Gtest: CreateVM is not called when isolated tests are run?

[Igor Ignatyev]

Dear Thomas,

I guess you are using TEST macro to define your tests. TEST is meant to be used for normal unit tests which does not depend on JVM, hence we do not call CreateVM, and your tests fail. you should use either TEST_VM or, if a test changes JVM state so it can’t be reused by other tests, TEST_OTHER_VM.

regarding LogStream tests, it’s a bug. these tests depend on JVM and should be changed to TEST_VM.

Thanks,
— Igor

> On Dec 7, 2016, at 1:52 PM, Thomas Stüfe <thomas.stuefe at gmail.com> wrote:
> 
> Hi all,
> 
> I am writing a metaspace-related gtest and was noticing my test to assert
> in Thread::current() because the test thread was not attached to the VM.
> When executed together with all other tests, it works.
> 
> Some other tests are affected as well, e.g. LogStream tests:
> 
> ./hotspot/variant-server/libjvm/gtest/gtestLauncher.exe -jdk ./images/jdk/
> --gtest_filter=LogStream*
> 
> [ RUN      ] LogStreamTest.from_log
> # To suppress the following error report, specify this argument
> # after -XX: or in .hotspotrc:  SuppressErrorAt=\runtime/thread.hpp:662
> #
> # A fatal error has been detected by the Java Runtime Environment:
> #
> #  Internal Error
> (c:\d031900\openjdk\jdk9-hs2\source\hotspot\src\share\vm\runtime/thread.hpp:662),
> pid=17512, tid=7260
> #  assert(current != 0LL) failed: Thread::current() called on detached
> thread
> 
> Did anyone else notice this behaviour?
> 
> I see that when I run all tests (when omitting the filter argument),
> Thread::create_vm is called, but not when the test is run isolated. So, the
> VM is not initialized and all but the most primitive tests will fail.
> 
> Is this intended or would this be a bug?
> 
> Kind Regards, Thomas