Add CodeQL analysis workflow configuration - #4
Merged
Merged
Conversation
|
You are seeing this message because GitHub Code Scanning has recently been set up for this repository, or this pull request contains the workflow file for the Code Scanning tool. What Enabling Code Scanning Means:
For more information about GitHub Code Scanning, check out the documentation. |
Pierre Malarme (pmalarme)
added a commit
that referenced
this pull request
Jul 9, 2026
Three non-blocking findings from an independent review (no blockers): 1. Drop the event-payload fallback in advance-on-push.yml guard #4. GitHub caps the push payload's .commits[] at 20 entries, so on a large force-push it could omit a delivery path and wrongly classify the push as machinery-only -> wrongly suppress a real advance (strand the participant). Use only git diff BEFORE..SHA (full-depth checkout, uncapped); when BEFORE is unreachable, treat the changed set as unknown and advance (legacy) rather than risk stranding. 3. Capture the classifier output in its own variable so a nonzero exit trips set -e (fail closed) instead of the `[[ "$(...)" == "true" ]]` comparison quietly evaluating false and advancing. 2. Add a drift test: every tracked top-level path in the real repo must classify as machinery (MACHINERY_PATHS) or known delivery (travel_assistant/ or a _root sibling), so a future root-level machinery file can't silently reintroduce the wrong-advance bug. 129 tests pass, lint 0 failures. Guard #4 re-simulated: machinery-only -> skip; mixed / overlay / unknown -> advance; classifier failure -> step aborts. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Pierre Malarme (pmalarme)
added a commit
that referenced
this pull request
Jul 9, 2026
* guard: don't advance on manual machinery-only pushes advance-on-push.yml only suppressed auto-advance for pushes carrying a [skip-advance] marker or a .workshop-state.json change. A participant who pulled the latest machinery from upstream and pushed it MANUALLY (no marker) was therefore bumped to the next step they never worked on. Add a fourth guard: classify the push's changed paths and skip advancing when they are ALL workshop machinery / platform files (.github/, .workshop/, Makefile, .devcontainer/, README.md, ...) and never the delivery (travel_assistant/ or its _root overlay siblings such as travel_toolbox/). The machinery list lives in advance_step.py as the single source of truth, exposed via a new `--is-machinery-only` mode that the workflow pipes the push's changed paths into, so the list is never duplicated in YAML. - advance_step.py: MACHINERY_PATHS, _is_machinery_path, _is_machinery_only_push, --is-machinery-only CLI mode - advance-on-push.yml: guard #4 using git diff (fallback: event payload union) - 00-intro.md: document the machinery-only skip in the troubleshooting note - tests: 12 new cases (classifier + CLI); 128 pass, lint 0 failures Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> * refactor: rename --is-machinery-only to --check-machinery-only The other advance_step.py modes are actions (--init, --reset, --on-push), so a boolean-adjective flag name read like a toggle. Rename the query-mode flag to a command-style --check-machinery-only for consistency; update the dest, the runner (_run_check_machinery_only), the workflow invocation, and the CLI tests. The boolean predicate helpers (_is_machinery_path / _is_machinery_only_push) keep the is_ prefix, which is correct for predicates. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> * harden machinery guard after rubber-duck review Three non-blocking findings from an independent review (no blockers): 1. Drop the event-payload fallback in advance-on-push.yml guard #4. GitHub caps the push payload's .commits[] at 20 entries, so on a large force-push it could omit a delivery path and wrongly classify the push as machinery-only -> wrongly suppress a real advance (strand the participant). Use only git diff BEFORE..SHA (full-depth checkout, uncapped); when BEFORE is unreachable, treat the changed set as unknown and advance (legacy) rather than risk stranding. 3. Capture the classifier output in its own variable so a nonzero exit trips set -e (fail closed) instead of the `[[ "$(...)" == "true" ]]` comparison quietly evaluating false and advancing. 2. Add a drift test: every tracked top-level path in the real repo must classify as machinery (MACHINERY_PATHS) or known delivery (travel_assistant/ or a _root sibling), so a future root-level machinery file can't silently reintroduce the wrong-advance bug. 129 tests pass, lint 0 failures. Guard #4 re-simulated: machinery-only -> skip; mixed / overlay / unknown -> advance; classifier failure -> step aborts. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> * ci: re-trigger checks (transient CodeQL runner-acquisition failure) Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Pierre Malarme (pmalarme)
added a commit
that referenced
this pull request
Jul 16, 2026
Address rubber-duck findings on the local --back backup engine: - Atomic publish: snapshot now renames the prior step-<N>/ aside, swaps the new snapshot in, and only deletes the old copy once the swap succeeds, with rollback on failure and staging cleanup on every failure path — so a mid-swap crash never leaves a step without a valid backup (#1). - Always capture the travel_assistant/ namespace, even a lone .gitkeep, so a root-only snapshot restores the placeholder instead of an empty dir (#2). - Protect repo machinery: _root overlay targets that resolve to MACHINERY_PATHS or .git are rejected up front, compared casefolded so case variants can't slip past on case-insensitive filesystems (#3). - Exact restore: clear each destination before copying so an undeclared stale target (e.g. left after a template sync) is replaced, not merged (#4). - Validate the backup manifest step is an integer before trusting it (#5). Add regression tests for each fix. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> Copilot-Session: fea011c8-1399-4f8a-a3bd-277408cae6c2
Pierre Malarme (pmalarme)
added a commit
that referenced
this pull request
Jul 16, 2026
* Add local --back step for the workshop engine Local learners who advance without committing each step had no way to move back. Add `advance_step.py --back` (plus a `make back` alias, README docs, and a `back-workshop.yml` CI pipeline) that restores the learner's saved work from the previous step's snapshot, or rebuilds the canonical step files when no snapshot exists. Namespace the backup layout so the agent snapshot and repo-root overlay targets never collide: `step-<N>/travel_assistant/`, `step-<N>/_root/`, and a `backup.json` completion manifest written last. This makes root-name collisions structurally impossible (#1) and lets `--back` restore structurally rather than by name. Snapshots are now published atomically (stage -> copy -> manifest -> swap) and fully replace any prior snapshot, so files deleted on a revisit are never resurrected (#3). Also: gate `--back` restore on a valid manifest (legacy/partial backups degrade to a canonical rebuild with a clear warning), reject `_root` overlay targets that shadow protected repo paths, and make reset/back/step backup dirs unique within the same second. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> Copilot-Session: fea011c8-1399-4f8a-a3bd-277408cae6c2 * Harden backup snapshot/restore per review Address rubber-duck findings on the local --back backup engine: - Atomic publish: snapshot now renames the prior step-<N>/ aside, swaps the new snapshot in, and only deletes the old copy once the swap succeeds, with rollback on failure and staging cleanup on every failure path — so a mid-swap crash never leaves a step without a valid backup (#1). - Always capture the travel_assistant/ namespace, even a lone .gitkeep, so a root-only snapshot restores the placeholder instead of an empty dir (#2). - Protect repo machinery: _root overlay targets that resolve to MACHINERY_PATHS or .git are rejected up front, compared casefolded so case variants can't slip past on case-insensitive filesystems (#3). - Exact restore: clear each destination before copying so an undeclared stale target (e.g. left after a template sync) is replaced, not merged (#4). - Validate the backup manifest step is an integer before trusting it (#5). Add regression tests for each fix. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> Copilot-Session: fea011c8-1399-4f8a-a3bd-277408cae6c2 --------- Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This pull request adds a new GitHub Actions workflow for CodeQL analysis to the repository. The workflow is set up to automatically scan the codebase for security vulnerabilities and code quality issues in both GitHub Actions and Python code. It will run on pushes, pull requests, and on a weekly schedule.
New CI/CD workflow for security analysis:
.github/workflows/codeql-analysis.ymlto enable automated CodeQL scanning for 'actions' and 'python' languages, triggered on push, pull request, manual dispatch, and a weekly schedule.