Bug in prevpow I guess

General Usage
1

Hi, what is the right place to report the following?

I think I stumbled onto a simple ‘bug’ (= very unexpected behavior) in prevpow (when solving the latest Project Euler problem). Using Julia 1.7.

prevpow(1234567890123456789)

gives -8446744073709551616 instead of the expected 1000000000000000000, even though the argument is well below typemax(Int). I suspect that the next power of 10 is involved, which causes an overflow (certain looks this way to me having a cursory glance at the source for prevpow.

Not sure whether this is worth fixing, as it is quite an edge case and can be attributed to overflow behavior, but as a Project Euler fan it is slightly irritating to me.

2

ah, yeah. The problem is the next power of 10 overflows since it uses an integer power internally. Fixing this properly is probably a little tricky. That said, I’ll make a PR with an attempt to fix this.

2 Likes
3

https://github.com/JuliaLang/julia/pull/43410#issuecomment-991984057

4 
julia> prevpow(2,1)
1

I would expect this to be 0 just looking at documentation:

prevpow(a, x)
The largest a^n not greater than x, where n is a non-negative integer.

5

Why? 0 isn’t a power of 2, and 1 is a power of 2 not greater than 1.

6

ah, I thought it returns the power itself because “previous power”… my bad