Infinity/NaN not supported

[mrjameshamilton]

If you compile the following:

public class Test  {
    public static final float POSITIVE_INFINITY = Float.POSITIVE_INFINITY;
    public static final float NEGATIVE_INFINITY = Float.NEGATIVE_INFINITY;
    public static final float NaN = Float.NaN;
}

the disassembler produces the following:

version 1.8;
public class TestInf extends java.lang.Object [
    SourceFile "TestInf.java";
] {
    public static final float POSITIVE_INFINITY = ∞float;
    public static final float NEGATIVE_INFINITY = -∞float;
    public static final float NAN = NaNfloat;

    public void <init>() {
        line 1
            aload_0
            invokespecial java.lang.Object#void <init>()
            return
    }

    public static void main(java.lang.String[]) {
        line 8
            getstatic java.lang.System#java.io.PrintStream out
            ldc ∞float
            invokevirtual java.io.PrintStream#void println(float)
        line 9
            getstatic java.lang.System#java.io.PrintStream out
            ldc -∞float
            invokevirtual java.io.PrintStream#void println(float)
        line 10
            getstatic java.lang.System#java.io.PrintStream out
            ldc NaNfloat
            invokevirtual java.io.PrintStream#void println(float)
        line 11
            return
    }

}

This is because the conversion from number to string uses DecimalFormat here.

The assembler also cannot read this back in, since it cannot parse these constants (they aren't numbers).

[shunf4]

try my fork which fixed this https://github.com/shunf4/proguard-assembler

though DecimalFormat may still cause problems like not accurately expressing double/float (disassembing and then assembling back may get a different value, e.g. Double.MIN_VALUE and Double.MAX_VALUE), not a clever way to precisely serialize & print decimal numbers