Request for review (M): 6998541: JSR 292 implement missing return-type conversion for OP_RETYPE_RAW (original) (raw)

John Rose john.r.rose at oracle.com
Wed May 11 09:12:26 PDT 2011

On May 11, 2011, at 8:36 AM, Tom Rodriguez wrote:

This doesn't seem right. 2 becomes false but 1 becomes true?

JSR 292 weirdness; see javadoc for explicitCastArguments:

If T0 and T1 are primitives and one is boolean, the boolean is treated as a one-bit unsigned integer. (This treatment follows the usage of the bytecode verifier.) A conversion from another primitive type behaves as if it first converts to byte, and then masks all but the low bit.

http://cr.openjdk.java.net/~jrose/pres/indy-javadoc-mlvm-old/2011-0510/java/lang/invoke/MethodHandles.html

Easier and more stable to check the low bit than remember the original length of the scalar and take all the bits into account. Otherwise long->int->boolean does something different than long->short->boolean.

-- John

More information about the hotspot-compiler-dev mailing list