Add a diagnostic attribute for special casing const bound errors for non-const impls #148641
Conversation
|
Some changes occurred in compiler/rustc_passes/src/check_attr.rs |
rustbot
added
A-attributes
This comment has been minimized.
This comment has been minimized.
oli-obk
force-pushed
the
push-olzwqxsmnxmz
branch
from
c0e67f0 to
e501aad
Compare
This comment has been minimized.
This comment has been minimized.
oli-obk
force-pushed
the
push-olzwqxsmnxmz
branch
from
e501aad to
d04c00d
Compare
This comment has been minimized.
This comment has been minimized.
| /// Checks if `#[diagnostic::on_const]` is applied to a trait impl | ||
| fn check_diagnostic_on_const(&self, attr_span: Span, hir_id: HirId, target: Target) { | ||
| if !matches!(target, Target::Impl { of_trait: true }) { | ||
| self.tcx.emit_node_span_lint( | ||
| MISPLACED_DIAGNOSTIC_ATTRIBUTES, | ||
| hir_id, | ||
| attr_span, | ||
| DiagnosticOnConstOnlyForTraitImpls, | ||
| ); | ||
| } | ||
| } |
There was a problem hiding this comment.
Should we also check that the impl is not const? Otherwise the annotation is fully inert in that case, right?
| //! This is an unusual feature gate test, as it doesn't test the feature | ||
| //! gate, but the fact that not adding the feature gate to | ||
| //! the aux crate will cause the diagnostic to not emit the | ||
| //! custom diagnostic message |
There was a problem hiding this comment.
Is that strictly necessary? I feel like at least for std, if we annotate with diagnostic::on_const we'd want that message to appear even in stable.
There was a problem hiding this comment.
std has the feature enabled. but in this case the aux crate doesn't have it. So if a user crate doesn't have the feature gate, their uses of diagnostic::on_const get ignored
There was a problem hiding this comment.
Ah! I see. So the feature must be enabled where the annotation is. Gotcha.
oli-obk
force-pushed
the
push-olzwqxsmnxmz
branch
from
d04c00d to
67b757b
Compare
|
This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed. Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers. |
5 participants
Somewhat of a follow-up to #144194
My plan is to resolve
https://github.com/rust-lang/rust/blob/f4e19c68786211f3c3cf2593442629599678800a/compiler/rustc_hir_typeck/src/callee.rs#L907-913
but doing so without being able to mark impls the way I do in this PR wrould cause all nice diagnostics about for loops and pointer comparisons to just be a
*const u32 does not implement [const] PartialEqerrors.