On Thu, 19 Mar 2026 16:44:21 GMT, Tobias Hartmann <[email protected]> wrote:

>> I wondered about that. Isn't there a risk that the test becomes fragile? 
>> What would you pick?
>
> I think if the difference of the heap size with and without your fix is 
> large, we can just pick a large enough value for the assert that would then 
> trigger without your fix and still be large enough to account for 
> fluctuations.

To avoid timeouts, I decreased the number of iterations and the test runs with 
`-Xmx20M`. I think it reports `2MB` being used with my patch. So if memory 
usage increases for some reason, it's very likely to cause an OOME. Do you 
still think an assert is needed?

-------------

PR Review Comment: 
https://git.openjdk.org/valhalla/pull/2062#discussion_r2961416888

Reply via email to