https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123589
--- Comment #3 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> I've done some debugging and came up with this patch. I tested it with the
> latest version of trunk on x86_64-w64-mingw32 and it works fine, and seems
> to fix the bug. Of course given that I'm not super familiar with the GNAT
> codebase it's possible this is more of a bodge than a root cause fix, but
> I'll submit it just in case it's useful.
Thanks! I think that "Has_Default_Aspect (Base)" would be slightly better,
although it's strictly equivalent because Has_Default_Aspect goes through
Base_Type internally (to be honest, the test Is_Public (Typ) looks bogus).
I'm going to test the following change for the sake of consistency:
diff --git a/gcc/ada/exp_ch3.adb b/gcc/ada/exp_ch3.adb
index af1b82d351d..df24be0a65a 100644
--- a/gcc/ada/exp_ch3.adb
+++ b/gcc/ada/exp_ch3.adb
@@ -5734,7 +5734,7 @@ package body Exp_Ch3 is
-- and do not need initialization procedures.
elsif Is_Standard_String_Type (Base)
- and then No (Default_Aspect_Component_Value (Base))
+ and then not Has_Default_Aspect (Base)
then
null;
@@ -5764,6 +5764,7 @@ package body Exp_Ch3 is
elsif (Present (Init_Proc (Component_Type (Base)))
and then No (Base_Init_Proc (Base)))
or else (Init_Or_Norm_Scalars and then Base = Typ)
+ or else Has_Default_Aspect (Base)
or else Is_Public (Typ)
then
Build_Array_Init_Proc (Base, N);
