RFR: 8351958: Some compile commands should be made diagnostic [v4]

[Tobias Hartmann]

On Fri, 16 May 2025 16:01:43 GMT, Marc Chevalier <mchevalier at openjdk.org> wrote:

>> Done (despite being out of the window so it won't show as outdated).
>
> Actually... That changes the behavior quite a bit. Currently, if the unlock is not supplied, we get a warning and the blackhole option is ignored. See this test:
> https://github.com/openjdk/jdk/blob/46a12e781edcbe9da7bd39eb9e101fc680053cef/test/hotspot/jtreg/compiler/blackhole/BlackholeExperimentalUnlockTest.java#L68-L73
> The exit value of the failing case is still 0, using `error_buf` changes it into a failure (exit 1 + nothing is run). I think it's more explicit and probably better, but quite a change in the behavior.
> It also changes this part of the test:
> https://github.com/openjdk/jdk/blob/46a12e781edcbe9da7bd39eb9e101fc680053cef/test/hotspot/jtreg/compiler/blackhole/BlackholeExperimentalUnlockTest.java#L105-L115
> It would not be possible to silent the warning and continue as if nothing happened. Again, I think the current behavior is not great: currently, an option is silently ignored (or semi-silently, with a warning). As much as I don't approve this behavior, it seems to be like that on purpose since tests exist for it. Wdyt?

Ah right, I missed that. Let's revert this part of the change then.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/25150#discussion_r2095273760