Specialize template functions

I have a templated utility that takes two numeric types and returns their average. Compile-time errors occur when one argument is an integer type and the other a floating type. Specialize the template so calls like average(3, 4.0) and average(3u, 4u) compile and produce the mathematically correct type and value; prefer the wider type and preserve signedness rules. Check common edge cases: overflow for integers, NaN for floats.

Before I commit this set of template overloads for our math library, make approval easy for reviewers in the repo: surface a one-line summary of the overload resolution rules up top, show example calls and outcomes for mixed int/float and signed/unsigned pairs, and flag any surprising conversions or potential undefined behavior like integer overflow. Suggest the minimal compile-time tests reviewers will expect.

I just got a build break from average(3, 4.0) in the core math file. The CI error says no matching function for call. I know template specialization exists but I’m not certain which combination of overloads and enable-ifs will keep overload resolution clear and not bloat instantiations. Recommend the likely cause and the best next implementation: either add a forwarding overload that promotes to common_type, or SFINAE-constrain the original template. Explain trade-offs and a one-change patch to try first.

Across several modules we keep adding single-purpose template specializations to silence compile errors from mixed types. That creates brittle overload resolution and slow compile times. Where are we losing time and credibility, and what habit should we change? Recommend a consistent rule for numeric templates — for example: always accept parameters by value, immediately std::common_type_t them, and write a small set of well-documented wrappers — plus a checklist for new template additions to prevent future sprawl.