mirror of https://github.com/svaarala/duktape.git
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
53 lines
1.8 KiB
53 lines
1.8 KiB
/*---
|
|
{
|
|
"knownissue": "corner case in floating point parse rounding"
|
|
}
|
|
---*/
|
|
|
|
/*===
|
|
1e+23
|
|
1e+23
|
|
1e+23
|
|
===*/
|
|
|
|
/* This integer constant parsed incorrectly in some version
|
|
* and would print, incorrectly:
|
|
*
|
|
* 1.0000000000000001e+23
|
|
*
|
|
* Looking at the IEEE representations:
|
|
*
|
|
* $ python
|
|
* Python 2.7.3 (default, Aug 1 2012, 05:14:39)
|
|
* [GCC 4.6.3] on linux2
|
|
* Type "help", "copyright", "credits" or "license" for more information.
|
|
* >>> import struct
|
|
* >>> struct.pack('>d', 1e+23).encode('hex')
|
|
* '44b52d02c7e14af6'
|
|
* >>> struct.pack('>d', 1.0000000000000001e+23).encode('hex')
|
|
* '44b52d02c7e14af7'
|
|
*
|
|
* It's apparent that the rounding of the last digit is incorrect.
|
|
* The Dragon4 generator output (when a large number of binary digits
|
|
* is requested explicitly) is:
|
|
*
|
|
* generated digits; k=77, .-- digit considered for rounding
|
|
* digits= v
|
|
* 10101001 01101000 00010110 00111111 00001010 01010111 10110100 00000000 00000000 00000000 00000000 00000000 ... (zeroes indefinitely)
|
|
* ^ ^
|
|
* `-- hidden bit `-- last mantissa digit
|
|
*
|
|
* The problem here is that the number is apparently exactly halfway between
|
|
* the two IEEE representations; the current rounding will only consider the
|
|
* immedate '1' digit and round up. This case is not trivial to fix because
|
|
* we don't know, based on just the one rounding digit, that the number is
|
|
* exactly halfway and we should round to even.
|
|
*/
|
|
|
|
try {
|
|
print(parseInt('0000100000000000000000000000', 10));
|
|
print(parseFloat('1e23', 10));
|
|
print(parseFloat('1e+23', 10));
|
|
} catch (e) {
|
|
print(e);
|
|
}
|
|
|