fix(e2e): correct stale provisioning docs + cli.rs drift, skip suite when staging unseeded #446

Merged

nrupard merged 1 commit from fix/PMS-656-e2e-provisioning-cleanups into main

2026-07-13 19:31:41 +02:00

nrupard commented

2026-07-13 19:25:22 +02:00

Owner

What

Three E2E-provisioning cleanups (PMS-656), surfaced while re-provisioning the shared E2E account after a staging data wipe.

1. Stale docs (docs/e2e.md, e2e/README.md)

Both instructed the operator to elevate the E2E account with UPDATE users SET role = 'admin' because "first SSO login JIT-creates the row as technician". Not true since MAPPS-330: place_bunyip_user (src/modules/auth/middleware.rs) JIT-creates the shadow row as admin in an auto-provisioned personal tenant, and effective_role_from_bunyip floors every bunyip role to mokosh admin. Rewrote the provisioning steps to describe the bunyip-provisioned account (just e2e-bootstrap), the automatic JIT-admin + personal tenant, and re-capturing E2E_TENANT_ID after a re-seed. The "everything admin-gated returns 403 -> elevate" troubleshooting entry is rewritten as no-longer-applicable.

2. cli.rs drift

src/cli.rs claimed the mokosh-bootstrap binary was "removed in PMS-493", but src/bin/mokosh-bootstrap.rs still exists and drives bootstrap-infisical. Corrected to say PMS-494 folded the subcommands into the main binary while the standalone binary remains (per the issue's YAGNI call: fix the comment, do not delete a working binary).

3. CI readiness gate

After a wipe the suite hard-failed opaquely at login (the incident that prompted this). e2e.yml now probes bunyip's GET /e2e-bootstrapped (e2e/scripts/check-e2e-bootstrapped.mjs) before running; on an explicit {bootstrapped:false} it sets skip=true so the suite is skipped with a notice pointing at the re-seed recipe and the job stays green, mirroring bunyip's own CI. Fail-open: a missing endpoint (production never enables it), a non-2xx, a parse error, or any network error all run the suite - the gate can only hide the known "staging not seeded" case, never a real failure.

Testing

just pre-commit green (the only code change is the cli.rs doc comment).
node --check on the new gate script; e2e.yml parses and the gate is wired (Check staging is E2E-seeded -> Run E2E suite gated on steps.seeded.outputs.skip != 'true').
Docs-only + CI-config change otherwise; no runtime behavior change to the server.

#PMS-656

## What Three E2E-provisioning cleanups (PMS-656), surfaced while re-provisioning the shared E2E account after a staging data wipe. ## 1. Stale docs (docs/e2e.md, e2e/README.md) Both instructed the operator to elevate the E2E account with `UPDATE users SET role = 'admin'` because "first SSO login JIT-creates the row as technician". Not true since MAPPS-330: `place_bunyip_user` (`src/modules/auth/middleware.rs`) JIT-creates the shadow row as `admin` in an auto-provisioned personal tenant, and `effective_role_from_bunyip` floors every bunyip role to mokosh admin. Rewrote the provisioning steps to describe the bunyip-provisioned account (`just e2e-bootstrap`), the automatic JIT-admin + personal tenant, and re-capturing `E2E_TENANT_ID` after a re-seed. The "everything admin-gated returns 403 -> elevate" troubleshooting entry is rewritten as no-longer-applicable. ## 2. cli.rs drift `src/cli.rs` claimed the `mokosh-bootstrap` binary was "removed in PMS-493", but `src/bin/mokosh-bootstrap.rs` still exists and drives `bootstrap-infisical`. Corrected to say PMS-494 folded the subcommands into the main binary while the standalone binary remains (per the issue's YAGNI call: fix the comment, do not delete a working binary). ## 3. CI readiness gate After a wipe the suite hard-failed opaquely at login (the incident that prompted this). `e2e.yml` now probes bunyip's `GET /e2e-bootstrapped` (`e2e/scripts/check-e2e-bootstrapped.mjs`) before running; on an explicit `{bootstrapped:false}` it sets `skip=true` so the suite is skipped with a notice pointing at the re-seed recipe and the job stays green, mirroring bunyip's own CI. **Fail-open**: a missing endpoint (production never enables it), a non-2xx, a parse error, or any network error all run the suite - the gate can only hide the known "staging not seeded" case, never a real failure. ## Testing - `just pre-commit` green (the only code change is the `cli.rs` doc comment). - `node --check` on the new gate script; `e2e.yml` parses and the gate is wired (`Check staging is E2E-seeded` -> `Run E2E suite` gated on `steps.seeded.outputs.skip != 'true'`). - Docs-only + CI-config change otherwise; no runtime behavior change to the server. #PMS-656

nrupard added 1 commit

2026-07-13 19:25:22 +02:00

fix(e2e): correct stale provisioning docs + cli.rs drift, skip suite when staging unseeded

E2E / Playwright against staging (pull_request) Successful in 1m42s

Details

Check / fmt + clippy + build + tests (pull_request) Successful in 6m15s

Details

Create release / Gate (release-branch merges only) (pull_request) Successful in 1s

Details

Create release / Create release from merged PR (pull_request) Has been skipped

Details

Integration / integration tests (pull_request) Successful in 8m40s

Details

abbe8c747a

Three E2E-provisioning cleanups surfaced while re-provisioning the shared account after a staging wipe.

Docs were stale: docs/e2e.md and e2e/README.md instructed the operator to elevate the E2E account with `UPDATE users SET role = 'admin'` because "first SSO login JIT-creates the row as technician". That has not been true since MAPPS-330: `place_bunyip_user` (src/modules/auth/middleware.rs) JIT-creates the shadow row as `admin` in an auto-provisioned personal tenant, and `effective_role_from_bunyip` floors every bunyip role to mokosh admin. Both docs now describe the bunyip-provisioned account (`just e2e-bootstrap`), the automatic JIT-admin + personal tenant, and re-capturing E2E_TENANT_ID after a re-seed; the "everything admin-gated returns 403 -> elevate" troubleshooting entry is rewritten as no-longer-applicable.

src/cli.rs claimed the `mokosh-bootstrap` binary was "removed in PMS-493", but src/bin/mokosh-bootstrap.rs still exists and drives `bootstrap-infisical`. Corrected to say PMS-494 folded the subcommands into the main binary while the standalone binary remains.

CI had no readiness gate: after a wipe the suite hard-failed opaquely at login (this incident). e2e.yml now probes bunyip's GET /e2e-bootstrapped (e2e/scripts/check-e2e-bootstrapped.mjs) before running; on an explicit {bootstrapped:false} it skips the suite with a notice pointing at the re-seed recipe and the job stays green, mirroring bunyip's own CI. Fail-open: a missing endpoint (production) or any probe error runs the suite, so the gate can only hide the known "staging not seeded" case, never a real failure.

#PMS-656

nrupard scheduled this pull request to auto merge when all checks succeed

2026-07-13 19:29:09 +02:00