Float Encoding

BasicJavaTypesMappingTest#testFloatMapping tests the serialization of Float.MAX_VALUE. The testing harness contains the following lines. The comments with "mtd" are mine and explain the problem:

// value Float.MAX_VALUE can have E38 or E+38, depending on impl
representation = representation.replace("E38", "E[\\+]?\\+38");
if (!(testValue instanceof Number)) {
    representation = Pattern.quote(representation);
    // quote numbers that do not fit into double decimal precision
} else {
    ...
}

There are several problems:

• After the replacement, the regex matches E+38 and E++38, but not E38 like the comment mentions.
• testValue is never assigned a value, i.e. it's always null, which means that !(testValue instanceof Number) is always true. This in itself is weird.
• Becuase the condition is always true, the regex is always quoted. That means that the replacement (which adds regex control characters like [, +, ]) become literal. This is also wrong.

Due to these issues, the TCK test does not accept a value of ...E38.

Start with BasicJavaTypesMappingTest#testFloatMapping and the shown testing-harness code. Reproduce the Float.MAX_VALUE case, inspect how testValue is initialized, and verify regex behavior for E38 and E+38. Done means the TCK accepts valid Float.MAX_VALUE representations without quoting regex controls.