summaryrefslogtreecommitdiffstats
path: root/testing/web-platform/tests/docs/manual-test.md
blob: 4b5469589fed8806d5887c7a8bc4b16183bef473 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
Some testing scenarios are intrinsically difficult to automate and
require a human to run the test and check the pass condition.

## When to Write Manual Tests

Whenever possible it's best to write a fully automated test. For a
browser vendor it's possible to run an automated test hundreds of
times a day, but manual tests are likely to be run a handful of times
a year. This makes them significantly less useful for catching
regressions than automated tests.

However, there are certain scenarios in which this is not yet
possible. For example:

* Tests that require interaction with browser security UI (e.g. a test
  in which a user refuses a geolocation permissions grant)

* Tests that require interaction with the underlying OS e.g. tests for
  drag and drop from the desktop onto the browser

* Tests that require non-default browser configuration e.g. images
  disabled

* Tests that require interaction with the physical environment
  e.g. tests that the vibration API causes the device to vibrate or
  that various sensor APIs respond in the expected way.

There are also some rare cases where it isn't possible to write a layout
test as a reftest, and a manual test must be written instead.

## Requirements for a Manual Test

Manual tests are distinguished by their filename; all manual tests
have filenames of the form `name-manual.ext` i.e. a `-manual`
suffix after the main filename but before the extension.

Manual tests must be fully
[self-describing](test-style-guielines.html#self-describing-tests). It
is particularly important for these tests that it is easy to determine
the result from the information presented to the tester, because a
tester may have hundreds of tests to get through, and little
understanding of the features that they are testing. Therefore
minimalism is a virtue. An ideal self-describing test will have:

* Step-by-step instructions for performing the test (if required)

* A clear statement of the test outcome (if it can be automatically
  determined after some setup) or of how to determine the outcome.

Any information other than this (e.g. quotes from the spec) should be
avoided.

## Using testharness.js for Manual Tests

A convenient way to present the results of a test that can have the
result determined by script after some manual setup steps is to use
testharness.js to determine and present the result. In this case one
must pass `{explicit_timeout: true}` in a call to `setup()` in order
to disable the automatic timeout of the test. For example:

```html
<!doctype html>
<title>Manual click on button triggers onclick handler</title>
<script src="/resources/testharness.js"></script>
<script src="/resources/testharnessreport.js"></script>
<script>
setup({explicit_timeout: true})
</script>
<p>Click on the button below. If a "PASS" result appears the test
passes, otherwise it fails</p>
<button onclick="done()">Click Here</button>
```