RFR: 8265750: Fatal error in safepoint.cpp after backport of 8258414

[Doerr, Martin]

Backout change looks good.
jdk-fix-request label should get added.

Best regards,
Martin

> Am 23.04.2021 um 19:34 schrieb Jaroslav Bachorík <jaroslav.bachorik at datadoghq.com>:
> 
> On Fri, Apr 23, 2021 at 7:22 PM Severin Gehwolf <sgehwolf at redhat.com> wrote:
>> 
>>> On Fri, 2021-04-23 at 18:57 +0200, Jaroslav Bachorík wrote:
>>> In 16 and 17 this crash is not happening.
>> 
>> Why?
> 
> I assume because the safepoint assumptions are slightly different?
> The fix for jdk head was pushed more than a month ago. The push was
> done by Markus Gronlund and I assume he wouldn't have pushed that
> without going through the whole shebang of Oracle internal test rigs.
> The backport for 16u is identical to 17. All the affected code in JFR
> is identical. Although there still might be a chance for error the
> same test that failed in 11u did pass for 16 as a part of the
> pre-commit GH actions.
> 
> AFAICT, the issue we are talking about was filed as a reaction to
> failure observed in 11u. The adjustments for the differences between
> JDK head and 11u are the most likely the cause of this failure. There
> is no indication this should apply to any other version than 11u
> unless I am missing something.
> 
> Cheers,
> 
> -JB-
> 
>> 
>> Thanks,
>> Severin
>> 
>>