I've been puzzling over this issue in relation to #1203 (comment) and think 🤔I have it sorted out!

In the test run referenced in the linked comment, the failing test was "get with reply callback returning default statusCode without body". It tested that .reply(() => [401]) produced a status code 200 and a body with a JSON-stringified array [401].

This was not the same test as the one added in PR #1203. In that PR the new test was "reply should send correct statusCode with array-notation and without body" and tested that .reply(() => [202]) produced a status code of 202.

It did not make assertions on the body, though the body behavior was unchanged: it sent a JSON-stringified array [202].

It's clear these tests could not have passed at the same time. It's not clear why the PR branch didn't reflect this. (Out of date with master? Flaky test suite? The test suite is more reliable now, in part because of #1465. And also #1077, though that was about fixing false positives rather than false negatives.)

While I agree with #1170, that .reply(() => [201]) should treat 201 as the status code, I disagree with simultaneously interpreting that to mean [201] should be the body. That seems like an unintended effect of that PR.

That behavior is in the process of being changed in #1520, which, if no objections, will cause .reply(() => [201]) to return an empty body.

The documentation is ambiguous, and you could argue this is a bug fix. Though, people may be depending on this have-your-cake-and-eat-it-too behavior that nock has had since 10.0.0, so we may as well label it a breaking change, which will ship (soon! see #1516) as part of v11.

/cc @mastermatt

Uh oh!

Fix test: "get with reply callback returning default statusCode without body" #1222

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions