Skip to content

test(e2e): record the oauth-software-publishers kind and settle its live checklist - #1360

Merged
vscheuber merged 1 commit into
mainfrom
test/oauth-software-publishers-e2e
Nov 16, 2026
Merged

vscheuber merged 1 commit into
mainfrom
test/oauth-software-publishers-e2e

Conversation

@vscheuber

@vscheuber vscheuber commented Nov 16, 2026 •

Copy link
Copy Markdown
Contributor

Part of #1357. This is the maintainer's recording and live-check session for the kind oauth-software-publishers (#1343), run on a development tenant after #1358 (agent files in the plain shape). The issue stays open for the maintainer to read the outcomes and decide; nothing in it is closed or ticked here.

What is in the pull request

One commit: the e2e tests config-manager-{push,pull,delete}-software-publishers, their three snapshots, nine recordings (three per command: am, openidm, environment) and the fixture test/e2e/exports/fr-config-manager/cloud/realms/alpha/realm-config/agents/SoftwarePublisher/zz-e2e-publisher.json, which is a real pull of the throw-away object (plain shape, example.com values, no key). Recorded in the order push, pull, delete (FRODO_TEST_NAME pushOne, pullOne, deleteOne, the commands are in the header comment of each test); afterwards the tenant no longer held zz-e2e-publisher. No production code, no other snapshot and no other recording changed. The pull all replay needs no new sentence: its comment already names this step (#1343).

Outcomes of the live checklist of #1357

Only the throw-away publishers zz-live-publisher and zz-e2e-publisher were written; the realm also holds a publisher that is not a throw-away and was only read; the listings of agents, trusted JWT issuers, CIMD profiles and OAuth2 clients are identical before and after. In short:

  • 1, 2, 3, 4, 6, 8, 9, 10: as expected. An empty realm answers No software publishers found. (exit 0), and -n of a missing one exits 1. A publisher is created from a hand-written file without _type. The pulled file is plain, has no _rev, and a named pull equals the pull of the realm byte for byte; two pulls give no diff; it equals the agent object of frodo agent export (which adds _rev and a volatile meta block). Delete, then push the pulled file: the object is created again and equals the earlier pull (the second write left nothing out). A file of another _type is refused before anything is sent. pull all has the publisher once and a second run gives the same publisher folder; pull all-static has no agents folder. pull oauth2-agents and push oauth2-agents skip the type with one warning and send no request to an agents path.
  • 5: overrides and variables. The pull writes the placeholder; a push without the variable fails naming it, exit 1, and sends nothing; with the variable in the process environment (feat(config-manager): pin frodo-lib 5.0.0-47, push fills placeholders from the process environment, a refused issuer attribute fails the push #1341) only the one value changes.
  • 7: a question for the maintainer, not decided here. AM accepts and stores a private key in jwkSet. The pull writes it into the file and warns, without ever printing it; an override placeholder takes it out of the file; a push of a file without jwkSet removes it from the object. Core 0013 says secrets are never exported in the clear: should the pull refuse or redact jwkSet that holds private key material by default? (The vendor documentation describes jwkSet as the publisher's public keys.)
  • 11: partly run. Named --dry-run, named delete and the second delete (exit 1, not found) behave as written. The option is --dry-run only: the -d in the issue text does not exist. Not run: the unnamed delete, the -y bulk form and the refusal without a terminal, because the realm holds a publisher that is not a throw-away; the unit tests and the delete guard tests cover them.
  • 12: partly. No tool reads an agent type's schema (CLI, library and MCP server have none). The single read of an existing publisher shows seven properties and no password-type one, and the documentation lists none; so there is no secret property, but see 7.

Decisions taken (reversible)

Verification

  • Replay of the three tests offline with a scratch HOME: 1 passed each, 2 + 3 + 2 snapshots written; every snapshot read (the pull snapshot is the plain fixture; push and delete show the experimental notice and the one deletion line).
  • npm run check exit 0; npx tsc --noEmit exit 0; npm run test:guidance 3 suites, 693 passed.
  • Recordings scanned for tokens, keys and passwords: the only credential-shaped header is the masked Bearer <bearer token>; no private key, JWT or password value. As in the existing recordings, the login read of the service account object contains its description (which names the person who created it) and its public key; no recording was edited.
  • Whole suite on this head under the machine-wide lock, with a scratch HOME: exit 0, 1100 suites passed (10 skipped), 10798 tests passed (149 skipped), 10286 snapshots passed, no failure and no obsolete snapshot. The pull request's own CI runs it as well.

Not verified

Unnamed and bulk deletes on a tenant; the schema of the type (no tool); inherited: true (none in a publisher); CIMD or issuer behaviour (other sessions).

…lishers kind

Recorded on a development tenant with the throw-away software publisher
zz-e2e-publisher in the alpha realm, in the order push, pull, delete; the
fixture is a real pull of that object (plain shape). Part of #1357.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
@vscheuber vscheuber added the no-auto-processing Concierge never touches this item; remove the label to hand it over label Nov 16, 2026
@vscheuber
vscheuber merged commit 74f0515 into main Nov 16, 2026
23 checks passed
@vscheuber
vscheuber deleted the test/oauth-software-publishers-e2e branch November 16, 2026 23:00
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

no-auto-processing Concierge never touches this item; remove the label to hand it over

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant