[jdk11u] Backport the fix of JDK-8180450 to JDK 11?

[Andrew Haley]

On 10/12/24 10:27, Ziyi Lin wrote:

 > It seems Spark and other concurrent applications will gain significant performance improvement from the fix.

I'm not sure. Looking at the rules for backporting, any bug fix "which contributes to the stability, security or performance of the affected line should be considered." So this definitely fits in the rule of what is allowed in 11u, but of course that doesn't mean that it will be backported, merely that it will be considered.

As of today, this fix is only available in JDK 23. I do intend to backport it to JDK 21 soon, and would welcome help backporting it to 17.

 > Is there a plan to backport the fix to JDK 11 in the future?

There is no strong technical reason why some version of the fix would not work on 11u, but it will require work. It's a complicated fix that touches many files, in effect rewriting interface type checking in C2, the server compiler.

Given the age of JDK 11, it's questionable whether backporting this will be wise. As JDK releases age, they become brittle to change, and breakage becomes more likely.

I confess I'm a bit baffled by the Apache discussion at https://lists.apache.org/thread/ptbg70k3tfq4ljz4rbz0y5hqlvof8y31 . They seem to be saying that "Targeting Java >11 excludes Java 11 users" but I don't understand why that should be. Can't they build on JDK 11 and also run on 17?

Maybe we should open a conversation with Apache Spark.

-- 
Andrew Haley  (he/him)
Java Platform Lead Engineer
Red Hat UK Ltd. <https://www.redhat.com>
https://keybase.io/andrewhaley
EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671