Um… when I said "remove the assert in print visitor" I thought you should remove it so we can see what parse tree is being generated, and then figure out what's wrong. The assert is otherwise valid and should be preserved.
I was also thinking of adding counters for printing with --log=time or somesuch (e.g. I was curious how many functions were skipped in parser). Maybe we'd need a low-level facility for that. On Sep 8, 2014, at 5:40 PM, Marcus Lagergren <[email protected]> wrote: > Please review: > > http://cr.openjdk.java.net/~lagergren/8057703/webrev/ > > https://bugs.openjdk.java.net/browse/JDK-8057703 > > Cleaned up Friday’s work and verified that no empty classes are installed > anymore. More to do bytecode generation wise. > Looks a lot better - removed the unnecessary mark used code generation phase. > > /M
