Let try this! /Staffan
> On 11 mar 2015, at 09:35, Jaroslav Bachorik <jaroslav.bacho...@oracle.com> > wrote: > > On 5.3.2015 11:37, Jaroslav Bachorik wrote: >> Please, review the following change >> >> Issue : https://bugs.openjdk.java.net/browse/JDK-8074041 >> Webrev: http://cr.openjdk.java.net/~jbachorik/8074041/webrev.00 >> >> This test fails very intermittently still and the failure is very hard >> to reproduce. Based on the symptoms the probable cause seems to be the >> asynchronous processing of a launched process stdout and stderr. >> >> The test is using ProcessTools.startProcess(...) to start the target >> process and process its output (stdout/stderr) to mark the occurrence of >> the expected string. When the target process has finished the test will >> check the mark to assert the correctness. However, it may happen that >> the assertion occurs in the time interval between the target process >> finishing (process.waitFor() returns) and the asynchronous stderr/stdout >> processors had the chance to dispatch the target process output to the >> processor checking for the occurrence of the expected test. >> >> The proposed fix is to wrap the Process instance into a wrapper >> enhancing the waitFor() method with the wait for the stdout/stderr >> processor being finished before proceeding. I ran the test in a tight >> loop for 1000 times after applying this patch without any failure. >> >> >> Thanks, >> >> -JB- >