On Thu, 13 May 2021 21:37:18 GMT, Kevin Rushforth <k...@openjdk.org> wrote:
>> src/java.desktop/macosx/native/libawt_lwawt/awt/AWTWindow.m line 1106: >> >>> 1104: { >>> 1105: JNI_COCOA_ENTER(env); >>> 1106: if (@available(macOS 10.12, *)) { >> >> @kevinrushforth said that since we set MIN_SDK (not sure of the exact >> variable name) to 10.12, that this is compiled down to a no-op .. which >> means it is useless and doesn't protect you from making the call on 10.11 >> So you might as well remove it. It won't prevent the crash that will happen >> on 10.11. >> @mrserb also pointed out people might then copy this pattern not realising >> it does not work, and there's a better way ... apparently ... > > Right. @johanvos discovered this fun fact about `@available` when he got a > crash report from a user. He filed > [JDK-8266743](https://bugs.openjdk.java.net/browse/JDK-8266743), which > describes this problem. > > The setting of minimum version of macOS is controlled by the > `-mmacosx-version-min` compile and link flag. The minimum version is defined > in > [make/autoconf/flags.m4](https://github.com/openjdk/jdk/blob/master/make/autoconf/flags.m4#L136) > and used in > [make/autoconf/flags-cflags.m4](https://github.com/openjdk/jdk/blob/master/make/autoconf/flags-cflags.m4#L555). > > One thing I don't know (and can't try, since I don't have access to a macOS > system that old) is whether the JDK will fail somewhere else anyway (e.g., if > they check for a minimum OS at start up). So this might be a moot point, but > as it stands, I think @mrserb is right that we should avoid this pattern. I > would probably just remove it, but you could decide to use something like > `respondsToSelector` (which is what I think Sergey was suggesting). Since OpenJFX does not have its own launcher and IIRC JDK only recently (JDK 17 b08 https://bugs.openjdk.java.net/browse/JDK-8260518 ) set the minimum to 10.12 it is possible that the submitter of the FX crash was using A JDK prior to that, in which case I am sure the Java Launcher would start up fine and you'd crash only when calling this code. So also I think very aguably *library* code has another reason to avoid this pattern. And verifying what happens on 10.11 might be best done with a launcher from JDK 17 b07 or later .. also @kevinrushforth you might want to add some of these thoughts to the FX bug. ------------- PR: https://git.openjdk.java.net/jdk/pull/3407