[clang] Use canonical type for substitution which might be incomplete

When checking deduction consistency, a substitution can be incomplete
such that only sugar parts refer to non-deduced template parameters.

This would not otherwise lead to an inconsistent deduction, so
this patch makes it so we canonicalize the types before substitution
in order to avoid that possibility, for now.

When we are able to produce substitution failure diagnostics for
partial ordering, we might want to improve the TemplateInstantiator
so that it does not fail in that case.

This fixes a regression on top of #100692, which was reported on
the PR. This was never released, so there are no release notes.

clang/lib/Sema/SemaTemplateDeduction.cpp

@@ -5505,8 +5505,11 @@ static TemplateDeductionResult CheckDeductionConsistency(
   Sema::ArgumentPackSubstitutionIndexRAII PackIndex(
       S, ArgIdx != -1 ? ::getPackIndexForParam(S, FTD, MLTAL, ArgIdx) : -1);
   bool IsIncompleteSubstitution = false;
-  QualType InstP = S.SubstType(P, MLTAL, FTD->getLocation(), FTD->getDeclName(),
-                               &IsIncompleteSubstitution);
+  // FIXME: A substitution can be incomplete on a non-structural part of the
+  // type. Use the canonical type for now, until the TemplateInstantiator can
+  // deal with that.
+  QualType InstP = S.SubstType(P.getCanonicalType(), MLTAL, FTD->getLocation(),
+                               FTD->getDeclName(), &IsIncompleteSubstitution);
   if (InstP.isNull() && !IsIncompleteSubstitution)
     return TemplateDeductionResult::SubstitutionFailure;
   if (!CheckConsistency)

clang/test/SemaTemplate/GH18291.cpp

@@ -112,3 +112,12 @@ namespace static_vs_nonstatic {
     }
   } // namespace explicit_obj_param
 } // namespace static_vs_nonstatic
+
+namespace incomplete_on_sugar {
+  template <unsigned P, class T> void f(T[P]) = delete;
+  template <unsigned P> void f(int[][P]);
+  void test() {
+    int array[1][8];
+    f<8>(array);
+  }
+} // namespace incomplete_on_sugar