Michael Warres

> Repro of g++-specific compile error with variadic templates, which compiles successfully with -std=c++17 but not with -std=c++20: https://godbolt.org/z/TT9a9YPf7 Changing [this constructor declaration](https://github.com/proxy-wasm/proxy-wasm-cpp-sdk/blob/95bb82ce45c41d9100fd1ec15d2ffc67f7f3ceee/proxy_wasm_api.h#L1210) from: ``` Counter(std::string_view name, MetricTagDescriptor... descriptors) ```...

clang-tidy failure may be https://github.com/llvm/llvm-project/issues/56709

> > Repro of g++-specific compile error with variadic templates, which compiles successfully with -std=c++17 but not with -std=c++20: https://godbolt.org/z/TT9a9YPf7 > > Changing [this constructor declaration](https://github.com/proxy-wasm/proxy-wasm-cpp-sdk/blob/95bb82ce45c41d9100fd1ec15d2ffc67f7f3ceee/proxy_wasm_api.h#L1210) [...] > appears to...

> I distinctly remember we had to appeal this way in envoy - there might be a patch there already. Indeed! https://github.com/envoyproxy/envoy/blob/6648db236f64218611f36e91285716c8bec67abb/bazel/proxy_wasm_cpp_sdk.patch

The only remaining CI failure is due to clang-tidy itself crashing, which I hope is transient. So I think this is ready for review, @martijneken and/or @PiotrSikora. Thanks!

This PR is obsoleted by #439.

I think this issue is obsolete now, runners are all updated.

I don't feel strongly, but as discussed offline, my preference would be for separate hostcalls, to make usage/dependency visible to static analysis, and also to leave open flexibility for variation...

Sorry, missed this earlier, taking a look

#143 discusses the same (or similar) issue, suggest deduping and reviving the discussion there, unless there is a different aspect to this issue. @wbpcode WDYT?