# mssql-django Gated PR Self-Check Use this prompt before merging any backend/compiler/schema PR. ## Objective Run a **merge gate** that prevents regressions from: - contract ambiguity, - double-escaping/double-transforms, - fast-path branches bypassing shared safety logic, - stale test exclusions, - insufficient integration proof. The output must be a clear **PASS / FAIL / BLOCKED** decision with evidence. ## Inputs - PR branch name - Target branch (usually `dev`) - Changed files (`git diff --name-only ...HEAD`) - Related Django test modules for changed areas ## Required Checks (in order) ### 1) Contract change declared For every behavior-changing helper/symbol, declare one explicit contract: - who returns raw values, - who escapes/normalizes, - where transformation happens, - and whether call sites must change. ### 2) All call sites audited Audit all consumers of changed symbols and verify: - contract applied exactly once, - no mixed old/new behavior, - no hidden secondary escaping/transform. ### 3) Fast-path invariants preserved If a fast-path exists (special-case `continue`/`return`/short-circuit), prove it does not bypass shared safety logic: - ORDER BY dedupe, - alias handling, - parameter ordering, - escaping rules, - pagination/order requirements. If it does bypass shared logic, fix by integrating with shared path, not by adding ad-hoc string post-processing. ### 4) Integration test proof Provide end-to-end proof for each behavior change: - Add local regression tests under `testapp/tests/` when applicable. - Run upstream Django tests that exercise the same path. - Include at least one regression that targets the **risky edge** (duplicate/equivalent path, alias path, mixed type path, etc.). ### 5) Doc/comments synced Ensure comments and PR description match current behavior: - remove stale “still failing” language if test now passes, - add concise reasons for any remaining exclusions. ### 6) Exclusion hygiene (strict) For any exclusion touched in `testapp/settings.py`: - Verify candidate removals under **normal settings** (`testapp.settings`), not only fast profile. - If using `settings_fast`, treat as exploratory only; confirm final decision with `testapp.settings`. - Keep only exclusions that are proven failing due to true SQL Server limitation or out-of-scope known issue. ### 7) Version matrix evidence Run targeted tests plus HOT guard matrix and report: - PASS lanes, - FAIL lanes, - BLOCKED lanes with concrete environment reason. ## Merge Gate (must all be true) - No unresolved contract ambiguity. - No double-escaping / double-transform pattern in call sites. - No fast-path branch bypassing shared invariants without explicit handling. - Integration regression proof exists for each behavior change. - Exclusion changes are validated under `testapp.settings`. - Targeted tests green + HOT guard green (or BLOCKED with environment reason). - PR description includes root cause, fix scope, and exclusions touched. ## Recommended Evidence Commands ```bash # Diff scope git --no-pager diff --name-status origin/dev...HEAD git --no-pager diff --stat origin/dev...HEAD # Local backend tests python manage.py test testapp.tests --verbosity 2 # Targeted local module python manage.py test testapp.tests.test_jsonfield --verbosity 2 # Upstream targeted test(s) cd django PYTHONPATH=/workspaces/mssql-django/django python tests/runtests.py --settings=testapp.settings # HOT matrix / CI guard # Run or verify the HOT matrix / CI guard workflow for this branch in your CI system # and record its status (PASS / FAIL / BLOCKED) in the gate summary. ``` ## Required Output Format ### Gate Summary - Overall: PASS / FAIL / BLOCKED - Scope: changed files + behavior areas ### Checklist Results - Contract change declared: PASS/FAIL (+ evidence) - Call-site audit: PASS/FAIL (+ evidence) - Fast-path invariants: PASS/FAIL (+ evidence) - Integration test proof: PASS/FAIL (+ tests run) - Doc/comments sync: PASS/FAIL (+ note) - Exclusion hygiene: PASS/FAIL (+ list kept/removed) - Version matrix evidence: PASS/FAIL/BLOCKED (+ lanes) ### Exclusions Decision Table For each touched exclusion: - test id - decision (keep/remove) - reason - proof command - result ### Merge Recommendation - Merge now / Needs fixes - If fixes needed, list minimal required changes. ## Notes specific to mssql-django - `testapp/runners.py` marks excluded tests as expected-failure (`x`) under normal settings. Do not treat `x` alone as proof of backend failure for exclusion decisions. - For exclusion removal decisions, always re-run candidate tests directly under `testapp.settings` and inspect actual pass/fail outcomes. - Prefer expression-level/compiler-level fixes over SQL string surgery.