[PATCH] D25816: Use descriptive message if list initializer is incorrectly parenthesized.
Serge Pavlov via cfe-commits
cfe-commits at lists.llvm.org
Fri Oct 21 03:34:56 PDT 2016
sepavloff marked 5 inline comments as done.
sepavloff added inline comments.
================
Comment at: include/clang/Basic/DiagnosticSemaKinds.td:1762
def err_init_incomplete_type : Error<"initialization of incomplete type %0">;
+def err_list_init_in_parens : Error<"list-initializer for non-class type "
+ "must not be parenthesized">;
----------------
arphaman wrote:
> Wouldn't it be better if we had a diagnostic with the type information? So, instead of showing `list-initializer for non-class type must not be parenthesized` clang would show `list-initializer for non-class type 'int' must not be parenthesized`. What do you think?
>
> As well as that, it seems that GCC issues a warning instead of an error for this diagnostic, so should this be a warning in clang as well?
Putting type name into the diagnostic message is a good idea, implemented.
As for making this message a warning instead of an error, I am in doubt.
Pros:
- it enhances GCC compatibility,
- wrong code may be easily fixed by compiler,
Cons:
- this is a violation of the standard,
- this extension does not add any benefits,
- wrong code can be easily fixed by user.
I chose to retain current clang behavior and reject questionable code. GCC [[ https://gcc.gnu.org/ml/gcc-patches/2011-08/msg00644.html | patch]] that introduced this message explains using warning by some uncertainty, 5 years passed, I think the standard is stable in viewpoint on such usage.
================
Comment at: lib/Sema/SemaExprCXX.cpp:1229
+ Diag(TInfo->getTypeLoc().getLocStart(), diag::err_list_init_in_parens)
+ << IList->getSourceRange()
+ << FixItHint::CreateRemoval(LParenLoc)
----------------
arphaman wrote:
> Formatting issue: clang-format replaces
>
> ```
> << IList->getSourceRange()
> << FixItHint::CreateRemoval(LParenLoc)
> ```
> with
> ```
> << IList->getSourceRange() << FixItHint::CreateRemoval(LParenLoc)
> ```
After insertion of type `Ty` this is no more the case.
https://reviews.llvm.org/D25816
More information about the cfe-commits
mailing list