RFR: 8331208: Memory stress test that checks OutOfMemoryError stack trace fails [v2]

[Doug Simon]

On Thu, 2 May 2024 01:38:47 GMT, David Holmes <dholmes at openjdk.org> wrote:

>> Doug Simon has updated the pull request incrementally with two additional commits since the last revision:
>> 
>>  - don't perform XX actions for OOME when in scope of an InternalOOMEMark
>>  - rename SandboxedOOMEMark to InternalOOMEMark
>
> src/hotspot/share/jvmci/jvmciRuntime.cpp line 114:
> 
>> 112:         if (ex->is_a(vmClasses::OutOfMemoryError_klass())) {
>> 113:           CLEAR_PENDING_EXCEPTION;
>> 114:         }
> 
> Just an observation but the original code will clear all exceptions except for an "async" exception, which these days is only the InternalError thrown by unsafe-access-errors. But the new code will only clear OOME thus allowing the (as expected) InternalError to remain, but also any other VirtualMachineErrors that may have arisen e.g. StackOverflowError. I actually think this is more correct, but it does seem a change in behaviour that we may need to be wary of.

In the context of Graal, it doesn't really make much of a difference as the Graal stub that calls this runtime routine will clear all exceptions anyway. But yes, I think limiting the clearing here to OOME is better.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/18925#discussion_r1587281572