The -r1 revision removed >=virtual/jdk-1.4 and added =virtual/jdk-1.4*. There is nothing in the ChangeLog as to why this was done. Given that -r0 compiled fine with jdk-1.5 I expect that -r1 should as well, so unless there is a good reason the depend should be reverted back to >=1.4.
-r0 wasn't generation-2 so compiling that with 1.5 would result in 1.5 bytecode - a way to break java packages depending on it :) It was never expected to be used with 1.5. But I agree that -r1 compiles fine with sun-jdk-1.5 and if it inherited java-ant-2, it would get build.xml rewritten to ensure 1.4 bytecode. There used to be a migrated -r1 ebuild in java-migrated-overlay which was doing that and worked fine.
Upstream uses 1.4, and to prevent any issues from cropping up, I set the depend to =1.4*. SWT is fairly complex, and some of the bugs are rather subtle. Until upstream switches to 1.5, Im keeping the depend on 1.4.
swt-3.2 for gtk supports embedding swing and awt components *only* under a 1.5 jdk. Eclipse (developed by the same upstream as swt) supports all features of J2SE 5.0 since eclipse-3.1m4. Upstream developers are definately using and testing against the 1.5 JDKs already.
(In reply to comment #3) > swt-3.2 for gtk supports embedding swing and awt components *only* under a 1.5 > jdk. Eclipse (developed by the same upstream as swt) supports all features of > J2SE 5.0 since eclipse-3.1m4. Upstream developers are definately using and > testing against the 1.5 JDKs already. > Reopening for re-evalution.
Fixed in CVS.