Files
supabase/apps/studio/tests
Andrew Valleteau da3b0cb3ec fix(cli): login creating duplicate tokens on re-renders (#46120)
## I have read the
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
file.

YES

## What kind of change does this PR introduce?

Bug fix

## What is the current behavior?

The CLI login page was firing `POST /platform/cli/login` multiple times
per page load due to an unstable `navigate` reference in the parent
component re-triggering the effect. This caused several duplicate
personal access tokens to be created for each browser sign-in attempt
(regression CLI-1491).

Additionally, error messages from non-Error rejection shapes (like those
from openapi-fetch) were being replaced with a generic "Unknown error"
message instead of surfacing the actual platform error.

## What is the new behavior?

1. **Prevents duplicate API calls**: Added a `useRef` guard
(`startedForSessionIdRef`) to ensure `createCliLoginSession` is only
called once per `sessionId`, even if the parent component re-renders
with a new `navigate` reference.

2. **Improved error handling**: Created a `getErrorMessage()` utility
that properly extracts error messages from both Error instances and
plain objects (e.g., `{ message: string, statusCode: number }`),
allowing platform error messages to surface instead of generic
fallbacks.

3. **Added comprehensive test coverage**:
- E2E test verifying the POST fires exactly once per page load with
realistic network latency
   - E2E test confirming platform error messages are displayed correctly
   - Unit test for non-Error rejection shapes
- Unit test verifying the effect doesn't re-trigger on parent re-renders

## Additional context

The fix addresses the root cause by:
- Using a ref to track which `sessionId` has already been processed,
preventing re-execution when deps change
- Extracting error message handling into a reusable utility that handles
both Error instances and plain objects
- Adding tests that specifically check for the regression (multiple POST
calls and error message display)

The E2E test includes a 400ms delay in the mock response to simulate
real-world conditions where the original bug only surfaced after React
committed multiple re-renders.

Closes: CLI-1491

https://claude.ai/code/session_01GAujw33MTRBRYSnS8cxcEa

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

* **Bug Fixes**
* Improved CLI login error handling to display platform-specific error
messages instead of generic fallback text.
  * Fixed duplicate login session creation during component rerenders.

* **Tests**
  * Added test coverage for error message display in CLI login.
* Added regression tests for single session creation and platform error
scenarios.

<!-- review_stack_entry_start -->

[![Review Change
Stack](https://storage.googleapis.com/coderabbit_public_assets/review-stack-in-coderabbit-ui.svg)](https://app.coderabbit.ai/change-stack/supabase/supabase/pull/46120?utm_source=github_walkthrough&utm_medium=github&utm_campaign=change_stack)

<!-- review_stack_entry_end -->

<!-- end of auto-generated comment: release notes by coderabbit.ai -->
2026-05-20 09:17:22 +02:00
..

UI Testing Notes

Rules

  • All tests should be run consistently (avoid situations whereby tests fails "sometimes")

  • Group tests in folders based on the feature they are testing. Avoid file/folder based folder names since those can change and we will forget to update the tests.

Examples: /logs /reports /projects /database-settings /auth

Custom Render and Custom Render Hook

customRender and customRenderHook are wrappers around render and renderHook that add some necessary providers like QueryClientProvider, TooltipProvider and NuqsTestingAdapter.

Generally use those instead of the default render and renderHook functions.

import { customRender, customRenderHook } from 'tests/lib/custom-render'

customRender(<MyComponent />)
customRenderHook(() => useMyHook())

Mocking API Requests

To mock API requests, we use the msw library.

Global mocks can be found in tests/lib/msw-global-api-mocks.ts.

To mock an endpoint you can use the addAPIMock function. Make sure to add the mock in the beforeEach hook. It won't work with beforeAll if you have many tests.

beforeEach(() => {
  addAPIMock({
    method: 'get',
    path: '/api/my-endpoint',
    response: {
      data: { foo: 'bar' },
    },
  })
})

API Mocking Tips:

  • Keep mocks in the same folder as the tests that use them
  • Add a test to verify the mock is working

This will make debugging and updating the mocks easier.

test('mock is working', async () => {
  const response = await fetch('/api/my-endpoint')
  expect(response.json()).resolves.toEqual({ data: { foo: 'bar' } })
})

Mocking Nuqs URL Parameters

To render a component that uses Nuqs with some predefined query parameters, you can use customRender with the nuqs prop.


customRender(<MyComponent />, {
  nuqs: {
    searchParams: {
      search: 'hello world',
    },
  },
})

<Popover> vs <Dropdown>

When simulating clicks on these components, do the following:

// for Popovers
import userEvent from '@testing-library/user-event'
await userEvent.click('Hello world')

// for Dropdowns
import clickDropdown from 'tests/helpers'
clickDropdown('Hello world')