Tests Frontend: Flaky ResolveForecastDialog test - handleCloseCalled is false

[omarkohl]

It failed once after about 320 executions and another time after about 60 executions of all frontend tests.

  ● forecast can be resolved with outcome

    expect(received).toBe(expected) // Object.is equality

    Expected: true
    Received: false

      171 |     await user.click(screen.getByRole("button", {name: "Save"}));
      172 |
    > 173 |     expect(handleCloseCalled).toBe(true);
          |                               ^
      174 |
      175 |     expect(requestBody).toBeTruthy();
      176 |     if (!requestBody) {

      at Object.<anonymous> (src/ResolveForecastDialog.test.tsx:173:31)

Probably a race condition when the dialog is closed the handleClose is not called immediately.

[omarkohl]

Script to run all tests until one fails.

#!/bin/bash
set -o errexit
set -o nounset
set -o pipefail

npm install

count=0

while true
do
    date
    echo $count
    # Seems to be the closest to what GitHub actions does
    npm run test -- --all --bail --ci --watchAll=false --runInBand || break
    echo
    echo
    echo
    sleep 2
    count=$((count+1))
done

[omarkohl]

Probably it's because the dialog only closes once the request to resolve the forecast has completed. Even in the test case with msw that might sometimes take a little longer.

With the following change the test immediately fails:

diff --git a/frontend/src/ResolveForecastDialog.test.tsx b/frontend/src/ResolveForecastDialog.test.tsx
index 8394aa0..e3ea2b9 100644
--- a/frontend/src/ResolveForecastDialog.test.tsx
+++ b/frontend/src/ResolveForecastDialog.test.tsx
@@ -80,6 +80,7 @@ test('forecast can be resolved with outcome', async () => {
     server.use(
         graphql.mutation("resolveForecast", async (req, res, ctx) => {
             requestBody = await req.json();
+            await new Promise(f => setTimeout(f, 1000));
             return res(
                 ctx.data({
                     "resolveForecast": {

[omarkohl]

After commit 8f34ec4 about 600 executions of all tests were successful, so it seems reasonable to conclude that this has been resolved.