Add compile-time safe exception macros for profiler telemetry

[bouwkast]

What does this PR do?

This was generated with help from Claude Code; please review accordingly.

Introduces custom exception classes for the profiler's C code to safely distinguish between constant error
messages (safe for telemetry) and dynamic content (excluded from telemetry), with compile-time enforcement to
prevent accidental PII leakage.

Implementation:
Two exception types:

• Datadog::Profiling::ProfilingError - Constant messages → included in telemetry for aggregation
• Datadog::Profiling::ProfilingInternalError - Dynamic content (libdatadog errors, system state) → excluded from
  telemetry, preserved locally for debugging

Compile-time safe macros using libdatadog's string concatenation technique:

#define RAISE_PROFILING_TELEMETRY_SAFE(msg) \
  rb_raise(datadog_profiling_constant_error_class, "" msg)

#define RAISE_PROFILING_TELEMETRY_UNSAFE(fmt, ...) \
  rb_raise(datadog_profiling_dynamic_error_class, fmt, ##__VA_ARGS__)

The "" msg trick ensures compilation fails if a non-literal string is passed to TELEMETRY_SAFE, preventing
accidental inclusion of dynamic content in telemetry.

Updated 51 error sites across 18 C files (26 constant → TELEMETRY_SAFE, 25 dynamic → TELEMETRY_UNSAFE) and
modified telemetry logging to selectively include only ProfilingError messages.

Motivation:

Following #4985 which removed the pii_safe parameter, we need to ensure profiler error messages in telemetry
contain only known-constant strings for compliance with intake requirements.

The profiler communicates errors to telemetry (Ruby code) via exception messages. Using exception types to distinguish safe vs. unsafe content allows the telemetry layer to filter appropriately.

The compile-time enforcement addresses maintainability concerns by making it impossible to accidentally pass
dynamic content to telemetry-safe exceptions through copy-paste errors or misunderstanding.

Change log entry

None.

Additional Notes:

This approach enables telemetry fingerprinting and aggregation for constant error messages while maintaining full
debugging context locally for dynamic errors. The two-tier system provides semantic clarity about what's safe for
telemetry vs. what should remain local-only.

Special handling for private_vm_api_access.c which cannot include ruby_helpers.h due to header conflicts with
private VM APIs—uses manual extern declarations with explanatory comments.

How to test the change?

Existing test coverage validates:

• ProfilingError messages are included in telemetry (spec/datadog/core/telemetry/logging_spec.rb)
• ProfilingInternalError messages are excluded from telemetry
• Both exception types can be raised from C code
• Telemetry filtering logic works correctly with both types

Compile-time enforcement can be verified by attempting to pass a variable to RAISE_PROFILING_TELEMETRY_SAFE
(compilation will fail as expected).

All CI tests passing.

[datadog-datadog-prod-us1]

⚠️ Tests

⚠️ Warnings

❄️ 1 New flaky test detected

tests.parametric.test_config_consistency.Test_Stable_Config_Default.test_invalid_files[/etc/datadog-agent/application_monitoring.yaml, parametric-ruby] from system_tests_suite (Datadog)

Timeout of 60 seconds exceeded waiting for HTTP server to start. Please check logs.

ℹ️ Info

🧪 All tests passed

🎯 Code Coverage
• Patch Coverage: 100.00%
• Total Coverage: 98.37% (-0.05%)

View detailed report

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: c31665f | Docs | Datadog PR Page | Was this helpful? Give us feedback!}

[github-actions]

Typing analysis

Note: Ignored files are excluded from the next sections.

Untyped methods

This PR introduces 1 partially typed method, and clears 1 partially typed method. It increases the percentage of typed methods from 54.67% to 54.7% (+0.03%).

Partially typed methods (+1-1)

❌ Introduced:

sig/datadog/profiling.rbs:20
└── def self.try_reading_skipped_reason_file: (?untyped file_api) -> ::String?

✅ Cleared:

sig/datadog/profiling.rbs:14
└── def self.try_reading_skipped_reason_file: (?untyped file_api) -> ::String?

If you believe a method or an attribute is rightfully untyped or partially typed, you can add # untyped:accept to the end of the line to remove it from the stats.

[pr-commenter]

Benchmarks

Benchmark execution time: 2025-11-07 21:32:47

Comparing candidate commit f04f01d in PR branch steven/error-logs-remediation-custom-profiler-code with baseline commit 9c11f64 in branch master.

Found 0 performance improvements and 2 performance regressions! Performance is the same for 42 metrics, 2 unstable metrics.

scenario:profiling - Allocations (baseline)

• 🟥 throughput [-323589.970op/s; -311052.662op/s] or [-6.146%; -5.908%]

scenario:tracing - Propagation - Datadog

• 🟥 throughput [-2946.528op/s; -2871.924op/s] or [-9.301%; -9.065%]

[ivoanjo]

So I believe this has been superseded by #5076 so should we close this one?