commit | d4275eb585156ed02f52d186a0cfb9eb8fa86819 | [log] [tgz] |
---|---|---|
author | John Ross <jwross@us.ibm.com> | Fri Nov 22 10:50:08 2013 -0600 |
committer | John Ross <jwross@us.ibm.com> | Fri Nov 22 12:10:30 2013 -0600 |
tree | 307136fa30eba7d79a80a65e254c23739f6014a1 | |
parent | 5bdaa4451d78c0dab69157169c53ee3ab40996c7 [diff] |
Bug 421487 - Check for null (again) when unwinding the thread local coordination stack. The previous fix still had a small window of opportunity for an NPE to occur. Each call to peek results in orphaned coordination processing, so despite the initial check for coordinator.peek() != null, the subsequent calls to peek() could result in null if a coordination on the thread local stack became orphaned during that time.