Skip to content

test: run clean-wheel dataset CLI acceptance at default export scale - #260

Merged
codeasier merged 5 commits into
mainfrom
test/250-clean-wheel-scale
Oct 11, 2026
Merged

codeasier merged 5 commits into
mainfrom
test/250-clean-wheel-scale

Conversation

@codeasier

@codeasier codeasier commented Oct 11, 2026 •

Copy link
Copy Markdown
Owner

Depends on #255; this branch includes its infrastructure commits so the mandatory nightly selection works independently. Review/merge #255 first.

Summary

Refs #250. Builds on the reviewed #248 infrastructure (remote equivalent parent 68d89670b188c93e41b1a5d8f41903057e5ada06). The delta after that infrastructure contains only #250 acceptance tests, required-node registration and related guidance/tests.

  • Install the built wheel and dependencies in a genuinely clean venv, isolate HOME/CWD and source/focus/budget environment variables, and run the actual installed pt-snap console script.
  • Import exactly 1,000,000 test-owned events into two full default-capacity 500,000-event shards. Exercise all catalog entries: 11 supported dataset templates succeed with installed CLI/API equivalence and independent oracles; event_frames and active_memory_frame_tree_at_event preserve explicit dataset rejection. Peak-memory report succeeds and memory-tree report preserves its corresponding rejection.
  • Retain the distinct 500,002-event producer regression: a full first shard plus exactly two final real events, cross-shard completed lifetime, final survivor, exact member lifecycle membership and CLI/API boundary parity.
  • Require both exact nightly node IDs; reject missing, duplicated, skipped, failed or errored results. Record wheel/source/member hashes, interpreter/dependency identities, subprocess commands/exit codes and trend-only timings. Use separate evidence directories so neither gate overwrites the other.
  • Keep source and exported files unchanged. Default query work budgets remain untouched. No paid model, user session, live hardware capture or upstream GUI is involved.

Verification

Frozen local source: 33b1b5211d7e4f7efafc7d59fe54cbb0140cb3f7.

  • Actual integrated python .github/scripts/run_dataset_acceptance.py --wheel ... --output-dir ...: 2 passed, zero skips, 2672 non-nightly cases deselected, total 126.68 seconds.
  • Million-event case: 97.927 seconds; restored 500002-event case: 28.000 seconds (including its separate venv/install, import and queries). Boundary import itself: 12.448 seconds. These are measured observations, not timing gates.
  • Focused provenance/governance/release/CI-runner/wheel/package/CLI-API/workload suites: 158 passed.
  • Ruff, Black, wheel/source byte audit and git diff checks passed.
  • basedpyright: 0 errors, 1745 warnings.
  • Wheel was built after freezing the commit. SHA-256: 9e3d25ab2faa9ccad55344f7618f1d8dc41b4a091f8d6f022d1813bdc95adf89.

The full generic nightly runner reported both exact node IDs once, matched this checkout SHA and supplied wheel hash, and retained independent JUnit properties and command logs. Remote GitHub nightly verification is still required after publication; local results are not a claim that the remote schedule has executed.

Review scope

The frozen #250 million-event test body and oracle are AST-identical to da63c1d9a6b85fc4ebb947ea0ab2eaaefb073fdb; wheel harness and fast workload files are byte-identical. The shared fixture only gains separate evidence-directory selection for the additional boundary case. The #248 cherry-pick conflict was its nightly marker beside the new oracle helper; the frozen file was retained, leaving exactly one marker on the million-event test. Incremental additions are the separate boundary case, its mandatory node and negative runner tests, and truthful scope guidance.

Snapshot provenance decision: no-update
Snapshot provenance no-update reason: Changes are confined to tests, CI required-node wiring and guidance; no src/pt_snap_cli/snapshot files or committed executable fixtures change.

Independent incremental review approved exact tree ca4ba01d275daa1c9a824ad39723a04a94f30001: 158 focused tests passed, original million-test/oracle equivalence verified, distinct short-shard regression and fail-closed required-node checks reviewed, all 113 packaged files matched the verified wheel. Remote CI and remote manual/nightly execution are pending; local evidence is not labeled remote CI.

Copy link
Copy Markdown
Owner Author

Published a docs-only follow-up 5f95f1de51c68ff88eb235f634368bf2ccb19dcc: the EN/ZH acceptance guides now describe both the million-event and 500,002-event gates, fresh dependency-isolated venvs, and optional offline wheelhouse. Only those two guides changed; 75 local light checks and independent review passed. Prior head 16e7031 completed Test CI successfully (7 applicable jobs, Python 3.13: 2658 passed / 14 skipped / 2 deselected, 93.18% coverage). Latest-head CI is pending; remote nightly remains unexecuted.

Copy link
Copy Markdown
Owner Author

Latest docs-corrected head 5f95f1de51c68ff88eb235f634368bf2ccb19dcc is now verified: Test run 38148110336 succeeded with seven applicable jobs passed. Python 3.13: 2658 passed, 14 skipped, 2 deselected; 93.18% coverage. Wheel audit and clean-wheel offline acceptance passed. The installed-wheel dataset/nightly job was skipped on this PR event; local million/short-shard gate evidence remains separate.

Copy link
Copy Markdown
Owner Author

检视结论:approve

对照 #250:干净 venv + 真 pt-snap;2×50 万 + short-final-shard;11 模板 CLI/API。无 src/ 产品改动。

合入注意:与 #259 / #263 在 REQUIRED_TEST_NODES 冲突。建议在 #255/#258/#261 之后合入本 PR,再让 #259/#263 依次 rebase 并集节点。本评论不 approve/merge。

@codeasier codeasier left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM,已转为正式 PR。GitHub 不允许作者账号提交 APPROVE。建议在 #255/#258/#261 之后合,再让 #259/#263 rebase 并集节点。

Preserve both mandatory wheel gates and the existing differential/platform contracts. Exact union matches independently reviewed and locally executed preview.

Copy link
Copy Markdown
Owner Author

已集成 #261 后的实际 main e5ce037498af115c39cf4a4b2740e4f2be2d9968,新 head c28a5a2be9c00c2c5d2c95c19465b7f7e41ff267,树 46c0015ade1db8866db4b93982f77987ff257ec1。

六处共享基础上下文冲突按独立审阅的 preview 解决:保留当前 CI/differential/platform,加入两个 mandatory clean-wheel 节点,保留 short-final boundary 负例,双语说明并集。实际 main 基树与预览基树一致,解决后的完整树也与已验证 preview 完全相同;生产代码不变,scale/workload 测试与原审阅 #260 字节相同。

该 exact-tree 本地验证源 f011e7ee71de1b1bed9a84d1e74d6100313d5251:Python 3.12/3.13 各 92 focused + 9 source-scale 通过;实际 generic runner 两个 wheel gates 通过、零跳过,133.84s,wheel SHA256 8772466700475887c021b50573e4942a9d748142852cbec1f326a7f2d08246eb。wheel audit、Ruff、Black、type check 通过。新远端 head 的托管 Test/Platform CI 尚待验证;本地预览证据不替代远端 CI。

@codeasier
codeasier merged commit 58ca234 into main Oct 11, 2026
13 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant