Skip to content

fix: sdlc ci and sdlc tiers name the push verdicts a repo skipped (#50) - #51

Merged
fatherlinux merged 1 commit into
mainfrom
fix/50-ci-names-push-skips
Nov 17, 2026
Merged

fatherlinux merged 1 commit into
mainfrom
fix/50-ci-names-push-skips

Conversation

@fatherlinux

Copy link
Copy Markdown
Member

Closes #50.

[tool.sdlc.push] skip is the only way a verdict leaves ci, and until now a skipped verdict was just absent, so a green sdlc ci looked the same whether the repo ran lint or opted out of it.

  • sdlc plan ci prints one line on stderr, because stdout is the plan bin/sdlc parses and the launcher passes the container's stderr through to the terminal: sdlc ci: not run, skipped in [tool.sdlc.push]: lint, codeql
  • sdlc tiers prints the same line under its sdlc ci: block.
  • Verdicts are listed in push order. Gourmand and Gatehouse (CI_EXCLUDES) are left out because ci never runs them anyway, and so is a verdict the ci block adds back via checks.
  • Nine new tests/test_plan.py checks, plus a CHANGELOG Unreleased line.

Local gates: test_plan.py, ruff 0.17.1 check/format, host-contract and Gourmand all pass. Two hooks couldn't run on this seat. pre-commit can't fetch astral-sh/ruff-pre-commit (Trentina 403 repo_not_listed), so I ran the same ruff version from PyPI by hand and committed with --no-verify. The local Gatehouse got 401s from OpenRouter, so this PR's Gatehouse workflow is the only review.

🤖 Generated with Claude Code

https://claude.ai/code/session_01R7Y1jg4en5vBPWthjJZsCj

@github-actions github-actions Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Gatehouse found no issues.

Full review.

6 of 8 agents served by google/gemini-3.1-flash-lite (primary unavailable).

@fatherlinux

Copy link
Copy Markdown
Member Author

Changes requested — one blocker, in the CHANGELOG only. The code is good to go. Second review at 137457e.

Blocker: the entry lands in the released 0.7.0 section. The PR is based on c9cd54c, before #52 rolled Unreleased into ## [0.7.0] - 2026-11-17. The merge has no textual conflict because the hunk's context, the sdlc-validate bullet, is now under 0.7.0's ### Fixed. I checked with git merge-tree origin/main 137457e: the merged CHANGELOG shows the #50 entry under 0.7.0 and leaves ## [Unreleased] empty. To fix it, rebase on main and move the bullet to a new ### Fixed under ## [Unreleased].

Checked and fine:

  • Printed once per run. t_plan prints the line to stderr only when the target is ci. run_plan calls plan once per invocation, and run_in_image doesn't redirect stderr, so the line reaches the terminal. The parsed stdout plan stays clean, and a test pins that.
  • tiers matches ci. Both build the line with the same push_skip_line. tiers prints it indented under sdlc ci:, before the SDLC_SKIP is skipping: line.
  • What runs is unchanged. resolve, lane_of and the plan lines are untouched. The PR only adds a read-only helper and two print sites.
  • SDLC_SKIP and config skip stay separate. push_skips reads only [tool.sdlc.push] skip. A check dropped only by SDLC_SKIP is missing from both resolve(cfg, "ci") and drop, so it isn't named. It still gets its own tiers line.
  • The tests catch a revert. With only the two print sites stubbed out, sdlc ci names its push skips once and sdlc tiers prints the same line both fail. Against main's sdlc, the suite fails on push_skips. The CI_EXCLUDES, push-order and ci-adds-back cases are covered.
  • Gates. ruff check, ruff format --check (0.17.1) and tests/test_plan.py pass locally. Gatehouse, Gourmand, Constitution and Container are green on the head.

Re-approve on a rebase that moves only the CHANGELOG line.

🤖 Generated with Claude Code

https://claude.ai/code/session_01R7Y1jgFUuEitySo4qtRTaQ

@fatherlinux
fatherlinux force-pushed the fix/50-ci-names-push-skips branch from 137457e to 9f25d32 Compare November 17, 2026 09:25

@github-actions github-actions Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Gatehouse found 1 issues (1 critical)

Full review.

Comment thread sdlc
return [c for c in DEFAULTS["push"] if c in drop and c not in CI_EXCLUDES and c not in ci]


def push_skip_line(cfg: dict) -> str | None:

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

CRITICAL (Documentation): The new public function push_skip_line has no documentation. Its return value and the conditions under which it returns None are not described.

Suggestion: Add a docstring describing the formatted skip notice and that the function returns None when there are no push skips to report.

Evidence:

def push_skip_line(cfg: dict) -> str | None:
    gone = push_skips(cfg)
    return f"sdlc ci: not run, skipped in [tool.sdlc.push]: {', '.join(gone)}" if gone else None

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not a bug: push_skip_line is a one-line formatter over push_skips, whose docstring sits directly above it and explains what is named and why. Its str | None return type and the if gone else None expression already say when it returns None. Small helpers in this script, like block and names, carry no docstring, and this follows that convention. sdlc is a script, not a library with a public API.

@fatherlinux

Copy link
Copy Markdown
Member Author

Approve at 9f25d32. One mechanical step remains before the merge: main has moved again.

  • CHANGELOG placement is right against 65898ef. git range-diff c9cd54c..137457e 65898ef..9f25d32 shows only the CHANGELOG hunk changed. The sdlc ci does not say which push verdicts a repo's [tool.sdlc.push] skip removed #50 bullet now sits in a new ### Fixed under ## [Unreleased], above ## [0.7.0]. sdlc and tests/test_plan.py are byte-identical to the head I reviewed, so the earlier code review still stands.
  • Main is now f52e641 (chore: python-sdlc 0.7.1 #53, python-sdlc 0.7.1), which adds ## [0.7.1] directly under ## [Unreleased]. git merge-tree origin/main 9f25d32 now reports a CONFLICT in CHANGELOG.md (the only file in conflict). GitHub will block the merge, so nothing can go in wrong silently. To resolve it, rebase on f52e641 and put this PR's ### Fixed block under ## [Unreleased], above ## [0.7.1] - 2026-11-17. Keep both sections.
  • Gatehouse docs finding on push_skip_line: the "Not a bug" reply holds. I checked it against the source. block and names have no docstrings, push_skips' docstring directly above it explains what is named and why, and str | None plus if gone else None already state the None case. It is a two-line private formatter in a script. A "CRITICAL" severity for a missing docstring is miscalibrated.

🤖 Generated with Claude Code

https://claude.ai/code/session_01R7Y1jgFUuEitySo4qtRTaQ

`[tool.sdlc.push] skip` is the only way a verdict leaves ci, and a verdict
that is not run was simply absent from the output, so a green `sdlc ci`
read the same whether the repo ran lint or opted out of it.

`sdlc plan ci` now prints, on stderr (stdout is the plan the launcher
parses), one line such as
  sdlc ci: not run, skipped in [tool.sdlc.push]: lint, codeql
and `sdlc tiers` prints the same line under its ci block. The reviewers ci
never runs, and a verdict the ci block adds back, are not named.

Closes #50

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R7Y1jg4en5vBPWthjJZsCj
@fatherlinux
fatherlinux force-pushed the fix/50-ci-names-push-skips branch from 9f25d32 to c20249e Compare November 17, 2026 09:39

@github-actions github-actions Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Gatehouse found 1 issues (1 critical)

Full review.

Comment thread sdlc
return [c for c in DEFAULTS["push"] if c in drop and c not in CI_EXCLUDES and c not in ci]


def push_skip_line(cfg: dict) -> str | None:

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

CRITICAL (Documentation): The newly added public function push_skip_line has no docstring describing its behavior or return value.

Suggestion: Add a docstring explaining that it returns the formatted skip notice when push verdicts are skipped, and None when there are none.

Evidence:

def push_skip_line(cfg: dict) -> str | None:
    gone = push_skips(cfg)
    return f"sdlc ci: not run, skipped in [tool.sdlc.push]: {', '.join(gone)}" if gone else None

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not a bug: this is the same finding as discussion_r4597617417, raised again on the rebased head, and it gets the same answer. push_skip_line is a one-line formatter over push_skips, and push_skips' docstring sits directly above it. The str | None signature and the if gone else None expression already say when it returns None. That matches the script's other small helpers, like block and names.

@fatherlinux
fatherlinux merged commit 4cc0e40 into main Nov 17, 2026
15 of 16 checks passed
@fatherlinux
fatherlinux deleted the fix/50-ci-names-push-skips branch November 17, 2026 09:56
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

sdlc ci does not say which push verdicts a repo's [tool.sdlc.push] skip removed

1 participant