I just debugged a test failure down to different r...

# random

[ephemient]

I just debugged a test failure down to different results on different JVMs:

import kotlin.math.*
sin(85.05112877980658 * PI / 180)

• 0.9962720762207499 (Double.fromBits(0x3fefe175fa29280f)) with Java 11 and Java 17 on Linux x86-64 as well as Java 11 on Apple Silicon
• 0.99627207622075 (Double.fromBits(0x3fefe175fa292810)) with Java 17 on Apple Silicon

[ephemient]

no, I'm not testing for equality, but I am testing that certain points remain in bounds after https://en.wikipedia.org/wiki/Web_Mercator_projection, and this one bit just happens to cause some other arithmetic to go wrong

[ephemient]

anyhow, I ended up rewriting it in a more numerically-stable fashion, but it would have been convenient if there were some way to use java.lang.StrictMath instead of java.lang.Math from `kotlin.math.*`…

[ephemient]

speaking of which. have there been any proposals to allow for hexadecimal floating point literals in Kotlin like in several other languages? it would be clearer to talk about

0x1.fe175fa29280fp-1
0x1.fe175fa292810p-1

than the raw IEEE 754 bit representation

[Paul Griffith]

looks like fractional hex literals were deliberately dropped (a while back)
https://youtrack.jetbrains.com/issue/KT-4528/Lexer-tries-to-read-a-fractional-hex-literal