Skip to content

Evidence publication fixture (keep open, do not merge) - #62

Closed
douglasjarquin wants to merge 1 commit into
mainfrom
evidence-fixture
Closed

douglasjarquin wants to merge 1 commit into
mainfrom
evidence-fixture

Conversation

@douglasjarquin

@douglasjarquin douglasjarquin commented Sep 6, 2026

Copy link
Copy Markdown
Owner

Evidence publication fixture (issue #35)

This draft pull request exists so evidence_publish.py has a real description to edit, verify, and inspect in a browser. Do not merge it.

Summary

Prose above the evidence block. A publish must keep this paragraph byte for byte.

Before/after evidence

Captured by the task worker (fixture) from candidate 46d424bebfbe against base 664df70cea6a; evidence run live-20260906T222651Z; verification run(s) live-second-publish. Worker media is a claim about the candidate build; it is not the coordinator's root verification.

counter.click (feature evidence): red-green

the base fails the user path and the candidate passes it

Before After
before screenshot 4758b9cfc85f after screenshot 75fb82a7d8fa
Before (screencast) After (screencast)
counter.click--before--f4455888a846.mp4
counter.click--after--a17d0128a607.mp4

Notes

Prose below the block. Also kept.

A reviewer added this paragraph by hand between publications.

douglasjarquin added a commit that referenced this pull request Sep 14, 2026
* docs: drop the manual evidence.publish-rendered row

Rendering of the published block is GitHub behavior; the offline suite
already covers the block format, receipts, prose preservation, and
failure handling against the fake gh. The draft fixture PR #62 is no
longer referenced.

* settings: an evidence block that opts out of automatic PR publication

Publication into the PR is on by default. The optional top-level
`evidence` block in .sum/settings.json carries one bool, `auto_publish`,
and `false` turns it off. `settings show` reports the effective value and
whether it came from the file or the default; `settings set` writes it
with --auto-publish-evidence true|false and removes it with
--clear-evidence.

* pr: publish a worker's before/after evidence into the reconciled PR

`sumctl pr evidence` was a stub. It now discovers every comparison.json
the worker's handoff lists, runs the runtime's own evidence_publish.py
once per run against the reconciled PR identity, and appends one
publication record per run carrying the outcome, destination, scenarios,
attachment count, and result path.

`--run` and `--visibility` are optional: absent, every discovered run is
published and the visibility comes from what gh reports for the
destination. The remaining flags that the command already declared
(--scenario, --evidence-root, --verification-run, --timeout, --dry-run,
--allow-head-mismatch, --replace-foreign-block) are honored.

A closed or merged PR is skipped, a recorded head mismatch is refused,
and a gh without --attach defers. Each is a record with a reason, and
none of them edits the PR.

* docs: evidence reaches the PR without a separate command

Record that publication is on by default and happens inside pr reconcile,
that the `evidence` settings block is the opt-out, and that every flag on
`pr evidence` is now optional. Name the Go test behind the two feature-map
rows.

* pr: drop the unread comparison list from the run model

---------

Co-authored-by: cs-test-runner <cs-test-runner@example.invalid>
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.

1 participant