RFR(M): 8031752: Failed speculative optimization should be reattempted when root of compilation is different

[Roland Westrelin]

I found a bug in this already reviewed but not yet pushed change. The method that is referenced in the SpeculativeTrapData can be unloaded while the method that owns the SpeculativeTrapData remains live. This is possible with reflection (at least). Here is a bug fix for that with a test case. The webrev only includes the added change on top of what was already reviewed to make reviewing it easier. Or should I have it reviewed with another bug id?

http://cr.openjdk.java.net/~roland/8031752/webrev.02-03/

Roland.