@jeffrey_analyst Profiling cpu+wall has been possible for quite some time:

RT @shipilev: Woot: now ships multi-JVM builds. You can pull a single JDK bundle, and then ask for a release/fastde…

@adwsingh

@spyced @forked_franz Could you please submit an issue at with environment details? You may also try the same without --cstack option. Thanks.

@dbg_nsk Да, конечно

@dbg_nsk Hooray! We have a winner! It's indeed "kill -11" with some extra tricks:

@kvadratyev Let's say the process is already started with "java -XX:-EnableDynamicAgentLoading You need to make a running Java app throw NullPointerException

@mbezoyan You may change JVM arguments or run shell commands, but debuggers/agents are not allowed. Changing the value of T to null in a debugger would be too trivial.

@barbartosh Any OpenJDK release from JDK 11 to JDK 22

@forked_franz @spyced 'asprof -e page-faults --cstack vm' will profile page faults and show complete stack traces with all Java, native, JVM and libc frames.

@nevgeniev @AndreyPechkurov @lemire @johnbind2 It's out of my hands: everything that links with JVMCI needs to be GPL

@ldubox Yes. Although launch4j is only for Windows, the same trick works on Linux and macOS.

@lemire In this statement, you are stepping on your own rake by blaming sampling profilers basing on a few random observations, while the actual problem is not in the tools at all.

@tagir_valeev @shipilev @intellijidea My eyes are bleeding every time I see Objects.hash in a code review. For me, it's literally *harder* to read :) In general, hashCode() should not have side effects and therefore should not allocate. It's not even an optimization but a code hygiene.

@tagir_valeev @romtsn String(byte[], int, int, int) constructor also does the trick. It's compatible with any Java version, despite having been deprecated for ages.

RT @rkennke: I‘m working on a whole new full-GC for Serial GC, because why not? 🤷‍♂️

Yeah, the biggest async-profiler release so far. Tons of new features and improvements!