Skip to content

D241 — K5 reads the holder's session on the merge that carries a taken-over branch - #440

Open
napkinstack-agent[bot] wants to merge 1 commit into
mainfrom
d241-k5-merge-sessions
Open

napkinstack-agent[bot] wants to merge 1 commit into
mainfrom
d241-k5-merge-sessions

Conversation

@napkinstack-agent

Copy link
Copy Markdown
Contributor

Why

A deliverable is taken over (nstack claim D<n> --take-over, on the decider's word). The new session merges the default branch into the former session's branch and opens the pull request. K5 refused it: sessions were read with git log --no-merges, every non-merge commit names the former session, and the holder's only commit — the merge — was never read. K5's way out, moving the hold back, would have undone the take-over. Downstream, a pull request otherwise ready (record, sheet, CI) was blocked at merge time.

flowchart LR
  F[former session<br/>commits: Agent-Session former] --> B[branch d1-…]
  M[main moved] --> X
  B --> X[merge by the holder<br/>Agent-Session holder]
  X --> K{K5}
  K -- before: reads non-merge commits only --> R[refused: former only]
  K -- now: also reads merges' sessions --> P[passes: holder found]
Loading

What

  • merge_sessions(): the sessions the pull request's merges name. Unioned into the sessions passed to check_hold (K5) only. A sheet's authors stay the non-merge commits': a verifier who merely merged main is not made an author.
  • The former sessions are not added to the hold: that would let the former session alone deliver after a take-over, the opposite of it. Its commits alone are still refused (the test asserts it).
  • project.md (Who holds a deliverable), RULES.md (K5 row), the module docstring and register row D241 say it.

Nothing green turns red.

Evidence

  • test_k5_reads_the_holder_s_session_on_the_merge_that_brings_the_former_work (through pull_request.run, as CI runs it): red before, green now; it also asserts the former session's commits alone still fail K5.
  • uv run bash platform/tests/run.sh: Platform tests OK.
  • Fresh verifier: no defect. It checked in a throwaway repository that main's own merges are outside base..head with an up-to-date base, that GitHub's Update branch merge carries no trailer, and that a merged feature branch brings its non-merge commits, which already counted: no new way past K5. One note left out on purpose: identities() (the person's git identity) still ignores merges; the opener covers the person.

Closes #439.

🤖 Generated with Claude Code

…n-over branch

A deliverable taken over, its new session merges the default branch
into the former session's branch and opens the pull request: its only
commit is that merge, which K5 never read (sessions came from
non-merge commits). K5 now also reads the sessions the merges name. A
sheet's authors stay the non-merge commits'.

Closes #439.

Agent-Session: nstack-os-d241

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

This branch has not been deployed

No deployments
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.

K5 refuses a taken-over deliverable's pull request when the new holder's only commits are merges

0 participants