fix(to_cpp1): improve recognition of dependent types and deducible parameters

[JohelEGP]

Resolves #534. Resolves #595.
Resolves #628 (as a discussion, this needs to be manually closed).
Adds what's needed to diagnose #572.

[JohelEGP]

Patch for merging with #596:

diff --git a/source/to_cpp1.h b/source/to_cpp1.h
index 5286c3ee..01cb93b5 100644
--- a/source/to_cpp1.h
+++ b/source/to_cpp1.h
@@ -2234,6 +2234,12 @@ public:
         {
             return self->is_dependent(*expr.result_type);
         }
+
+        auto operator()(requires_expression_node const&) const
+            -> bool
+        {
+            return false;
+        }
     };
 
     auto is_dependent(

[JohelEGP]

In type-only contexts, we can also allow a space between the identifier and the template-argument-list opener.

assert <_>(); // OK, `<` never an operator.
_ = std::vector <int>(); // Error, `<` interpreted as an operator.
_ = :std::vector <int> = (); // Error, same as above, but as a type-only context, we know it's not an operator.

[hsutter]

Hi! Sorry it took me so long to get to this one... but this looks like it might still be current. It's a bigger diff, but part of it is the stack_ helpers that got merged from I think another one of your PRs.

Is this up to date, and would you like me to review it?

[JohelEGP]

Yes, and yes, please do.

[hsutter]

Is it intended that these test cases fail?

• pure2-bugfix-for-dependent-types-recursion
• pure2-bugfix-for-dependent-types

For some reason the latter works on GCC 14?

[JohelEGP]

pure2-bugfix-for-dependent-types-recursion should fail.
It tests that the algorithm detects recursio,
and returns false (i.e., "not a dependent type"),
to let the Cpp1 compiler give its diagnostic.

pure2-bugfix-for-dependent-types shouldn't fail on a modern compiler.
I see some are failing because they don't yet support optional typename.