Uniform header inclusion

[AntoinePrv]

No description provided.

[serge-sans-paille]

That's cool, but you also need to write some rule to ensure we don't regress on those, otherwise we're going to repeatedly hit this.

I think I prefer having standard header include last, as a way to detect some missing includes hidden by transitive inclusion. This can be enforced by the following clang-format option:

IncludeBlocks: Regroup
IncludeCategories:
  - Regex:           '"\.\.?/.*"'
    Priority:        2
    SortPriority:    2
  - Regex:           '<.*'
    Priority:        3

[AntoinePrv]

@serge-sans-paille this should be good now, though slightly larger.
I added the clang-format rule.

[DiamonDinoia]

I recently also added support for clang-tidy include what you use: flatironinstitute/finufft@67f0e7c

What so you think about incorporating it here?

If this is noise, feel free to discard the suggestion :-)

[serge-sans-paille]

how did we miss that one?

[AntoinePrv]

Indeed ! That why maybe randomized CI settings might be helpful to explore the combinatorial mix of settings. It could be done in a CRON job.

[serge-sans-paille]

I'm very happy with the automation of that task \o/

[AntoinePrv]

I recently also added support for clang-tidy include what you use: flatironinstitute/finufft@67f0e7c

What so you think about incorporating it here?

@DiamonDinoia I also very much want to have this in xsimd, but I am logging out for the next couple days. I'll merge this one so that it does not get outdated. Feel free to send a PR if mine does not come soon enough :)

[DiamonDinoia]

I recently also added support for clang-tidy include what you use: flatironinstitute/finufft@67f0e7c
What so you think about incorporating it here?

@DiamonDinoia I also very much want to have this in xsimd, but I am logging out for the next couple days. I'll merge this one so that it does not get outdated. Feel free to send a PR if mine does not come soon enough :)

No worries! I am about to logout for ~2 weeks :)