On Thu, 29 Aug 2024 at 09:12, Dhruv Chawla <dhr...@nvidia.com> wrote: > > On 28/08/24 15:40, Jonathan Wakely wrote: > > External email: Use caution opening links or attachments > > > > > > On Wed, 28 Aug 2024 at 06:47, Dhruv Chawla <dhr...@nvidia.com> wrote: > >> > >> version.syn#2 requires <string> to define > >> __cpp_lib_allocator_traits_is_always_equal. > >> > >> The attached patch therefore defines the > >> __glibcxx_want_allocator_traits_is_always_equal macro to get the > >> definition of the feature-test macro from <bits/version.h>. > >> > >> This is not isolated only to <string> though. The standard requires the > >> following (as per version.syn#2): > >> > >> #define __cpp_lib_allocator_traits_is_always_equal 201411L > >> // freestanding, also in <memory>, <scoped_allocator>, <string>, > >> <deque>, > >> // <forward_list>, <list>, <vector>, <map>, <set>, <unordered_map>, > >> // <unordered_set> > >> > >> After fixing <string>, all the above headers except <memory> still fail to > >> define the macro. Should similar fixes be created for these headers as > >> well? > > > > Yes please, this is a regression since GCC 13. It was previously > > defined in <bits/alloc_traits.h> which is included by all the headers > > above. In GCC 14 we changed how we define feature test macros, and we > > accidentally stopped defining it in all those headers. > > > > I have updated the headers now, please have a look. > > Bootstrapped and regtested on aarch64-linux-gnu.
Tests added for all headers, regtested successfully, DCO sign-off present - this looks great, thanks. I'll push it to trunk and gcc-14. > > >> > >> Signed-off-by: Dhruv Chawla <dhr...@nvidia.com> > >> > >> -- > >> Regards, > >> Dhruv > > > > -- > Regards, > Dhruv
