API documentation

Authentication

All requests to the API must provide a valid API key in the Auth-Key HTTP header. The only exception are the endpoints that manipulate the API keys themselves, in which case the requests must be authenticated using another method (HTTP basic authentication, client certificates, etc).

Example:

$ KEY='00000000-0000-0000-0000-000000000000'
$ curl --header "Auth-Key: $KEY" https://host/api/v1/auth

GET /api/v1/auth

This endpoint can be used to test your API key. It returns a 200 (OK) HTTP status if your API key is valid, and 403 (Forbidden) otherwise. The response also includes the username corresponding to the API key in the Auth-User header.

POST /api/v1/getkey

Gets a new API key. Any existing API key is invalidated after a new one is obtained.

This endpoint does not require an existing API key, but does require proper authentication with a client certificate (e.g. Debian SSO)

GET /api/v1/getkey

Displays a user-friendly HTML page that can be used by users to get an API key using this existing in-browser client certificate.

This endpoint does not require an existing API key, but does require proper authentication with a client certificate (e.g. Debian SSO)

POST /api/v1/test/:suite/:arch

URL parameters:

  • :suite: which suite to test

  • :arch: which architecture to test

Other parameters:

  • tests: a JSON object decribing the tests to be executed. This parameter can be either a file upload or a regular POST parameter.

The tests JSON object must be an Array of objects. Each object represents a single test request, and can contain the following keys:

  • package: the (source!) package to be tested

  • trigger: a string that identifies the reason why this test is being requested. debci only stores this string, and it does not handle this in any special way.

  • pin-packages: an array describing packages that need to be obtained from different suites than the main one specified by the suite parameter. This is used e.g. to run tests on testing with a few packages from unstable, or on unstable with a few packages from experimental. Each item of the array is another array with 2 elements: the first is the package, and the second is the source. Examples:

  • ["foo", "unstable"]: get foo from unstable

  • ["src:bar", "unstable"]: get all binaries built from bar from unstable

In the example below, we are requesting to test debci and vagrant from testing, but with all binaries that come from the ruby-defaults source coming from unstable:

$ cat tests.json
[
  {
    "package": "debci",
    "trigger": "ruby/X.Y",
    "pin-packages": [
      ["src:ruby-defaults", "unstable"]
    ]
  },
  {
    "package": "vagrant",
    "trigger": "ruby/X.Y",
    "pin-packages": [
      ["src:ruby-defaults", "unstable"]
    ]
  }
]

# tests as a file upload
$ curl --header "Auth-Key: $KEY" --form tests=@tests.json \
    https://host/api/v1/test/testing/amd64

# tests as a regular POST parameter
$ curl --header "Auth-Key: $KEY" --data="$(cat tests.json)" \
    https://host/api/v1/test/testing/amd64

POST /api/v1/test/:suite/:arch/:package

This is a shortcut to request a test run for a single package.

URL parameters:

  • :suite: which suite to test

  • :arch: which architecture to test

  • :package: which (source!) package to test

Example:

$ curl --header "Auth-Key: $KEY" --data '' https://host/api/v1/test/unstable/amd64/debci

GET /api/v1/test

Retrieves results for your test requests.

Parameters:

  • since: UNIX timestamp; tells the API to only retrieve results that are newer then the given timestamp.

Some test results may be updated after being created, for example while a test is still running, it will be returned, but it's status will be null. After it is completed, it will be updated to have the correct status. So, if you are processing test results, make sure you support receiving the same result more than once, and updating the corresponding data on your side.

The response is a JSON object containing the following keys:

  • until: UNIX timestamp that represents the timestamp of the latest results available. can be used as the since parameter in subsequent requests to limit the list of results to only the ones newer than it.

  • results: a list of test results, each of each will containing at least the following items:

  • trigger: the same string that was provided in the test submission. (string)

  • package: tested package (string)

  • arch: architecture where the test ran (string)

  • suite: suite where the test ran (string)

  • version: version of the package that was tested (string)

  • status: “pass”, “fail”, or “tmpfail” (string), or null if the test didn't finish yet.

  • run_id: an id for the test run, generated by debci (integer)

Note that this endpoint will only list requests that were made by the same API key that is being used to call it.

Example:

$ curl --header "Auth-Key: $KEY" https://host/api/v1/test?since=1508072999
{
  "until": 1508159423,
  "results": [
    {
      "trigger": "foo/1.2",
      "package": "bar",
      "arch": "amd64",
      "suite": "testing",
      "version": "4.5",
      "status": "fail",
      "run_id": 12345
    },
    {
      "trigger": "foo/1.2",
      "package": "baz",
      "arch": "amd64",
      "suite": "testing",
      "version": "2.7",
      "status": "pass",
      "run_id": 12346
    }
  ]
}

POST /api/v1/retry/:run_id

This endpoint can be used to reschedule a test that has already been performed, e.g. because the reason of the failure has been solved.

URL parameters:

  • :run_id: which Job ID to retry