RFR: 8336912: G1: Undefined behavior for G1ConfidencePercent=0
Thomas Schatzl
tschatzl at openjdk.org
Wed Jul 31 08:10:35 UTC 2024
On Mon, 29 Jul 2024 11:11:37 GMT, Albert Mingkun Yang <ayang at openjdk.org> wrote:
>> Hello,
>>
>> please review this change to workaround an unintentional division by zero found by ubsan.
>>
>> The global option `G1ConfidencePercent` allowed zero as confidence for predictions which this change simply disallows: not sure what the response of the VM would be if the user told it that it should have absolutely no confidence (i.e. confidence is zero) in the predictions (not "very low" confidence).
>>
>> Testing: after this change, the ubsan failure goes away.
>>
>> Thanks,
>> Thomas
>
> Marked as reviewed by ayang (Reviewer).
Thanks @albertnetymk @sangheon for your reviews
-------------
PR Comment: https://git.openjdk.org/jdk/pull/20371#issuecomment-2259914277
More information about the hotspot-gc-dev
mailing list