RFR: 8244146: javac changes for JEP 306

[Joe Darcy]

On Mon, 3 May 2021 17:27:58 GMT, Brian Goetz <briangoetz at openjdk.org> wrote:

> 
> 
> I'd like to better understand what the future plan for VM treatment of the `ACC_STRICT` flag. It would be nice to have a plan to eventually reclaim this flag. Right now, it seems we're not setting the flag any more in the static compiler (and eliminating it from the reflective view?), but the VM still accepts the flag. Is there a plan for eventually warning on, and then rejecting, and then repurposing this valuable real estate?

For the purposes of javac and JEP 306, in the changeset the ACC_STRICT bit is no longer emitted for -source 17/--target 17. This also implies any strictfp-ness is lost from a reflective representation constructed from a class file, which includes annotation processing from a class file (as directly tested for in TestStrictfpRetention.java) and would include core reflection as well.

Some details of the VM handling of ACC_STRICT are still under discussion. A necessary condition to reclaiming the bit for other purposes is having JVM validation checks dependent on class file version since ACC_STRICT to mean strict floating-point evaluation is valid for class file versions 46.0 through 60.0. Presumably accepting ACC_STRICT for class files 46.0 through 60.0 is desirably for compatibility.

Warning of ACC_STRICT on version 61.0 and higher class files is possible.

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

PR: https://git.openjdk.java.net/jdk/pull/3831