20062014 Ericsson AB. All Rights Reserved. The contents of this file are subject to the Erlang Public License, Version 1.1, (the "License"); you may not use this file except in compliance with the License. You should have received a copy of the Erlang Public License along with this software. If not, it can be retrieved online at http://www.erlang.org/. Software distributed under the License is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License for the specific language governing rights and limitations under the License. Code Coverage Analysis Peter Andersson cover_chapter.xml
General

Although Common Test was created primarly for the purpose of black box testing, nothing prevents it from working perfectly as a white box testing tool as well. This is especially true when the application to test is written in Erlang. Then the test ports are easily realized by means of Erlang function calls.

When white box testing an Erlang application, it is useful to be able to measure the code coverage of the test. Common Test provides simple access to the OTP Cover tool for this purpose. Common Test handles all necessary communication with the Cover tool (starting, compiling, analysing, etc). All the Common Test user needs to do is to specify the extent of the code coverage analysis.

Usage

To specify what modules should be included in the code coverage test, you provide a cover specification file. Using this file you can point out specific modules or specify directories that contain modules which should all be included in the analysis. You can also, in the same fashion, specify modules that should be excluded from the analysis.

If you are testing a distributed Erlang application, it is likely that code you want included in the code coverage analysis gets executed on an Erlang node other than the one Common Test is running on. If this is the case you need to specify these other nodes in the cover specification file or add them dynamically to the code coverage set of nodes. See the ct_cover page in the reference manual for details on the latter.

In the cover specification file you can also specify your required level of the code coverage analysis; details or overview. In detailed mode, you get a coverage overview page, showing you per module and total coverage percentages, as well as one HTML file printed for each module included in the analysis that shows exactly what parts of the code have been executed during the test. In overview mode, only the code coverage overview page gets printed.

You can choose to export and import code coverage data between tests. If you specify the name of an export file in the cover specification file, Common Test will export collected coverage data to this file at the end of the test. You may similarly specify that previously exported data should be imported and included in the analysis for a test (you can specify multiple import files). This way it is possible to analyse total code coverage without necessarily running all tests at once.

To activate the code coverage support, you simply specify the name of the cover specification file as you start Common Test. This you do either by using the -cover flag with ct_run. Example:

$ ct_run -dir $TESTOBJS/db -cover $TESTOBJS/db/config/db.coverspec

You may also pass the cover specification file name in a call to ct:run_test/1, by adding a {cover,CoverSpec} tuple to the Opts argument. Also, you can of course enable code coverage in your test specifications (read more in the chapter about using test specifications).

Stopping the cover tool when tests are completed

By default the Cover tool is automatically stopped when the tests are completed. This causes the original (non cover compiled) modules to be loaded back in to the test node. If a process at this point is still running old code of any of the modules that are cover compiled, meaning that it has not done any fully qualified function call after the cover compilation, the process will now be killed. To avoid this it is possible to set the value of the cover_stop option to false. This means that the modules will stay cover compiled, and it is therefore only recommended if the erlang node(s) under test is terminated after the test is completed or if cover can be manually stopped.

The option can be set by using the -cover_stop flag with ct_run, by adding {cover_stop,true|false} to the Opts argument to ct:run_test/1, or by adding a cover_stop term in your test specification (see chapter about test specifications).

The cover specification file

These are the terms allowed in a cover specification file:

      %% List of Nodes on which cover will be active during test.
      %% Nodes = [atom()]
      {nodes, Nodes}.       

      %% Files with previously exported cover data to include in analysis.
      %% CoverDataFiles = [string()]
      {import, CoverDataFiles}.

      %% Cover data file to export from this session.
      %% CoverDataFile = string()
      {export, CoverDataFile}.

      %% Cover analysis level.
      %% Level = details | overview
      {level, Level}.       

      %% Directories to include in cover.
      %% Dirs = [string()]
      {incl_dirs, Dirs}.

      %% Directories, including subdirectories, to include.
      {incl_dirs_r, Dirs}.

      %% Specific modules to include in cover.
      %% Mods = [atom()]
      {incl_mods, Mods}.

      %% Directories to exclude in cover.
      {excl_dirs, Dirs}.

      %% Directories, including subdirectories, to exclude.
      {excl_dirs_r, Dirs}.

      %% Specific modules to exclude in cover.
      {excl_mods, Mods}.

      %% Cross cover compilation
      %% Tag = atom(), an identifier for a test run
      %% Mod = [atom()], modules to compile for accumulated analysis
      {cross,[{Tag,Mods}]}.
    

The incl_dirs_r and excl_dirs_r terms tell Common Test to search the given directories recursively and include or exclude any module found during the search. The incl_dirs and excl_dirs terms result in a non-recursive search for modules (i.e. only modules found in the given directories are included or excluded).

Note: Directories containing Erlang modules that are to be included in a code coverage test must exist in the code server path, or the cover tool will fail to recompile the modules. (It is not sufficient to specify these directories in the cover specification file for Common Test).

Cross cover analysis

The cross cover mechanism allows cover analysis of modules across multiple tests. It is useful if some code, e.g. a library module, is used by many different tests and the accumulated cover result is desirable.

This can of course also be achieved in a more customized way by using the export parameter in the cover specification and analysing the result off line, but the cross cover mechanism is a build in solution which also provides the logging.

The mechanism is easiest explained via an example:

Let's say that there are two systems, s1 and s2, which are tested in separate test runs. System s1 contains a library module m1 which is tested by the s1 test run and is included in s1's cover specification:

s1.cover: {incl_mods,[m1]}.

When analysing code coverage, the result for m1 can be seen in the cover log in the s1 test result.

Now, let's imagine that since m1 is a library module, it is also used quite a bit by system s2. The s2 test run does not specifically test m1, but it might still be interesting to see which parts of m1 is actually covered by the s2 tests. To do this, m1 could be included also in s2's cover specification:

s2.cover: {incl_mods,[m1]}.

This would give an entry for m1 also in the cover log for the s2 test run. The problem is that this would only reflect the coverage by s2 tests, not the accumulated result over s1 and s2. And this is where the cross cover mechanism comes in handy.

If instead the cover specification for s2 was like this:

s2.cover: {cross,[{s1,[m1]}]}.

then m1 would be cover compiled in the s2 test run, but not shown in the coverage log. Instead, if ct_cover:cross_cover_analyse/2 is called after both s1 and s2 test runs are completed, the accumulated result for m1 would be available in the cross cover log for the s1 test run.

The call to the analyse function must be like this:

ct_cover:cross_cover_analyse(Level, [{s1,S1LogDir},{s2,S2LogDir}]).

where S1LogDir and S2LogDir are the directories named <TestName>.logs for each test respectively.

Note the tags s1 and s2 which are used in the cover specification file and in the call to ct_cover:cross_cover_analyse/2. The point of these are only to map the modules specified in the cover specification to the log directory specified in the call to the analyse function. The name of the tag has no meaning beyond this.

Logging

To view the result of a code coverage test, click the button labled "COVER LOG" in the top level index page for the test run.

Prior to Erlang/OTP 17.1, if your test run consisted of multiple tests, cover would be started and stopped for each test within the test run. Separate logs would be available via the "Coverage log" link on the test suite result pages. These links are still available, but now they all point to the same page as the button on the top level index page. The log contains the accumulated results for the complete test run. See the release notes for more information about this change.

The buttonc takes you to the code coverage overview page. If you have successfully performed a detailed coverage analysis, you find links to each individual module coverage page here.

If cross cover analysis has been performed, and there are accumulated coverage results for the current test, then the - "Coverdata collected over all tests" link will take you to these results.