run2000 / compactbeans Goto Github PK

First Java-8 based release.

This makes a smaller DescriptorData object, as each boolean is currently a separate 32-bit value. Combining them into a single int makes for a single 32-bit value. Makes it easter to justify a separate flag for isInDefaultEventSet().

Copy across more tests from OpenJDK test suite, add them to the Ant suite.

See http://hg.openjdk.java.net/jdk8/jdk8/jdk/rev/2ad01ecbaf90 and http://mail.openjdk.java.net/pipermail/beans-dev/2014-May/000098.html

On top of explicit bean info. This means merging of explicit/introspected bean info with additional bean info.

This Java 1.5ifies this part of the API.

This is the bit that introspects *BeanInfo classes to provide explicit bean information. Look at ways of making the descriptors immutable once they have been created by the introspector.

Some of these might be ported back to JDK8u40 or something, but haven't appeared yet.

For instance, 8034164, 8034085. See http://mail.openjdk.java.net/pipermail/beans-dev/2014-February/000073.html

From the BeanInfoFinder package, see if we still need the DEFAULT and DEFAULT_NEW constants. These would only ever find Sun internal BeanInfo classes, which would never be shipped (or found usefully) with Compactbeans.

Spin up a VM with various compact profile JREs set up. See how many tests work in compact profiles in practise. If there are failures, see how many can be adapted to pass.

Currently this is hard-coded to true.

This doesn't appear in DecscriptorData, as that only contains data from the FeatureDescriptor abstract class.

Options are:

1. Do nothing. Leave it as always true. This may be ok as it appears to be a design-time property.
2. Add it as a parameter to constructors. This adds yet more combinatorial variations to the existing constructors. Don't like this idea.
3. Add explicitly to DescriptorData with the understanding that the property only applies to EventSetDescriptor objects.
4. Add implicitly to DescriptorData as a special key in the attribute map. Similar to how Transient is handled in java.beans package.

The indexed property test reports a failure on the compact JREs (1.8.0 32-bit Debian), but passes okay on the full JRE (1.8.0_05 32-bit Windows). Investigate where the difference lies.

First go at updating from one OpenJDK release to another. Work out the kinks in doing this in the future.

Make sure nothing completely dumb is in the Javadoc.

Similar enough to the property change framework to be manageable.