Propagate restrictions against struct literals to the RHS of assignment #17290
Conversation
| RESTRICT_NO_STRUCT_LITERAL | ||
| } else { | ||
| UNRESTRICTED | ||
| }; |
There was a problem hiding this comment.
It would probably be cleaner code to give the enum variants here int values and treat them like flags.
There was a problem hiding this comment.
Looking at the RESTRICT_NO_BAR_OR_DOUBLEBAR_OP variant, it seems like a reasonable way to refactor, but it might be better as a separate PR.
bkoropoff
force-pushed
the
issue-17283
branch
2 times, most recently
from
2adc5c8 to
b04a566
Compare
|
New version force-pushed, now with more bitflags. |
| static Unrestricted = 0b0000, | ||
| static ResStmtExpr = 0b0001, | ||
| static ResNoBarOp = 0b0010, | ||
| static ResNoDoubleBarOp = 0b0100, |
There was a problem hiding this comment.
Please spell out Restrict for the flags and don't lose the BarOr from this one - longer names ftw
There was a problem hiding this comment.
Losing BarOr was intentional, as it can just set both flags to indicate it doesn't want either.
Although, looking at it now, it doesn't actually seem to be set anywhere in the parser. I guess I'll just drop it.
|
Nice, thanks for the bitflags conversion! r=me with the changes requested. |
This makes having multiple restrictions at once cleaner. Also drop NO_DOUBLEBAR restriction since it is never used.
This prevents confusing errors when accidentally using an assignment
in an `if` expression. For example:
```rust
fn main() {
let x = 1u;
if x = x {
println!("{}", x);
}
}
```
Previously, this yielded:
```
test.rs:4:16: 4:17 error: expected `:`, found `!`
test.rs:4 println!("{}", x);
^
```
With this change, it now yields:
```
test.rs:3:8: 3:13 error: mismatched types: expected `bool`, found `()` (expected bool, found ())
test.rs:3 if x = x {
^~~~~
```
Closes issue rust-lang#17283
bkoropoff
force-pushed
the
issue-17283
branch
from
b04a566 to
0e230c0
Compare
This prevents confusing errors when accidentally using an assignment in an
ifexpression. For example:Previously, this yielded:
With this change, it now yields:
Closes issue #17283
r? @nick29581