On Fri, 5 Jun 2026 17:43:19 GMT, Coleen Phillimore <[email protected]> wrote:
>> src/hotspot/share/runtime/sharedRuntime.cpp line 2847:
>>
>>> 2845: }
>>> 2846:
>>> 2847: bool CompiledEntrySignature::check_supers_and_deoptimize(int arg_num)
>>> {
>>
>> compute_calling_conventions() checks the thread state vs the link_time
>> argument, but shouldn't this method, check_supers_and_deoptimize(), be
>> protected by an assert ensuring current thread state is _thread_in_vm?
>
> Yeah, I could move the assert down.
> And run the valhalla-comp-stress tests too. I ran local tests with
> -Xlog:continuations=trace:rt.log since freezing and thawing calls
> frame::describe a lot.
I moved the assert to check_supers_and_deoptimize() for thread_in_vm, but added
one that asserted not thread_in_native since that was the problematic case.
The valhalla-comp-stress test group passed.
-------------
PR Review Comment:
https://git.openjdk.org/valhalla/pull/2517#discussion_r3364891828
