RFR: 8362203: assert(state == nullptr || state->get_thread_oop() != nullptr) failed: incomplete state

[Serguei Spitsyn]

On Mon, 14 Jul 2025 23:42:03 GMT, Leonid Mesnik <lmesnik at openjdk.org> wrote:

> The problem happens if post_early_vm_start is triggered.
> The fix is to complete initialization once threadObj become available.
> 
> I'vent observed that recompute_enable and hitting assertion might happens before this step. However it might be makes sense to check state in assertion if needed.

Changes requested by sspitsyn (Reviewer).

src/hotspot/share/prims/jvmtiExport.cpp line 678:

> 676:   EVT_TRIG_TRACE(JVMTI_EVENT_VM_START, ("Trg VM start event triggered" ));
> 677: 
> 678:   // The thread state might be incomplete if initialized in post_early_vm_start

Nit: Better to say "The JvmtiThreadState might be incomplete" to avoid confusion with other means of thread state.

src/hotspot/share/prims/jvmtiThreadState.cpp line 1045:

> 1043: 
> 1044: void JvmtiThreadState::update_thread_oop(JavaThread* thread) {
> 1045:   assert(thread->threadObj() != nullptr, "Should threadObj be already initialized");

Nit: Better just to say: "Sanity check".

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

PR Review: https://git.openjdk.org/jdk/pull/26303#pullrequestreview-3018124707
PR Review Comment: https://git.openjdk.org/jdk/pull/26303#discussion_r2206025058
PR Review Comment: https://git.openjdk.org/jdk/pull/26303#discussion_r2206028317