I can confirm that, with this patch, Clang successfully compiles my actual code base.

My few cases format consistently with the condition case:

diff
diff --git a/tests/integration/test.waarudo.io.dynamic.cpp b/tests/integration/test.waarudo.io.dynamic.cpp
index be5a626..e1265b9 100644
--- a/tests/integration/test.waarudo.io.dynamic.cpp
+++ b/tests/integration/test.waarudo.io.dynamic.cpp
@@ -21,7 +21,7 @@ public:
   }

In D120806#3354119, @MyDeveloperDay wrote:

LFTM

Do we also need to consider this in the context of "AllowShortIfStatments"? (that would be in a different review in my view)

Thank you.

Thank you for doing this.

I have another case:
.clang-format:

BraceWrapping:
  AfterStruct: true
  SplitEmptyRecord: true
BreakBeforeBraces: Custom

Thank you.

Yes, here's my reply: https://bugs.llvm.org/show_bug.cgi?id=47161#c2.

I was a bit late, but thanks for everything!

Could you do a rebase? The patch no longer applies:

In D79773#2141164, @MyDeveloperDay wrote:

This is actually failing in the handling of parsing the <> in (parseAngle() actually parseBrace() inside of that), it is not really concepts specific except it may be seen now because of the presence of the ;

I found another case. See how everything after the requires-clause is indented for constructors with just the constructor's name (it works otherwise, maybe because it looks like a function).

class [[nodiscard]] data_t
{
public:
    template <std::ranges::input_range R>
        requires std::same_as<std::ranges::range_value_t<R>, flippable_tile_id>
        /*explicit*/ data_t(R&& ids, const format_t f = encoding_t::csv)
          : format_t{f}, tile_ids_{ranges::to_vector(std::forward<R>(ids))}
        {
        }

Ah, that makes sense. The very definition of Allman is to always break before braces. I suppose I'll need to use BreakBeforeBraces: Custom and BraceWrapping:. I did some testing and noticed that the weird format comes with any of these:

Yes, it's valid C++ (is my hope). It's from WIP code, which I made available to show you: https://github.com/johelegp/jge.

Another weird formatting when direct-initializing a variable with a non-parenthesized requires-expression:

Thank you. Everything I reported works fine now.

Using parameter packs within non-parenthesized requires-clauses results in weird formatting:

template <std::semiregular F, std::semiregular... Args>
    requires std::invocable < F, std::invoke_result_t<Args>
... > struct constant;

vs

template <std::semiregular F, std::semiregular... Args>
    requires(std::invocable<F, std::invoke_result_t<Args>...>)
struct constant;

In D79773#2125040, @MyDeveloperDay wrote:

Just so I'm clear for this and your Allman coment would you show me what you are seeing and what you expect to see? just so I understand.

constexpr plane(const plane& other) noexcept(
    (detail::default_allocator_is_nothrow &&
     std::is_nothrow_copy_constructible_v<T>)) requires(std::copyable<T>)
  : plane{to1d(other), other.sz.w}
{
}

Without the parentheses in trailing requires-clauses, all code following the constructor initializer list is further indented.

BreakBeforeBraces: Allman isn't respected.

template <class D, class U, class Rep2>
friend constexpr auto
operator*(const size2d& l, const units::quantity<D, U, Rep2>& r) noexcept(
    noexcept(l.w* r)) requires(requires { l.w* r; }) //
{
    return size2d<decltype(l.w() * r)>{l.w * r, l.h * r};
}

When a trailing requires-clause is right before the function body, its opening brace should go on its own line, as dictated by my config file.