[8u] Apache Spark workloads failures: no mismatched stores, except on raw memory

[Severin Gehwolf]

On Fri, 2019-10-25 at 14:57 +0200, Roland Westrelin wrote:
> > I don't understand, here it is in 8u:
> > 
> >   8140309: [REDO] failed: no mismatched stores, except on raw memory: StoreB StoreI
> >   Summary: Mismatched stores on same slice possible with Unsafe.Put*Unaligned methods
> >   Reviewed-by: kvn, thartmann
> > 
> >   http://hg.openjdk.java.net/jdk8u/jdk8u-dev/hotspot/rev/0ffee573412b
> 
> Ah. That piece was not backported:
> 
> @@ -2393,7 +2400,8 @@
>               st->Opcode() == Op_StoreVector ||
>               Opcode() == Op_StoreVector ||
>               phase->C->get_alias_index(adr_type()) == Compile::AliasIdxRaw ||
> -             (Opcode() == Op_StoreL && st->Opcode() == Op_StoreI), // expanded ClearArrayNode
> +             (Opcode() == Op_StoreL && st->Opcode() == Op_StoreI) || // expanded ClearArrayNode
> +             (is_mismatched_access() || st->as_Store()->is_mismatched_access()),
>               "no mismatched stores, except on raw memory: %s %s", NodeClassNames[Opcode()], NodeClassNames[st->Opcode()]);
>  
>        if (st->in(MemNode::Address)->eqv_uncast(address) &&

Sounds to me we should file an 8u-only bug to get it included. Would
that be about right?

Thanks,
Severin