> When using -Xcomp, the liveness of some objects the test allocates is more > precisely known, allowing the objects to be collected before the test > expects. This became an issue in the loom repo because it has changes that > result in a full GC when the codecache is swept. This is fixed by using > statics to reference the objects. Also, if a GC does happen, this seems to > get rid of the Survivor region, so the test was updated to no longer check > for it. > > I'm choosing to fix this in the jdk repo rather than the loom repo since it > is a latent bug that theoretically could occur even without the loom changes, > and also to help reduce the amount of changes to be reviewed when loom is > integrated into jdk.
Chris Plummer has updated the pull request incrementally with one additional commit since the last revision: Use Reference.reachabilityFence() instead of static field. ------------- Changes: - all: https://git.openjdk.java.net/jdk/pull/7257/files - new: https://git.openjdk.java.net/jdk/pull/7257/files/99235d74..c69d7411 Webrevs: - full: https://webrevs.openjdk.java.net/?repo=jdk&pr=7257&range=01 - incr: https://webrevs.openjdk.java.net/?repo=jdk&pr=7257&range=00-01 Stats: 7 lines in 1 file changed: 4 ins; 2 del; 1 mod Patch: https://git.openjdk.java.net/jdk/pull/7257.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/7257/head:pull/7257 PR: https://git.openjdk.java.net/jdk/pull/7257
