RFR: 8263108: Class initialization deadlock in java.lang.constant [v2]

[Jaikiran Pai]

On Tue, 16 Mar 2021 02:34:39 GMT, Jaikiran Pai <jpai at openjdk.org> wrote:

>>> If you and others think that we can ignore this case, then your proposed approach of using this lazy holder for initialization, IMO, is much cleaner and natural to read and I will go ahead and update this PR to use it.
>> 
>> For me, at least, the holder pattern is clearer. I'm happy with that approach.   ( I don't have an objection to the alternative, just a mild preference for the holder )
>
> Hello Chris, using the holder pattern sounds fine to me then. I've updated this PR accordingly. The test continues to pass with this fix.
> 
> Peter, I didn't get a chance to try out the `@Stable` for the previous approach but given that we decided to use this holder pattern, we will no longer need the performance tweaks.

Failure in Linux x86 build in GitHub Actions is unrelated to this change and looks like an environmental issue:

2021-03-16T02:35:22.3488438Z Err:59 https://dl.bintray.com/sbt/debian  Release.gpg
2021-03-16T02:35:22.3489341Z   502  Bad Gateway [IP: 35.161.90.47 443]
2021-03-16T02:35:30.2615937Z Reading package lists...
2021-03-16T02:35:30.2842714Z E: The repository 'https://dl.bintray.com/sbt/debian  Release' is no longer signed.

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

PR: https://git.openjdk.java.net/jdk/pull/2893