This change sets the job condition to false, effectively disabling this pipeline leg for all PRs/rolling builds. Even if intended for short-term investigation, it will prevent CI from validating unrelated changes in this pipeline and is very risky to merge. Please gate this behind a dedicated variable/parameter (e.g., a temporary feature flag) or revert before merging.

These platform matrix changes remove previously covered OSes (e.g., Windows/macOS) for this leg. That materially reduces test coverage and can hide regressions unrelated to the userevents investigation. Please avoid narrowing platforms in the main pipeline, or gate it behind a temporary variable so normal PR validation coverage is preserved.

DOTNET_DiagnosticServerDiag is being unconditionally enabled for every tracee run. This can significantly increase stderr output and potentially affect timing/behavior of the scenario under test, making results harder to interpret once the investigation is done. Consider enabling it only when a separate opt-in env var/command-line switch is set (or only under Helix/CI) so local/normal runs stay quiet by default.

ep_sess_diag_enabled is lazily initialized without any synchronization. session_tracepoint_write_event can be called from multiple threads, so concurrent first-use can cause a data race/UB in C. Consider using an atomic/once-init pattern (e.g., store the flag in a volatile int64_t and initialize via ep_rt_atomic_compare_exchange_int64_t, or always compute from getenv without caching) to make this thread-safe.

This code calls strerror(errno) but ep-session.c doesn't include the header that declares strerror (typically <string.h>). With warnings-as-errors this can fail the build due to an implicit declaration. Please include the appropriate header (and/or follow the repo’s existing include pattern for strerror).

ep_sp_diag_enabled is lazily initialized without synchronization. If this file’s logging is reachable from multiple threads during session setup/teardown, the first-use initialization can race and is undefined behavior in C. Please use an atomic/once-init pattern (or avoid caching) so the env-var check is thread-safe.

ds_diag_enabled is lazily initialized without synchronization. If multiple threads can reach this logging (e.g., server loop interactions), the first-use initialization can race and is undefined behavior in C. Please switch to an atomic/once-init pattern (or avoid caching) to make the env-var check thread-safe.

The record-trace stderr handler writes to stdout (Console.WriteLine) even though the prefix indicates stderr. This makes it harder to distinguish errors and may break any tooling that relies on stderr. Please write these lines to Console.Error (keeping the timestamp/prefix if desired).

ep_sess_diag_enabled is a non-atomic static that is lazily initialized from getenv and accessed on the event write path, which can be called from multiple threads. This is a C data race (undefined behavior). Please make the flag initialization thread-safe using the eventpipe runtime volatile/atomic helpers or a one-time init primitive.

ep_sp_diag_enabled is a non-atomic static lazily initialized via getenv and can be accessed concurrently during session/provider setup. This is a C data race (undefined behavior). Please switch to a thread-safe initialization (volatile/atomic helper or one-time init) for the diagnostic flag.

pr: none disables PR validation for this diagnostics pipeline. If this is only for local investigation, it should not be merged as-is; please restore the PR trigger (or gate it behind an opt-in variable) so normal PRs continue to get coverage.

pr: none disables PR runs of the linker-tests pipeline. This removes trimming safety validation for PRs and should be reverted before merging.

pr: none disables PR validation for the hardware-intrinsics pipeline. This should not be merged; please restore the PR trigger configuration so changes affecting JIT/HW intrinsics continue to get covered on PRs.

The tracee stderr handler writes to stdout (Console.WriteLine) even though the prefix indicates stderr. Please write these lines to Console.Error so failures are surfaced correctly and stderr remains meaningful in CI logs.

This matrix was narrowed to fewer platforms for investigation purposes. Merging this would permanently reduce PR coverage (e.g., dropping previously covered OS/arch legs) and can let platform-specific breaks slip through. Please revert this narrowing (or gate it behind an opt-in variable) before merging.

pr: none disables this pipeline for all PRs, removing an important validation signal. Temporary investigation changes like this should not land in mainline; please restore the previous PR trigger configuration before merging.