Dedup Fluent invalid ptr errors in const eval

[workingjubilee]

I believe these forms aren't unbearable to work with, and Fluent's ability to use selectors allows us to avoid enumerating every case laboriously.

Closes #116764 if accepted.

r? @fee1-dead

[workingjubilee]

This is an extremely conservative refactor to achieve the goal, admittedly, so perhaps @RalfJung would want further changes along these lines. In any case, this is the kind of thing that is permissible in Fluent, so we don't have to buckle under having a thousand little messages for the same thing, and this way the translator can retain the full context.

[workingjubilee]

Apparently this kind of form is the kind of thing Project Fluent prefers?

[RalfJung]

There are 7 more messages that duplicate the box/ref cases:

const_eval_validation_box_to_mut = {$front_matter}: encountered a box pointing to mutable memory in a constant
const_eval_validation_box_to_static = {$front_matter}: encountered a box pointing to a static variable in a constant
const_eval_validation_box_to_uninhabited = {$front_matter}: encountered a box pointing to uninhabited type {$ty}

const_eval_validation_ref_to_mut = {$front_matter}: encountered a reference pointing to mutable memory in a constant
const_eval_validation_ref_to_static = {$front_matter}: encountered a reference pointing to a static variable in a constant
const_eval_validation_ref_to_uninhabited = {$front_matter}: encountered a reference pointing to uninhabited type {$ty}

const_eval_validation_unaligned_box = {$front_matter}: encountered an unaligned box (required {$required_bytes} byte alignment but found {$found_bytes})
const_eval_validation_unaligned_ref = {$front_matter}: encountered an unaligned reference (required {$required_bytes} byte alignment but found {$found_bytes})

const_eval_validation_null_box = {$front_matter}: encountered a null box
const_eval_validation_null_ref = {$front_matter}: encountered a null reference

const_eval_validation_invalid_box_meta = {$front_matter}: encountered invalid box metadata: total size is bigger than largest supported object
const_eval_validation_invalid_ref_meta = {$front_matter}: encountered invalid reference metadata: total size is bigger than largest supported object

const_eval_validation_invalid_box_slice_meta = {$front_matter}: encountered invalid box metadata: slice is bigger than largest supported object
const_eval_validation_invalid_ref_slice_meta = {$front_matter}: encountered invalid reference metadata: slice is bigger than largest supported object

This affects all ValidationError variants that take a PtrKind.

[workingjubilee]

It could use a selector, but it's not clear defaulting between "box" and "reference" yields a correct result, and selectors require defaults.

[RalfJung]

Also we'd have to repeat the selector 7 times, no? That's still O(n*m) for what should be O(n+m)

[fee1-dead]

Also we'd have to repeat the selector 7 times, no? That's still O(n*m) for what should be O(n+m)

Then we'd need to create an identifier for "reference", and another for "box", and then we pass that as a parameter to these messages

[fee1-dead]

@rustbot author

[bors]

☔ The latest upstream changes (presumably #119621) made this pull request unmergeable. Please resolve the merge conflicts.

[Dylan-DPC]

@workingjubilee any updates on this? Thanks

[JohnCSimon]

@workingjubilee
Ping from triage: I'm closing this due to inactivity, Please reopen when you are ready to continue with this.
Note: if you are going to continue please open the PR BEFORE you push to it, else you won't be able to reopen - this is a quirk of github.

@rustbot label: +S-inactive

[workingjubilee]

Yeah, sorry, I lost track of this unfortunately and the issue probably should be reassessed in light of #132181