Re: Add clang-format for libstdc++

[Jonathan Wakely <jwakely@redhat.com>]

On Tue, 7 Feb 2023 at 21:02, Thomas Rodgers via Libstdc++
<libstdc++@gcc.gnu.org> wrote:
>
> I am potentially interested in this, with an additional twist -
>
> Is it possible to get clang-format to enforce the libstdc++ uglification
> rules (probably requires a clang-format plugin though)?
>
> The concrete example I am thinking of here is when we adopt something like
> the PSTL. It was a substantial manual effort to do the uglification on that
> code in the first place, and despite multiple reviews by myself and
> jwakely, we still managed to miss some identifiers. To complicate matters,
> the upstream also makes ongoing changes while making a best effort to
> preserve the uglifcation. They also miss some uglfications. The main
> contributors in this case are shipping the PSTL as part of oneTBB and so
> don't have the same requirements to 'get it right' as libstdc++, so this
> often goes undetected.
>
> I am looking to rebase libstdc++'s PSTL in the near future so I'd be
> willing to try something based on clang-format, if it existed, for this
> particular purpose (I'm less interested in enforcing the indentation rules
> in this case, as I want to preserve the upstream's formatting).
>
> I also think it is potentially useful to have both the uglification and
> indentation combined. The case I'm thinking of is something like the <simd>
> contribution (or ranges) which are similarly large and need to get the
> uglification checks and in those cases, the correct libstdc++ formatting as
> part of the adoption process.


The best we have for this right now is the 17_intro/names.cc test.
Names that are used in the PSTL (e.g. "binary_op", "exec", etc.)
should be added to that file, so that if they appear without the "__"
prefix we get test failures.