float tests: deduplicate min, max, and rounding tests

[RalfJung]

Part of #141726

Best reviewed commit-by-commit.

• Use assert_biteq! in the mod.rs tests. This requires some trickery to make shadowing macros with imports work.
• The min, max, minimum, maximum tests in tests/floats/f*.rs are entirely subsumed by what we already have in tests/float/mod.rs, so I just removed them.
• The rounding tests (floor etc) in f*.rs had more test points, so I copied them over. They didn't have 0.5 and -0.5 though which seem like interesting points in particular regarding the sign of the resulting zero if that's what it sounds to, and they didn't max min/max/inf/nan tests, so this was really a merger of both tests.

r? @tgross35

[RalfJung]

The old test seemed to assume that the result of MIN.midpoint(MAX) should be -0.0, but actually we return 0.0.

[Urgau]

The midpoint function in C++ with clang and GCC also seems to return +0.0, so this should be fine. We don't guarantee the sign of zero anyway for midpoint.

[RalfJung]

A bunch of the -Float::MIN_POSITIVE rounding tests were missing the - as well.

[tgross35]

LGTM, thanks for the cleanup. One question then r=me.

[tgross35]

What is the reason for the trailing _ rather than just pub(crate) use assert_approx_eq;? I assume there may be some name conflicts.

[RalfJung]

I got an error about name clashes:

error[E0659]: `assert_approx_eq` is ambiguous
   --> library/coretests/tests/floats/mod.rs:681:9
    |
681 |         assert_approx_eq!((-1.7 as Float).fract(), -0.7);
    |         ^^^^^^^^^^^^^^^^ ambiguous name
    |
    = note: ambiguous because of a conflict between a `macro_rules` name and a non-`macro_rules` name from another module

I think I'll pick actually proper names for both runtime and const macros and use imports for name management, that's probably clearer.

[RalfJung]

I restructured the macros a bit to make it more clear (I think), please take a look. Putting the const ones in a module doesn't work out that well since (a) macros have their own special global namespace, and (b) we can't glob-import them anyway.

[tgross35]

Yeah that looks good to me. Maybe just add a breadcrumb comment at the use {assert_*_const as assert_*} remapping? Also if you don't mind grabbing it while you're here, I notice the preexisting doc example for the float_test macro is missing its closing /// ``` .

I still can't wrap my head around the way macro namespaces work sometimes...

[RalfJung]

There's already a comment there. What would you like to have added to it?

I still can't wrap my head around the way macro namespaces work sometimes...

Yeah.^^ I wonder if we can get rid of it in a future edition...

[tgross35]

There is a comment at the assert_approx_eq_rt as assert_approx_eq version mentioning that the rt version is used by default so const can shadow it, I was just thinking to mirror this at the place we shadow it (lines 193-194). But it's noncritical, r=me with or without it.

[tgross35]

Oops forgot the ✅