[7ttp]

TL;DR

fixes the stack replaying supabase/roles.sql on a later start and running it as supabase_admin

what's missing?

adding or dropping a service between starts ran roles.sql again, so a plain create role failed with role "…" already exists

it also ran as supabase_admin instead of postgres, so a migration using a schema from roles.sql failed with permission denied for schema

now fixed by?

the stack runs it once on first setup, as postgres, like the legacy backend

visual rep

flowchart TD
  A["roles.sql<br/>create role app_user; create schema app"] --> B["first supabase start"]
  B -- before --> C["runs as supabase_admin<br/>app owned by supabase_admin"]
  C -.-> D["migration: create table app.t<br/>permission denied for schema app"]
  B -- now --> E["runs as postgres<br/>app owned by postgres"]
  E --> F["migration: create table app.t<br/>ok"]
  B --> G["stop, then start with another service set"]
  G -- before --> H["roles.sql runs again"]
  H -.-> I["role app_user already exists<br/>start fails"]
  G -- now --> J["roles.sql already applied<br/>start ok"]
  classDef t fill:transparent
  classDef bad fill:transparent,stroke-dasharray:5 4
  class A,B,C,E,F,G,H,J t
  class D,I bad

Loading

ref:

• missed in: feat(stack): replace the experimental runtime and integrate CLI commands #6684

[github-actions]

CLI benchmark (Lightweight Commands): no changes found

Commit 66dc64f against develop at 74a9afc · 10 runs each · report only, never blocks merging

Command	Workload	Before	After	Change
supabase --version	version	188 ms	188 ms	no real change
supabase --help	help	218 ms	211 ms	no real change
supabase init	init	201 ms	200 ms	no real change
supabase migration new bench	migration-new	197 ms	206 ms	no real change
supabase config convert	config-convert	224 ms	223 ms	no real change
supabase gen types --lang typescript --project-id replayfixtureprojref	gen-types	260 ms	254 ms	no real change
supabase projects list	projects-list	226 ms	222 ms	no real change
supabase functions list --project-ref replayfixtureprojref	functions-list	224 ms	226 ms	no real change

API calls: unchanged for all 8 commands.

Binary size: 135.0 MB, unchanged.

Details: timing ranges, API calls per command, slowest phases

Timing ranges

Command	Workload	Before range	After range	Change	Verdict
supabase --version	version	179-209 ms	172-199 ms	+0 ms (+0%)	within noise
supabase --help	help	199-247 ms	168-246 ms	-7 ms (-3%)	within noise
supabase init	init	180-230 ms	192-225 ms	-1 ms (0%)	within noise
supabase migration new bench	migration-new	190-221 ms	177-261 ms	+9 ms (+5%)	within noise
supabase config convert	config-convert	220-254 ms	199-236 ms	-1 ms (0%)	within noise
supabase gen types --lang typescript --project-id replayfixtureprojref	gen-types	213-278 ms	220-289 ms	-5 ms (-2%)	within noise
supabase projects list	projects-list	208-267 ms	205-271 ms	-4 ms (-2%)	within noise
supabase functions list --project-ref replayfixtureprojref	functions-list	210-257 ms	202-284 ms	+2 ms (+1%)	within noise

API calls per command

Command	Workload	Before	After	Endpoint changes
supabase --version	version	0	0
supabase --help	help	0	0
supabase init	init	0	0
supabase migration new bench	migration-new	0	0
supabase config convert	config-convert	0	0
supabase gen types --lang typescript --project-id replayfixtureprojref	gen-types	2	2
supabase projects list	projects-list	1	1
supabase functions list --project-ref replayfixtureprojref	functions-list	2	2

Slowest phases

No command moved beyond noise, so there are no phase deltas.

Workflow run

Comment /bench to run the lightweight benchmark (this comment) again on the latest commit, or /bench full to also run the integration-heavy benchmark, which posts in its own comment.

[github-actions]

🤖 AI Review

Both independent reviews completed with no findings. Inspection of the diff and surrounding code confirmed the overlay lifecycle, postgres credentials and ownership, webhook refresh, and shadow-cache invalidation. No actionable issues or fundamental conflicts with the pending next-branch work were found. Tests were inspected but not executed.

Findings

No issues found.

Stats

Claude findings: 0 · Codex findings: 0 · Confirmed: 0 · Refuted: 0 · Uncertain: 0

---

Models: claude-opus-5-5 + gpt-6.1-sol · Trigger: auto · Workflow run

This review runs once per PR. A maintainer can request another with a /ai-review comment.

[jgoux]

Looks good: the overlay now runs once on first setup as postgres, later starts still reapply webhooks, and the shadow cache key bump drops baselines built under the old owner.