangerous, just like any other cases without clear spec.
Geir Magnusson Jr wrote:
What do you suggest then?
Paulex Yang wrote:
Geir Magnusson Jr wrote:
Then I guess we just fix the test.
Agree, just think we should not "fix" the test by removing the assertion.
geir
Paulex Yang wrote:
Tim Ellison wrote:
Geir Magnusson Jr wrote:
Now, on my windows box I got a clean build, no failures.... hm.
The test works on en_US locale and fails on en_UK. I'm guessing that
your machine is set up as en_US.
Richard offered a patch that sets the locale to en_US for all
MessageFormatTest-s, but I suggested that was not a suitable solution.
For now I suggest we remove the assertion, since it is beyond the spec
requirements for this type.
Tim,
Should we also consider the principle of "follow the RI" here? I think
it is a sample of unclear spec, and we should assert if our
implementation follows RI, i.e., the assert about return value of
toPattern() is necessary. The issue here is just that the testcase
assumes the return value is locale-independent, so I think Richard's
patch is OK. Further, from the test perspective, maybe(just maybe) test
case on only one locale is not enough to check Harmony's toPattern()
logic, tests on more different locale(especially the 'exceptional' case
like en_UK) are better.
Also FYI, I tried the original test case with RI on en_UK locale, and it
failed exactly as Harmony.
Regards,
Tim
Geir Magnusson Jr wrote:
This is nuts. We need to chase down the commit that broke this, and
reverse it. We can't have a broken build persisting this long.
geir
Tim Ellison wrote:
(1) Linux build/tests are passing again, but for some reason the
'BUILD SUCCESSFUL' note didn't go to the commit list?
(2) Windows build/test is still failing with:
Wrong full date pattern expected:<...full...> but was:<...long...>
junit.framework.ComparisonFailure: Wrong full date pattern
expected:<...full...> but was:<...long...> at
org.apache.harmony.text.tests.java.text.MessageFormatTest.test_applyPatternLjava_lang_String(MessageFormatTest.java:244)
at
java.lang.reflect.AccessibleObject.invokeV(AccessibleObject.java:205)
Regards,
Tim
---------------------------------------------------------------------
Terms of use : http://incubator.apache.org/harmony/mailing.html
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail:
[EMAIL PROTECTED]
---------------------------------------------------------------------
Terms of use : http://incubator.apache.org/harmony/mailing.html
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
Terms of use : http://incubator.apache.org/harmony/mailing.html
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
Terms of use : http://incubator.apache.org/harmony/mailing.html
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
--
Paulex Yang
China Software Development Lab
IBM
---------------------------------------------------------------------
Terms of use : http://incubator.apache.org/harmony/mailing.html
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
