On Wed, 19 Mar 2025 at 11:48, Tomasz Kaminski <tkami...@redhat.com> wrote: > > > > > On Wed, Mar 19, 2025 at 12:36 PM Jonathan Wakely <jwak...@redhat.com> wrote: >> >> On 19/03/25 09:32 +0100, Tomasz Kamiński wrote: >> >This patch implements LWG2713 by adding missing allocator aware version of >> >unordered associative containers constructors accepting only "range" >> >(pair of iterators, initializer_list, or from_range), and corresponding >> >deduction guides. >> > >> >In addition the std::ranges::to<std::unordered_set>(alloc) is well-formed, >> >likewise for rest of unordered containers. >> > >> >libstdc++-v3/ChangeLog: >> > >> > * include/bits/unordered_map.h (unordered_map): >> > Define constructors accepting: >> > (_InputIterator, _InputIterator, const allocator_type&), >> > (initializer_list<value_type>, const allocator_type&), >> > (from_range_t, _Rg&&, const allocator_type&) >> > (unordered_multimap): Likewise. >> > * include/bits/unordered_set.h (unordered_set): >> > Define constructors and deduction guide accepting: >> > (_InputIterator, _InputIterator, const allocator_type&), >> > (initializer_list<value_type>, const allocator_type&). >> > Define constructor (from_range_t, _Rg&&, const allocator_type&). >> > (unordered_multiset): Likewise. >> > * testsuite/23_containers/unordered_map/cons/66055.cc: New tests. >> > * testsuite/23_containers/unordered_map/cons/deduction.cc: New tests. >> > * testsuite/23_containers/unordered_map/cons/from_range.cc: New >> > tests. >> > * testsuite/23_containers/unordered_multimap/cons/66055.cc: New >> > tests. >> > * testsuite/23_containers/unordered_multimap/cons/deduction.cc: >> > New tests. >> > * testsuite/23_containers/unordered_multimap/cons/from_range.cc: >> > New tests. >> > * testsuite/23_containers/unordered_multiset/cons/66055.cc: New >> > tests. >> > * testsuite/23_containers/unordered_multiset/cons/deduction.cc: >> > New tests. >> > * testsuite/23_containers/unordered_multiset/cons/from_range.cc: >> > New tests. >> > * testsuite/23_containers/unordered_set/cons/66055.cc: New tests. >> > * testsuite/23_containers/unordered_set/cons/deduction.cc: New tests. >> > * testsuite/23_containers/unordered_set/cons/from_range.cc: New >> > tests. >> > * testsuite/std/ranges/conv/1.cc: New tests. >> >--- >> >I have added ranges::to test, as they provide additional motivation. >> >Tested on x86_64-linux. OK for trunk? >> >> I think we should be a little cautious here, because we're very close >> to the GCC 15 release and this is changing C++11/14/17 code as well. >> We don't want a last-minute regression this close to the release. >> >> Although these constructors all look correct, they introduce the >> possibility of new ambiguities just by changing the overload set. >> >> The C++23 from_range_t constructors are OK to add now. For the rest of >> them, I think I'd prefer to guard them with #if __cplusplus >= 202002L >> for now, and open a bugzilla bug reminding us to remove those #if >> guards for GCC 16. > > I think it would be better to just reduce the GCC15 patch to only new > from_range constructors. I think if we are concerned with new ambiguities > for old constructors we should not add them conditionally. > Otherwise we are risking breaking old code when compiled with new standard.
Yes, but the new standards are considered experimental so a bit of experimental breakage is acceptable ;-) But I'm fine with waiting for GCC 16 and then adding them for all standards >= C++11, and if it breaks things we'll have nearly a whole year to find out before making a release (and we can give that feedback to LWG). > > I will prepare a partial patch. Thanks.
