GH-50487: [C++] Extra semicolon warning from ARROW_SUPPRESS_DEPRECATION_WARNING macro with -Wpedantic

[thisisnic]

Rationale for this change

When building Arrow 25.0.0 from source and then building the R package, we see:

  /home/nic/arrow/r/libarrow/arrow-25.0.0/include/arrow/compute/api
  _vector.h:277:3: note: in expansion of macro
  'ARROW_SUPPRESS_DEPRECATION_WARNING'
       277 |   ARROW_SUPPRESS_DEPRECATION_WARNING
           |   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

The ARROW_SUPPRESS_DEPRECATION_WARNING and ARROW_SUPPRESS_MISSING_DECLARATIONS_WARNING macros have trailing semicolons after _Pragma directives, which produce empty statements and trigger -Wpedantic warnings wherever the macros are used.

What changes are included in this PR?

Remove unnecessary semicolons after _Pragma calls in the clang, GCC, and missing-declarations macro variants in macros.h.

Are these changes tested?

Existing tests cover these macros. The change only removes empty statements.

Are there any user-facing changes?

No.

• GitHub Issue: [C++] Extra semicolon warning from ARROW_SUPPRESS_DEPRECATION_WARNING macro with -Wpedantic #50487

[github-actions]

⚠️ GitHub issue #50487 has been automatically assigned in GitHub to PR creator.

[Copilot]

Pull request overview

This PR fixes -Wpedantic warnings caused by empty statements emitted from warning-suppression macros in Arrow C++ headers. It updates cpp/src/arrow/util/macros.h so _Pragma(...) uses no trailing semicolons, preventing “extra ‘;’” warnings at macro expansion sites.

Changes:

• Remove trailing semicolons after _Pragma(...) in the Clang and GCC ARROW_SUPPRESS_DEPRECATION_WARNING macro variants.
• Remove trailing semicolons after _Pragma(...) in ARROW_SUPPRESS_MISSING_DECLARATIONS_WARNING for GCC.

[pitrou]

I had no idea that it was possible to concatenate pragma statements like this. As long as it passes CI I'm fine with it :)

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit 96d8eb1.

There were 2 benchmark results indicating a performance regression:

• Commit Run on test-mac-arm at 2026-07-20 21:22:05Z
  • tpch (R) with engine=arrow, format=parquet, language=R, memory_map=False, query_id=TPCH-11, scale_factor=1
  • tpch (R) with engine=arrow, format=native, language=R, memory_map=False, query_id=TPCH-05, scale_factor=1

The full Conbench report has more details. It also includes information about 117 possible false positives for unstable benchmarks that are known to sometimes produce them.