[PATCH] D72498: [clangd] Print underlying type for decltypes in hover

Sam McCall via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Mon Jan 13 06:54:08 PST 2020


sammccall added a comment.

In D72498#1816957 <https://reviews.llvm.org/D72498#1816957>, @kadircet wrote:

> In D72498#1816785 <https://reviews.llvm.org/D72498#1816785>, @ilya-biryukov wrote:
>
> > In D72498#1816424 <https://reviews.llvm.org/D72498#1816424>, @lh123 wrote:
> >
> > > Currently, I think that in most cases, showing both expanded (canonical) and spelled types is sufficient.
> > >
> > > > This has been used in ycmd for ~4 years without complaint. https://github.com/clangd/clangd/issues/58#issuecomment-507800970
> >
> >
> > That actually doesn't look bad. Maybe let's try doing that and see whether we'll get negative feedback?
> >  That seems to give useful information in **all** cases, so at least it'll cover all use-cases even it's more verbose.
> >
> > What do others think?
>
>
> SGTM, happy to update all types in `HoverInfo` to contain both a `pretty-printed` and `canonical` version. Where pretty-printed would just means desugared, so keywords like auto/decltype would've
>  been stripped away and it would be the type as written in all other cases, while canonical would refer to `clang canonical` with all of the type aliases etc. resolved. Ofc.
>  Does that SG to you as well @sammccall ?


No, I think printing *both* is at least somewhat likely to be too verbose, especially since the previous release showed no types at all.
And we're out of time to iterate on the behavior and presentation for this cycle. I think we should do something more conservative and then experiment in the next release cycle.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D72498/new/

https://reviews.llvm.org/D72498





More information about the cfe-commits mailing list