[ https://issues.apache.org/jira/browse/SYSTEMML-541?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15207571#comment-15207571 ]
Glenn Weidner commented on SYSTEMML-541: ---------------------------------------- The testReblockGroupedAggregate(org.apache.sysml.test.integration.functions.recompile.ReblockRecompileTest) appears to fail consistently similar to original description when using seed value of 296304710250949L. For example, the test failed when run under Eclipse and outside via mvn command line after replacing line 102 of ReblockRecompileTest.runReblockTest: //long seed = System.nanoTime(); long seed = 296304710250949L; Example of seed returned from System.nanoTime that allowed test to run successfully: long seed = 92558998986232L; Note this was done using local Windows development environment. > Investigate intermittent ReblockRecompile unit test failure > ----------------------------------------------------------- > > Key: SYSTEMML-541 > URL: https://issues.apache.org/jira/browse/SYSTEMML-541 > Project: SystemML > Issue Type: Bug > Components: Test > Reporter: Glenn Weidner > Assignee: Glenn Weidner > Priority: Minor > > Recent nightly build reported failure: > Failed tests: > ReblockRecompileTest>AutomatedTestBase.tearDown:1402 Detailed matrices > characteristics: > ! DML<->R # stored values in DML: 20 > ! DML<->R # stored values in R: 3 > ! DML<->R identical values(z=0): 17 > ! DML<->R wrong values(z=1.0E-10): 3 > ! DML<->R min error: 0.0 > ! DML<->R max error: 8.0 > This also was observed once during PR 77. -- This message was sent by Atlassian JIRA (v6.3.4#6332)