Fix macOS + Determinism check CI on master

[offdev]

What was red on master

Run 35240883610 (CI merge, 2026-09-17) had three failed jobs, two distinct root causes:

Job	Failure
macOS Intel + macOS arm64 (AppleClang)	samples/hello/hello-baseline.cpp:179: error: no matching function for call to 'min'
Determinism check (ubuntu-24.04)	src/laige-core/logging.cpp:500: error: ignoring return value of 'ssize_t write(...)' [-Werror=unused-result] — only in the job's Release g++ tree (build-det-release)

Root causes

macOS — std::min type deduction. std::min(emitted, baseline_.size()) mixes std::uint64_t with std::size_t. On macOS uint64_t is unsigned long long while size_t is unsigned long, so template deduction is a hard AppleClang error. On Linux/Windows the two types alias, which is why every other P0 job compiled the same source (and why this only surfaced on the first merge that runs the M1-DET-04 baseline machinery on macOS). Fix: cast the size_t operand to uint64_t so both arguments are one type on every platform.

Determinism check — warn_unused_result under fortification. glibc declares write() with warn_unused_result, and the attribute is attached only under fortification — i.e. optimized builds (that's the exact Debug-passes/Release-fails split: -g leaves __wur empty, -O3 enables it). A (void) cast does not suppress warn_unused_result, and the engine's -Werror makes it fatal. Clang/AppleClang/MSVC don't have this GCC diagnostic, so only the Release g++ tree failed. Fix: consume the result — the crash notice is best-effort, and there is no async-signal-safe fallback if the write fails (LOG-007).

Verification (local: g++ 16.2.1 / clang++ 22.1.8, x86-64 Linux)

• logging.cpp compiles clean under -Wall -Werror in Debug, Release, and Release + _FORTIFY_SOURCE=2 (reproducing the CI trigger before/after); hello sample TUs clean under all four compiler/flag sets.
• Full ctest: 82/82 — incl. hello_baseline_fpx/hello_baseline_fp32 + the mismatch/truncated/malformed/missing fixtures, replay and detcheck suites.
• The CI detcheck flow reproduced end to end: four targeted builds (g++ Debug, clang++ Debug, clang++ Debug+ASan, g++ Release), reference-baseline sanity on both backends, synthetic self-check, and pairs A (g++ Debug vs clang++ Debug) + B (ASan Debug vs Release) on both backends — all result=OK ticks=301.
• The macOS deduction error cannot be reproduced off-AppleClang (uint64_t == size_t on Linux/Windows); the P0 macOS jobs are the regression guard for it, as for the prior macOS CI fixes ([M1-HEAD-01] Fix macOS CI: platform-dependent zero-warn assertion in SecondRunFailsWithoutLogging #33, [M1-DET-01] Fix Windows CI: quote the trait_compile fixture compiler invocation #35, [M1-DET-02] Fix macOS/Windows CI: replay test fopen C4996; exact tick count under frame budget 1 #37, [M1-DET-03] Fix macOS/Windows CI: replay tool strncpy C4996; exact tick count for bounded laige-run; CRLF-tolerant replay check script #39).

Out of scope / follow-up

• docs/benchmarks/determinism-matrix.md marks macOS/Windows float_pinned_32 as pending first merge run; those rows update from the first green merge (matrix policy: entries change only by matrix results).
• If a pending platform's hello_baseline_fp32 desyncs, the documented policy is to mark it unsupported + skip — fixed_point_16_16 failures are never skipped.