Github user paulk-asert commented on a diff in the pull request:
https://github.com/apache/groovy/pull/338#discussion_r64549052
--- Diff:
src/main/org/codehaus/groovy/transform/sc/transformers/BooleanExpressionTransformer.java
---
@@ -134,15 +134,19 @@ public void visit(final GroovyCodeVisitor visitor) {
// int on stack
} else if (top.equals(ClassHelper.long_TYPE)) {
MethodVisitor mv =
controller.getMethodVisitor();
- mv.visitInsn(L2I);
+ mv.visitInsn(LCONST_0);
+ mv.visitInsn(LCMP);
controller.getOperandStack().replace(ClassHelper.boolean_TYPE);
} else if (top.equals(ClassHelper.float_TYPE)) {
MethodVisitor mv =
controller.getMethodVisitor();
- mv.visitInsn(F2I);
+ mv.visitInsn(F2D);
--- End diff --
We probably could. In general, comparisons with floats are notoriously
ugly, e.g. see:
https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/
But given we are comparing against 0.0, it is probably an edge case that
isn't as bad as the general case (but see 'Infernal zero' in the aforementioned
link). A general recommendation is to use doubles instead of floats but I don't
really know for sure whether it does in this case. My rationale was to look at
what bytecode was produced from `assert floatValue != 0.0f` and then
replicating that - which is where the `F2D` and `DCMPG` came from.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
