| [aspectj-dev] compiler status for 1.1.1 |
The following 5 bugs have been fixed in the core compiler for the 1.1.1 release: https://bugs.eclipse.org/bugs/show_bug.cgi?id=39458 https://bugs.eclipse.org/bugs/show_bug.cgi?id=39460 https://bugs.eclipse.org/bugs/show_bug.cgi?id=39462 https://bugs.eclipse.org/bugs/show_bug.cgi?id=39479 https://bugs.eclipse.org/bugs/show_bug.cgi?id=39993 There are two outstanding bugs that need some further evaluation of both their severity and the difficulty of implementing a small fix before a final 1.1.1 decision is made: https://bugs.eclipse.org/bugs/show_bug.cgi?id=39470 https://bugs.eclipse.org/bugs/show_bug.cgi?id=39959 There are two bugs that I'm unsure whether or not to address for 1.1.1. These bugs are both important things to fix, but they are larger fixes than the previous ones with the resulting increased risk of causing unexpected regressions. I think that the amount of testing we decide to do should be a major factor in whether or not to include these in 1.1.1. https://bugs.eclipse.org/bugs/show_bug.cgi?id=39711 This bug is about field-access join points matching synthetic fields. I believe the correct rule is that accesses of synthetic fields (based on the JVM's SYNTHETIC attribute) should not be considered AspectJ join points. This is a fairly simple rule to implement. The reason not to do this is that there might be programs out there that depend on synthetic field accesses being treated as join points and this change would break them. The reason to do it is that the longer we go without plugging this hole, then the greater the chance that there will be programs that depend on this "feature". https://bugs.eclipse.org/bugs/show_bug.cgi?id=37621 This is a JDT-core bug that has also been reported in ajc. It is a somewhat obscure bug, but it can result in completely legal bytecode that just does the wrong thing -- executing both branch of an if/else is possible. This bug and several much less severe compiler bugs are fixed in the 2.1.1 JDT-core release. Moving our underlying compiler from 2.1.0 to 2.1.1 is an easy task because of the small number of changes made to JDT-core for this maintenance task. I've completed all of the work for this move except for the last step of actually replacing the version of JDT-core we build the ajc release from. -Jim