EricWF added a comment. In http://reviews.llvm.org/D12247#233595, @yiranwang wrote:
> In http://reviews.llvm.org/D12247#233349, @EricWF wrote: > > > In http://reviews.llvm.org/D12247#233323, @davidxl wrote: > > > > > We certainly need a fix without breaking ABI. Is there a ABI conformance > > > test for libcxx? > > > > > > Well this patch definitely breaks the ABI. This was my attempt at fixing > > the problem in `std::function` that might not be ABI breaking.. > > https://gist.github.com/EricWF/3a35b140a66d4826a9 > > > I think following your thought, we should abandon the usage of > "aligned_storage" in "function" totally. Instead, just say "void > *__buf_[3];", and check alignment_of _Fp (can be arbitrarily big, but usually > small), and sizeof the two to guard use of __buf_. The alignment bump to 16 > by aligned_storage on AARCH64 is not what we want, performance wise. Sort of. I wouldn't abandon the use of `aligned_storage` in function but I would change it from `aligned_storage<3*sizeof(void*)>::type` to `aligned_storage<3*sizeof(void*), alignof(void*)>::type`. That would have eventually the same behavior your describing. The reason that I haven't made that change is that it is ABI breaking. > Also, such kind of change need to applied for some more places, at least in > libc++, such like <libcxx/include/future> Yes it would. http://reviews.llvm.org/D12247 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
