RFR: 8299683: [S390X] Problems with -XX:+VerifyStack [v2]

[Martin Doerr]

On Tue, 31 Jan 2023 08:45:04 GMT, sid8606 <duke at openjdk.org> wrote:

>> src/hotspot/cpu/s390/sharedRuntime_s390.cpp line 2693:
>> 
>>> 2691: 
>>> 2692:   // let the unpacker layout information in the skeletal frames just allocated.
>>> 2693:   offs = __ offset();
>> 
>> Shouldn't we use the offset from above instead of the current one? (PPC64 does that.) I think this line should get removed.
>
> I have tried that using same offset from above instead current one but it's not  pointing to correct PC for oop Maps.

Setting last_Java_pc = pc() + current offset looks weird. Could it be that this breaks something which prevents the failure from happening? E.g. pointing to somewhere outside the deopt blob?

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

PR: https://git.openjdk.org/jdk/pull/12161