[libc++][test] Make narrowing in nasty_char_traits::to_char_type more explicit

[frederick-vs-ja]

Previously, the cast was allowed due to P0960R3, which made narrowing implicitly done in the parenthesized aggregate initialization. It seems that MSVC doesn't like such an implicit manner and emits warning C4242.

This patch makes the narrowing more explicit. Follows up #135338.

[llvmbot]

@llvm/pr-subscribers-libcxx

Author: A. Jiang (frederick-vs-ja)

Changes

Previously, the cast was allowed due to P0960R3, which made narrowing implicitly done in the parenthesized aggregate initialization. It seems that MSVC doesn't like such an implicit manner and emits warning C4242.

This patch makes the narrowing more explicit. Follows up #135338.

---

Full diff: https://github.com/llvm/llvm-project/pull/138375.diff

1 Files Affected:

• (modified) libcxx/test/support/nasty_string.h (+1-1)

diff --git a/libcxx/test/support/nasty_string.h b/libcxx/test/support/nasty_string.h
index c2968f52e3a00..6b17bbc3f2232 100644
--- a/libcxx/test/support/nasty_string.h
+++ b/libcxx/test/support/nasty_string.h
@@ -70,7 +70,7 @@ struct nasty_char_traits {
 
   static constexpr int_type not_eof(int_type c) noexcept { return eq_int_type(c, eof()) ? ~eof() : c; }
 
-  static constexpr char_type to_char_type(int_type c) noexcept { return char_type(c); }
+  static constexpr char_type to_char_type(int_type c) noexcept { return {static_cast<char>(c)}; }
 
   static constexpr int_type to_int_type(char_type c) noexcept { return int_type(c.c); }
 

[philnik777]

How is this more explicit? It replaces one cast with a different one.

[frederick-vs-ja]

How is this more explicit? It replaces one cast with a different one.

It seems intended that parenthesized (direct-non-list) aggregate initialization behaves like initialization via constructors, and MSVC considers narrowing in ctor-init-list to be "not explicit enough".