[jdk18] RFR: 8279597: [TESTBUG] ReturnBlobToWrongHeapTest.java fails with -XX:TieredStopAtLevel=1 on machines with many cores

[Hao Sun]

On Wed, 12 Jan 2022 11:17:19 GMT, Hao Sun <haosun at openjdk.org> wrote:

> > > You can also set small value with `-XX:CICompilerCount=n' for this test. Increasing CodeCachecsize to 64M may help your case but we can have machines with core count even larger.
> > 
> > 
> > Yes. Agree. I think your solution would be better. Will test more and update soon. Thanks.
> 
> @vnkozlov I'm afraid it doesn't work either.
> 
> If we set CICompilerCount to a small value, e.g., 2, the test case
> 
> 1. passed on server/client builds as expected
> 2. but **failed** on non-compiler JVM build, e.g., zero, with the following error msg:  `CICompilerCount (2) cannot be greater than 0 because there are no compilers`
> 
> Because there are constraints for the valid values of CICompilerCount. See https://github.com/openjdk/jdk18/blob/master/src/hotspot/share/runtime/flags/jvmFlagConstraintsCompiler.cpp#L54

On a second thought, I suppose it might be okay for this test case to **FAIL** with **ZERO** JVM build.

1) this test case is located under `test/hotspot/jtreg/compiler/*`, and I think those test cases are designed to test JVM **WITH compilers**.
2) there exist a number of test cases where CICompilerCount is specified, e.g.,
  `test/hotspot/jtreg/compiler/loopopts/superword/TestNegBaseOffset.java
  test/hotspot/jtreg/compiler/classUnloading/methodUnloading/TestOverloadCompileQueues.java`
  And in my local test, these test cases **failed as expected** with ZERO JVM build.

Hence, my point is that we could follow your solution, i.e. setting `CICompilerCout` to a small value. @vnkozlov 
Please check the new commit.

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

PR: https://git.openjdk.java.net/jdk18/pull/93