Adjust the standards for proto target to align with those of Protocol Buffers.

[owent]

Fixes #3799

Changes

Please provide a brief description of the changes here.

For significant contributions please make sure you have completed the following items:

• CHANGELOG.md updated for non-trivial changes
• Unit tests have been added
• Changes in public API reviewed

Analysis

The generated .pb.cc codes

inline constexpr EntityRef::Impl_::Impl_(
    ::_pbi::ConstantInitialized) noexcept
      : id_keys_{},
        description_keys_{},
        schema_url_(
            &::google::protobuf::internal::fixed_address_empty_string,
            ::_pbi::ConstantInitialized()),
        type_(
            &::google::protobuf::internal::fixed_address_empty_string,
            ::_pbi::ConstantInitialized()),
        _cached_size_{0} {}

port.h in protobuf:

// Take advantage of C++20 constexpr support in std::string.
class alignas(8) GlobalEmptyStringConstexpr {
 public:
  const std::string& get() const { return value_; }
  // Nothing to init, or destroy.
  std::string* Init() const { return nullptr; }

  template <typename T = std::string, bool = (T(), true)>
  static constexpr std::true_type HasConstexprDefaultConstructor(int) {
    return {};
  }
  static constexpr std::false_type HasConstexprDefaultConstructor(char) {
    return {};
  }

 private:
  std::string value_;
};

using GlobalEmptyString = std::conditional_t<
    GlobalEmptyStringConstexpr::HasConstexprDefaultConstructor(0),
    const GlobalEmptyStringConstexpr, GlobalEmptyStringDynamicInit>;

PROTOBUF_EXPORT extern GlobalEmptyString fixed_address_empty_string;

port.cc in protobuf

PROTOBUF_ATTRIBUTE_NO_DESTROY PROTOBUF_CONSTINIT
    PROTOBUF_ATTRIBUTE_INIT_PRIORITY1 GlobalEmptyString
        fixed_address_empty_string{};

If protobuf is built with C++17 but the generated .pb.cc files are built with C++20, protobuf exports GlobalEmptyStringDynamicInit fixed_address_empty_string; while .pb.cc expects GlobalEmptyStringConstexpr fixed_address_empty_string;, causing an unresolved external symbol.

On GCC/Clang, constexpr typically doesn’t require an extra symbol, so this mismatch doesn’t surface; MSVC’s behavior is right because constexpr is not a force rule, but it exposes the issue.

This didn’t happen before v30 because only the dynamic GlobalEmptyString existed. It was fixed in protobuf v32 by always using GlobalEmptyStringDynamicInit on MSVC, so the only affected case is MSVC with protobuf v31.

[Copilot]

Pull request overview

This PR adds CI validation to ensure OpenTelemetry C++ can be built with a different C++ standard than the protobuf dependency it uses, addressing issue #3799 where MSVC linking errors occurred when protobuf was compiled with C++17 but OpenTelemetry was built with C++20.

Key Changes:

• Adds new cmake.different_std.test CI target for both Linux (GCC) and Windows (MSVC) platforms
• Linux GCC test installs dependencies with C++17, then builds OpenTelemetry with C++23
• Windows MSVC test validates similar scenario with C++20 for OpenTelemetry

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 4 comments.

File	Description
ci/do_ci.sh	Adds new cmake.different_std.test target for Linux/GCC that builds with C++23 STL and enables OTLP file support
ci/do_ci.ps1	Adds new cmake.different_std.test target for Windows/MSVC that builds with C++20 STL, maintainer mode, and OTLP file support
.github/workflows/ci.yml	Adds two new CI jobs: one for GCC (installs protobuf with C++17, builds with C++23) and one for MSVC (tests different C++ standards)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.00%. Comparing base (25c923a) to head (5426ace).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3800      +/-   ##
==========================================
+ Coverage   89.94%   90.00%   +0.06%     
==========================================
  Files         225      225              
  Lines        7167     7167              
==========================================
+ Hits         6446     6450       +4     
+ Misses        721      717       -4     

see 2 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ThomsonTan]

nit: remove this commented out test?

[owent]

This test currently triggers an MSVC bug and causes an internal compiler error. Should we enable it once the MSVC compiler on GitHub runners is updated? I’m leaving it here so we can re-enable it quickly. It’s the only test that reproduces #3799; I can reproduce it locally with VS 2026, but that isn’t usable on gihutb right now. Alternatively, should we remove it and add it back later?