On Thu, 2005-06-09 at 21:14 +0100, robert burrell donkin wrote: > On Thu, 2005-06-09 at 10:51 +1200, Simon Kitching wrote: > > On Wed, 2005-06-08 at 23:22 +0100, robert burrell donkin wrote: > > > > > > > ======================= other > > > > > > > > The src distro builds fine with Java1.5 on Linux (debian 3.1). > > > > > > > > Unfortunately, there are ?? unit test failures when using java 1.3.1: > > > > TEST org.apache.commons.betwixt.strategy.TestConversionFlavour FAILED > > > > TEST org.apache.commons.betwixt.schema.TestSchemaGeneration FAILED > > > > TEST org.apache.commons.betwixt.io.read.TestBindTimeTypeMapping FAILED > > > > TEST org.apache.commons.betwixt.io.read.TestMaps FAILED > > > > TEST org.apache.commons.betwixt.recursion.TestSharedIDGeneration FAILED > > > > TEST org.apache.commons.betwixt.derived.TestWriteClass FAILED > > > > > > > > They look like real failures; some of the output is shown below. > > > > > > not sure whether they are real failures or not. a lot of the tests are > > > *very* sensitive to changes in undocumented introspection symantics > > > (such as ordering of methods) and have had a history of failing on > > > different JVM's. i would have expected to have had some feedback from > > > users by now if there were major problems on 1.5. > > > > > > i'm inclined to push this release out as is since there are users who > > > need a formal release with the extra functionality urgently and worry > > > about 1.5 compatibility later. > > > > The info above shows that the tests *work* on 1.5, but *fail* on 1.3. > > knew i should have turned in earlier :) > > > As you say, these are probably due to method ordering and are problems > > with the tests rather than with the code. But it's hard to be sure. > > > > At the very least, I think the release notes need to state that this > > release may not work on java 1.3. > > i think thomas is right and probably this should be the 1.7 release > (rather than 1.6.1) which means a vote. so, everything will be delayed a > few days. which will probably give me a chance to take a look into this > issue (i hope).
all the test failures were due to changes to the algorithm used by sun to calculate default orders for property descriptors between 1.3.1 and 1.4. i've added code that should allow the tests to cope better. unfortunately, it's not a very satisfactory solution since the tests are now very hard to read. (there is a feature that's been needed for a while which would sidestep this issue but i didn't want to introduce code changes...) - robert --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
