openjdk 22, openjdk@21 21.0.2 (new formula)

[WangEdward]

Prepare for moving to OpenJDK 22 #166788

[ZhongRuoyu]

Thanks @WangEdward, can you do this (and all the migrations to openjdk@21) in #166788 instead? Otherwise we'd have two identical openjdk formulae before #166788 gets merged.

[WangEdward]

@chenrui333 @ZhongRuoyu
The commits in #166788 are too messy, so I moved everything here. Could you please add long build and CI-long-timeout to this? Thanks!

[ZhongRuoyu]

Squashed b749377 and 9f5cb0a.

[ZhongRuoyu]

These changes should help to fix the build failures. But please avoid pushing until CI finishes because it takes quite a long time to do one run. Thanks!

[ZhongRuoyu]

@WangEdward, we seek your cooperation by avoid pushing to this branch before CI is completed. Due to CI-build-dependents-from-source Pass --build-dependents-from-source to brew test-bot. , the tests take a very long time and we don't want to start over after hours of run like this: https://github.com/Homebrew/homebrew-core/actions/runs/8602702349/job/23584693893?pr=168294.

[WangEdward]

@WangEdward, we seek your cooperation by avoid pushing to this branch before CI is completed. Due to CI-build-dependents-from-source Pass --build-dependents-from-source to brew test-bot. , the tests take a very long time and we don't want to start over after hours of run like this: https://github.com/Homebrew/homebrew-core/actions/runs/8602702349/job/23584693893?pr=168294.

Sorry about that. I will make sure to push after all jobs are finished next time.

I thought that the test results were likely to be the same for arm and intel machines, and if there are any extra failing items for intel, it's not likely to be an OpenJDK-related problem. The build time on intel machines is extensively long (>10 hrs), and if we are eventually going to run the whole test for all the machines, why bother wasting time on testing it at the first build?

[ZhongRuoyu]

I thought that the test results were likely to be the same for arm and intel machines, and if there are any extra failing items for intel, it's not likely to be an OpenJDK-related problem.

Although that's likely the case in general, it also comes with the assumption that the OpenJDK build itself performs as expected. We are usually more cautious with compilers, considering how they might affect users and also the dependent formulae. (And OpenJDK is a bit special in that formulae built with it also depend on it at runtime.)

The build time on intel machines is extensively long (>10 hrs), and if we are eventually going to run the whole test for all the machines, why bother wasting time on testing it at the first build?

Because subsequent tests will not be run with CI-build-dependents-from-source Pass --build-dependents-from-source to brew test-bot. , once we have a complete list of what's broken.

[landsman]

Guys, can you please create an issue to describe what has to be done to introduce openjdk@22? I'm a bit lost in this communication.

[ZhongRuoyu]

At this moment all we need is patience. Many PRs are waiting for a long-timeout CI slot and we'll rerun CI here when the backlog is lighter.

[calvinit]

openjdk@21 21.0.3 has been released. Can we update this formula first? 22 non-LTS, I think can be dealt with later.

PS: JDK Releases.

[WangEdward]

openjdk@21 21.0.3 has been released. Can we update this formula first? 22 non-LTS, I think can be dealt with later.

PS: JDK Releases.

#170043

[ZhongRuoyu]

Thanks @WangEdward @SMillerDev @iMichka and everyone!

[github-actions]

@ZhongRuoyu has requested bottles to be published to this PR.

[calvinit]

openjdk@21 21.0.3 has been released. Can we update this formula first? 22 non-LTS, I think can be dealt with later.
PS: JDK Releases.

#170043

@WangEdward Hi guy, now this PR has been merged. Would you mind adding a new PR to update the latest openjdk 22.0.1? Thanks.

[ZhongRuoyu]

Would you mind adding a new PR to update the latest openjdk 22.0.1?

• openjdk 22.0.1 #173964