Selenium is an umbrella project for various tools and libraries that enable automation of web browsers. Amongst other things it provides the support infrastructure for the W3C WebDriver specification, that lets you write interchangable code for all major web browsers.
The project is made possible by volunteer contributors who have put in thousands of hours of their own time, and made the source code freely available under the Apache 2.0 license.
Narrative documentation:
- User Manual
- New Handbook (work in progress)
API documentation:
See CONTRIBUTING.md.
Selenium uses a custom build system aptly named crazyfun available on all fine platforms (Linux, Mac, Windows). We are in the process of replacing this with buck, so don't be alarmed if you see some directories carrying multiple build directive files. crazyfun's build files are named build.desc, while buck's are named simply BUCK.
To build Selenium, in the same directory as this file:
./go
The order of building modules is determined by the build system. If you want to build an individual module (assuming all dependent modules have previously been built) try something like:
./go //javascript/atoms:test:run
In this case, javascript/atoms
is the module directory,
test
is a target in that directory's build.desc
file,
and run
is the action to run on that target.
As you see build targets scroll past in the log,
you may want to run them individually.
crazyfun can run them individually,
by target name as long as :run
is appended (see above).
To list all available targets, you can append the -T
flag:
./go -T
Although the plan is to return to a vanilla build of Buck as soon as possible, we currently use a fork, hosted at https://github.com/shs96c/buck To build using Buck, first clone that repo and build using ant. Then add Buck's "bin" directory to your PATH.
To obtain a list of all available targets:
buck targets
And build a particular file:
buck build //java/client/src/org/openqa/selenium:webdriver-api
There are aliases for commonly invoked targets in the .buckconfig
file, and these aliases can be invoked directly:
buck build htmlunit
All buck output is stored under "buck-out", with the outputs of build
rules in buck-out/gen
.
If you are doing a number of incremental builds, then you may want to
use buckd
, which starts a long-lived buck process to watch outputs
and input files. If you do this, consider using watchman
too, since
the Java 7 file watcher isn't terribly efficient. This can be cloned
from https://github.com/facebook/watchman
- Java 6 JDK
java
andjar
on the PATH
Although the build system is based on rake it's strongly advised
to rely on the version of JRuby in third_party/
that is invoked by
go
. The only developer type who would want to deviate from this is
the “build maintainer” who's experimenting with a JRuby upgrade.
Note that all Selenium Java artefacts are built with Java 6 (mandatory). Those will work with any Java >= 6.
- Python 2.6.x to 2.7 (without this, Python tests will be skipped)
- Ruby 1.9.3
If you plan to compile the IE driver you also need:
- Visual Studio 2008
- 32 and 64 bit cross compilers
The build will work on any platform, but the tests for IE will be skipped silently, if you are not building on Windows.
For an express build of the binaries we release run the following from the
directory containing the Rakefile
:
./go clean release
All build output is placed under the build
directory. The output can
be found under build/dist
. If an error occurs while running this
task complaining about a missing Albacore gem, the chances are you're
using rvm
. If this is the case, switch to the system ruby:
rvm system
Of course, building the entire project can take too long. If you just want to build a single driver, then you can run one of these targets:
./go chrome
./go firefox
./go htmlunit
./go ie
As the build progresses, you'll see it report where the build outputs are being placed. Of course, just building isn't enough. We should really be able to run the tests too. Try:
./go test_chrome
./go test_firefox
./go test_htmlunit
./go test_ie
Note that the test_chrome
target requires that you have the separate
Chrome Driver
binary available on your PATH
.
If you are interested in a single language binding, try one of:
./go test_java
./go test_dotnet
./go test_rb
./go test_javascript
To run all the tests just run:
./go test
This will detect your OS and run all the tests that are known to be stable for every browser that's appropriate to use for all language bindings. This can take a healthy amount of time to run.
To run the minimal logical Selenium build:
./go test_javascript test_java
As a side note, none of the developers run tests using Cygwin. It is very unlikely that the build will work as expected if you try and use it.
The code base is generally segmented around the languages used to write the component. Selenium makes extensive use of JavaScript, so let's start there. Working on the JavaScript is easy. First of all, start the development server:
./go debug-server
Now navigate to
http://localhost:2310/javascript.
You'll find the contents of the javascript/
directory being shown.
We use the Closure
Library for
developing much of the javascript, so now navigate to
http://localhost:2310/javascript/atoms/test.
The tests in this directory are normal HTML files with names ending
with _test.html
. Click on one to load the page and run the test. You
can run all the javascript tests using:
./go test_javascript
Ignore the Maven POM file present in the same directory. It is only used for releasing to jars to Maven Repository (public or local), and is not considered the main build mechanism.
Here is the public Selenium Maven repository.
./go
only makes a top-level build
directory. Outputs are placed
under that relative to the target name. Which is probably best
described with an example. For the target:
./go //java/client/src/org/openqa/selenium:selenium-api
The output is found under:
build/java/client/src/org/openqa/selenium/selenium-api.jar
If you watch the build, each step should print where its output is
going. Java test outputs appear in one of two places: either under
build/test_logs
for JUnit or in
build/build_log.xml
for TestNG
tests. If you'd like the build to be chattier, just append log=true
to the build command line.
More general, but basic, help for go…
./go --help
Remember, go is just a wrapper around
Rake, so you can use the standard
commands such as rake -T
to get more information about available
targets.
If it is not clear already, Selenium is not built with Maven, it is built with Crazy-Fun though that is invoked with go as outlined above so you do not really have to learn too much about that.
That said, it is possible to relatively quickly build selenium pieces
for Maven to use. You are only really going to want to do this when
you are testing the cutting-edge of Selenium development (which we
welcome) against your application. Here is the quickest way to build
and deploy into you local maven repository (~/.m2/repository
), while
skipping Selenium's own tests.
./go release
cd maven
mvn clean install
This sequence will push some seven or so jars into you local Maven repository with something like 'selenium-server-2.0-SNAPSHOT.jar' as the name.
Refer to the Building Web Driver wiki page for the last word on building the bits and pieces of Selenium.