[PATCH] D41217: [Concepts] Concept Specialization Expressions

[Saar Raz via Phabricator via cfe-commits]

saar.raz added inline comments.


================
Comment at: lib/Sema/SemaTemplateInstantiate.cpp:679-681
+      Diags.Report(Active->PointOfInstantiation,
+                   diag::note_constraint_substitution_here)
+          << Active->InstantiationRange;
----------------
saar.raz wrote:
> rsmith wrote:
> > Is this note ever useful? It will presumably always point into the same concept definition that the prior diagnostic also pointed at, and doesn't seem to add anything in the testcases.
> > 
> > Maybe we could keep the CodeSynthesisContext around as a marker that we've entered a SFINAE context, but not have any corresponding diagnostic. (The note produced for the enclosing `ConstraintsCheck` context covers that.) Or we could remove this and store a flag on the `ConstraintsCheck` to indicate whether we're in a SFINAEable portion of it.
> If the concept definition is multiline/contains macros, this would point at the exact place where the problematic constraint occured, we should probably add tests for this case though.
> Maybe we can omit the diagnostic when the concept and the constraint are on the same line or something?
> 
Correction - I just now realized you were referring to the 'in instantiation of template class' note and not the 'while checking the satisfaction...' note.
In this case - the only case I can think of where the problematic instantiation and the constraint expression would be in very different places is indeed multiline constraint expressions or macros in a constraint expression (but you know better than I do - can you think of any other cases? or maybe there are other non-SFINAE errors that can result from substitution and would not produce a note?). Maybe these cases are remote enough to warrant removing this diagnostic or as I said earlier omit them in cases where they are unhelpful.


Repository:
  rC Clang

https://reviews.llvm.org/D41217