================
@@ -35,6 +35,13 @@ AST_MATCHER_P(Stmt, nextStmt,
ast_matchers::internal::Matcher<Stmt>,
return InnerMatcher.matches(**I, Finder, Builder);
}
+
+AST_MATCHER(Expr, isUnsupportedRangeInit) {
+ const Expr *E = Node.IgnoreParenImpCasts();
+ if (Finder->getASTContext().getLangOpts().CPlusPlus20)
+ return isa<CXXStdInitializerListExpr>(E);
----------------
zeyi2 wrote:
> I would've expected:
> ```c++
> const Expr *E = Node.IgnoreParenImpCasts();
> if (Finder->getASTContext().getLangOpts().CPlusPlus20)
> // the ranges versions will handle temporaries correctly
> return false;
>
> return E->isPRValue();
> ```
IMHO return `false` directly may be too broad, please see the comment above.
> Maybe what should happen instead is if the init is a temporary, the message
> is slightly different hinting at the lifetime, or do the same with a note?
Thanks for the suggestion!
A quick draft: `note: reusing the temporary range directly in the replacement
may be unsafe; consider materializing it in a local variable first`?
That said, I'm a bit unsure whether this should be done in this PR or as a
follow-up, as this PR is intended as a narrow fix for the current
false-positive. Changing the diagnostic wording/notes would expand the scope
into a design question for the check. IMO this is better tracked separately.
If you think this should be part of this PR, I can update the patch
accordingly. WDYT?
https://github.com/llvm/llvm-project/pull/185791
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
