RFR: 8273928: Use named run ids when problem listing tests [v3]
Kim Barrett
kbarrett at openjdk.java.net
Sat Sep 18 19:10:49 UTC 2021
On Fri, 17 Sep 2021 08:54:16 GMT, Stefan Karlsson <stefank at openjdk.org> wrote:
>> Today when you have multiple jtreg run sections in a test, each run gets an automated id that match the location in the file. For example:
>> Passed: gc/stringdedup/TestStringDeduplicationAgeThreshold.java#id0
>> Passed: gc/stringdedup/TestStringDeduplicationAgeThreshold.java#id1
>> Passed: gc/stringdedup/TestStringDeduplicationAgeThreshold.java#id2
>> Passed: gc/stringdedup/TestStringDeduplicationAgeThreshold.java#id3
>>
>> The path to the test plus the id can be used to problem lists the test. Say that id3 matches a run done with ZGC, and we need to problem list this test with ZGC, then the problem list would contain:
>> gc/stringdedup/TestStringDeduplicationAgeThreshold.java#id3
>>
>> The problem is when someone adds a new run section before that, then all the ids will be shifted and #id3 doesn't correspond to the ZGC run anymore. A similar problem occurs if two run sections are swapped.
>>
>> I propose that we refrain from using the automatically generated ids when problem listing tests. Instead we add explicit ids like this:
>> * @test id=Z
>>
>> An additional benefit of doing this is that it will be easier to see what was actually run:
>> Passed: gc/stringdedup/TestStringDeduplicationAgeThreshold.java#G1
>> Passed: gc/stringdedup/TestStringDeduplicationAgeThreshold.java#Parallel
>> Passed: gc/stringdedup/TestStringDeduplicationAgeThreshold.java#Serial
>> Passed: gc/stringdedup/TestStringDeduplicationAgeThreshold.java#Z
>>
>> I've gone through the tests in the HotSpot problem lists + some that affects ZGC. There are probably more tests that would benefit from getting explicit ids, but I started with a small set to begin with.
>
> Stefan Karlsson has updated the pull request incrementally with one additional commit since the last revision:
>
> Fix TestReferenceClearDuringReferenceProcessing
Changes requested by kbarrett (Reviewer).
test/hotspot/jtreg/ProblemList.txt line 104:
> 102: runtime/InvocationTests/invokevirtualTests.java#old-int 8271125 generic-all
> 103: runtime/jni/terminatedThread/TestTerminatedThread.java 8219652 aix-ppc64
> 104: runtime/os/TestTracePageSizes.java#no-options 8267460 linux-aarch64
I don't think the whitespace changes to align fields should be made. That's not the style of this file. And I think it shouldn't be the style, since maintaining the alignment will frequently introduce otherwise uninteresting diffs.
test/hotspot/jtreg/serviceability/dcmd/gc/HeapDumpCompressedTest.java line 86:
> 84:
> 85: /*
> 86: * @test id=SHenandoah
Lowercase the first "H".
-------------
PR: https://git.openjdk.java.net/jdk/pull/5557
More information about the hotspot-dev
mailing list