On Wed, 25 Mar 2026 18:56:49 GMT, Tom Rodriguez <[email protected]> wrote:

> Maybe interpreter safepoints should only be installed for bytecodes for which 
> async delivery is ok?
I don't think we can do that because the interpreter safepoints are either for 
single stepping (safepoint_table where each bytecode calls 
InterpreterRuntime::at_safepoint()) so we want them all, or backwards branches 
and returns.  The return bytecodes are in the case of cannot throw an async 
exception because of generateOopMap::do_exception_edge().  Maybe not allowing 
async exceptions on those is causing the problem.

Yes, it's making this test fail more often.

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

PR Comment: https://git.openjdk.org/jdk/pull/30171#issuecomment-4129524938

Reply via email to