Repository navigation
fix(ci) + test(observability): bundle CI/QG fixes with PR #774 drain - #846
Merged
Merged
Conversation
statistics.quantiles() requires at least 2 data points and raises StatisticsError when called with a 1-element list. The existing guard (`if not rows`) only covered the empty-list case; a single-element corpus passed through and crashed. Extend the guard to `if len(rows) < 2` so both zero-row and one-row inputs return the 3.0 fallback. Exposed by test_single_row_falls_back in ai/tests/test_calibrate_phase_f_recipes_unit.py (shipped in #845). The production bug predated #845 (present at f21d9bc). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Raise statement coverage from 89.1 % to 91.3 % by targeting the specific uncovered branches identified via `go test -coverprofile`: - WaitForShutdown SIGTERM signal-delivery path (was only context-cancel path tested). - InitOTel with OTEL_TRACES_SAMPLER_ARG in [0.0, 1.0]: covers the `sampleRatio = parsed` assignment branch (0.0, 0.5, 1.0 boundary cases). - SetControllerSources half-nil cases: q=non-nil+r=nil registers only queue gauges; q=nil+r=non-nil registers only the node gauge. - Prometheus registry isolation: two independent Metrics instances on separate registries do not bleed counter increments into each other (ADR-1014 isolation property). - OTel attribute key strings locked to ADR-0782 schema values to catch any accidental rename before it reaches OTLP consumers. ADR-0108 deliverables: - no digest needed: trivial (pure test coverage fill-in) - no alternatives: only-one-way fix (add the missing test branches) - AGENTS.md: invariant notes added for SIGTERM test, half-nil cases, and attr-key schema lock - changelog.d/added/observability-coverage-gaps.md - docs/rebase-notes.md: no rebase impact (test + doc only) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This was referenced Jun 8, 2026
lusoris
added a commit
that referenced
this pull request
Jun 8, 2026
…ts + ffmpeg patch style (#847) * chore(changelog): backfill 30 missing fragments for PRs #806–#846 Adds one-line changelog.d fragments for 30 merged PRs in the #806–#846 range that shipped without a fragment. PRs #805, #807, and #825 were already covered by existing ADR-numbered fragments (1090, 1093, 1099); PRs #836, #839, and #840 were covered by topic-named fragments (go-rust-ci-red-bundle, cuda-done-path-double-unref-ort-coverage, 0840-pic-pool-odr-cuda-gpumask-cov-floor). All 30 new files verified by scripts/release/concat-changelog-fragments.sh with no new warnings (pre-existing chore/perf/refactor/ section warnings are unrelated to this change). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * chore(coverage): ratchet ort_backend.c floor back to 83% (PR #844 error-injection completed) PR #840 lowered the ort_backend.c per-file coverage floor from 83 to 79 because the ADR-0922 target was above the structural ceiling: ORT error paths are unreachable without error injection. PR #844 added 8 ORT error-injection tests, bringing measured coverage to 84%. The floor is now safely ratcheted back to 83 (1 pp slack against 84% measured). Also updates the T-COVERAGE-ORT-FLOOR-OVERSHOOT-2026-06-08 row in docs/state.md to reflect the completed two-step fix (PR #840 temporary reset + PR #844 error-injection + this ratchet). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * chore(ci): raise pytest timeouts for coverage-gpu/mcp-smoke/dnn (avoid silent hangs) coverage-gpu: add --timeout=300 --timeout-method=signal to CUDA and SYCL pytest calls (previously bare; GPU kernel init can stall silently within 12-min budget). mcp-smoke: add --timeout=60 --timeout-method=signal to socket-opening tests (previously bare; 12-min job budget gives no per-test guard). dnn: add --timeout=60 --timeout-method=signal (non-critical hygiene). Also install pytest-timeout in the dnn, mcp-smoke, and coverage-gpu venvs. coverage job --timeout=180 already correct (PR #840 precedent, unchanged). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * chore(ffmpeg-patches): align 0010 include path to <libvmaf/libvmaf_cuda.h> (matches SYCL/HIP/Metal pattern) Patch 0010 used bare `libvmaf_cuda.h` in both the configure pkg_config probe lines and the C #include guard in vf_libvmaf.c. Every other backend patch (0003 SYCL, 0011 HIP, 0012/0013 Metal) uses the subdirectory-prefixed form `libvmaf/<name>.h`, which is the installed path under the include prefix that pkg-config and compilers actually see. The bare form works today only because the CUDA header happens to be installed both ways on the current dev machine; it would silently break on a stricter sysroot (e.g. the container build, a cross-compile host) where only `libvmaf/libvmaf_cuda.h` is present. Three occurrences updated (no functional change, style-consistency fix): configure: -enabled libvmaf ... libvmaf_cuda.h → libvmaf/libvmaf_cuda.h configure: +enabled libvmaf_cuda ... libvmaf_cuda.h → libvmaf/libvmaf_cuda.h configure: +enabled libvmaf ... libvmaf_cuda.h → libvmaf/libvmaf_cuda.h vf_libvmaf.c: #include <libvmaf_cuda.h> → #include <libvmaf/libvmaf_cuda.h> Series-replay verification (git am --3way against pristine n8.1) is deferred; patch 0010 is not standalone-applicable — it builds on 0001-0009. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> --------- Co-authored-by: Lusoris <lusoris@pm.me> Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Bundles 2 CI fixes (from #845 tip reds) with PR #774's Go observability coverage tests. Avoids 2 CI cycles.
CI Fixes
fix(ai): extend_ugc_tight_interval_widthguard fromif not rowstoif len(rows) < 2—statistics.quantiles()requires at least 2 data points; a single-element corpus raisedStatisticsError. Pre-existing production bug first exposed bytest_single_row_falls_back(added in chore(bundle): drain 8 rebased DRAFT PRs (docs + helm + tests-coverage) #845 at commit a42e4ea).PR #774 — Go observability coverage 89.1% → 91.3%
branch-coverage fill-ins for pkg/observability: SIGTERM signal-delivery path, OTEL sampler ratio assignment, SetControllerSources half-nil cases, Prometheus registry isolation, and OTel attribute key schema lock (ADR-0782).
Test plan
_ugc_tight_interval_width([])→ 3.0,_ugc_tight_interval_width([single_row])→ 3.0, multi-row → in [1.5, 3.5] — all PASSgit grep '^<<<<<<'andgit grep '^>>>>>>'both cleanDeep-dive deliverables (ADR-0108)
python -c "import statistics; statistics.quantiles([1.0], n=4)"→ StatisticsError (confirms the bug); after fix,_ugc_tight_interval_width([row])→ 3.0state.md touch