Open Mesbah-Alam opened 6 years ago
Hi @chandrams
Since you had originally developed the use case for the modularity serviceloaders tests (if I am not mistaken), could you please shed some light on the following?
Two of the 3 Modularity Service loaders tests fail, consistently, both on hotspot and openj9 jdk9 and jdk10 sdks. I am wondering if the tests have become invalid, or if it is pointing to a legitimate issue.
Test 2
The first failure happens in Test 2, T2
(as coded in TestServiceLoaders2.java), where we are basically running the same test as T1 except here the requires and uses are added through addReads() and addUses() programmatically instead of adding them via requires and uses clauses in module-info.java.
As you can see in the error message in the description of the issue, it lists only one display service (DisplayServiceImpl4) instead of all 4, as listed in test1.
Test 3
I commented out test 2 for testing test 3 , and that too failed at the same validation as T2, but interestingly, this time we get 2 services in the 'actual' list, instead of 1 as seen in test2.
https://urv-jenkins.canlab.ibm.com:9095/view/SV_Tests/job/test.personal/256/console
STF 16:19:18.124 - Monitoring processes: T3
T3 JUnit version 4.12
T3 .Added readability edge via addReads() API from com.test.serviceloaders2 to displayService
T3 Added uses edge via addUses() API from com.test.serviceloaders2 to displayService
T3 DisplayServiceImpl3
T3 DisplayServiceImpl4
T3 E
T3 Time: 0.033
T3 There was 1 failure:
Have these two tests become invalid ?
The failing tests have been excluded from running in the build : https://github.com/AdoptOpenJDK/openjdk-systemtest/pull/138
Hi @Mesbah-Alam
As discussed, these issues were seen when the tests were written for which an internal bug was raised. But this was during initial days of modularity, we will need to validate the test scenario and expected test behavior and raise a bug if required.
@chandrams , I've dug up the internal RTC defect that you had raised on this issue against the JCL team in August 2016 (forwarded to you on Slack) and noticed that we have indeed had the exact failure since during early days. What was the source (modularity jep, etc) of the original use case, based on which this test was written? Could you point me to that?
@Mesbah-Alam
This is the modularity source link - http://openjdk.java.net/projects/jigsaw/spec/sotms/#services I had forwarded both the above link and RTC defect link during our chat on slack last month as well. Let me know if you need help understanding the use case. It is related to service loaders and automatic modules.
@chandrams, yes, I could use some help! Since you wrote it originally based on the spec, could you please explain the failure that has been happening? Was the test written correctly? Did it ever pass on any Oracle SDK?
Modularity service loaders test fails with following error on all platforms on both hotspot and openj9 jdk9 and jdk10 builds. This test was written based on old modularity implementations and likely has become invalid. The check at which the test fails may need to be updated/ removed.
https://ci.adoptopenjdk.net/view/System%20tests/job/openjdk9_hs_systemtest_s390x_linux/27/tapResults/