Fix ICE in mir when evaluating SizeOf on unsized type

[osa1]

Not quite ready yet. This tries to fix #80742 as discussed on Zulip topic,
by using delay_span_bug.

I don't understand what delay_span_bug does. It seems like my error message
is never used. With this patch, in this program:

#![allow(incomplete_features)]
#![feature(const_evaluatable_checked)]
#![feature(const_generics)]

use std::fmt::Debug;
use std::marker::PhantomData;
use std::mem::size_of;

struct Inline<T>
where
    [u8; size_of::<T>() + 1]: ,
{
    _phantom: PhantomData<T>,
    buf: [u8; size_of::<T>() + 1],
}

impl<T> Inline<T>
where
    [u8; size_of::<T>() + 1]: ,
{
    pub fn new(val: T) -> Inline<T> {
        todo!()
    }
}

fn main() {
    let dst = Inline::<dyn Debug>::new(0); // line 27
}

these errors are printed, both for line 27 (annotated line above):

• "no function or associated item named new found for struct Inline<dyn Debug> in the current scope"
• "the size for values of type dyn Debug cannot be known at compilation time"

Second error makes sense, but I'm not sure about the first one and why it's
even printed.

Finally, I'm not sure about the span passing in const_eval.

[rust-highfive]

r? @oli-obk

(rust-highfive has picked a reviewer for you, use r? to override)

[oli-obk]

I don't understand what delay_span_bug does. It seems like my error message
is never used.

That means it's working 😆 . delay_span_bug is a way to make sure we poison the compiler to cause an ICE if no errors are reported. The message and span would be used in that case, but as long as an error is reported elsewhere, it is not shown. While we could report an error here, that would just add to the noise and likely be a less useful error than the one emitted by typeck.

Second error makes sense, but I'm not sure about the first one and why it's
even printed.

The first error occurs because we fail to satisfy the generic bounds, so we don't find the new method that you wrote. Basically you have to treat it as if the compiler goes through all impl blocks to find methods, but it doesn't even look into the impl block if it can't satisfy the bounds on the impl block.

[osa1]

Thanks for the detailed answer @oli-obk , very helpful.

[osa1]

Interesting.. I updated by branch (https://github.com/osa1/rust/tree/fix_80742) but the PR is not updated, it's still showing the old version.

[osa1]

@oli-obk This is ready for reviews & merge.

[oli-obk]

Thanks!

@bors r+ rollup

[bors]

📌 Commit 3fb53c2 has been approved by oli-obk

[RalfJung]

I think we should still abort execution. The result that we return here is wrong. So there should be a new error variant, similar to TransmuteSizeDiff, and that should be raised here.

[oli-obk]

@bors r-

[RalfJung]

(Too late, it was rolled up.)

[oli-obk]

Either the rollup succeeds, then this can be fixed in a follow up, or it does not, then it can be fixed here

[osa1]

I'll submit another PR.

[osa1]

#81243