rsmith added a comment. In https://reviews.llvm.org/D46834#1102395, @dexonsmith wrote:
> In https://reviews.llvm.org/D46834#1102391, @rsmith wrote: > > > The policy certainly seems designed around the CLI use case. For serialized > > diagnostics, it would make sense to either serialize the snippet or enough > > information that the snippet can be reconstructed. And if that can't be > > done, or fails to satisfy some other use case, then we should discuss how > > we proceed -- for instance, we could consider having different diagnostic > > messages for the case where we have a snippet and for the case where we do > > not. > > > Right. There are places in the IDE where there is a condensed view of all > diagnostics (like a Vim location list), and others where the diagnostics are > shown inline with the sources. I think what we want is an optional auxiliary > record/field in a diagnostic with that contains context for when the source > context is missing, and then the IDE can choose which to display. It's > optional because most diagnostics are good enough as is for location lists. That sounds good to me. I think it would also make sense to use the alternate form for the CLI case if the user is using `-fno-caret-diagnostics` for some reason. Repository: rC Clang https://reviews.llvm.org/D46834 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
