PR1266 Retrospective

[TsayAdobe]

The incident in question occurred when this PR introduced a flawed line of code into the production environment. This situation underscores the limitations of code review processes, and illustrates how easily undesirable code can be inadvertently pushed to the production.

  loadStyle(`https://www.adobe.com/special/chimera/caas-libs/${version}/app.css`);
  await loadScript(`https://www.adobe.com/special/chimera/caas-libs/${version}/react.umd.js`);
  await loadScript(`https://www.adobe.com/special/chimera/caas-libs/${version}/react.dom.umd.js`);
  // await loadScript(`https://www.adobe.com/special/chimera/caas-libs/${version}/main.min.js`);
  await loadScript('http://localhost:5000/dist/main.js');

Unit tests are designated as the initial safeguard against error introduction. The relevant test code is housed in this file. The functions loadScript and loadStyle are mocked, with verification being carried out to check the count of function calls. Given the constraints inherent to unit tests, this verification was deemed reasonable.

          expect(loadScript.callCount).to.equal(3);
          expect(loadStyle.callCount).to.equal(1);

However, the CasS block primarily serves as an interface for the Chimera service, making the unit test somewhat ineffective in identifying bugs that arise from the integration between Milo and Chimera.

To better address this gap, it is proposed that a suite of integration tests be developed to ensure the block is accurately rendered within a Milo page. Through such tests, a more thorough validation of the integration points can be achieved, substantially mitigating the risk of similar bugs reaching the production environment in the future.