[NativeAOT] Migrate GC bridge logging to printf

[jonathanpeppers]

Dependency

Depends on #12210 and targets its jonathanpeppers-native-printf-logging-next branch.

Scope

Migrates the five remaining easy NativeAOT-reachable formatted logging calls in src/native/clr/host/gc-bridge.cc to the canonical shared printf APIs introduced by #12155.

Behavior preservation

• Retains the LOG_GC category, existing gc_spew_enabled() and Util::should_log() gates, and the no-check write after the explicit gate.
• Preserves message text and control flow.
• Uses %zu for size_t values and PRIxPTR for handle-width hexadecimal values, including the explicit 0x prefix corresponding to {:#x}.
• Preserves null handling for Java class names with optional_string().

Non-goals

Does not migrate the three nearby std::format calls in bridge-processing.cc that feed OSBridge::_monodroid_gref_log; those require a separate API design.

Validation

• dotnet build src\native\native-nativeaot.csproj -p:AndroidSupportedTargetJitAbis=arm64-v8a -p:AndroidSupportedTargetAotAbis=arm64 -v:minimal
• dotnet build src\native\native-clr.csproj -p:AndroidSupportedTargetJitAbis=arm64-v8a -p:AndroidSupportedTargetAotAbis=arm64 -v:minimal
• git diff --check

[Copilot]

Pull request overview

This PR continues the NativeAOT/CoreCLR native-host logging migration by converting the remaining GC-bridge formatted log sites in src/native/clr/host/gc-bridge.cc from std::format-style logging to the shared printf-style logging helpers introduced earlier in the series.

Changes:

• Replace log_info(...) / log_info_nocheck_fmt(...) formatted calls with log_infof(...) and log_writef(...).
• Preserve existing GC-spew/category gating (Logger::gc_spew_enabled() and Util::should_log()), including the explicit “no-check” write after the gate.
• Use PRIxPTR + uintptr_t for handle-width hex formatting and add <cinttypes> for the format macro.

[jonathanpeppers]

/review

[github-actions]

✅ Android PR Reviewer completed successfully!

[github-actions]

✅ LGTM — clean, behavior-preserving migration

Reviewed the migration of the five remaining NativeAOT-reachable formatted logging calls in src/native/clr/host/gc-bridge.cc from the std::format-style APIs to the printf-style shared APIs.

Correctness verified against the API definitions:

Old call	New call	Gating behavior
log_info (macro → DO_LOG_FMT)	log_infof	Both check log_categories & category before writing — ✅ equivalent
log_info_nocheck_fmt	log_writef (..., LogLevel::Info, ...)	Both write unconditionally — ✅ equivalent, and correctly placed after the explicit Util::should_log (LOG_GC) gate

Format specifier mapping is correct:

• {} on size_t → %zu ✅
• {:#x} on handle → 0x%" PRIxPTR " with <cinttypes> added ✅ (and improved by using uintptr_t — see inline)
• {} on char* → %s with optional_string() guard ✅ (a real nullptr-crash fix — see inline)

Positive callouts:

• The LOG_GC category, gc_spew_enabled() / Util::should_log() gates, message text, and control flow are all preserved exactly as the PR description claims — verified against the full source.
• <cinttypes> (the C++ header, per repo convention) is the correct include for PRIxPTR.
• Two of the changes (optional_string guard and signed→unsigned handle) are genuine robustness/correctness improvements over the pre-migration code, not just mechanical swaps.

Notes:

• CI on the dnceng-public pipeline was still pending at review time (no checks reported yet) — this is not a blocker but should be green before merge, since the change is native code that only the build validates.
• As the PR states, the three std::format calls in bridge-processing.cc feeding _monodroid_gref_log are intentionally out of scope.

No blocking issues. Two 💡 suggestions posted inline highlighting the subtle improvements for future reference.

Generated by Android PR Reviewer for #12211 · 92.3 AIC · ⌖ 18.9 AIC · ⊞ 6.8K
_{Comment /review to run again}

[github-actions]

🤖 💡 Native memory — Good, necessary change: wrapping class_name in optional_string() guards the %s conversion. Host::get_java_class_name_for_TypeManager() genuinely returns nullptr on several paths (null/failed Class.getName, OOM on UTF8 conversion — see host-shared.cc), and passing a raw nullptr to a printf %s is undefined behavior (crash on most C libraries), whereas the old std::format {} on a char* silently formatted it. This migration is more robust than the code it replaces, not just equivalent.

Rule: Guard printf %s against nullptr (repo optional_string convention)

[github-actions]

🤖 💡 Native C++ — Nice correctness improvement: switching the handle from intptr_t to uintptr_t here means large 64-bit handle addresses (high bit set) now format as plain 0x... hex. The old std::format {:#x} on a signed intptr_t would render negative values as -0x..., which is misleading for a pointer/gref. The explicit 0x prefix + %" PRIxPTR " correctly mirrors the previous {:#x} output for typical values.

Rule: Use unsigned width-matched types for pointer hex formatting