This is a regression tester for http.
1) start the server, with python server.py
2) Get a file of all tests available by getting http://localhost:8000/
3) Use that file as an argument to mozilla -f (this requires either a
debug build or a build with --enable-perf-metrics)
4) Look at the results of http://localhost:8000/id/report
Each run gets a unique ID number, which is used to map a test run to a
database of results. Every test (which lives in the tests/ directory) is
enumerated in step 2. Adding a new test is as simple as just creating the
directory and addiing the appropriate files.
Each test must have a config file, which is simply a list of all the files
which will be requested, in order. Each file (x) can either be present in
the directory as that name, in which case the contents will be sent to the
browser, or as x.headers and (optionally) x.body.
The headers file has a first line which is the response code, and then any
subsequent lines are header: value pairs. If the body file is not present,
the test server will construct a one line response. This is useful for
testing mozilla's response to headers (see the responses/ subdirectory of
tests for examples)
The default definition of a tests's 'success' is that all the files are
retreived in the correct order. This can be overridden on a per test
basis. For example, the referer test additionally checks that a
referer header was sent.
To do this, a file called tester.py must exist in the test's directory.
This class must provide an object called tester, which is used by the
driver code instead of the default tester. This default tester is
automatically present in the file's namespace under the name 'BaseTester'.
This means that it can be inherited from in order to only override certain
The routines in the default tester are documented in BaseTest.py. The most
common routine which needs to be extended in verify_request. For an
example, see the referer test.
Bradley Baetz <email@example.com>, <firstname.lastname@example.org>