[danielrazavi]

What changes are you trying to make?

Restore Assignment 1 SQL execution feedback alongside the existing Assignment 2 grader. October 9 commit 3fe0023c0af279b275f746948ad08134a631a25f replaced the old shared runner. This restores A1 in a separate workflow and runner, leaving all existing A2 files unchanged.

• Automatic A1 feedback for changed assignment1.sql files on student-owned repositories, on assignment-one / assignment-1 and A2 branch aliases. Mixed A1/A2 submissions on A2 branches run both workflows independently.
• Separate A1 report, artifact, and PR feedback, plus manual branch/PR inputs. Multiple cohort files are rejected rather than selecting one arbitrarily.
• Restore historical multistatement execution and three-row previews; handle missing answers, malformed markers, fresh failure reports, quoted semicolons, and bounded query execution on an isolated database.

What did you learn from the changes?

The March 24–April 5 runner (cf333a0, 0349f73, 3445382, f7e32c2, ae700d8) executed SQL and displayed results; it did not check correctness or award marks. Immediately before October 9, it was manual-only and prioritized A2 over A1 when both files changed. Restored feedback explicitly labels executable SQL as requiring instructor review.

Was there another approach?

A separate A1 workflow avoids modifying the current A2 workflow, grading rules, regression tests, or report format. Shared-workflow routing would have required changing current A2 behavior.

Challenges and conflicts

No merge conflicts against main at cf65aa97590eba771031c052e08db3b62f5b5d8f. This is a new branch adding four files; no current files are overwritten.

Automatic runs retain A2's student-repository owner restriction; upstream UofT-DSI use is manual. A2's existing branch restrictions are unchanged, so A2 does not automatically run from an A1 branch. GitHub token restrictions can prevent PR comments; manual runs with appropriate permissions are the fallback. The selected submission ref must include the restored runner/tests, and manual selection requires an unmerged assignment1.sql change against main.

How were these changes tested?

• Local regression suite: 14 passed, 14 skipped (7 new A1 checks and 7 available A2 checks; 14 existing A2 tests require A2_SOLUTION_RUBRIC, which was unavailable).
• Both checked-in A1 cohort templates produced 12-question reports.
• Existing A2 runner end-to-end check produced a complete 12-question report; its pytest invocation passed.
• New workflow YAML parsed; embedded feedback script tested for normal, malformed-report, and oversized-feedback cases.
• Database fingerprints verified unchanged; external attachment denied; long SQL interrupted; mixed A1/A2 changed-file selection verified, including a path containing spaces.
• git diff --check passed. Local tests used Python 3.9.6; workflow specifies Python 3.11. GitHub-hosted execution and actual PR comment delivery have not been exercised.

Related task

Urgent Trello restoration task

See .github/scripts/assignment1-feedback.md for provenance, usage, and limitations.

Checklist

• Local checks confirm restored execution feedback and unchanged A2 behavior within the available tests.
• Verify a student-repository GitHub Actions run and PR feedback delivery before deployment.

[github-actions]

Thanks for your contribution! 🎉

Please remember to tag or request a review from the DSI team. Give us up to 72 hours to review your pull request. We appreciate your patience and efforts.