On 8/21/06, Henri Yandell <[EMAIL PROTECTED]> wrote:
On 8/21/06, Henri Yandell <[EMAIL PROTECTED]> wrote: > On 8/16/06, Henri Yandell <[EMAIL PROTECTED]> wrote: > > On 8/16/06, Gary Gregory <[EMAIL PROTECTED]> wrote: > > > FWIW, here are some test results for test.time I have: > > > > > > Compiled with Sun Java 1.4.2_12: > > > > > > - test.time has 1 failure with Sun Java 1.3.1_15 > > > - test.time passes with Sun Java 1.4.2_12 > > > - test.time passes with Sun Java 1.5.0_08 > > > > > > Doing the same above with the code compiled with Sun Java 1.3.1_15 yield > > > the same results. This is all on Windows XP SP2 and Ant 1.6.5. > > > > Wish I could get JDK 1.3 to work on a Linux box. Both my Debian Sarge > > (hardly a new distro) and Ubuntu laptop have glibc errors. Must grab > > the IBM version and see how it goes. > > Grabbed the IBM SDK for 1.3 ....Compiling under 1.4.x and 1.5 and then running ant test gives: test.builder: [junit] Running org.apache.commons.lang.builder.BuilderTestSuite [junit] Tests run: 261, Failures: 8, Errors: 0, Time elapsed: 0.131 sec [junit] Test org.apache.commons.lang.builder.BuilderTestSuite FAI test.time: [junit] Running org.apache.commons.lang.time.TimeTestSuite [junit] Some date rounding tests not run since the current version is 1.3.1 [junit] Tests run: 55, Failures: 10, Errors: 0, Time elapsed: 5.255 sec [junit] Test org.apache.commons.lang.time.TimeTestSuite FAILED
Digging into these 1.3 runtime bugs: One of the time ones is the one you reported Gary. It's to do with http://issues.apache.org/jira/browse/LANG-59 and is linked to what we think is a bug within the JDK (really ought to try that out with Java 6 and see if it's still there). Looks like things were worse in the 1.3 JVM. The other failures within the time package are a timezone problem in the DateUtilsTest assertDate method. I'm 7 hours off of GMT, so the values are 7 hours off. Probably not a problem when running under GMT. The IBM date code is probably the same as the Sun code. The Builder package errors are linked to two types of error. A bunch of HashCodeBuilder errors in which the hashcode created is not the same as we expect; and the second set are to do with ToStringBuilder returning object attributes in a different order than expected. Both are explained if the reflection libraries are returning things in a different order than the Sun libraries. So a whole bunch of bugs, but I don't see much we can do about any of them. Maybe we could add alphabetical sorting to the attribute names for the builder classes if that turns out to be the case. Not sure we care if hashcodes and toStrings are different between JVMs. Hen --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
