Johan Mabille

Yes the old connection pattern will still be supported (but marked as deprecated).

> Hi Johan, I feel I owe a response here and I apologize for the delay. I've been even worse, sorry for that! > Requiring all kernels to be updated...

> `The devil is in the details`. I totally agree, that's why I precised *implementation* detail; we should focus on the specification only. What I meant is as long as...

https://github.com/jupyter-xeus/xeus/pull/363 introduces a new `get_starting_message` and deprecates `print_starting_message`

@IsabelParedes sorry I missed your comment, `get_started_message` is indeed a typo. Both other options are fine with me, I let you pick up one ;)

Indeed, we are missing a bunch of definitions with `XSIMD_NO_SUPPORTED_ARCHITECTURE`. I am working on a fix, but this is not trivial.

We had the same issue with emscripten, using an [alternative implementation](https://github.com/xtensor-stack/xsimd/blob/master/include/xsimd/arch/generic/xsimd_generic_math.hpp#L981) fixed the issue. It is less efficient, but more accurate (although I'm not sure to understand why, `sincos(a)` should...

IIRC debugpy (or ptvsd back in the days ;)) has been added as a dependency because users complained about it not being installed while they wanted to use the debugger....

Hi, Thanks for reporting! Indeed ABI breaks from nlohmann::json are quite rare, and usually on minor updates, not patch updates. I guess we should explicitely specify the abi namespace in...

After diving into the details: - [inline namespaces](https://json.nlohmann.me/features/namespace/) have been added to nlohmann_json in 3.11.0 to detect this kind of issues at link time (which is exactly what you got...