<table border="1" cellspacing="0" cellpadding="8">
    <tr>
        <th>Issue</th>
        <td>
            <a href=https://github.com/llvm/llvm-project/issues/142707>142707</a>
        </td>
    </tr>

    <tr>
        <th>Summary</th>
        <td>
            clang evaluates is_constant_evaluated incorrectly with narrowing conversion
        </td>
    </tr>

    <tr>
      <th>Labels</th>
      <td>
            clang
      </td>
    </tr>

    <tr>
      <th>Assignees</th>
      <td>
      </td>
    </tr>

    <tr>
      <th>Reporter</th>
      <td>
          efriedma-quic
      </td>
    </tr>
</table>

<pre>
    Consider the following example:

```
#include <type_traits>
struct X { unsigned u; };
void f(X x);
void g(int a) {
    f({0xFFFFFFFFLL + std::is_constant_evaluated()});
}
```

The standard says the conversion is narrowing if "the source is a constant expression whose value after integral promotions will fit into the target type".  gcc correctly rejects because of this: it treats the expression as a constant expression, then it fails to narrow.  clang incorrectly accepts because isCXX11ConstantExpr() doesn't treat the evaluation as a constant context.

We have to be a little careful here... consider the following variant:

```
#include <type_traits>
struct X { unsigned u; };
void f(X x);
void g(int a) {
 f({0x100000000LL - std::is_constant_evaluated()});
}
```

clang incorrectly rejects.  gcc miscompiles: it checks the narrowing in a constant context, but then doesn't evaluate the actual expression in a constant context, so the integer overflows to 0.

(I don't expect anyone to intentionally do this... wrote some synthetic testcases while looking at isCXX11ConstantExpr() for other reasons.)
</pre>
<img width="1" height="1" alt="" src="http://email.email.llvm.org/o/eJzMlE9v4zYQxT8NfRlEoCg7tg462Nk1UCDHAs1tQZEji12aVDkj__n2BSWlcbvZnnqoIMCGyZn3hu9naiJ3CoiN2BzE5stKj9zH1GCXHNqzfvpjdGbVRntvXmIgZzEB9whd9D5eXTgB3vR58CiqvZDT-yyXV-6FqlwwfrQIonrh-4DfOGnHJKqvQu6J02gY3kBsDzCGyYiFUVQHENsvojoIub9EZ6ETavcGN6Hqhx9PQu1cYNBC1bmDkHsAmPaK7UHejsvz-gpCHYDYZo_V3tE3EwOxDvwNL9qPmtHmIlVn1XeN_P0f08j9rz1CrrQ6WSB9p-kwTAwXTORiAEcQdErz0bgOhFJ5B8UxGcyrGt7FAW9DQprKrn0khGwGQXeMCVxgPCXtYUjxHNnFQHB13kPnOC_GSZl1OiFDPlmhVAFwMgZMTAkN-zsk_B0NE7Ro9EgIsQPuHYlqD46BE2qeJ3iwon_iUaiXvDXkyk47T8BxmbUAMF7ngcOHtjYGhwdtRy9vb2X5snT-ehvSfOhgI1IQars4mg3NwfxoyMTAeONizuM3hF5fMFtpETR4x-wRjE7YjR56TFgUxVT9CboXnZwO_P9A9y9uS7k8r6_w9N9g-2M6CxkLMWdHJp4H5_GdDdOj-T6z8cBz-CSJzEU78szGR5TvFqcW2vCo_SNlP21FM9gT_pggXjB1Pl4n2uSSulC7X8DGReg2oGHQ4R7DBEIuDZkc7f0dbJyQzxRcU-T8Vzwj0D1wj-wMMBIbTUhw7Z1H8DF-z7Nq_hdiu5ggco8JEmqKgQqh6pVtKltXtV5hU27Xu7relRu56ptNh6bc6Xpr5G5tS8S6rEqrsHw2VaVau3KNkmojn-VaynVdyqLedptOrRHLZ9nathZriWftfOH95VzEdFo5ohGbcq22crvyukVP0w2u1JS1UCpf5qnJBU_teCKxlt4R00cLduyxmdF4T4vgU8z-hs7Vcf8AxcfltxqTb3rmIUMk1FGo48lxP7aFiWehjll5-XgaUswACnWcJiGhjsswl0b9GQAA__-F0Szt">