jtreg update

I've been working on some bug fixes and features for the OpenJDK Regression Test Harness, jtreg. This new version has just been released in the standard place on the OpenJDK website. Here is a quick summary of the changes.
  1. New option -noreport.

    An issue with the underlying JT Harness is that it always generates reports about all the tests in the test suite, instead of just the tests you selected to run. This means that there may be a noticeable pause when the tests have completed while jtreg writes out reports. For those folk that never read the reports anyway, you can now disable the reports with this new option.

  2. Fix -exclude option in -samevm mode.

    This combination of options did not work correctly. Now, it should.

  3. Improve version detection in jtreg.sh script.

    The wrapper shell script tries to verify it is running on JDK 1.5 or later. The version detection used to be confused if you were running JDK in headless mode, when a spurious message could appear on System.out. Such messages are now filtered out.

  4. Allow System.setIO calls in -samevm mode.

    jtreg tries to insulate tests from each other, so that side effects in one test cannot affect the functioning of another. Previously, tests were not allowed to call System.setIO when running in samevm mode. This prevented a number of tests from running in samevm mode. However, since jtreg only supports running one test at a time, there is no reason to prohibit this call.

  5. Fix the initialization of jtreg-specific system properties for samevm tests to use a space separated list, not the default List.toString() format ("[a,b,c]").

    It was an oversight that a suitable toString() method was not provided on an internal class, resulting in the default toString() for a List being used, instead of the intended format of a space-separated list.

  6. Set the classpath for -samevm tests to include standard jtreg directories, the same as for -othervm tests.

    This is to reduce any unnecessary differences between othervm mode and samevm mode.

  7. Enable XML report via system property.

    While some folk don't want reports at all, other folk want more reports. The system property javatest.report.kinds can be used to select which report types are generated by the underlying JT Harness report facilities. The property can be set to a space separated list of any of the values "text", "html", "xml". The default is "text html". The XML format will be useful for a possible upcoming new plugin for Hudson.

  8. Reduce false positives when setting keywords from test description.

    Some users reported problems with -ignore:quiet and -k:!ignore incorrectly ignoring tests with -XDignore.symbol.file=true. This fixes that issue.

  9. Recategorize the -e option in the "Main" group instead of the "JDK" group.

  10. Update FAQ with reference to -e as a way to override the PATH setting.

  11. (jtdiff) Add convenience 'jtdiff' script.

  12. (jtdiff) Refactor and add new "super-diff" mode.

    "Super-diff" mode is a way of analyzing a matrix of test runs, organized by platform and date. It runs a standard jtdiff for each row in the matrix, thus comparing test results across platforms for each date, and for each column in the matrix, thus comparing the history of the test results for each platform. See jtdiff -help -super for more details.

Comments:

Post a Comment:
Comments are closed for this entry.
About

Discussions about technical life in my part of the OpenJDK world.

Search

Archives
« April 2014
SunMonTueWedThuFriSat
  
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
   
       
Today