NetSurf Integration Testing =========================== [TOC] # Overview The monkey frontend is used to perform complex tests involving operating the browser as a user might (opening windows, navigating to websites and rendering the contents etc.) A test is written as a set of operations in a yaml file. These files are parsed and executed by the monkey driver script. There are very few tests within the NetSurf repository. The large majority of integration tests are instead held within the [netsurf-test](http://source.netsurf-browser.org/netsurf-test.git/) repository. To allow more effective use of these tests additional infrastructure has been constructed to allow groupings of tests to be run. This is used extensively by the CI system to perform integration testing on every commit. The infrastructure also provides for special CGI scripts which have known behaviours such as delays or returning specific content to extend test capabilities. # Running a test An individual test can be run using the monkey_driver.py python script from within the NetSurf repository $ make TARGET=monkey $ ./test/monkey_driver.py -m ./nsmonkey -t test/monkey-tests/start-stop.yaml The command actually executed can be augmented using the wrapper switch, this allows the test to be run under a debugger or profiler. For example to wrap execution under valgrind memory checker $ ./test/monkey_driver.py -m ./nsmonkey -w 'valgrind -v --track-origins=yes' -t test/monkey-tests/start-stop.yaml # Running more than one test Each test is a member of a group and the tests within each group are run together. Groups are listed within division index files. A group of tests may occur within more than one division. To run the integration tests the monkey-see-monkey-do python script is used. It downloads the test plan for a division from the NetSurf test infrastructure and executes it. $ ./test/monkey-see-monkey-do Fetching tests... Parsing tests... Running tests... Start group: initial [ Basic checks that the browser can start and stop ] => Run test: start-stop-no-js.yaml => Run test: basic-navigation.yaml => Run test: start-stop.yaml Start group: no-networking [ Tests that require no networking ] => Run test: resource-scheme.yaml Start group: ecmascript [ ECMAScript tests ] PASS If no division is specified on the command line the "default" division is used. Other divisions are specified with the d switch for example to specify the "short-internet" division: $ ./test/monkey-see-monkey-do -d short-internet Additionally the g switch allows the limiting of tests within a single group to be executed. $ ./test/monkey-see-monkey-do -g no-networking Fetching tests... Parsing tests... Running tests... Start group: no-networking [ Tests that require no networking ] => Run test: resource-scheme.yaml PASS # Test files Each test is a individual [YAML](https://en.wikipedia.org/wiki/YAML) file and consists of associative arrays (key/value pairs), lists and comments. As a minimum a test must contain an associative array with keys for `title`, `group` and `steps`. The `steps` key must contain a list of test operations as a value. Each operation is an associative list and must, as a minimum, contain an action key with a suitable value. A minimal test that simply starts the browser without JavaScript and then quits it without ever opening a window to browse to a page title: start and stop browser without JS group: initial steps: - action: launch options: - enable_javascript=0 - action: quit # Actions ## launch Start a browser instance. A test will generally have a single launch action paired with a quit action. Additional command line parameters may be set using the `launch-options` key the value of which must be a list of command line arguments to be passed to the browser (without their leading hyphens) The environment of the browser can be altered with the `environment` key the value is an associative array of environment variables which will be added to the browsers environment variables. User options may be set using the `options` key with a value containing a list of options to set. These options allow operation with differing user choices to be tested without a separate Choices file. The `language` key sets the LANGUAGE environment variable which controls the browsers user interface language. Note this is distinct from the language the browser requests from HTTP servers which is controlled with the `accept_language` user option. The following launch action would start a browser: * Passing `--verbose` on the commandline * The `NETSURFRES` environment variable set to `/home/netsurf/resources` * The user options `enable_javascript` and `send_referer` set to false. * The `LANGUAGE` environment variable set to `en` ``` - action: launch launch-options: - verbose environment: NETSURFRES: /home/netsurf/resources options: - enable_javascript=0 - send_referer=0 language: en ``` ## window-new Open a new browser window. The test may open as many browser windows as necessary and they are usually paired with a `window-close` action The browser must have been previously launched or this action will assert the test with a failure. The `tag` key *must* also be present with a value (unique for all window-new actions). The value is used to identify subsequent operations in this window. As an example this will open a new window which can subsequently be referred to with the win1 identifier: - action: window-new tag: win1 ## window-close Closes a previously opened window. The window is identified with the `window` key, the value of this must be a previously created window identifier or an assert will occur. - action: window-close window: win1 ## navigate Cause a window to start navigating to a new URL. The window to be navigated is identified with the `window` key, the value of this must be a previously created window identifier or an assert will occur. The URL to navigate to navigate to is controlled either by the `url` or `repeaturl` key. The `url` value is directly used as the address to navigate to. - action: navigate window: win1 url: about:about The `repeaturl` value is used as a repeat action identifier allowing navigation in a loop with different values. - action: repeat values: - https://www.google.com/ - https://apple.com/ - https://microsoft.com/ tag: urls - action: navigate window: win1 repeaturl: urls ## reload Cause a window to (re)navigate to the current URL The window to be navigated is identified with the `window` key, the value of this must be a previously created window identifier or an assert will occur. - action: reload window: win1 ## stop Cause a window to immediately stop any navigation. The window to be navigated is identified with the `window` key, the value of this must be a previously created window identifier or an assert will occur. - action: stop window: win1 ## sleep-ms Wait for time to pass before continuing to next action. The value of the `time` key is either the duration to wait for in milliseconds or a `repeat` action identifier. The optional `conditions` key may contain a list of conditionals used to terminate the delay early. If a `repeat` action identifier is in use the loop is terminated if a condition is met. For example to wait 10 seconds: - action: sleep-ms time: 10000 if a repeat action identifier is used the delay duration is the current iteration value and the delay is timed from when the current iteration started. The `sleep-ms` action here delays by 50 milliseconds more each iteration until the window navigation is complete when the `sleep-ms` action is delaying. - action: repeat min: 0 step: 50 tag: sleepytimer steps: - action: launch - action: window-new tag: win1 - action: navigate window: win1 url: about:about - action: sleep-ms time: sleepytimer conditions: - window: win1 status: complete - action: quit ## block Wait for conditions to be met before continuing. This is similar to the `sleep-ms` action except that it will wait forever for the conditions to be met. The `conditions` key must contain a list of conditionals used to terminate the block. - action: block conditions: - window: win1 status: complete valid `status` values are `complete` or `loading`. ## repeat Repeat a set of actions. The identifier of the repeat action is set with the `tag` key and must be present and unique among `repeat` action identifiers. The actions to be repeated are placed in the `steps` list which may include any action and behaves just like the top level list. An iterator context is created for the repeat loop. The iterator can either be configured as a numeric value or as a list of values. The numeric iterator is controlled with the `min` ,`step` and `max` keys. All these keys are integer values and their presence is optional. The `min` value is the initial value of the iterator which defaults to 0. The `step` value controls how much the iterator is incremented on every loop with default value of 1. The loop terminates if the `max` value is exceeded. If no `max` value is specified the loop is infinite (i.e. no default) but may still be terminated by the `sleep-ms` action - action: repeat min: 0 step: 50 max: 100 tag: loopvar steps: - action: launch - action: quit A value iterator has a `values` key containing a list. On each iteration of the loop a new value is available and can be used by the `navigate` action. Note that `min` ,`step` and `max` are ignored if there is a `values` key - action: repeat values: - https://www.google.com/ - https://www.blogger.com/ - https://apple.com/ - https://microsoft.com/ tag: urls steps: - action: navigate window: win1 repeaturl: urls - action: block conditions: - window: win1 status: complete ## timer-start Start a timer. The identifier for the timer is set with the `timer` key. - action: timer-start timer: timer1 ## timer-restart Re-start a timer The identifier for the timer is set with the `timer` key. - action: timer-restart timer: timer1 ## timer-stop Stop a timer The identifier for the timer is set with the `timer` key. - action: timer-stop timer: timer1 ## timer-check Check a timer meets a condition. The identifier for the timer is set with the `timer` key. The conditional is set with the `condition` key which must be present. ## plot-check Perform a plot of a previously navigated window. The window to be rendered is identified with the `window` key, the value of this must be a previously created window identifier or an assert will occur. The `area` key allows control of the area to be redraw. The parameters are on two forms: * A sequence of four numbers in the form `x0 y0 x1 y1` * The keyword extent which attempt to plot the entire extent of the canvas An optional list of checks may be specified with the `checks` key. If any check is not satisfied an assert will occur and the test will fail. Multiple checks can be specified and all most pass successfully. The checks available are: * The key `text-contains` where the text must occur somewhere in the plotted output. * The key `text-not-contains` where the text must not occur in the plotted output. * The key `bitmap-count` which specifies the number of images that must be present. - action: plot-check window: win1 area: extent checks: - text-contains: NetSurf - text-contains: Browser - text-not-contains: Chrome - bitmap-count: 1 ## click Perform a user mouse click on a specified window. The window to be clicked is identified with the `window` key, the value of this must be a previously created window identifier or an assert will occur. The `target` key contains an associative array which is used to select the location of the mouse operation in the window. The key `text` can be used to select text to be operated upon which matches the first occurrence of the text. The key `bitmap` has an integer value to select the index of the image to click. The optional `button` key selects which button is pressed it can take the value `left` or `right`. The default if not specified is `left` The optional `kind` key selects which button operation to be performed it can take the value `single`, `double` or `triple`. The default if not specified is `single` - action: click window: win1 target: text: "about:Choices" ## wait-loading Wait for the navigated page to start loading before moving to the next action. The window to be waited upon is identified with the `window` key, the value of this must be a previously created window identifier or an assert will occur. - action: wait-loading window: win1 ## add-auth Add basic authentication details for a navigation. The keys `url`, `realm`, `username` and `password` must be given. When a basic authentication challenge occurs that matches the url and realm parameters the associated username and password are returned to answer the challenge. - action: add-auth url: http://test.netsurf-browser.org/cgi-bin/auth.cgi?user=foo&pass=bar realm: Fake Realm username: foo password: bar ## remove-auth Remove a previously added authentication details. - action: remove-auth url: http://test.netsurf-browser.org/cgi-bin/auth.cgi?user=foo&pass=bar realm: Fake Realm username: foo password: bar ## clear-log Clear log for a window. The window to be cleared is identified with the `window` key, the value of this must be a previously created window identifier or an assert will occur. ## wait-log Wait for string to appear in log output. The window to be waited upon is identified with the `window` key, the value of this must be a previously created window identifier or an assert will occur. The keys `source` `foldable` `level` and `substring` must be specified ## js-exec Execute javascript in a window. The window to be execute within is identified with the `window` key, the value of this must be a previously created window identifier or an assert will occur. The `cmd` key contains the javascript to execute. ## page-info-state Check the page information status matches an expected value. The window to be checked is identified with the `window` key, the value of this must be a previously created window identifier or an assert will occur. The value of the `match` key is compared to the windows page information status and an assert occurs if there is a mismatch. ## quit This causes a previously launched browser instance to exit cleanly.