Debugging and event-correlation integration validation — November 15, 2026

Executed on native aarch64 Linux in an Ubuntu 22.04 container:
  Linux 6.12.76-linuxkit, GCC 11.4.0, GNU GDB 12.1.
  python3 scripts/verify-debugging-labs.py --require-linux --gdb
The container used --output /evidence, bound to the local
output/debugging-validation-linux directory.

Passed:
  - Published C++11 source compiled with -Wall -Wextra -Wpedantic; no diagnostics.
  - direct/queued modes, each with baseline/source-gap/dispatch-wait scenarios.
  - Eight complete invocations and 48 records per capture.
  - Expected one-TID direct and two-TID queued relationships.
  - Controlled source and caller waits observed at the intended boundaries.
    A broad lower bound checks the injected wait, not a portable latency target.
  - Synthetic known stage durations and cross-thread interleaved sequences.
  - Fourteen malformed, incomplete, mixed-run, or wrong-thread traces rejected.
  - Invalid CLI arguments and /dev/full output failure rejected.
  - Published C++ member GDB session: Worker::process -> Dispatcher::dispatch
    -> main, seq=1, completed=0 at entry and 1 after return; normal exit.
  - Live procfs sample against a finite timing_lab child.
  - Counter-delta, process/task identity replacement, reversed-counter, and
    parenthesized task-name fixtures.

The live kernel reported scheduler statistics disabled or unknown. Runtime and
raw samples were retained; runqueue_wait_ns was null. No positive live runqueue
latency measurement or enabling of accounting is claimed. Enabled-counter
subtraction was checked with synthetic data only.

Portable analyzer/procfs fixtures also passed on macOS; Linux execution is
explicitly skipped there. --require-linux turns that skip into a failure.

Site checks passed with /execmap and empty URL prefixes: 20 tests each.
Existing browser suite: 13 passed and one intentional duplicate-viewport skip
per prefix. Four new reading pages also passed 320/390/1440 px overflow checks;
download requests succeeded and published assets match their tested sources.

Exact commands, compiler/OS versions, source/binary hashes, CSVs, raw procfs
samples and GDB logs are in output/debugging-validation-linux/report.json and
its companion files. Generated measurements are not part of the publication.
CI is configured to run the verifier; a hosted x86-64 CI run is not claimed as
already observed locally.

Not executed: perf/BPF capture or attach, remote GDB/IDE flows, optional
watchpoints and x86-64 entry-register checks, sanitizers, scheduler/frequency
changes, or lockup reproduction. Private application history and measurements
from the supplied report were not imported as laboratory evidence.