CDT_3: ship <tl/expected.hpp> (#9145)

[atta-ullah01]

Replaces the FetchContent mechanism with a vendored copy of tl/expected.hpp (v1.1.0).

The header is placed in CGAL/tl/expected.hpp to avoid namespace pollution.
This also removes the CMake module CGAL_setup_tl-excepted.cmake.

Release Management

• Affected package(s): Constrained_triangulation_3
• Issue(s) solved (if any): CDT_3: ship <tl/expected.hpp> #9145
• License and copyright ownership: CC0 1.0 Universal (tl/expected.hpp)

[sloriot]

Thanks for the PR. I'd move the tl/expected.hpp file into Installation package.

[atta-ullah01]

I added the SPDX tag to the header. I cannot run the detect_packages_licenses script locally without modifying unrelated files (version mismatch). Could you please update the license.txt manifest before merging?

[afabri]

@lrineau and @sloriot : Is there not a problem if someone already uses this header? Should "our" version go into a CGAL::tl namespace? Is it really needed that we use not yet standard language features?

[lrineau]

@lrineau and @sloriot : Is there not a problem if someone already uses this header? Should "our" version go into a CGAL::tl namespace?

That is true. It could go in the CGAL::cpp23 namespace. I will make and test the change.

Is it really needed that we use not yet standard language features?

std::excepted was standardized in 2022, and part of C++23. That is a vocabulary type, like std::any, or std::optional. It is not really new. It can be seen as an improved version of std::optional: in case the object is void, an error object is in place, to describe why.

[afabri]

And does std::optional not offer enough ? It seems we add a lot of machinery for little.

[lrineau]

And does std::optional not offer enough ? It seems we add a lot of machinery for little.

Obviously I agree. And the C++ committee as well. The "lot of machinery" is just one added file under licence CC0, after this PR.

[afabri]

And what about the define TL_... ? Will that not trigger already defined errors?

[lrineau]

And what about the define TL_... ? Will that not trigger already defined errors?

I have renamed the macros in commit 0be2691.

And the new namespace is now CGAL::cpp23:: instead of tl:: (commit 7cf93c0).

[afabri]

Let's also add it to the official documentation.

[sloriot]

Successfully tested in CGAL-6.2-Ic-82

[lrineau]

Let's also add it to the official documentation.

That part was not yet done. There are probably other things added to STL_Extension that I had not yet documented.

[sloriot]

you can still open an issue and doc it when ready. There is no need to delay the integration that can be the cause of a failure in the testsuite during the fetch content.