Reconcile 30 unpushed local commits; quality_gates.yml has never run #112

Activity

1. jeremymanning

   added

   P0-criticalBlocks everything; safety or correctness landmineBlocks everything; safety or correctness landmine

   on Aug 17, 2026
2. jeremymanning

   added a parent issue on Aug 17, 2026

   • Master Plan: Get Clustrix Production-Ready #108
3. jeremymanning

   added a commit that references this issue on Aug 17, 2026

   Issue #98: Preserve in-flight epic working files for #103

   ...

   43bbfaa
4. jeremymanning commented on Aug 17, 2026

   

   jeremymanning

   on Aug 17, 2026

   MemberAuthor

   More actions

   Decision recorded: option (b) — pushed to a branch for review. PR #128.

   29 commits are now on origin/epic/test-coverage-90-percent (3fa3627..aa9d80d). The work is off a single laptop and reviewable. PR #128 is explicitly marked do not merge as-is, so Phase 1/2 (#114, #117) can triage which tests survive rather than merging ~2,500 mock occurrences into master immediately before the issue that removes them.

   A live PyPI token blocked the first push

   GitHub push protection rejected it:

   remote: - GITHUB PUSH PROTECTION
   remote:     - Push cannot contain secrets
   remote:       —— PyPI API Token ——
   remote:        commit: 6c918ca
   remote:        path: .ccpm_backup/.claude_20250828_140214/settings.local.json:30
   

   Question	Answer
   Ever public?	No — every commit reachable from origin/master was scanned; absent from all
   Blast radius	11 local commits from 6c918ca onward (tracked file, so every later tree carried it)
   Source	.ccpm_backup/ — an accidental 245-file, 1.2 MB backup of local .claude/ settings, never tracked on origin
   Fix	Rewrite scoped to origin/master..HEAD only, so public history was never touched; .ccpm_backup/ now gitignored

   .git was backed up before the rewrite, and refs/original/refs/heads/master -> 4feda53 holds the pre-rewrite state.

   This is a gap in the #108 audit worth noting: the security track scanned the working tree and origin/master history for AWS/HF/generic patterns, but not the unpushed commits, and PyPI tokens were not in the pattern set. Push protection caught what the audit did not — which is itself the argument for enabling secret scanning (#111).

   Remaining AC on this issue

   • Decision recorded with rationale
   • Work no longer exists only on the laptop (lives on a pushed branch, per the AC's second clause)
   • quality_gates.yml still returns 404 — it is on the epic branch, not master, so it will not run until Issue #112: Surface the test-coverage epic work for review (do not merge as-is) #128 (or a subset of it) lands
   • .claude/epics/test-coverage-90-percent/ progress files still claim completion; reconcile during Issue #112: Surface the test-coverage epic work for review (do not merge as-is) #128 review
   • Untracked WIP (103-*.md, updates/103/) committed to the branch rather than lost

   Local master still sits ahead of origin/master with the same (rewritten) commits. That is now redundant with the pushed branch — say the word and I will reset it to origin/master, or leave it if you prefer to keep working there.
5. jeremymanning commented on Aug 17, 2026

   

   jeremymanning

   on Aug 17, 2026

   MemberAuthor

   More actions

   Resolved — the work is off the laptop and reviewed.

   Sequence:

   1. 29 commits pushed to origin/epic/test-coverage-90-percent (blocked first by a live PyPI token; .ccpm_backup/ stripped from the 11 affected local commits, public history never touched).
   2. PR Issue #112: Surface the test-coverage epic work for review (do not merge as-is) #128 opened as the triage surface.
   3. Reviewed in depth, then closed without merging — see the review comment on Issue #112: Surface the test-coverage epic work for review (do not merge as-is) #128.
   4. Local master reset to origin/master (now a9393b7, with tests/integration/ is unmarked, so the documented unit-test command provisions billable AWS EKS clusters #109 merged). git rev-list --count origin/master..master → 0.

   The branch is retained on the remote, so nothing is lost.

   Why it will not merge

   • All 637 new tests live outside tests/unit/, so none ever run in CI — the epic's own goal is unmet, and the PR edits the README coverage badge 10% → 5%.
   • It adds ~2,678 lines of production modules with zero importers (notebook_magic_{aws,azure,gcp,ssh,enhanced}.py) — precisely what Delete ~5,100 lines of orphaned modules (13% of the package has zero importers) #122 deletes — which also import the mock shim Remove mock-awareness from shipped code (production branches on isinstance(..., Mock)) #116 removes.
   • It contains an unreviewed behavior change to @cluster: comprehensions become auto-parallelization candidates (loop_analysis.py, 787→1712 lines). Now tracked separately as Unreviewed behavior change: comprehensions become @cluster auto-parallelization candidates #132.
   • +695 mock occurrences, ~28% growth on the 2,513 that Replace assertion-free mock tests with tests that actually execute (2,513 mock occurrences) #117 must remove.

   Follow-ups filed

   • pytest.ini is dead config that silently shadows pyproject.toml (--strict-markers, testpaths, markers all inert) #130 — pytest.ini is dead config shadowing pyproject.toml (found while red-teaming; explains why the addopts story differed between branches)
   • Extract the salvageable pieces from the closed test-coverage epic (PR #128) #131 — extraction plan for the salvageable pieces
   • Unreviewed behavior change: comprehensions become @cluster auto-parallelization candidates #132 — the comprehension-parallelization behavior change

   Remaining AC on this issue

   • Decision recorded with rationale
   • No unpushed work on the laptop
   • Untracked WIP preserved on the branch
   • quality_gates.yml still 404s — it lives only on the closed branch. Tracked in Extract the salvageable pieces from the closed test-coverage epic (PR #128) #131, to land non-blocking (it has never actually executed: it needs two jobs that both fail).
   • .claude/epics/test-coverage-90-percent/ progress files still claim completion for work that will not merge — reconcile or archive.
6. jeremymanning commented on Aug 17, 2026

   

   jeremymanning

   on Aug 17, 2026

   MemberAuthor

   More actions

   Closing: the P0 concern — 30 commits existing only on one laptop — is resolved. They are on origin/epic/test-coverage-90-percent and git rev-list --count origin/master..master is 0.

   The two leftover ACs are not P0 and are tracked where they belong:

   • quality_gates.yml landing non-blocking → Extract the salvageable pieces from the closed test-coverage epic (PR #128) #131
   • .claude/epics/ progress files reconciled → Extract the salvageable pieces from the closed test-coverage epic (PR #128) #131

   Keeping this open as P0 would misrepresent what is actually urgent.
7. jeremymanning

   closed this as completedon Aug 17, 2026