media/tools/layout_tests/README

Layout test analyzer script.

* Prerequisite:

This script requires PySVN (http://pysvn.tigris.org/) to be installed on the
machine.

* How to execute

python layouttest_analyzer.py (Please run with '-h' for available command-line
                               options)

Example: python layouttest_analyzer.py -r imasaki@chromium.org 
-t /var/www/analyzer/graph.html

* File/Directory structure

layout_tests/
  README: this file.
  *.py: python scripts.
  tmp/: temp files.
  test_data/: data for unit tests.
  graph/: the default location for graph files.
  anno/: the default location for bug annotation files.
  result/: the default location for the analyzer results.
  testname/: the default location for the test group definition files in CSV.

* Execution overview.

1) using PySVN, the script collects layout test of interest
(specified in command-line parameter and testnames/*.csv file) from
Webkit SVN repository as well as test description. Then, it gets the
latest test expectation file and does the simple parsing. The
parsed test expectation is joined with the layout tests using test name
as its join key. The tests are classified into 'whole', 'skip',
'nonskip' test groups. 'Whole' contains all tests, 'skip' contains
tests that are skipped as specified in the test expectation file, 'nonskip'
tests are in the test expectation file and not skipped (these are tests
that need attention).

2) the script reads the previous latest analyzer results, then,
compares it with current results for each test group ('whole', 'skip',
and 'nonskip'). It also looks into the SVN test expectation history
diff for the time period for the test names of interest.

3) all obtained information is sent out to a email list specified
in the parameters in pretty HTML format.

4) the trend graph is updated using the current results.

* Components

** LayoutTests:

A class to store test names in layout tests. The test names (including
regular expression patterns) are read from a CSV file and used for
getting layout test names from Webkit SVN as well as test description.

** TestExpectations:

A class to model the content of test expectation file for analysis.
The location of the test expectations file can be found in
|TEST_EXPECTATIONS_LOCATIONS|. It is necessary to parse this
file and store meaningful information for the analysis (joining with
existing layout tests using a test name).  Instance variable
|all_test_expectation_info| is used.  A test name such as
'media/video-source-type.html' is used for the key to store
information. However, a test name can appear multiple times in the
test expectation file. So, the map should keep all the occurrence
information. For example, the current test expectation file has the
following two entries: 

BUGWK58587 LINUX DEBUG GPU : media/video-zoom.html = IMAGE 
BUGCR86714 MAC GPU : media/video-zoom.html = CRASH IMAGE 

In this case, all_test_expectation_info['media/video-zoom.html'] will
have a list with two elements, each of which is the map of the test
expectation information.

** TestExpectationsHistory:

A class to represent history of the test expectation file. The history
is obtained by calling PySVN.log()/diff() APIs using the specified time
period.

** TrendGraph: 

A class to manage trend graph which is using Google Visualization
  APIs. Google Visualization API
  (http://code.google.com/apis/chart/interactive/docs/gallery/annotatedtimeline.html)
  is used to present the historical analyzer result. Currently, data
  is directly written to JavaScript file using file in-place
  replacement for simplicity.