RFR: JDK-8009152 - A number of jtreg tests need review/improvement (original) (raw)

Gary Collins GARY.COLLINS at ORACLE.COM
Wed Mar 6 16:43:08 PST 2013

Request for Review

Changes to allow better test utilization of jtreg hotspot regression tests 1: REPLACED JAVA_HOME with TESTJAVA 2: REPLACED TESTJAVA with COMPILEJAVA Some tests needed tools like (javac,jar,path to native libraries) to run. NOTE: THIS IS NEEDED WHEN TESTJAVA is a JRE. If TESTJAVA is a JDK then COMPILEJAVA is not needed. NOTE: FALLBACK when COMPILEJAVA is not defined for JTREG is TESTJAVA (with implied JDK being tested) 3: Replaced most of the BIT_FLAG with TESTVMOPTS NOTE: This allows the tests where possible to run with multiple compilers and switches. Asked for by Quality team 4: Added TESTJAVACOPTS for all javac compile lines. Suggestion from tools folks. 5: Added a new shell env template. Phase 2 of hotspot test cleanup will utilize this file NOTE: This test_env.sh file will allow engineers to remove allot of the copy/paste of test setup from each shell testcase. 6: Fixed a gc testcase that was timing out on a 24 core box.. Simple fix , add a timeout. 7: I filed many bugs around native case issues. This changeset it not to address broken tests. Only to conform them in a unified way of running.
8: Fixing tests as defined above allows us to run the tests from a JRE. embedded use cases this is a stop gap. 9: In order to run jtreg with a JRE you need to define TESTJAVA (JRE - jdk/testjdk) & COMPILEJAVA (JDK - compilejdk) If TESTJAVA is a JDK no need to set COMPILEJAVA.
NOTE: Most shell scripts from the command line may fail with the following issues: TESTSRC not defined, TESTCLASSES not defined, COMPILEJAVA path not found because we don't check for them at runtime. #5 above will fix this in the next phase of improving these tests to be more robust. 

Fully tested on linux-i586 builds with server/client jvm.

Webrev: http://cr.openjdk.java.net/~collins/8009152/webrev-rev0/

Please let me know if you have any suggestions. Thanks Gary

More information about the hotspot-dev mailing list