Max Sagebaum

Ok, I had a closer look. List-initialization for `z` still compiles https://godbolt.org/z/YY6fhbbe7. But gives the warning. If we remove the move in the cpp2 example we get the ``same'' warning,...

Nice to see the update. I would have a closer look, but I am currently on vecation. Maybe I will have time at the beginning of September.

> I wouldn't be against a composition-over-inheritance approach, but this does mean big changes for the API, and subsequently, to the existing metafunctions. If we decide to go ahead with...

It also seems that it is not possible to declare it in any working way current: `typename` does not work: (https://cpp2.godbolt.org/z/x4oYsj1E5) ``` f: () std::vector(); main: () = { }...

Created a pull request https://github.com/hsutter/cppfront/pull/1153.

Yes, that would also solve my problem. We could enable it just for clang and gcc and enable it when MSVC supports it.

On a second thought, it would solve the problem in cppfront but depending on how cppfront outputs concatinated strings I would then have the problem in the cpp1 compiler. -...

This is actually a valid point. So to summarize: - Cppfront forwards `\x{...}` escapes - Cppfront code is not allowed to use it since not all targeted compilers support it.

In my experience gcc uses the temporary directory when compiling from source to an executable. When using intermediate objects `*.o` files, now temporary files should be generated.

If you do not use UFCS, it starts to work. https://cpp2.godbolt.org/z/vePsbM7bj It seems that for a regular function call the analysis is not as elaborate. In order to get a...