We're seeing some monitoring failures (mainly jstat-related) in the tmtools suite when running with G1. There are three separate reasons for the failures:
a) the G1 monitoring data is wrong
b) the G1 monitoring data is correct but different/inconsistent compared to what the other GCs do
c) the tests do not know how to handle G1 when needing to get GC-specific data
This CR will try to fix the failures that fall into the a) category and as many failures that fall into the b) category as it's practical to do from within G1 (for the latter we'll have to "massage" the way G1 generates data to look more consistent with what the other GCs do). The rest will be covered by separate CRs.
The Comments section has more detailed information on specific test failures.