From: <tho...@us...> - 2014-04-01 20:14:28
|
Revision: 8023 http://sourceforge.net/p/bigdata/code/8023 Author: thompsonbry Date: 2014-04-01 20:14:25 +0000 (Tue, 01 Apr 2014) Log Message: ----------- Partial fix for #871. This addresses the most likely code path in which we would fail to clear the interrupt status of a thread. However, the PipelineJoin runs on a worker thread in the Journal's executor service. That executor service should clear the interrupt status of the worker thread before assigning another task. Therefore, this does not provide an explanation for how an interrupt would escape scope of the invocation context in which the interrupt occurred. Thus, it does not provide an explanation for the behavior reported on that ticket. Modified Paths: -------------- branches/BIGDATA_RELEASE_1_3_0/bigdata/src/java/com/bigdata/bop/join/PipelineJoin.java Modified: branches/BIGDATA_RELEASE_1_3_0/bigdata/src/java/com/bigdata/bop/join/PipelineJoin.java =================================================================== --- branches/BIGDATA_RELEASE_1_3_0/bigdata/src/java/com/bigdata/bop/join/PipelineJoin.java 2014-04-01 20:12:42 UTC (rev 8022) +++ branches/BIGDATA_RELEASE_1_3_0/bigdata/src/java/com/bigdata/bop/join/PipelineJoin.java 2014-04-01 20:14:25 UTC (rev 8023) @@ -1781,7 +1781,7 @@ if (bindex++ % 50 == 0) { // Periodically check for an interrupt. - if (Thread.currentThread().isInterrupted()) + if (Thread.interrupted()) throw new InterruptedException(); } This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site. |