[PATCH FOR REVIEW] 7194035: update tests for upcoming changes for jtreg
serguei.spitsyn at oracle.com
serguei.spitsyn at oracle.com
Mon Aug 27 10:09:27 PDT 2012
Jon,
The fix looks fine.
Thanks,
Serguei
On 8/24/12 4:51 PM, Jonathan Gibbons wrote:
> Currently, jtreg incorrectly confuses the concept of the /directory/
> in which the test's class will be written with the /classpath/ used to
> locate all of the test's classes, including any library classes. It
> provides env variable TESTCLASSES and system property test.classes,
> and tests use these to mean both a directory and a classpath.
>
> jtreg will be fixed to separate the two concepts, such that
> TESTCLASSES/test.classes will continue to mean the /directory/ in
> which the test's class will the written, and will introduce
> TESTCLASSPATH/test.class.path for the classpath used to locate all the
> test's classes.
>
> This change only affects a small number of tests, which use both
> @library and TESTCLASSES. A number of serviceability tests are affected,
> FAILED: sun/tools/jcmd/jcmd-big-script.sh
> FAILED: sun/tools/jcmd/jcmd-f.sh
> FAILED: sun/tools/jcmd/jcmd-help-help.sh
> FAILED: sun/tools/jcmd/jcmd-pid.sh
> FAILED: sun/tools/jinfo/Basic.sh
> FAILED: sun/tools/jmap/Basic.sh
> FAILED: sun/tools/jps/jps-m_2.sh
> FAILED: sun/tools/jps/jps-Vvml_2.sh
> FAILED: sun/tools/jstack/Basic.sh
> However, all the failures are caused by issues in shared library code,
> meaning that only that code needs to be updated.
>
> The fix is to update the tests where they are using
> TESTCLASSES/test.classes as a classpath.
>
> To ease migration onto the new jtreg, it is suggested that the tests
> check to see if the new variables are defined (i.e. new jtreg) and to
> fall back on the old variables if they are not defined (i.e. old
> jtreg). In shell terms, this can be written
> ${TESTCLASSPATH:-${TESTCLASSES}}
>
> The patch can be seen here:
> http://cr.openjdk.java.net/~jjg/7194035/webrev.00/
>
> See also
> http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=7194035
>
> -- Jon
More information about the serviceability-dev
mailing list