RFR: 8310183: Update GitHub Actions to use boot JDK for building jtreg

[Christian Stein]

On Fri, 16 Jun 2023 08:45:01 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

> Wait, so why does it fix the bug? Is it a MSYS path conversion bug?

It does not fix the bug, it works around it. Something (in MSYS) failes to work (calling `javac`) that used to work without problem until some week ago (new version of MSYS?), and looking into the workflow step log, it worked (calling same `javac`) just a second before the failing call. I tried to reproduce it locally and also on a Codespaces instance, but never get the `Bad address` error to appear.

Do you have a better idea how to trace to the underlying issue on GHA?

>  If so, could we "just" `cp` everything from `JAVA_HOME_17_X64` to local path and try that?

I didn't try that out. Might work, too. Yet, there is an installation available at a local path: `bootjdk/jdk`.

> My concern with hooking jtreg to boot JDK would require jtreg to be buildable with STSes, like JDK 22 for eventual JDK 23 builds. Is there a plan/commitment for jtreg to do it?

Copied from https://github.com/openjdk/jtreg/blob/master/doc/building.md

> The script is intended to be run in a Unix-like shell, such as bash on Linux or Mac OS X, or with Cygwin, MSYS2, or WSL on Windows. At a minimum, you must either set the JAVA_HOME environment variable or specify the location of the JDK to be used to build jtreg with the --jdk command-line option. **It must be a recent build of JDK 11 or later.**

So yes, "JDK 11 or later" includes 22 and following releases.

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

PR Comment: https://git.openjdk.org/jdk/pull/14507#issuecomment-1594419317