RFR: 8313395: LotsUnloadTest.java fails with OOME transiently with libgraal
Leonid Mesnik
lmesnik at openjdk.org
Tue Aug 20 19:42:03 UTC 2024
On Tue, 20 Aug 2024 18:50:14 GMT, Doug Simon <dnsimon at openjdk.org> wrote:
>> test/hotspot/jtreg/runtime/cds/appcds/dynamicArchive/LotsUnloadTest.java line 29:
>>
>>> 27: * @bug 8278602 8313395
>>> 28: * @summary Lots of classes being unloaded while we try to dump a dynamic archive
>>> 29: * @requires vm.cds & !(vm.compMode == "Xcomp" & vm.graal.enabled)
>>
>> Any reasons to don't use probelmlist for this test exclusion?
>> The 'requires' us usually for permanent test filtering.
>
> Permanent test filtering makes sense in this case since "the purpose of this test is neither to test libgraal nor Xcomp".
ough, I thought that it is temporary.
I don't think "the purpose of this test is neither to test libgraal nor Xcomp" is good enough reason to don't run test. There are a lot of tests that have not a purpose to tests graal/Xcomp or any other specific mode. However sometimes they are failing and finding product issues. The test dump archive and try to reuse it . I don't see why this shouldn't be tested with Graal + Xcomp in general.
If test requires more memory with Graal then it makes sense to split itge Graal-specific testcase with higher Xmx.
If test couldn't be fixed to be work with Graal (or any other particular mode) or you think it is unreasonably hard to fix it then it is fine to use requires. Please just add explanation in the bug or PR.
The another valid reason is to exclude specific configuration if there is no corresponding problemlist.
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/20648#discussion_r1723871417
More information about the hotspot-runtime-dev
mailing list