RFR: 8325444: GHA: JDK-8325194 causes a regression

[Magnus Ihse Bursie]

On Wed, 7 Feb 2024 19:50:51 GMT, Christoph Langer <clanger at openjdk.org> wrote:

> The [change](https://github.com/openjdk/jdk/commit/d1c82156ba6ede4b798ac15f935289cfcc99d1a0) for [JDK-8325194](https://bugs.openjdk.org/browse/JDK-8325194) broke get-jtreg because the way to determine the build jdk was not correct.
> 
> I then tried to use the bootstrap JDK with the correct wiring but it turned out that the build of JTReg fails with the current JDK, see [here](https://github.com/RealCLanger/jdk/actions/runs/7820130826/job/21334120478).
> 
> So I propose to fix it by going to the originally proposed solution using `$JAVA_HOME_17_<arch>`.

I don't like this approach. There must be better ways to achieve this, like inputting the correct value as input to the action, or setting it as a global gh variable. Where does even the `$JAVA_HOME_17_arm64` come from? Is it provided by GH? I don't recall setting variables with that weird case style in our GHA files.

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

PR Comment: https://git.openjdk.org/jdk/pull/17756#issuecomment-1933499696