[8u] RFR 8236179: C1 register allocation error with T_ADDRESS

[Aleksey Shipilev]

On 3/25/21 11:49 AM, Andrew Haley wrote:
> On 3/25/21 9:45 AM, Aleksey Shipilev wrote:
>> I agree with Felix, it is low-risk enough to consider for 8u292 rampdown.
> 
> Sure. I was just baffled that this is the first time we've noticed this
> bug in almost ten years. I guess if it's triggered by Shenandoah that
> explains it.

Yeah, AFAICS, you need to be running on AArch64, with Shenandoah actually entering the LRB slowpath 
under active GC, with C1 actually spilling the address argument on stack (?) due to a register 
pressure (?). This is pretty hard to come by unless you are (un)lucky.

I have not been able to reproduce it just now.

Well, maybe that's an argument to leave it for 8u302, actually.

-- 
Thanks,
-Aleksey