diff options
| author | Jason Merrill <jason@redhat.com> | 2023-06-27 05:15:01 -0400 |
|---|---|---|
| committer | Jason Merrill <jason@redhat.com> | 2023-06-28 00:42:12 -0400 |
| commit | abdf0b6cdff5783b97f35ad61ae31433f0569dfd (patch) | |
| tree | 883f684665687209e8f4b227992d40086420227c /gcc/cp/method.cc | |
| parent | 83f69969c0f3eb93302f6ef7714884db9079e803 (diff) | |
| download | gcc-abdf0b6cdff5783b97f35ad61ae31433f0569dfd.zip gcc-abdf0b6cdff5783b97f35ad61ae31433f0569dfd.tar.gz gcc-abdf0b6cdff5783b97f35ad61ae31433f0569dfd.tar.bz2 | |
c++: inherited constructor attributes
Inherited constructors are like constructor clones; they don't exist from
the language perspective, so they should copy the attributes in the same
way. But it doesn't make sense to copy alias or ifunc attributes in either
case. Unlike handle_copy_attribute, we do want to copy inlining attributes.
The discussion of PR110334 pointed out that we weren't copying the
always_inline attribute, leading to poor inlining choices.
PR c++/110334
gcc/cp/ChangeLog:
* cp-tree.h (clone_attrs): Declare.
* method.cc (implicitly_declare_fn): Use it for inherited
constructor.
* optimize.cc (clone_attrs): New.
(maybe_clone_body): Use it.
gcc/testsuite/ChangeLog:
* g++.dg/cpp1z/nodiscard-inh1.C: New test.
Diffstat (limited to 'gcc/cp/method.cc')
| -rw-r--r-- | gcc/cp/method.cc | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/gcc/cp/method.cc b/gcc/cp/method.cc index 91cf943..8ed967d 100644 --- a/gcc/cp/method.cc +++ b/gcc/cp/method.cc @@ -3294,6 +3294,8 @@ implicitly_declare_fn (special_function_kind kind, tree type, /* Copy constexpr from the inherited constructor even if the inheriting constructor doesn't satisfy the requirements. */ constexpr_p = DECL_DECLARED_CONSTEXPR_P (inherited_ctor); + /* Also copy any attributes. */ + DECL_ATTRIBUTES (fn) = clone_attrs (DECL_ATTRIBUTES (inherited_ctor)); } /* Add the "this" parameter. */ |