JDK-8231118

[Sergey Nazarkin]

Thanks, Alexey! I’ve looked at the patch and found aarch32 port is not affected by this bug. 

Notes for other reviewers: informative patch code is

src/hotspot/cpu/arm/sharedRuntime_arm.cpp
@@ -462,11 +457,13 @@
   // Pushing an even number of registers for stack alignment.
   // Selecting R9, which had to be saved anyway for some platforms.
   __ push(RegisterSet(R0, R3) | R9 | LR);
+  __ fpush_hardfp(FloatRegisterSet(D0, 8));
 
   __ mov(R0, Rmethod);
   __ mov(R1, LR);
   __ call(CAST_FROM_FN_PTR(address, SharedRuntime::fixup_callers_callsite));
 
+  __ fpop_hardfp(FloatRegisterSet(D0, 8));
   __ pop(RegisterSet(R0, R3) | R9 | LR);


Sergey Nazarkin




> On Mar 6, 2020, at 18:23, Aleksey Shipilev <shade at redhat.com> wrote:
> 
> See "8231118: ARM32: Math tests failures":
>  https://bugs.openjdk.java.net/browse/JDK-8231118
> 
> I suspect 8u port is also affected by this. I have no cycles to verify it, though, so maybe someone
> on this list wants to pick it up.
> 
> -- 
> Thanks,
> -Aleksey