RFR: 8336410: runtime/NMT/TotalMallocMmapDiffTest.java Total malloc diff is incorrect. Expected malloc diff range

[Thomas Stuefe]

On Thu, 25 Jul 2024 11:12:40 GMT, Johan Sjölen <jsjolen at openjdk.org> wrote:

> Hi,
> 
> We've seen intermittent failures in this test because of the change being out of the pre-defined range. The largest "out of bounds" we've observed is 1.5MiB, so increasing the fudge factor to allow for 25MiBs more/less should be sufficient for us to not see this error for a while. This kind of adjustment is an unfortunate necessity when developing these kinds of tests.

Ping 

> Should we consider splitting the fudge factor into 2 - one for lower bound, the other for upper one?
> 
> I can see how the lower bound should be tighter, whereas the upper one might be harder to define and we might need to keep upping?

That sounds reasonable. 

In addition, consider running the test with C1 only, to remove C2 and its arenas out of the equation.

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

PR Comment: https://git.openjdk.org/jdk/pull/20326#issuecomment-2254381299