gecko/js/tests/ecma/GlobalObject/15.1.2.3-1.js

442 lines
26 KiB
JavaScript
Raw Normal View History

/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* ***** BEGIN LICENSE BLOCK *****
* Version: MPL 1.1/GPL 2.0/LGPL 2.1
*
* The contents of this file are subject to the Mozilla Public License Version
* 1.1 (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
* http://www.mozilla.org/MPL/
*
* Software distributed under the License is distributed on an "AS IS" basis,
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
* for the specific language governing rights and limitations under the
* License.
*
* The Original Code is Mozilla Communicator client code, released
* March 31, 1998.
*
* The Initial Developer of the Original Code is
* Netscape Communications Corporation.
* Portions created by the Initial Developer are Copyright (C) 1998
* the Initial Developer. All Rights Reserved.
*
* Contributor(s):
*
* Alternatively, the contents of this file may be used under the terms of
* either the GNU General Public License Version 2 or later (the "GPL"), or
* the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
* in which case the provisions of the GPL or the LGPL are applicable instead
* of those above. If you wish to allow use of your version of this file only
* under the terms of either the GPL or the LGPL, and not to allow others to
* use your version of this file under the terms of the MPL, indicate your
* decision by deleting the provisions above and replace them with the notice
* and other provisions required by the GPL or the LGPL. If you do not delete
* the provisions above, a recipient may use your version of this file under
* the terms of any one of the MPL, the GPL or the LGPL.
*
* ***** END LICENSE BLOCK ***** */
gTestfile = '15.1.2.3-1.js';
/**
File Name: 15.1.2.3.js
ECMA Section: 15.1.2.3 Function properties of the global object:
parseFloat( string )
Description: The parseFloat function produces a number value dictated
by the interpretation of the contents of the string
argument defined as a decimal literal.
When the parseFloat function is called, the following
steps are taken:
1. Call ToString( string ).
2. Remove leading whitespace Result(1).
3. If neither Result(2) nor any prefix of Result(2)
satisfies the syntax of a StrDecimalLiteral,
return NaN.
4. Compute the longest prefix of Result(2) which might
be Resusult(2) itself, that satisfies the syntax of
a StrDecimalLiteral
5. Return the number value for the MV of Result(4).
Note that parseFloate may interpret only a leading
portion of the string as a number value; it ignores any
characters that cannot be interpreted as part of the
notation of a decimal literal, and no indication is given
that such characters were ignored.
StrDecimalLiteral::
Infinity
DecimalDigits.DecimalDigits opt ExponentPart opt
.DecimalDigits ExponentPart opt
DecimalDigits ExponentPart opt
Author: christine@netscape.com
Date: 28 october 1997
*/
var SECTION = "15.1.2.3-1";
var VERSION = "ECMA_1";
var TITLE = "parseFloat(string)";
var BUGNUMBER="none";
startTest();
writeHeaderToLog( SECTION + " "+ TITLE);
new TestCase( SECTION, "parseFloat.length", 1, parseFloat.length );
new TestCase( SECTION, "parseFloat.length = null; parseFloat.length", 1, eval("parseFloat.length = null; parseFloat.length") );
new TestCase( SECTION, "delete parseFloat.length", false, delete parseFloat.length );
new TestCase( SECTION, "delete parseFloat.length; parseFloat.length", 1, eval("delete parseFloat.length; parseFloat.length") );
new TestCase( SECTION, "var MYPROPS=''; for ( var p in parseFloat ) { MYPROPS += p }; MYPROPS", "", eval("var MYPROPS=''; for ( var p in parseFloat ) { MYPROPS += p }; MYPROPS") );
new TestCase( SECTION, "parseFloat()", Number.NaN, parseFloat() );
new TestCase( SECTION, "parseFloat('')", Number.NaN, parseFloat('') );
new TestCase( SECTION, "parseFloat(' ')", Number.NaN, parseFloat(' ') );
new TestCase( SECTION, "parseFloat(true)", Number.NaN, parseFloat(true) );
new TestCase( SECTION, "parseFloat(false)", Number.NaN, parseFloat(false) );
new TestCase( SECTION, "parseFloat('string')", Number.NaN, parseFloat("string") );
new TestCase( SECTION, "parseFloat(' Infinity')", Infinity, parseFloat("Infinity") );
new TestCase( SECTION, "parseFloat(' Infinity ')", Infinity, parseFloat(' Infinity ') );
new TestCase( SECTION, "parseFloat('Infinity')", Infinity, parseFloat("Infinity") );
new TestCase( SECTION, "parseFloat(Infinity)", Infinity, parseFloat(Infinity) );
new TestCase( SECTION, "parseFloat(' +Infinity')", +Infinity, parseFloat("+Infinity") );
new TestCase( SECTION, "parseFloat(' -Infinity ')", -Infinity, parseFloat(' -Infinity ') );
new TestCase( SECTION, "parseFloat('+Infinity')", +Infinity, parseFloat("+Infinity") );
new TestCase( SECTION, "parseFloat(-Infinity)", -Infinity, parseFloat(-Infinity) );
new TestCase( SECTION, "parseFloat('0')", 0, parseFloat("0") );
new TestCase( SECTION, "parseFloat('-0')", -0, parseFloat("-0") );
new TestCase( SECTION, "parseFloat('+0')", 0, parseFloat("+0") );
new TestCase( SECTION, "parseFloat('1')", 1, parseFloat("1") );
new TestCase( SECTION, "parseFloat('-1')", -1, parseFloat("-1") );
new TestCase( SECTION, "parseFloat('+1')", 1, parseFloat("+1") );
new TestCase( SECTION, "parseFloat('2')", 2, parseFloat("2") );
new TestCase( SECTION, "parseFloat('-2')", -2, parseFloat("-2") );
new TestCase( SECTION, "parseFloat('+2')", 2, parseFloat("+2") );
new TestCase( SECTION, "parseFloat('3')", 3, parseFloat("3") );
new TestCase( SECTION, "parseFloat('-3')", -3, parseFloat("-3") );
new TestCase( SECTION, "parseFloat('+3')", 3, parseFloat("+3") );
new TestCase( SECTION, "parseFloat('4')", 4, parseFloat("4") );
new TestCase( SECTION, "parseFloat('-4')", -4, parseFloat("-4") );
new TestCase( SECTION, "parseFloat('+4')", 4, parseFloat("+4") );
new TestCase( SECTION, "parseFloat('5')", 5, parseFloat("5") );
new TestCase( SECTION, "parseFloat('-5')", -5, parseFloat("-5") );
new TestCase( SECTION, "parseFloat('+5')", 5, parseFloat("+5") );
new TestCase( SECTION, "parseFloat('6')", 6, parseFloat("6") );
new TestCase( SECTION, "parseFloat('-6')", -6, parseFloat("-6") );
new TestCase( SECTION, "parseFloat('+6')", 6, parseFloat("+6") );
new TestCase( SECTION, "parseFloat('7')", 7, parseFloat("7") );
new TestCase( SECTION, "parseFloat('-7')", -7, parseFloat("-7") );
new TestCase( SECTION, "parseFloat('+7')", 7, parseFloat("+7") );
new TestCase( SECTION, "parseFloat('8')", 8, parseFloat("8") );
new TestCase( SECTION, "parseFloat('-8')", -8, parseFloat("-8") );
new TestCase( SECTION, "parseFloat('+8')", 8, parseFloat("+8") );
new TestCase( SECTION, "parseFloat('9')", 9, parseFloat("9") );
new TestCase( SECTION, "parseFloat('-9')", -9, parseFloat("-9") );
new TestCase( SECTION, "parseFloat('+9')", 9, parseFloat("+9") );
new TestCase( SECTION, "parseFloat('3.14159')", 3.14159, parseFloat("3.14159") );
new TestCase( SECTION, "parseFloat('-3.14159')", -3.14159, parseFloat("-3.14159") );
new TestCase( SECTION, "parseFloat('+3.14159')", 3.14159, parseFloat("+3.14159") );
new TestCase( SECTION, "parseFloat('3.')", 3, parseFloat("3.") );
new TestCase( SECTION, "parseFloat('-3.')", -3, parseFloat("-3.") );
new TestCase( SECTION, "parseFloat('+3.')", 3, parseFloat("+3.") );
new TestCase( SECTION, "parseFloat('3.e1')", 30, parseFloat("3.e1") );
new TestCase( SECTION, "parseFloat('-3.e1')", -30, parseFloat("-3.e1") );
new TestCase( SECTION, "parseFloat('+3.e1')", 30, parseFloat("+3.e1") );
new TestCase( SECTION, "parseFloat('3.e+1')", 30, parseFloat("3.e+1") );
new TestCase( SECTION, "parseFloat('-3.e+1')", -30, parseFloat("-3.e+1") );
new TestCase( SECTION, "parseFloat('+3.e+1')", 30, parseFloat("+3.e+1") );
new TestCase( SECTION, "parseFloat('3.e-1')", .30, parseFloat("3.e-1") );
new TestCase( SECTION, "parseFloat('-3.e-1')", -.30, parseFloat("-3.e-1") );
new TestCase( SECTION, "parseFloat('+3.e-1')", .30, parseFloat("+3.e-1") );
// StrDecimalLiteral::: .DecimalDigits ExponentPart opt
new TestCase( SECTION, "parseFloat('.00001')", 0.00001, parseFloat(".00001") );
new TestCase( SECTION, "parseFloat('+.00001')", 0.00001, parseFloat("+.00001") );
new TestCase( SECTION, "parseFloat('-0.0001')", -0.00001, parseFloat("-.00001") );
new TestCase( SECTION, "parseFloat('.01e2')", 1, parseFloat(".01e2") );
new TestCase( SECTION, "parseFloat('+.01e2')", 1, parseFloat("+.01e2") );
new TestCase( SECTION, "parseFloat('-.01e2')", -1, parseFloat("-.01e2") );
new TestCase( SECTION, "parseFloat('.01e+2')", 1, parseFloat(".01e+2") );
new TestCase( SECTION, "parseFloat('+.01e+2')", 1, parseFloat("+.01e+2") );
new TestCase( SECTION, "parseFloat('-.01e+2')", -1, parseFloat("-.01e+2") );
new TestCase( SECTION, "parseFloat('.01e-2')", 0.0001, parseFloat(".01e-2") );
new TestCase( SECTION, "parseFloat('+.01e-2')", 0.0001, parseFloat("+.01e-2") );
new TestCase( SECTION, "parseFloat('-.01e-2')", -0.0001, parseFloat("-.01e-2") );
// StrDecimalLiteral::: DecimalDigits ExponentPart opt
new TestCase( SECTION, "parseFloat('1234e5')", 123400000, parseFloat("1234e5") );
new TestCase( SECTION, "parseFloat('+1234e5')", 123400000, parseFloat("+1234e5") );
new TestCase( SECTION, "parseFloat('-1234e5')", -123400000, parseFloat("-1234e5") );
new TestCase( SECTION, "parseFloat('1234e+5')", 123400000, parseFloat("1234e+5") );
new TestCase( SECTION, "parseFloat('+1234e+5')", 123400000, parseFloat("+1234e+5") );
new TestCase( SECTION, "parseFloat('-1234e+5')", -123400000, parseFloat("-1234e+5") );
new TestCase( SECTION, "parseFloat('1234e-5')", 0.01234, parseFloat("1234e-5") );
new TestCase( SECTION, "parseFloat('+1234e-5')", 0.01234, parseFloat("+1234e-5") );
new TestCase( SECTION, "parseFloat('-1234e-5')", -0.01234, parseFloat("-1234e-5") );
new TestCase( SECTION, "parseFloat(0)", 0, parseFloat(0) );
new TestCase( SECTION, "parseFloat(-0)", -0, parseFloat(-0) );
new TestCase( SECTION, "parseFloat(1)", 1, parseFloat(1) );
new TestCase( SECTION, "parseFloat(-1)", -1, parseFloat(-1) );
new TestCase( SECTION, "parseFloat(2)", 2, parseFloat(2) );
new TestCase( SECTION, "parseFloat(-2)", -2, parseFloat(-2) );
new TestCase( SECTION, "parseFloat(3)", 3, parseFloat(3) );
new TestCase( SECTION, "parseFloat(-3)", -3, parseFloat(-3) );
new TestCase( SECTION, "parseFloat(4)", 4, parseFloat(4) );
new TestCase( SECTION, "parseFloat(-4)", -4, parseFloat(-4) );
new TestCase( SECTION, "parseFloat(5)", 5, parseFloat(5) );
new TestCase( SECTION, "parseFloat(-5)", -5, parseFloat(-5) );
new TestCase( SECTION, "parseFloat(6)", 6, parseFloat(6) );
new TestCase( SECTION, "parseFloat(-6)", -6, parseFloat(-6) );
new TestCase( SECTION, "parseFloat(7)", 7, parseFloat(7) );
new TestCase( SECTION, "parseFloat(-7)", -7, parseFloat(-7) );
new TestCase( SECTION, "parseFloat(8)", 8, parseFloat(8) );
new TestCase( SECTION, "parseFloat(-8)", -8, parseFloat(-8) );
new TestCase( SECTION, "parseFloat(9)", 9, parseFloat(9) );
new TestCase( SECTION, "parseFloat(-9)", -9, parseFloat(-9) );
new TestCase( SECTION, "parseFloat(3.14159)", 3.14159, parseFloat(3.14159) );
new TestCase( SECTION, "parseFloat(-3.14159)", -3.14159, parseFloat(-3.14159) );
new TestCase( SECTION, "parseFloat(3.)", 3, parseFloat(3.) );
new TestCase( SECTION, "parseFloat(-3.)", -3, parseFloat(-3.) );
new TestCase( SECTION, "parseFloat(3.e1)", 30, parseFloat(3.e1) );
new TestCase( SECTION, "parseFloat(-3.e1)", -30, parseFloat(-3.e1) );
new TestCase( SECTION, "parseFloat(3.e+1)", 30, parseFloat(3.e+1) );
new TestCase( SECTION, "parseFloat(-3.e+1)", -30, parseFloat(-3.e+1) );
new TestCase( SECTION, "parseFloat(3.e-1)", .30, parseFloat(3.e-1) );
new TestCase( SECTION, "parseFloat(-3.e-1)", -.30, parseFloat(-3.e-1) );
new TestCase( SECTION, "parseFloat(3.E1)", 30, parseFloat(3.E1) );
new TestCase( SECTION, "parseFloat(-3.E1)", -30, parseFloat(-3.E1) );
new TestCase( SECTION, "parseFloat(3.E+1)", 30, parseFloat(3.E+1) );
new TestCase( SECTION, "parseFloat(-3.E+1)", -30, parseFloat(-3.E+1) );
new TestCase( SECTION, "parseFloat(3.E-1)", .30, parseFloat(3.E-1) );
new TestCase( SECTION, "parseFloat(-3.E-1)", -.30, parseFloat(-3.E-1) );
// StrDecimalLiteral::: .DecimalDigits ExponentPart opt
new TestCase( SECTION, "parseFloat(.00001)", 0.00001, parseFloat(.00001) );
new TestCase( SECTION, "parseFloat(-0.0001)", -0.00001, parseFloat(-.00001) );
new TestCase( SECTION, "parseFloat(.01e2)", 1, parseFloat(.01e2) );
new TestCase( SECTION, "parseFloat(-.01e2)", -1, parseFloat(-.01e2) );
new TestCase( SECTION, "parseFloat(.01e+2)", 1, parseFloat(.01e+2) );
new TestCase( SECTION, "parseFloat(-.01e+2)", -1, parseFloat(-.01e+2) );
new TestCase( SECTION, "parseFloat(.01e-2)", 0.0001, parseFloat(.01e-2) );
new TestCase( SECTION, "parseFloat(-.01e-2)", -0.0001, parseFloat(-.01e-2) );
// StrDecimalLiteral::: DecimalDigits ExponentPart opt
new TestCase( SECTION, "parseFloat(1234e5)", 123400000, parseFloat(1234e5) );
new TestCase( SECTION, "parseFloat(-1234e5)", -123400000, parseFloat(-1234e5) );
new TestCase( SECTION, "parseFloat(1234e+5)", 123400000, parseFloat(1234e+5) );
new TestCase( SECTION, "parseFloat(-1234e+5)", -123400000, parseFloat(-1234e+5) );
new TestCase( SECTION, "parseFloat(1234e-5)", 0.01234, parseFloat(1234e-5) );
new TestCase( SECTION, "parseFloat(-1234e-5)", -0.01234, parseFloat(-1234e-5) );
// hex cases should all return 0 (0 is the longest string that satisfies a StringDecimalLiteral)
new TestCase( SECTION, "parseFloat('0x0')", 0, parseFloat("0x0"));
new TestCase( SECTION, "parseFloat('0x1')", 0, parseFloat("0x1"));
new TestCase( SECTION, "parseFloat('0x2')", 0, parseFloat("0x2"));
new TestCase( SECTION, "parseFloat('0x3')", 0, parseFloat("0x3"));
new TestCase( SECTION, "parseFloat('0x4')", 0, parseFloat("0x4"));
new TestCase( SECTION, "parseFloat('0x5')", 0, parseFloat("0x5"));
new TestCase( SECTION, "parseFloat('0x6')", 0, parseFloat("0x6"));
new TestCase( SECTION, "parseFloat('0x7')", 0, parseFloat("0x7"));
new TestCase( SECTION, "parseFloat('0x8')", 0, parseFloat("0x8"));
new TestCase( SECTION, "parseFloat('0x9')", 0, parseFloat("0x9"));
new TestCase( SECTION, "parseFloat('0xa')", 0, parseFloat("0xa"));
new TestCase( SECTION, "parseFloat('0xb')", 0, parseFloat("0xb"));
new TestCase( SECTION, "parseFloat('0xc')", 0, parseFloat("0xc"));
new TestCase( SECTION, "parseFloat('0xd')", 0, parseFloat("0xd"));
new TestCase( SECTION, "parseFloat('0xe')", 0, parseFloat("0xe"));
new TestCase( SECTION, "parseFloat('0xf')", 0, parseFloat("0xf"));
new TestCase( SECTION, "parseFloat('0xA')", 0, parseFloat("0xA"));
new TestCase( SECTION, "parseFloat('0xB')", 0, parseFloat("0xB"));
new TestCase( SECTION, "parseFloat('0xC')", 0, parseFloat("0xC"));
new TestCase( SECTION, "parseFloat('0xD')", 0, parseFloat("0xD"));
new TestCase( SECTION, "parseFloat('0xE')", 0, parseFloat("0xE"));
new TestCase( SECTION, "parseFloat('0xF')", 0, parseFloat("0xF"));
new TestCase( SECTION, "parseFloat('0X0')", 0, parseFloat("0X0"));
new TestCase( SECTION, "parseFloat('0X1')", 0, parseFloat("0X1"));
new TestCase( SECTION, "parseFloat('0X2')", 0, parseFloat("0X2"));
new TestCase( SECTION, "parseFloat('0X3')", 0, parseFloat("0X3"));
new TestCase( SECTION, "parseFloat('0X4')", 0, parseFloat("0X4"));
new TestCase( SECTION, "parseFloat('0X5')", 0, parseFloat("0X5"));
new TestCase( SECTION, "parseFloat('0X6')", 0, parseFloat("0X6"));
new TestCase( SECTION, "parseFloat('0X7')", 0, parseFloat("0X7"));
new TestCase( SECTION, "parseFloat('0X8')", 0, parseFloat("0X8"));
new TestCase( SECTION, "parseFloat('0X9')", 0, parseFloat("0X9"));
new TestCase( SECTION, "parseFloat('0Xa')", 0, parseFloat("0Xa"));
new TestCase( SECTION, "parseFloat('0Xb')", 0, parseFloat("0Xb"));
new TestCase( SECTION, "parseFloat('0Xc')", 0, parseFloat("0Xc"));
new TestCase( SECTION, "parseFloat('0Xd')", 0, parseFloat("0Xd"));
new TestCase( SECTION, "parseFloat('0Xe')", 0, parseFloat("0Xe"));
new TestCase( SECTION, "parseFloat('0Xf')", 0, parseFloat("0Xf"));
new TestCase( SECTION, "parseFloat('0XA')", 0, parseFloat("0XA"));
new TestCase( SECTION, "parseFloat('0XB')", 0, parseFloat("0XB"));
new TestCase( SECTION, "parseFloat('0XC')", 0, parseFloat("0XC"));
new TestCase( SECTION, "parseFloat('0XD')", 0, parseFloat("0XD"));
new TestCase( SECTION, "parseFloat('0XE')", 0, parseFloat("0XE"));
new TestCase( SECTION, "parseFloat('0XF')", 0, parseFloat("0XF"));
new TestCase( SECTION, "parseFloat(' 0XF ')", 0, parseFloat(" 0XF "));
// hex literals should still succeed
new TestCase( SECTION, "parseFloat(0x0)", 0, parseFloat(0x0));
new TestCase( SECTION, "parseFloat(0x1)", 1, parseFloat(0x1));
new TestCase( SECTION, "parseFloat(0x2)", 2, parseFloat(0x2));
new TestCase( SECTION, "parseFloat(0x3)", 3, parseFloat(0x3));
new TestCase( SECTION, "parseFloat(0x4)", 4, parseFloat(0x4));
new TestCase( SECTION, "parseFloat(0x5)", 5, parseFloat(0x5));
new TestCase( SECTION, "parseFloat(0x6)", 6, parseFloat(0x6));
new TestCase( SECTION, "parseFloat(0x7)", 7, parseFloat(0x7));
new TestCase( SECTION, "parseFloat(0x8)", 8, parseFloat(0x8));
new TestCase( SECTION, "parseFloat(0x9)", 9, parseFloat(0x9));
new TestCase( SECTION, "parseFloat(0xa)", 10, parseFloat(0xa));
new TestCase( SECTION, "parseFloat(0xb)", 11, parseFloat(0xb));
new TestCase( SECTION, "parseFloat(0xc)", 12, parseFloat(0xc));
new TestCase( SECTION, "parseFloat(0xd)", 13, parseFloat(0xd));
new TestCase( SECTION, "parseFloat(0xe)", 14, parseFloat(0xe));
new TestCase( SECTION, "parseFloat(0xf)", 15, parseFloat(0xf));
new TestCase( SECTION, "parseFloat(0xA)", 10, parseFloat(0xA));
new TestCase( SECTION, "parseFloat(0xB)", 11, parseFloat(0xB));
new TestCase( SECTION, "parseFloat(0xC)", 12, parseFloat(0xC));
new TestCase( SECTION, "parseFloat(0xD)", 13, parseFloat(0xD));
new TestCase( SECTION, "parseFloat(0xE)", 14, parseFloat(0xE));
new TestCase( SECTION, "parseFloat(0xF)", 15, parseFloat(0xF));
new TestCase( SECTION, "parseFloat(0X0)", 0, parseFloat(0X0));
new TestCase( SECTION, "parseFloat(0X1)", 1, parseFloat(0X1));
new TestCase( SECTION, "parseFloat(0X2)", 2, parseFloat(0X2));
new TestCase( SECTION, "parseFloat(0X3)", 3, parseFloat(0X3));
new TestCase( SECTION, "parseFloat(0X4)", 4, parseFloat(0X4));
new TestCase( SECTION, "parseFloat(0X5)", 5, parseFloat(0X5));
new TestCase( SECTION, "parseFloat(0X6)", 6, parseFloat(0X6));
new TestCase( SECTION, "parseFloat(0X7)", 7, parseFloat(0X7));
new TestCase( SECTION, "parseFloat(0X8)", 8, parseFloat(0X8));
new TestCase( SECTION, "parseFloat(0X9)", 9, parseFloat(0X9));
new TestCase( SECTION, "parseFloat(0Xa)", 10, parseFloat(0Xa));
new TestCase( SECTION, "parseFloat(0Xb)", 11, parseFloat(0Xb));
new TestCase( SECTION, "parseFloat(0Xc)", 12, parseFloat(0Xc));
new TestCase( SECTION, "parseFloat(0Xd)", 13, parseFloat(0Xd));
new TestCase( SECTION, "parseFloat(0Xe)", 14, parseFloat(0Xe));
new TestCase( SECTION, "parseFloat(0Xf)", 15, parseFloat(0Xf));
new TestCase( SECTION, "parseFloat(0XA)", 10, parseFloat(0XA));
new TestCase( SECTION, "parseFloat(0XB)", 11, parseFloat(0XB));
new TestCase( SECTION, "parseFloat(0XC)", 12, parseFloat(0XC));
new TestCase( SECTION, "parseFloat(0XD)", 13, parseFloat(0XD));
new TestCase( SECTION, "parseFloat(0XE)", 14, parseFloat(0XE));
new TestCase( SECTION, "parseFloat(0XF)", 15, parseFloat(0XF));
// A StringNumericLiteral may not use octal notation
new TestCase( SECTION, "parseFloat('00')", 0, parseFloat("00"));
new TestCase( SECTION, "parseFloat('01')", 1, parseFloat("01"));
new TestCase( SECTION, "parseFloat('02')", 2, parseFloat("02"));
new TestCase( SECTION, "parseFloat('03')", 3, parseFloat("03"));
new TestCase( SECTION, "parseFloat('04')", 4, parseFloat("04"));
new TestCase( SECTION, "parseFloat('05')", 5, parseFloat("05"));
new TestCase( SECTION, "parseFloat('06')", 6, parseFloat("06"));
new TestCase( SECTION, "parseFloat('07')", 7, parseFloat("07"));
new TestCase( SECTION, "parseFloat('010')", 10, parseFloat("010"));
new TestCase( SECTION, "parseFloat('011')", 11, parseFloat("011"));
// A StringNumericLIteral may have any number of leading 0 digits
new TestCase( SECTION, "parseFloat('001')", 1, parseFloat("001"));
new TestCase( SECTION, "parseFloat('0001')", 1, parseFloat("0001"));
new TestCase( SECTION, "parseFloat(' 0001 ')", 1, parseFloat(" 0001 "));
// an octal numeric literal should be treated as an octal
new TestCase( SECTION, "parseFloat(00)", 0, parseFloat(00));
new TestCase( SECTION, "parseFloat(01)", 1, parseFloat(01));
new TestCase( SECTION, "parseFloat(02)", 2, parseFloat(02));
new TestCase( SECTION, "parseFloat(03)", 3, parseFloat(03));
new TestCase( SECTION, "parseFloat(04)", 4, parseFloat(04));
new TestCase( SECTION, "parseFloat(05)", 5, parseFloat(05));
new TestCase( SECTION, "parseFloat(06)", 6, parseFloat(06));
new TestCase( SECTION, "parseFloat(07)", 7, parseFloat(07));
new TestCase( SECTION, "parseFloat(010)", 8, parseFloat(010));
new TestCase( SECTION, "parseFloat(011)", 9, parseFloat(011));
// A StringNumericLIteral may have any number of leading 0 digits
new TestCase( SECTION, "parseFloat(001)", 1, parseFloat(001));
new TestCase( SECTION, "parseFloat(0001)", 1, parseFloat(0001));
// make sure it's reflexive
new TestCase( SECTION, "parseFloat(Math.PI)", Math.PI, parseFloat(Math.PI));
new TestCase( SECTION, "parseFloat(Math.LN2)", Math.LN2, parseFloat(Math.LN2));
new TestCase( SECTION, "parseFloat(Math.LN10)", Math.LN10, parseFloat(Math.LN10));
new TestCase( SECTION, "parseFloat(Math.LOG2E)", Math.LOG2E, parseFloat(Math.LOG2E));
new TestCase( SECTION, "parseFloat(Math.LOG10E)", Math.LOG10E, parseFloat(Math.LOG10E));
new TestCase( SECTION, "parseFloat(Math.SQRT2)", Math.SQRT2, parseFloat(Math.SQRT2));
new TestCase( SECTION, "parseFloat(Math.SQRT1_2)", Math.SQRT1_2, parseFloat(Math.SQRT1_2));
new TestCase( SECTION, "parseFloat(Math.PI+'')", Math.PI, parseFloat(Math.PI+''));
new TestCase( SECTION, "parseFloat(Math.LN2+'')", Math.LN2, parseFloat(Math.LN2+''));
new TestCase( SECTION, "parseFloat(Math.LN10+'')", Math.LN10, parseFloat(Math.LN10+''));
new TestCase( SECTION, "parseFloat(Math.LOG2E+'')", Math.LOG2E, parseFloat(Math.LOG2E+''));
new TestCase( SECTION, "parseFloat(Math.LOG10E+'')", Math.LOG10E, parseFloat(Math.LOG10E+''));
new TestCase( SECTION, "parseFloat(Math.SQRT2+'')", Math.SQRT2, parseFloat(Math.SQRT2+''));
new TestCase( SECTION, "parseFloat(Math.SQRT1_2+'')", Math.SQRT1_2, parseFloat(Math.SQRT1_2+''));
test();