[libcxx-commits] [PATCH] D74870: Split _LIBCPP_STRING_EXTERN_TEMPLATE_LIST up into a V1 and UNSTABLE version.

Louis Dionne via Phabricator via libcxx-commits libcxx-commits at lists.llvm.org
Mon Feb 24 09:08:08 PST 2020


ldionne added a comment.

In D74870#1888514 <https://reviews.llvm.org/D74870#1888514>, @EricWF wrote:

> In D74870#1884581 <https://reviews.llvm.org/D74870#1884581>, @ldionne wrote:
>
> > In D74870#1883765 <https://reviews.llvm.org/D74870#1883765>, @EricWF wrote:
> >
> > > My personal preference is to not split the lists, my rational is:
> > >
> > > - It encourages unstable to deviate from stable more. I would like us to keep the code and instantiations between the two as similar as possible. The more they differ, the more it is to maintain and the more likely it is that one contains bugs or starts to rot.
> >
> >
> > I actually think that acknowledging that they are different instantiation lists makes everything clearer and easier to maintain. We don't "pretend" that they are close when they are really two different things. As we embrace changes in the unstable version of the ABI, I don't see why we'd strive to keep it close to the stable ABI. We should simply setup testers for both ABIs, since both seem to be in use.
>
>
> Ideally we'll keep the unstable and stable ABI lists as close as possible, with the goal of moving all unstable instantiations into stable.


Are you saying that we'd want to put explicit instantiations used by the unstable ABI into the `.so` in the stable ABI, too? I don't understand why we'd do that?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D74870/new/

https://reviews.llvm.org/D74870





More information about the libcxx-commits mailing list