This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] PR c/13134
On Thu, 4 Dec 2003 17:17:48 -0800, Richard Henderson <rth@redhat.com> wrote:
> On Thu, Dec 04, 2003 at 01:57:15PM -0800, Matt Austern wrote:
>> + /* If either declaration has a nondefault visibility, use it.
>> + (Should we give an error if we have two different nondefault
>> + visibilities?) */
>
> We should do something, yes.
I think an error would be appropriate.
This reminds me of the bit from the shared library discussion in Kona where
Lawrence touched on the Sun implementation of Windows-like symbol binding,
whereby symbols are normally hidden, dllexport makes them protected, and
dllimport makes them global (i.e. gives them default visibility).
I don't think that such a scheme would require two different nondefault
visibilities in the same TU; at most you would have a dllimport (default)
decl and a dllexport (protected) decl.
>> + __attribute__((visibility ("hidden")))
>> + void foo() { }
> ...
>> + __attribute__((visibility ("hidden")))
>> + int xyzzy = 5;
>
> Misplaced attributes. They should go immediately before the name.
Attributes can now also appear as declspecs.
Jason