Recover non-list repr attr with a span for later attr validation #143533
Conversation
|
|
rustbot
added
A-attributes
|
Some changes occurred in compiler/rustc_passes/src/check_attr.rs Some changes occurred in compiler/rustc_attr_parsing |
| @@ -2049,12 +2049,6 @@ impl<'tcx> CheckAttrVisitor<'tcx> { | |||
| if item.is_some() { | |||
| match target { | |||
| Target::Struct | Target::Union | Target::Enum => continue, | |||
| Target::Fn | Target::Method(_) => { | |||
There was a problem hiding this comment.
This was telling us that #[repr] was actually #[repr(align(..))] which is wrong.
There was a problem hiding this comment.
I think this fix is not included in #143522, we should make sure this does not get lost
| | ^^^^^^^^ | ||
| | | | ||
| | expected this to be a list | ||
| | help: must be of the form: `#[repr(C | Rust | align(...) | packed(...) | <integer type> | transparent)]` |
There was a problem hiding this comment.
This help is a bit misleading at the crate level, but seems inevitable unless we are aware of the attr target when we emit this error, which seems unnecessary.
There was a problem hiding this comment.
We might soon start emitting warnings like this earlier which regains us that info. Not for this pr but am aware of it :)
| @@ -567,32 +576,59 @@ LL | | #[coroutine = 63] || {} | |||
| LL | | } | |||
| | |_- not a `const fn` | |||
|
|
|||
| error[E0517]: attribute should be applied to a struct, enum, or union | |||
There was a problem hiding this comment.
This is:
#[repr]
fn foo() {}
Previously we didn't emit an error at all here. See comment above about it being mistakenly considered like repr(align(..)), which is fixed by just removing that arm from the ReprEmpty case.
compiler-errors
force-pushed
the
crate-empty-repr
branch
from
d7ee65c to
d668c53
Compare
|
@compiler-errors I just approved a pr this morning by @JonathanBrouwer which I think conflicts with this one so you might need to do a rebase I'm afraid |
|
Apart from that these changes look nice, but I'll wait till tomorrow with the r+ since I have a feeling some of the impl needs to change a bit after the other one. Wait let me link it |
|
That PR just fixes #143522, so I'll close this one. |
rustbot
removed
the
S-waiting-on-review
|
Oh awesome, didn't expect the other one to do that much for this issue but I guess it makes sense. okay! |
Fixes #143522
This seems like an obvious solution, but I'm open to other approaches. Pardon the span cascade in the test file, but that's unfortunately a consequence of adding a
#![]attr to the test :)r? @jdonszelmann