Repository navigation
fix: shared fixes so the dbt Power User feature PRs merge in any order - #34
Conversation
…ort, so they hold whichever features are built A fake CLI in a shared test now answers the data tools the features ask for (comments, health, defer state, cost estimate, run results, docs coverage) with an empty result before it records its own calls. The chain and feature index tests check that the call passes through unchanged and is not held, with no log noise, rather than that every feature is a stub. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… what the CLI printed Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ar time Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…t records their failures The test now builds its own engine with a recording log, working state, a CLI that answers only the data tools with an empty result, and the session, clock, store and ui calls the features make. A control in the same test checks that a throwing feature function is logged through it, so an empty log means the features ran without failing. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… failure logged Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
review-loop: APPROVED at 6d4e6da (reviewers: gpt-6-astra high, gpt-6.1-sol high) |
Code Review SummaryStatus: No Issues Found | Recommendation: Merge Incremental review of Overview
Files Reviewed (1 file in this increment)
Previous Review Summaries (2 snapshots, latest commit 590a869)Current summary above is authoritative. Previous snapshots are kept for context only. Previous review (commit 590a869)Status: No Issues Found | Recommendation: Merge Previous findings at Note: Files Reviewed (7 files)
Previous review (commit 6d4e6da)Status: 5 Issues Found | Recommendation: Address before merge Overview
Issue Details (click to expand)WARNING
SUGGESTION
Files Reviewed (9 files)
Reviewed by deepseek-v4.1-flash-1017 · Input: 34.7K · Output: 3K · Cached: 341K |
… blanks before a cd; fakes answer feature calls before recording; the no-failure test checks failures only Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
review-loop: APPROVED at 590a869 (reviewers: gpt-6-astra high, gpt-6.1-sol high) |
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
review-loop: APPROVED at 267104a (reviewers: gpt-6-astra high, gpt-6.1-sol high) |
saravmajestic
left a comment
There was a problem hiding this comment.
Approving. The nothingToReport helper gives each feature's CLI call an empty answer, so the shared tests no longer assume features are stubs. The strict "no feature fails in any phase" test with its positive control is a good guard. The runTool first-line fallback and the linear-time cd pattern both look right.
Merge this before the feature PRs. You already listed the remaining cross-PR test interactions (#27 with #24 in runs.test.tsx, #26 with #33 in comments.test.tsx); those need fixing in the feature PRs' own test files before they're merged together.
Summary
One change to files shared by the eight dbt Power User feature PRs (#24, #25, #26, #27, #28, #31, #32, #33). With it, each feature PR merges onto main in any order with the shared tests green, and none of them needs to touch a shared file.
What changes
hooks/inert-features.ts(nothingToReport), gives each data tool a feature asks the CLI for an empty answer:docs_share_comments,dbt_health,dbt_defer_state,dbt_cost_estimate,dbt_run_results,dbt_docs_coverage.lineage.test.tsx,features/sql-graph.test.tsxandtask-impact.test.tsxanswer those calls first, so the calls they record are only their own.features/index.test.tsdrives every phase through a hand-built engine that has what the built features call. It records each failure throughui.log, and a positive control checks that a throwing feature is logged. A separate strict test asserts that no feature fails in any phase. That assertion sits where no feature PR's edit to the file can reach it, so a throwing feature can't go unseen after a merge.tool-chain.test.ts, a dbt command and a non-dbt command both reach the tool unchanged and neither is refused. The non-dbt one comes back with no added context. A dbt command may get a note from a feature that watches dbt, so that isn't asserted for it.altimate-cli.tsrunTool). A crash or a plain-text failure has no answer object; it now shows its first line (up to 500 characters) instead of empty text. A JSON answer still speaks for itself.lineage.tsx). Thecdpattern's blanks stop at a newline. A 64,000-line blank run went from about 2.5 s to about 1 ms. There are tests for each kind ofcdand for the long run.Skipped
tool-chain.ts. It's correct and was tested, but it edits the same lines as feat(runs): record each dbt run's results and show them in /runs #27 and would conflict with it. It can land after feat(runs): record each dbt run's results and show them in /runs #27.plugin.json, the README settings and health lines, and a/to-modelsection. They describe features as working. That is false on main and on most combinations of the feature PRs, so the release package documents each one when its feature ships.integrations.tsxandtask-impact.tsx, and a comment inaltimate-cli.tsthat is still true on main. None changes behaviour.Verification
git merge-tree --write-treewith each of the eight feature branches: no conflicts.claude plugin test .,tests/*.shandpytest tests/all pass on main plus this branch: 405 tests, 7/7 shell tests, 65 pytest. They also all pass with each of the eight feature branches merged on its own.Merging all eight feature PRs together fails 33 tests, all in two feature-owned test files. These are interactions between feature PRs that this PR can't fix without conflicting with them:
runs.test.tsxfails 31 when runs-pane (feat(runs): record each dbt run's results and show them in /runs #27) and cost-gate (feat: hold dbt run, build and show over the BigQuery scan limit until allowed #24) are both merged. Cost adds a note to every dbt command, and runs asserts its context lines exactly.comments.test.tsxfails 2 when team-comments (feat: team comments from shared dbt docs in Claude's context and a /comments pane #26) and health (feat: project health: new DataPilot issues on changed models, in the band, with an optional push gate #33) are both merged. Health's session-startdbt_healthcall shows up in the CLI calls comments records.Each fix belongs in that PR's own test file, for example answering feature calls with
nothingToReport. feat(runs): record each dbt run's results and show them in /runs #27 can also drop itsruns ·log filter infeatures/index.test.ts.Merge this PR before the feature PRs.
🤖 Generated with Claude Code