#rust-lang

eddyb, I'm surprised (but happy) to see this works: https://is.gd/kcgw0k (because of the type-dependent const issue).

Also surprised (a little confused) that T has to be bound 'static.

Is this all correct?

This limitation also seems a bit off: https://is.gd/av0L1m

woboats: the 'static thing might be a bug, I forget

woboats: yeah that drop thing will go away

eddyb, cool. It happens to be exactly as expressive as I need right now :D

I'm a little concerned that it's kind of jagged though, seems hard to figure out what the rules are if for some reason you care.

woboats: the weird rules are a result of carving a useful niche in a very limited system

i.e. we hope to simplify by extension, not add more rules

that makes sense

0xFFFFFFFFu32 >> 32 panics / fails to compile with "bitshift exceeds the type's number of bits". Is that correct? I don't think 32 _exceeds_ 32, right?

Should I open an issue for that or is that expected behaviour?

0-31, no?

32 is bigger than a u32

(in this sense)

0xFFFFFFFFu32 >> 31 == 1

If would expect

0xFFFFFFFFu32 >> 32 == 0

and then 0xFFFFFFFFu32 >> 33 to be an overflow

I'm pretty sure "the number of bits" in a u32 is 32, right? ;)

:)

kosta: shifting out all the bits is undefined on some hardware

Hm ok

you might be able to work around it with wrapping_shr, I think?

playbot: 0xFFFFFFFFu32.wrapping_shr(32)

not

playbot: 0xFFFFFFFFu32.overflowing_shr(32) # might help though :)

playbot: 0xFFFFFFFFu32.overflowing_shr(32) // might help though :)

playbot: 0xFFFFFFFFu32.checked_shr(32).unwrap_or(0)

hm ok that looks more generic than the "if" condition i did

but I think the error message should be more accurate

Would you accept a PR to fix the error message?

thanks for the checked_shr() code!

PR would most likely get accepted

most likely some phrasing involving shifting out all the bits

yeah ok

I'll put it on my todo list, but I can't promise anything :)

I found an interesting way to cheat the lack of generic associated types: https://is.gd/aPf3cc

It 'leaks' the bound, which is unfortunate, but it doesn't even use HRTB which means it works for type params too

Actually implementing it might be a problem

ergh I need ATC so bad