RFR: 8332717: ZGC: Division by zero in heuristics
Aleksey Shipilev
shade at openjdk.org
Wed Jun 19 09:30:10 UTC 2024
On Tue, 18 Jun 2024 12:15:04 GMT, Erik Österlund <eosterlund at openjdk.org> wrote:
> Before we have run the first major collection, the "time to perform GC" is considered not stable. We missed once such check when computing how many old GC threads we should use compared to young GC threads, resulting in a division by zero when starting the first "warmup" major collection. This division by zero is detected by UBSan. I added a check for unstable times like we have in many other places, which I have verified removes the problem.
Looks okay.
I think you want to change fix-versions in JIRA, so that no backport is created when you push. Then we need to consider pulling this in JDK 21+.
-------------
Marked as reviewed by shade (Reviewer).
PR Review: https://git.openjdk.org/jdk/pull/19768#pullrequestreview-2127655364
More information about the hotspot-gc-dev
mailing list