I have a typical search where the user types some text in an input, async work is done and a table with the results is properly updated.
I have tests that must wait for this search step and then assert business rules regarding the results, like if the table records are eligible for edit.
Every time I ran a complete test battery (something like 80 test files), one or two of the tests involving that search inevitably fail. But if immediately after that, I run the same test alone, the test passes. It's excruciating and makes e2e testing in CI/CD pointless for the project.
I've read the Cypress documentation about flaky tests and searched questions in StackOverflow and GitHub with only complete failure. It's a drama.
Here is one of the tests:
import { searchList } from '../helpers';
import { createFluxoIniciado, randomFluxoNome } from './common';
import { fluxoSelectors } from './selectors';
describe('fluxos finish', () => {
it('can manually finish a fluxo INICIADO', () => {
// feed data to be searched
const fluxoNome = randomFluxoNome();
createFluxoIniciado({ fluxoNome });
// search
searchList(fluxoNome);
// do something with search results
fluxoSelectors.fluxos.view().click();
fluxoSelectors.finish().click();
fluxoSelectors.confirm().click();
// serach again
searchList(fluxoNome);
cy.contains('FINALIZADO');
});
});
The code in searchList is where trouble emerge sometimes. It uses the callback strategy recommended here. The code attempts to cause retries if not all rows have the searched text.
export function searchList (text) {
cy.get('#searchText')
.scrollIntoView()
.type(text)
.blur();
cy.get('tbody tr').should($trs => {
$trs.each((i, $tr) => {
expect($tr).to.contain(text);
});
}, { timeout: 15000 });
}
Here is an example of a test failure inside a run all test execution:
