On 3/31/25 4:59 PM, Marek Polacek wrote:
On Thu, Mar 27, 2025 at 11:27:04AM -0400, Jason Merrill wrote:
On 3/25/25 3:37 PM, Marek Polacek wrote:
Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk/14?
-- >8 --
Since r15-8011 cp_build_indirect_ref_1 won't do the *&TARGET_EXPR ->
TARGET_EXPR folding not to change its value category. That fix is
correct but it made us stop extending the lifetime in this testcase,
causing a wrong-code issue -- extend_ref_init_temps_1 did not see
through the extra *& because it doesn't use a tree walk. It is not
hard to fix that, but there may be other places that need this
adjustment. :/
Hmm, this suggests to me that we should revert the 117512 patch and instead
fix it in build_over_call, perhaps by forcing 'val' to be an lvalue after
it's built rather than relying on 'to' being an lvalue already.
Thanks for the suggestion.
Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk/14?
@@ -3870,13 +3870,11 @@ cp_build_indirect_ref_1 (location_t loc, tree ptr,
ref_operator errorstring,
return error_mark_node;
}
else if (do_fold && TREE_CODE (pointer) == ADDR_EXPR
- && same_type_p (t, TREE_TYPE (TREE_OPERAND (pointer, 0)))
- /* Don't let this change the value category. '*&TARGET_EXPR'
- is an lvalue, but folding it into 'TARGET_EXPR' would turn
- it into a prvalue of class type. */
- && lvalue_p (TREE_OPERAND (pointer, 0)))
+ && same_type_p (t, TREE_TYPE (TREE_OPERAND (pointer, 0))))
/* The POINTER was something like `&x'. We simplify `*&x' to
- `x'. */
+ `x'. This change the value category: '*&TARGET_EXPR'
"This might change" or "can change". OK with that tweak.
Jason
