[Clang] SFINAE on mismatching pack length during constraint satisfaction checking

[cor3ntin]

If a fold expanded constraint would expand packs of different size, it is not a valid pack expansion and it is not satisfied. This should not produce an error.

Fixes #99430

[llvmbot]

@llvm/pr-subscribers-clang

Author: cor3ntin (cor3ntin)

Changes

If a fold expanded constraint would expand packs of different size, it is not a valid pack expansion and it is not satisfied. This should not produce an error.

Fixes #99430

---

Full diff: https://github.com/llvm/llvm-project/pull/101879.diff

2 Files Affected:

• (modified) clang/lib/Sema/SemaConcept.cpp (+4)
• (modified) clang/test/SemaCXX/cxx2c-fold-exprs.cpp (+30)

diff --git a/clang/lib/Sema/SemaConcept.cpp b/clang/lib/Sema/SemaConcept.cpp
index 9e16b67284be4..c34d32002b5ad 100644
--- a/clang/lib/Sema/SemaConcept.cpp
+++ b/clang/lib/Sema/SemaConcept.cpp
@@ -531,6 +531,10 @@ static ExprResult calculateConstraintSatisfaction(
 
     std::optional<unsigned>
     EvaluateFoldExpandedConstraintSize(const CXXFoldExpr *FE) const {
+
+      // We should ignore errors in the presence of packs of different size.
+      Sema::SFINAETrap Trap(S);
+
       Expr *Pattern = FE->getPattern();
 
       SmallVector<UnexpandedParameterPack, 2> Unexpanded;
diff --git a/clang/test/SemaCXX/cxx2c-fold-exprs.cpp b/clang/test/SemaCXX/cxx2c-fold-exprs.cpp
index 1e0bc7bcfb4e7..0674135aac483 100644
--- a/clang/test/SemaCXX/cxx2c-fold-exprs.cpp
+++ b/clang/test/SemaCXX/cxx2c-fold-exprs.cpp
@@ -275,3 +275,33 @@ static_assert(S<int>::g<int>() == 2); // expected-error {{call to 'g' is ambiguo
 
 
 }
+
+namespace GH99430 {
+
+template <class _Ty1, class _Ty2>
+using _Synth_three_way_result = int;
+
+template <class... _Types>
+class tuple;
+
+template <int _Index>
+struct tuple_element;
+
+template <class, int...>
+struct _Three_way_comparison_result_with_tuple_like {
+  using type = int;
+};
+
+template <class... _TTypes, int... _Indices>
+  requires(requires {
+    typename _Synth_three_way_result<_TTypes, tuple_element<_Indices>>;
+  } && ...)
+
+struct _Three_way_comparison_result_with_tuple_like<tuple<_TTypes...>, _Indices...>{
+    using type = long;
+};
+
+static_assert(__is_same_as(_Three_way_comparison_result_with_tuple_like<tuple<int>, 0, 1>::type, int));
+static_assert(__is_same_as(_Three_way_comparison_result_with_tuple_like<tuple<int>, 0>::type, long));
+
+}

[cor3ntin]

The intent is for this to be backported to 19 (therefore no changelog)

This was a fairly easy fix thanks to @Endilll reduction! Very much appreciated :)

[Endilll]

LGTM

[Endilll]

The intent is for this to be backported to 19 (therefore no changelog)

Yes, this is very important to backport.

[AaronBallman]

LGTM! This should be backported, thank you for the quick fix!

[cor3ntin]

/cherry-pick da380b2

[llvmbot]

/pull-request #101967