Move GraalVM job to JDK 11.

[mbien]

GraalVM 8 and 11 are EOL + PRs are blocked by the graal test job atm which runs on JDK 8 still.

lets try to upgrade to GraalVM 11 as intermediate step

meta issue #4904

[mbien]

20.3.2 and 22.3.3 don't have the python component and 21.1.0 gets a SIGSEGV, lets randomly try more versions

... can't reproduce SIGSEGV locally

what I've learned so far:

• graal already dropped JDK 11 support, but it would be good to move to JDK 11 first, since JDK 17 will likely require extra steps due to modules. Lets try to find a release which has all the components, ships JDK 11 and doesn't crash.
• 22.0.0.2 language component download fails completely
• not all releases come with all components, example error Component js requires to upgrade GraalVM Core to at least 21.3.3, which isn't great since 21.3.3.1 and 22.3.3 don't have python and 22.1.0 has no js, which means the last graal JDK 11 release which supports all tested languages would be 22.3.1 -> lets try that version
• components can easily take the VM down, esp the experimental ones

[mbien]

DebugAllBaseTest: JPDADebugger#getCurrentThread() is null when python is tested while js, r and ruby do seem to work on Graal 22.3.0 and 22.3.1.

update:
this is caused by a JVM crash on the other end of the debugging session, can be resolved by installing libxcrypt-compat on arch based systems. (watch out for this line: com.oracle.truffle.api.CompilerDirectives$ShouldNotReachHere: Unable to load native posix support library )

The CI timeouts can be solved by splitting the tests up into smaller units instead of iterating through all scripts in one test method. This makes it also easier to debug, esp when CI fails. In future this can be easily updated to use parameterized tests of junit 4+ or left as is.

PolyglotTest is working fine until ruby is loaded when the test app JVM locks one core to 100% until the test times out. Thread dumps showed that the main thread is still doing something while running the ruby script but I am not familiar enough with graal to figure out what it is (this might be a looping compiler task which is using the suspended main thread?). I ran this for ~15mins and the program did not progress.

Initially I thought this might have a similar cause as the issue above but recompiling the ssl binding using graalvm-ce-java11-22.3.1/languages/ruby/lib/truffle/post_install_hook.sh as recommended by the ruby component didn't resolve it.
-> The test is disabled for now.

[lkishalmi]

Looks good to me!

[mbien]

would be great if any of the graal gurus could take a look at this too @sdedic @entlicher @dbalek

[sdedic]

LGTM

[mbien]

thanks! squashing and merging.