toPrecision() in Edge gives incorrect output in rare cases

[dajp1]

My apologies if this is the wrong place to submit this.

I have found what I suspect is a bug in Edge. In JavaScript, the lines:

a = 0.0003;
b = a.toPrecision(3);

result in b containing "0.000200". (IE, Chrome and Firefox all give the "0.000300" that I was expecting.)

See http://jsfiddle.net/dajp/2b47qLg0/

I'm using Edge 40.15063.0.0.
This is causing me some difficulties. Many thanks.

[MikeHolman]

Definitely seems like a real bug and something we should fix. That code is crazy complicated, so I can't immediately understand the problem.

Looks like something is going wrong inside NumberUtilities::FDblToStr's call to FDblToRgbPrecise though.

[obastemur]

0.0003 IEEE754 => 0.0003000000142492353916168212890625

So, toPrecision(3) indeed should not return 0.0002

[obastemur]

@dajp1 the expected output for your sample should be 0.00 not 0.0003 (since the precision is 3 and there are three zeroes pre any number.) When precision is set to 4 ChakraCore calculates the correct result.

However ChakraCore (as reported) results in 0.0002 when precision is 3.

[dajp1]

@obastemur I don't agree. As far as I can see, the specification says that toPrecision() should return a string to the specified number of significant digits, and examples always start with the first non-zero digit.

In my case I was expecting "0.000300" since that has three significant digits starting with the first non-zero digit, and this is what I get on every browser I have tried except Edge. If precision was set to 4, I would expect "0.0003000", which is indeed what I get (on Edge, IE, Firefox and Chrome). The bug only appears when the number of significant digits is set to three or fewer.

[dajp1]

Sorry - newbie mistake, I pressed the wrong button. Didn't mean to close this issue.

[obastemur]

As far as I can see, the specification says that toPrecision() should return a string to the specified number of significant digits, and examples always start with the first non-zero digit.

I agree on that. A counting issue(mixed with toFixed :/ ) on my end. Well, either way ChakraCore was returning the wrong result :)

[redchair123]

Stumbled upon a similar issue with (-6.9e-11).toExponential(4). Edge 13/14 give the correct string '-6.9000e-11' but 15/16 give '-6.8000e-11'. @obastemur does the patch also cover this case?

[obastemur]

Just tested. Yes, the fix covers that case too.