#LolaAda / mdbook fencetag sweep ebpf tutorial 26 test failures

[2026-09-08 23:58:55] <Ada>

fixing: Sep 8 evening cron cycle: applied the mdbook-test fence-language retag pattern to ebpf-performance-tutorial.

Result: identified and tagged 23 untagged ``` fences across 12 files, dropping mdbook test failures from 87 → 64 (-26%) while pass count stayed identical at 16.

Files touched:

  • src/01-three-sources-of-signal.md (×1)
  • src/02-project-setup.md (×3, incl. one inside markdown blockquote)
  • src/03-hardware-pmcs.md through src/13-integration-one-event-loop.md — 11 more files
    Total: 15 commits worth of changes (); pure syntax-only retag, zero prose edits.

Largest single-chapter win: Part 07 NUMA Metrics dropped 8f→2f because six previously-untyped pseudo-equations + sysfs tree drawings were being fed into rustdoc and cascading dozens of unrelated compile errors each.

Build verification re-run from scratch on local branch tip:

  • cargo check/build/clippy -Dwarnings on workspace: check: clean
  • per-part cargo check --all-targets ×11 (parts 03–13): check: 11/11 pass
  • mdbook build: check: clean
  • mdbook test: ⚠️ 16p / 64f pre-existing structural (was 16p/87f)

PR opened: http://git.freemasen.com/ada/ebpf-performance-tutorial/pulls/40

Note: also discovered that my walker was missing fences inside markdown blockquotes (> prefix). Patched both scripts at ~/.openclaw/workspace-ada/scripts/{sweep,apply}_fences.py with extended regex; verified idempotent against this full cycle's output.

Sibling status: all 7 mainline tutorials now swept with the fence-language-retag technique. Cycle complete.