Bugs

Bugs in the built-in Math namespace.

• Numeric instability of transcendental functions
• Sine and cosine in V8
• Trigonometric functions in V8
• V8 not IEEE 754-2008 compliant
• Mozilla discussion on sine and cosine in V8
• V8 replaced a lookup table by computing Math.tan as Math.sin/Math.cos
• Browser math accuracy issues
• Mozilla attempt to address precision in new Math functions
• Mozilla's previous lack of tolerance tests for Math functions
• Mozilla Math.expm1 accuracy
• Mozilla thread on implementing ES6 math functions
• V8 Math.atanh issues
• V8 Math.acosh issues
• V8 Math.asinh issues
• V8 numeric issues in hyperbolic functions
• ES6 accuracy of special functions
• Accuracy of Math.exp in V8
• TC39 meeting discussing Math accuracy issues
• Spreadsheet showing trigonometric results across browsers
• Accuracy of Math.pow in V8
• Accumulation of errors in Mozilla Math.pow
• Accuracy of hyperbolic trigonometric functions in V8
• ES6 shim accuracy issues
• TC39 testing of Math built-ins
• V8 Math.random PRNG quality
• V8 trig function precision issues
• WebKit Math.random repeat values in separate workers
• Predictable Math.random in WebKit
• Predictable Math.random in Chrome
• Desire for fastmath implementations
• Mozilla Math.round issues
• Mozilla Math.pow issue
• Mozilla Math.pow issue
• V8 deoptimizes if Math.floor provided -0
• V8 deoptimizes if Math.ceil provided 0
• V8 Math.round accuracy
• Observability of distinguishable NaNs (see also get-nans and ECMA262 issue 635)
• NaN converted to -NaN
• After 10 years, Mozilla updates Math.random
• WebKit bug to use a better PRNG
• WebKit 32-bit Math bug
• Mozilla optimizations break commutativity
• Mozilla debate on speed versus precision
• Mozilla division bug to due C library delegation
• Fixit to unify Math implementations in V8
• Fixit to unify Math.pow implementation in V8

Notes

• To search for V8 Math bugs, see V8 issue tracker.
• To search for WebKit Math bugs, see WebKit issue tracker.
• To search for Mozilla Math bugs, see Bugzilla.