Description

I noticed some sporadic failures of the new comptask test (#1293) on JDK8.

The reason for this failure is that sometimes C2Compiler::compile_method;java/lang/Thread.<init> is not found in the stacktrace. I had a look at the code, and noticed that Thread.<init> was relatively simple in JDK8, while some complexity was delegated to Thread.init. Indeed, I get this from -XX:+PrintCompilation:

    278  106    b  3       java.lang.Thread::<init> (49 bytes)
    278  107    b  4       java.lang.Thread::<init> (49 bytes)
    279  106       3       java.lang.Thread::<init> (49 bytes)   made not entrant
    279  108    b  3       java.lang.Thread::init (214 bytes)
    280  109    b  4       java.lang.Thread::init (214 bytes)
    283  108       3       java.lang.Thread::init (214 bytes)   made not entrant

Example failure here.

Related issues

#1293

Motivation and context

The test should not be flaky.

How has this been tested?

GHA

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.