Fix1215 #4516
Fix1215 #4516
Conversation
|
Please follow the existing dmd code style:
You can use |
|
Thanks for the review. Style fixed, please take another look. |
|
Is this supported? |
|
It's not indeed, thanks for the catch ! Here's a fixed version. |
| } | ||
| else if (t) | ||
| { | ||
| index->error(loc, "Expected an expression as index, got a type (%s)", t->toChars()); |
There was a problem hiding this comment.
DMD error messages (user-visible ones, anyway) generally start with a lower-case letter.
|
Sorry for pestering you with yet another round of stylistic nit-picks, but here you go. ;) I'll defer to Kenji for the actual technical review. |
|
Please do comment, this is my first contribution to dmd, so this is useful :) |
|
Sorry for late reply. I reviewed code, and confirmed that the code is enough to support a indexed type with dot identifier. It looks not smart, but I know that comes from the current dmd internal design. To make things more clean, we need to improve the Type structure design around TypeQualifier, and I think it's not so important at this first step. LGTM after the code style and test case tweaks. @legrosbuffle To support the indexed type with dot identifier, we also need to update D grammer. Can you also work for that? If you cannot open a grammer update PR in parallel, I'll file a website bugzilla issue after merging this. |
| } | ||
|
|
||
| case DYNCAST_EXPRESSION: | ||
| sm = takeTypeTupleIndex(loc, sc, s, id, (Expression*)id); |
There was a problem hiding this comment.
For all casts, insert a space before *.
s/(Expression*)id/(Expression *)id/
See the corresponding pull request here: dlang/dmd#4516
| @@ -0,0 +1,118 @@ | |||
| // PERMUTE_ARGS: | |||
| // REQUIRED_ARGS: | |||
There was a problem hiding this comment.
Normally compilable test is compiled with -c and generates objfile. But the test for 1215 does not have to do it.
So, you can list -o- in REQUIRED_ARGS: to stop backend codegen.
There was a problem hiding this comment.
Done, thanks for the suggestion.
|
Thanks for the comments ! I've also created a PR for the grammar on dlang.org. |
|
LGTM. Thanks for your awesome work! |
|
Auto-merge toggled on |
|
This should have been squashed down to one or two commits before it was merged. |
…iled. If a TypeQualifier is not trivial dot-chain of scope and member symbols, it will be resolved as the equivalent expression. The new TypeQualified structure introduced in dlang#4516 for issue 1215 is now fully in `TypeQualified::resolveExprType`, and the ICE issue will be solved in it.
…iled. If a TypeQualifier is not trivial dot-chain of scope and member symbols, it will be resolved as the equivalent expression. The new TypeQualified structure introduced in dlang#4516 for issue 1215 is now fully in `TypeQualified::resolveExprType`, and the ICE issue will be solved in it.
Parse Args[0].T as a type in:
struct S(Args...) {
Args[0].T m;
alias T = Args[0].T;
}
See bug 1215 for context.