Valhalla JTREG test failures #13182
Comments
tajila
added
comp:vm
project:valhalla
|
In order to run the tests I had to:
|
|
This issue can be broken up into smaller issues to address the specific failures |
|
Looks like some cases should be captured by verifier, but they are not. |
Yes, I would look at those issues last since the spec isn't fully finalized. The first priority are failures relating to field layouts or initialization since that is the most settled part of the spec. |
|
As mentioned in the meeting today, I discovered some layout differences between OpenJ9 and RI on Atomic VT and volatile VT fields. More details can be found here: #13130 (comment). Result of some tests here could be affected by such difference. |
pshipton
modified the milestones:
Release 0.31 (Java 18),
Release 0.32 (Java 8, 11, 17, 18) Apr refresh
pshipton
modified the milestones:
Release 0.32 (Java 8, 11, 17, 18) Apr refresh,
Tracking
|
As the verifier change is already merged, we can re-run these tests. @ehrenjulzert |
|
When running using The following tests are passing on openj9: All the tests are passing on RI. |
|
Talked to Ehren to also try with flattenning enabled. Some tests might pass with with flattenning enabled. |
|
When flattening is enabled (with The following command was used to run the tests with |
Correctly generate method signatures for lambda functions that take Q type arguments For eclipse-openj9#13182 (fixes error in WithFieldAccessorTest) Signed-off-by: Ehren Julien-Neitzert <[email protected]>
Some entries in the FCC have their clazz field set to null, these null classes should not be dereferenced For eclipse-openj9#13182 (fixes error in UninitializedInlineFieldsTest, as well as other tests) Signed-off-by: Ehren Julien-Neitzert <[email protected]>
|
@ehrenjulzert can you update in #13182 (comment) at the beginning of each test if it has been fixed and is now passing ? sth like: |
I updated it, although I decided to format it differently (I listed the passing and failing tests in two separate code blocks). I figured this would make it easier to compare to in the future (since when the tests are run, this is the format that the results are printed in). |
- Create VM_ValueTypeHelpers::copyFlattenableArray function and call it instead of VM_ArrayCopyHelpers::referenceArrayCopy when copying flattened or primitive arrays - Throw a NPE instead of an ASE in doReferenceArrayCopy when an attempt is made to copy a null object into an array of primitive value types for eclipse-openj9#13182 Signed-off-by: Ehren Julien-Neitzert <[email protected]>
- Create VM_ValueTypeHelpers::copyFlattenableArray function and call it instead of VM_ArrayCopyHelpers::referenceArrayCopy when copying flattened or primitive arrays - Throw a NPE instead of an ASE in doReferenceArrayCopy when an attempt is made to copy a null object into an array of primitive value types for eclipse-openj9#13182 Signed-off-by: Ehren Julien-Neitzert <[email protected]>
- Create VM_ValueTypeHelpers::copyFlattenableArray function and call it instead of VM_ArrayCopyHelpers::referenceArrayCopy when copying flattened or primitive arrays - Throw a NPE instead of an ASE in doReferenceArrayCopy when an attempt is made to copy a null object into an array of primitive value types for eclipse-openj9#13182 Signed-off-by: Ehren Julien-Neitzert <[email protected]>
|
I reran these tests locally (mac) and will start going through them. The RI passes all except field_layout/NullMarkersTest.java. OpenJ9 is failing the following runtime tests. OpenJ9 passing tests: |
|
I created a draft pr in the extensions repository that includes a script to run these tests, some minimal changes to get them working, and an exclude list of the failures that I'm working on. |
|
After a first pass I excluded 4 tests that are not compatible with OpenJ9 and was able to get a few more passing by removing checks for hotspot specific error messages. There are 22 remaining failures, see ProblemList-hotspotjtreg.txt for more details. |
|
The following tests are still failing. Some of them are not tests that I would expect to pass with J9 (for example c1 tests) however they are not failing in a way I would expect (crashing or unexpected exception) so I've left them on the list to investigate. compiler/valhalla/inlinetypes/TestArrayCopyWithOops.java #20522 Try again once extensions repo is updated for #20372 These tests expect behavior that is inconsistent with https://openjdk.org/jeps/8316779. Ignore until this is resolved. |
The following are the jtreg tests failures. They can be found in https://github.com/openjdk/valhalla/tree/lworld/test/hotspot/jtreg/runtime/valhalla/inlinetypes
JT Harness : Tests that failed
The text was updated successfully, but these errors were encountered: