>
congratulations on this milestone reached.
Thanks!
> Regarding the build: jobs get stuck after success during "Recording test results"
> Can we be sure that this is unrelated?
I haven't made any changes to the maven or ant scripts, but I wouldn't be surprised either way. I'm hoping they were just transient temporary problems with the hudson machine. I'll retriggering the builds tonight to see if they produce a different outcome.
> Should we contact some releng expert to be sure we know what's happening?
Absolutely, but who? The last time the builds were locking up for the newindex branch, I posted on every mailing list and IRC channel I could think of and didn't get any help. Do you know anyone who could assist? It would be very helpful.
I eventually solved it last time by pushing some self-test code that logged stack traces if a unit test locked up. I did something similar for these builds and didn't see the stack traces in the log - that suggests that things may be locking up in the build scripts rather than in actual Eclipse code.
> Several newindex jobs are affected, whereas JDT/UI jobs still complete.
Yes, I agree that's worrying. I'll do what I can to investigate, but if you can suggest a releng expert that would be much better.
> Another question: seeing changes around name environments:
> could this create trouble for BETA_JAVA9 where we have different
> changes in this area?
The newindex branch pretty much contains a rewrite of JavaSearchNameEnvironment. If BETA_JAVA9 contains a lot of changes in this file, the merge will be complicated. The remaining changes to are fairly minimal.
Fortunately, it still contains the old implementation and uses a single line to flip between the new and old implementations. If the merge gets too hairy, the index should still work with the old implementation. It may make sense to have me and someone working on the BETA_JAVA9 branch to do some pair programming during the merge if it gets bad.
- Stefan