8  Running Tests and Analyzing Results

8 Running Tests and Analyzing Results

The Common Test framework provides a high-level operator interface for testing, providing the following features:

  • Automatic compilation of test suites (and help modules)
  • Creation of extra HTML pages for improved overview.
  • Single-command interface for running all available tests
  • Handling of configuration files specifying data related to the System Under Test (SUT) (and any other variable data)
  • Mode for running multiple independent test sessions in parallel with central control and configuration

When Common Test starts, it automatically attempts to compile any suites included in the specified tests. If particular suites are specified, only those suites are compiled. If a particular test object directory is specified (meaning all suites in this directory are to be part of the test), Common Test runs function make:all/1 in the directory to compile the suites.

If compilation fails for one or more suites, the compilation errors are printed to tty and the operator is asked if the test run is to proceed without the missing suites, or be aborted. If the operator chooses to proceed, the tests having missing suites are noted in the HTML log. If Common Test is unable to prompt the user after compilation failure (if Common Test does not control stdin), the test run proceeds automatically without the missing suites. This behavior can however be modified with the ct_run flag -abort_if_missing_suites, or the ct:run_test/1 option {abort_if_missing_suites,TrueOrFalse}. If abort_if_missing_suites is set to true, the test run stops immediately if some suites fail to compile.

Any help module (that is, regular Erlang module with name not ending with "_SUITE") that resides in the same test object directory as a suite, which is part of the test, is also automatically compiled. A help module is not mistaken for a test suite (unless it has a "_SUITE" name). All help modules in a particular test object directory are compiled, no matter if all or only particular suites in the directory are part of the test.

If test suites or help modules include header files stored in other locations than the test directory, these include directories can be specified by using flag -include with ct_run, or option include with ct:run_test/1. Also, an include path can be specified with an OS environment variable, CT_INCLUDE_PATH.

Example (bash):

$ export CT_INCLUDE_PATH=~testuser/common_suite_files/include:~testuser/common_lib_files/include

Common Test passes all include directories (specified either with flag/option include, or variable CT_INCLUDE_PATH , or both, to the compiler.

Include directories can also be specified in test specifications, see Test Specifications.

If the user wants to run all test suites for a test object (or an OTP application) by specifying only the top directory (for example, with start flag/option dir), Common Test primarily looks for test suite modules in a subdirectory named test. If this subdirectory does not exist, the specified top directory is assumed to be the test directory, and test suites are read from there instead.

To disable the automatic compilation feature, use flag -no_auto_compile with ct_run, or option {auto_compile,false} with ct:run_test/1. With automatic compilation disabled, the user is responsible for compiling the test suite modules (and any help modules) before the test run. If the modules cannot be loaded from the local file system during startup of Common Test, the user must preload the modules before starting the test. Common Test only verifies that the specified test suites exist (that is, that they are, or can be, loaded). This is useful, for example, if the test suites are transferred and loaded as binaries through RPC from a remote node.

The ct_run program can be used for running tests from the OS command line, for example, as follows:

  • ct_run -config <configfilenames> -dir <dirs>
  • ct_run -config <configfilenames> -suite <suiteswithfullpath>
  • ct_run -userconfig <callbackmodulename> <configfilenames> -suite <suiteswithfullpath>
  • ct_run -config <configfilenames> -suite <suitewithfullpath> -group <groups> -case <casenames>

Examples:

 $ ct_run -config $CFGS/sys1.cfg $CFGS/sys2.cfg -dir $SYS1_TEST $SYS2_TEST
 $ ct_run -userconfig ct_config_xml $CFGS/sys1.xml $CFGS/sys2.xml -dir $SYS1_TEST $SYS2_TEST
 $ ct_run -suite $SYS1_TEST/setup_SUITE $SYS2_TEST/config_SUITE
 $ ct_run -suite $SYS1_TEST/setup_SUITE -case start stop
 $ ct_run -suite $SYS1_TEST/setup_SUITE -group installation -case start stop

The flags dir, suite, and group/case can be combined. For example, to run x_SUITE and y_SUITE in directory testdir, as follows:

 $ ct_run -dir ./testdir -suite x_SUITE y_SUITE

This has the same effect as the following:

 $ ct_run -suite ./testdir/x_SUITE ./testdir/y_SUITE

For details, see Test Case Group Execution.

The following flags can also be used with ct_run:

Lists all available start flags.

Specifies where the HTML log files are to be written.

Associates the test run with a name that gets printed in the overview HTML log files.

Refreshes the top-level HTML index files.

Starts interactive shell mode (described later).

Steps through test cases using the Erlang Debugger (described later).

Uses test specification as input (described later).

Allows user-specific terms in a test specification (described later).

, tells Common Test to suppress printouts for specified connections (described later).

Points out a user HTML style sheet (described later).

To perform code coverage test (see Code Coverage Analysis).

To specify if the cover tool is to be stopped after the test is completed (see Code Coverage Analysis).

To install event handlers.

To install event handlers including start arguments.

To install Common Test Hooks including start arguments.

To enable or disable Built-in Common Test Hooks. Default is true.

Specifies include directories (described earlier).

Disables the automatic test suite compilation feature (described earlier).

Aborts the test run if one or more suites fail to compile (described earlier).

Extends timetrap time-out values.

Enables automatic timetrap time-out scaling.

Tells Common Test to repeat the tests n times (described later).

Tells Common Test to repeat the tests for duration of time (described later).

Tells Common Test to repeat the tests until stop_time (described later).

On time-out, the test run is aborted when the current test job is finished. If skip_rest is provided, the remaining test cases in the current test job are skipped (described later).

Provides a decryption key for encrypted configuration files.

Points out a file containing a decryption key for encrypted configuration files.

Switches off HTML enhancements that can be incompatible with older browsers.

Enables modification of the logging behavior, see Log options.

Sets verbosity levels for printouts.

Disables automatic escaping of special HTML characters. See the Logging chapter.

Note

Directories passed to Common Test can have either relative or absolute paths.

Note

Any start flags to the Erlang runtime system (application ERTS) can also be passed as parameters to ct_run. It is, for example, useful to be able to pass directories to be added to the Erlang code server search path with flag -pa or -pz. If you have common help- or library modules for test suites (separately compiled), stored in other directories than the test suite directories, these help/lib directories are preferably added to the code path this way.

Example:

$ ct_run -dir ./chat_server -logdir ./chat_server/testlogs -pa $PWD/chat_server/ebin

The absolute path of directory chat_server/ebin is here passed to the code server. This is essential because relative paths are stored by the code server as relative, and Common Test changes the current working directory of ERTS during the test run.

The ct_run program sets the exit status before shutting down. The following values are defined:

  • 0 indicates a successful testrun, that is, without failed or auto-skipped test cases.
  • 1 indicates that one or more test cases have failed, or have been auto-skipped.
  • 2 indicates that the test execution has failed because of, for example, compilation errors, or an illegal return value from an information function.

If auto-skipped test cases do not affect the exit status. The default behavior can be changed using start flag:

 -exit_status ignore_config
Note

Executing ct_run without start flags is equal to the command: ct_run -dir ./

For more information about the ct_run program, see module ct_run and section Installation.

Common Test provides an Erlang API for running tests. The main (and most flexible) function for specifying and executing tests is ct:run_test/1. It takes the same start parameters as ct_run, but the flags are instead specified as options in a list of key-value tuples. For example, a test specified with ct_run as follows:

$ ct_run -suite ./my_SUITE -logdir ./results

is with ct:run_test/1 specified as:

1> ct:run_test([{suite,"./my_SUITE"},{logdir,"./results"}]).

The function returns the test result, represented by the tuple {Ok,Failed,{UserSkipped,AutoSkipped}}, where each element is an integer. If test execution fails, the function returns the tuple {error,Reason}, where the term Reason explains the failure.

The default start option {dir,Cwd} (to run all suites in the current working directory) is used if the function is called with an empty list of options.

During execution of tests started with ct:run_test/1, the Erlang shell process, controlling stdin, remains the top-level process of the Common Test system of processes. Consequently, the Erlang shell is not available for interaction during the test run. If this is not desirable, for example, because the shell is needed for debugging purposes or for interaction with the SUT during test execution, set start option release_shell to true (in the call to ct:run_test/1 or by using the corresponding test specification term, described later). This makes Common Test release the shell immediately after the test suite compilation stage. To accomplish this, a test runner process is spawned to take control of the test execution. The effect is that ct:run_test/1 returns the pid of this process rather than the test result, which instead is printed to tty at the end of the test run.

Note

To use the functions ct:break/1,2 and ct:continue/0,1, release_shell must be set to true.

For details, see ct:run_test/1 manual page.

With the ct_run flag, or ct:run_test/1 option group, one or more test case groups can be specified, optionally in combination with specific test cases. The syntax for specifying groups on the command line is as follows:

 $ ct_run -group <group_names_or_paths> [-case <cases>]

The syntax in the Erlang shell is as follows:

 1> ct:run_test([{group,GroupsNamesOrPaths}, {case,Cases}]).

Parameter group_names_or_paths specifies one or more group names and/or one or more group paths. At startup, Common Test searches for matching groups in the group definitions tree (that is, the list returned from Suite:groups/0; for details, see section Test Case Groups.

Given a group name, say g, Common Test searches for all paths leading to g. By path is meant a sequence of nested groups, which must be followed to get from the top-level group to g. To execute the test cases in group g, Common Test must call the init_per_group/2 function for each group in the path to g, and all corresponding end_per_group/2 functions afterwards. This is because the configuration of a test case in g (and its Config input data) depends on init_per_testcase(TestCase, Config) and its return value, which in turn depends on init_per_group(g, Config) and its return value, which in turn depends on init_per_group/2 of the group above g, and so on, all the way up to the top-level group.

This means that if there is more than one way to locate a group (and its test cases) in a path, the result of the group search operation is a number of tests, all of which are to be performed. Common Test interprets a group specification that consists of a single name as follows:

"Search and find all paths in the group definitions tree that lead to the specified group and, for each path, create a test that does the following, in order:

  1. Executes all configuration functions in the path to the specified group.
  2. Executes all, or all matching, test cases in this group.
  3. Executes all, or all matching, test cases in all subgroups of the group."

The user can specify a specific group path with parameter group_names_or_paths. With this type of specification execution of unwanted groups (in otherwise matching paths), and/or the execution of subgroups can be avoided. The command line syntax of the group path is a list of group names in the path, for example:

$ ct_run -suite "./x_SUITE" -group [g1,g3,g4] -case tc1 tc5

The syntax in the Erlang shell is as follows (requires a list within the groups list):

1> ct:run_test([{suite,"./x_SUITE"}, {group,[[g1,g3,g4]]}, {testcase,[tc1,tc5]}]).

The last group in the specified path is the terminating group in the test, that is, no subgroups following this group are executed. In the previous example, g4 is the terminating group. Hence, Common Test executes a test that calls all init configuration functions in the path to g4, that is, g1..g3..g4. It then calls test cases tc1 and tc5 in g4, and finally all end configuration functions in order g4..g3..g1.

Note

The group path specification does not necessarily have to include all groups in the path to the terminating group. Common Test searches for all matching paths if an incomplete group path is specified.

Note

Group names and group paths can be combined with parameter group_names_or_paths. Each element is treated as an individual specification in combination with parameter cases. The following examples illustrates this.

Examples:

 -module(x_SUITE).
 ...
 %% The group definitions:      
 groups() ->
   [{top1,[],[tc11,tc12,
	      {sub11,[],[tc12,tc13]},
	      {sub12,[],[tc14,tc15,
			 {sub121,[],[tc12,tc16]}]}]},

    {top2,[],[{group,sub21},{group,sub22}]},
    {sub21,[],[tc21,{group,sub2X2}]},
    {sub22,[],[{group,sub221},tc21,tc22,{group,sub2X2}]},
    {sub221,[],[tc21,tc23]},
    {sub2X2,[],[tc21,tc24]}].

The following executes two tests, one for all cases and all subgroups under top1, and one for all under top2:

 $ ct_run -suite "x_SUITE" -group all
 1> ct:run_test([{suite,"x_SUITE"}, {group,all}]).

Using -group top1 top2, or {group,[top1,top2]} gives the same result.

The following executes one test for all cases and subgroups under top1:

 $ ct_run -suite "x_SUITE" -group top1
 1> ct:run_test([{suite,"x_SUITE"}, {group,[top1]}]).

The following runs a test executing tc12 in top1 and any subgroup under top1 where it can be found (sub11 and sub121):

 $ ct_run -suite "x_SUITE" -group top1 -case tc12
 1> ct:run_test([{suite,"x_SUITE"}, {group,[top1]}, {testcase,[tc12]}]).

The following executes tc12 only in group top1:

 $ ct_run -suite "x_SUITE" -group [top1] -case tc12
 1> ct:run_test([{suite,"x_SUITE"}, {group,[[top1]]}, {testcase,[tc12]}]).

The following searches top1 and all its subgroups for tc16 resulting in that this test case executes in group sub121:

 $ ct_run -suite "x_SUITE" -group top1 -case tc16
 1> ct:run_test([{suite,"x_SUITE"}, {group,[top1]}, {testcase,[tc16]}]).

Using the specific path -group [sub121] or {group,[[sub121]]} gives the same result in this example.

The following executes two tests, one including all cases and subgroups under sub12, and one with only the test cases in sub12:

 $ ct_run -suite "x_SUITE" -group sub12 [sub12]
 1> ct:run_test([{suite,"x_SUITE"}, {group,[sub12,[sub12]]}]).

In the following example, Common Test finds and executes two tests, one for the path from top2 to sub2X2 through sub21, and one from top2 to sub2X2 through sub22:

 $ ct_run -suite "x_SUITE" -group sub2X2
 1> ct:run_test([{suite,"x_SUITE"}, {group,[sub2X2]}]).

In the following example, by specifying the unique path top2 -> sub21 -> sub2X2, only one test is executed. The second possible path, from top2 to sub2X2 (from the former example) is discarded:

 $ ct_run -suite "x_SUITE" -group [sub21,sub2X2]
 1> ct:run_test([{suite,"x_SUITE"}, {group,[[sub21,sub2X2]]}]).

The following executes only the test cases for sub22 and in reverse order compared to the group definition:

 $ ct_run -suite "x_SUITE" -group [sub22] -case tc22 tc21
 1> ct:run_test([{suite,"x_SUITE"}, {group,[[sub22]]}, {testcase,[tc22,tc21]}]).

If a test case belonging to a group (according to the group definition) is executed without a group specification, that is, simply by (using the command line):

$ ct_run -suite "my_SUITE" -case my_tc

or (using the Erlang shell):

1> ct:run_test([{suite,"my_SUITE"}, {testcase,my_tc}]).

then Common Test ignores the group definition and executes the test case in the scope of the test suite only (no group configuration functions are called).

The group specification feature, as presented in this section, can also be used in Test Specifications (with some extra features added).

You can start Common Test in an interactive shell mode where no automatic testing is performed. Instead, Common Test starts its utility processes, installs configuration data (if any), and waits for the user to call functions (typically test case support functions) from the Erlang shell.

The shell mode is useful, for example, for debugging test suites, analyzing and debugging the SUT during "simulated" test case execution, and trying out various operations during test suite development.

To start the interactive shell mode, start an Erlang shell manually and call ct:install/1 to install any configuration data you might need (use [] as argument otherwise). Then call ct:start_interactive/0 to start Common Test.

If you use the ct_run program, you can start the Erlang shell and Common Test in one go by using the flag -shell and, optionally, flag -config and/or -userconfig.

Examples:

  • ct_run -shell
  • ct_run -shell -config cfg/db.cfg
  • ct_run -shell -userconfig db_login testuser x523qZ

If no configuration file is specified with command ct_run, a warning is displayed. If Common Test has been run from the same directory earlier, the same configuration file(s) are used again. If Common Test has not been run from this directory before, no configuration files are available.

If any functions using "required configuration data" (for example, functions ct_telnet or ct_ftp) are to be called from the Erlang shell, first require configuration data with ct:require/1,2. This is equivalent to a require statement in the Test Suite Information Function or in the Test Case Information Function.

Example:

 
 1> ct:require(unix_telnet, unix).
 ok
 2> ct_telnet:open(unix_telnet).
 {ok,<0.105.0>}
 4> ct_telnet:cmd(unix_telnet, "ls .").
 {ok,["ls .","file1  ...",...]}

Everything that Common Test normally prints in the test case logs, are in the interactive mode written to a log named ctlog.html in directory ct_run.<timestamp>. A link to this file is available in the file named last_interactive.html in the directory from which you execute ct_run. Specifying a different root directory for the logs than the current working directory is not supported.

If you wish to exit the interactive mode (for example, to start an automated test run with ct:run_test/1), call function ct:stop_interactive/0. This shuts down the running ct application. Associations between configuration names and data created with require are consequently deleted. Function ct:start_interactive/0 takes you back into interactive mode, but the previous state is not restored.

Using ct_run -step [opts], or by passing option {step,Opts} to ct:run_test/1, the following is possible:

  • Get the Erlang Debugger started automatically.
  • Use its graphical interface to investigate the state of the current test case.
  • Execute the test case step-by-step and/or set execution breakpoints.

If no extra options are specified with flag/option step, breakpoints are set automatically on the test cases that are to be executed by Common Test, and those functions only. If step option config is specified, breakpoints are also initially set on the configuration functions in the suite, that is, init_per_suite/1, end_per_suite/1, init_per_group/2, end_per_group/2, init_per_testcase/2 and end_per_testcase/2.

Common Test enables the Debugger auto-attach feature, which means that for every new interpreted test case function that starts to execute, a new trace window automatically pops up (as each test case executes on a dedicated Erlang process). Whenever a new test case starts, Common Test attempts to close the inactive trace window of the previous test case. However, if you prefer Common Test to leave inactive trace windows, use option keep_inactive.

The step functionality can be used together with flag/option suite and suite + case/testcase, but not together with dir.

The most flexible way to specify what to test, is to use a test specification, which is a sequence of Erlang terms. The terms are normally declared in one or more text files (see ct:run_test/1), but can also be passed to Common Test on the form of a list (see ct:run_testspec/1). There are two general types of terms: configuration terms and test specification terms.

With configuration terms it is, for example, possible to do the following:

  • Label the test run (similar to ct_run -label).
  • Evaluate any expressions before starting the test.
  • Import configuration data (similar to ct_run -config/-userconfig).
  • Specify the top-level HTML log directory (similar to ct_run -logdir).
  • Enable code coverage analysis (similar to ct_run -cover).
  • Install Common Test Hooks (similar to ct_run -ch_hooks).
  • Install event_handler plugins (similar to ct_run -event_handler).
  • Specify include directories to be passed to the compiler for automatic compilation (similar to ct_run -include).
  • Disable the auto-compilation feature (similar to ct_run -no_auto_compile).
  • Set verbosity levels (similar to ct_run -verbosity).

Configuration terms can be combined with ct_run start flags or ct:run_test/1 options. The result is, for some flags/options and terms, that the values are merged (for example, configuration files, include directories, verbosity levels, and silent connections) and for others that the start flags/options override the test specification terms (for example, log directory, label, style sheet, and auto-compilation).

With test specification terms, it is possible to state exactly which tests to run and in which order. A test term specifies either one or more suites, one or more test case groups (possibly nested), or one or more test cases in a group (or in multiple groups) or in a suite.

Any number of test terms can be declared in sequence. Common Test compiles by default the terms into one or more tests to be performed in one resulting test run. A term that specifies a set of test cases "swallows" one that only specifies a subset of these cases. For example, the result of merging one term specifying that all cases in suite S are to be executed, with another term specifying only test case X and Y in S, is a test of all cases in S. However, if a term specifying test case X and Y in S is merged with a term specifying case Z in S, the result is a test of X, Y, and Z in S. To disable this behavior, that is, to instead perform each test sequentially in a "script-like" manner, set term merge_tests to false in the test specification.

A test term can also specify one or more test suites, groups, or test cases to be skipped. Skipped suites, groups, and cases are not executed and show up in the HTML log files as SKIPPED.

When multiple test specification files are specified at startup (either with ct_run -spec file1 file2 ... or ct:run_test([{spec, [File1,File2,...]}])), Common Test either executes one test run per specification file, or joins the files and performs all tests within one single test run. The first behavior is the default one. The latter requires that start flag/option join_specs is provided, for example, run_test -spec ./my_tests1.ts ./my_tests2.ts -join_specs.

Joining a number of specifications, or running them separately, can also be accomplished with (and can be combined with) test specification file inclusion.

With the term specs, a test specification can include other specifications. An included specification can either be joined with the source specification or used to produce a separate test run (as with start flag/option join_specs above).

Example:

 %% In specification file "a.spec"
 {specs, join, ["b.spec", "c.spec"]}.
 {specs, separate, ["d.spec", "e.spec"]}.
 %% Config and test terms follow
 ...

In this example, the test terms defined in files "b.spec" and "c.spec" are joined with the terms in source specification "a.spec" (if any). The inclusion of specifications "d.spec" and "e.spec" results in two separate, and independent, test runs (one for each included specification).

Option join does not imply that the test terms are merged, only that all tests are executed in one single test run.

Joined specifications share common configuration settings, such as the list of config files or include directories. For configurations that cannot be combined, such as settings for logdir or verbosity, it is up to the user to ensure there are no clashes when the test specifications are joined. Specifications included with option separate do not share configuration settings with the source specification. This is useful, for example, if there are clashing configuration settings in included specifications, making it them impossible to join.

If {merge_tests,true} is set in the source specification (which is the default setting), terms in joined specifications are merged with terms in the source specification (according to the description of merge_tests earlier).

Notice that it is always the merge_tests setting in the source specification that is used when joined with other specifications. Say, for example, that a source specification A, with tests TA1 and TA2, has {merge_tests,false} set, and that it includes another specification, B, with tests TB1 and TB2, that has {merge_tests,true} set. The result is that the test series TA1,TA2,merge(TB1,TB2) is executed. The opposite merge_tests settings would result in the test series merge(merge(TA1,TA2),TB1,TB2).

The term specs can be used to nest specifications, that is, have one specification include other specifications, which in turn include others, and so no

When a test case group is specified, the resulting test executes function init_per_group, followed by all test cases and subgroups (including their configuration functions), and finally function end_per_group. Also, if particular test cases in a group are specified, init_per_group and end_per_group, for the group in question, are called. If a group defined (in Suite:groups/0) as a subgroup of another group, is specified (or if particular test cases of a subgroup are), Common Test calls the configuration functions for the top-level groups and for the subgroup in question (making it possible to pass configuration data all the way from init_per_suite down to the test cases in the subgroup).

The test specification uses the same mechanism for specifying test case groups through names and paths, as explained in section Test Case Group Execution, with the addition of element GroupSpec.

Element GroupSpec makes it possible to specify group execution properties that overrides those in the group definition (that is, in groups/0). Execution properties for subgroups might be overridden as well. This feature makes it possible to change properties of groups at the time of execution, without having to edit the test suite. The same feature is available for group elements in the Suite:all/0 list. For details and examples, see section Test Case Groups.

Test specifications can be used to run tests both in a single test host environment and in a distributed Common Test environment (Large Scale Testing). The node parameters in term init are only relevant in the latter (see section Test Specifications in Large Scale Testing). For details about the various terms, see the corresponding sections in the User's Guide, for example, the following:

Configuration terms:

 {merge_tests, Bool}.

 {define, Constant, Value}.

 {specs, InclSpecsOption, TestSpecs}.

 {node, NodeAlias, Node}.

 {init, InitOptions}.
 {init, [NodeAlias], InitOptions}.

 {label, Label}.
 {label, NodeRefs, Label}.

 {verbosity, VerbosityLevels}.
 {verbosity, NodeRefs, VerbosityLevels}.

 {stylesheet, CSSFile}.
 {stylesheet, NodeRefs, CSSFile}.

 {silent_connections, ConnTypes}.
 {silent_connections, NodeRefs, ConnTypes}.

 {multiply_timetraps, N}.
 {multiply_timetraps, NodeRefs, N}.

 {scale_timetraps, Bool}.
 {scale_timetraps, NodeRefs, Bool}.

 {cover, CoverSpecFile}.
 {cover, NodeRefs, CoverSpecFile}.

 {cover_stop, Bool}.
 {cover_stop, NodeRefs, Bool}.

 {include, IncludeDirs}.
 {include, NodeRefs, IncludeDirs}.

 {auto_compile, Bool},
 {auto_compile, NodeRefs, Bool},

 {abort_if_missing_suites, Bool},
 {abort_if_missing_suites, NodeRefs, Bool},

 {config, ConfigFiles}.
 {config, ConfigDir, ConfigBaseNames}.
 {config, NodeRefs, ConfigFiles}.
 {config, NodeRefs, ConfigDir, ConfigBaseNames}.

 {userconfig, {CallbackModule, ConfigStrings}}.
 {userconfig, NodeRefs, {CallbackModule, ConfigStrings}}.

 {logdir, LogDir}.                                        
 {logdir, NodeRefs, LogDir}.

 {logopts, LogOpts}.
 {logopts, NodeRefs, LogOpts}.

 {create_priv_dir, PrivDirOption}.
 {create_priv_dir, NodeRefs, PrivDirOption}.

 {event_handler, EventHandlers}.
 {event_handler, NodeRefs, EventHandlers}.
 {event_handler, EventHandlers, InitArgs}.
 {event_handler, NodeRefs, EventHandlers, InitArgs}.

 {ct_hooks, CTHModules}.
 {ct_hooks, NodeRefs, CTHModules}.

 {enable_builtin_hooks, Bool}.

 {basic_html, Bool}.
 {basic_html, NodeRefs, Bool}.

 {esc_chars, Bool}.
 {esc_chars, NodeRefs, Bool}.

 {release_shell, Bool}.

Test terms:

 {suites, Dir, Suites}.                                
 {suites, NodeRefs, Dir, Suites}.

 {groups, Dir, Suite, Groups}.
 {groups, NodeRefs, Dir, Suite, Groups}.

 {groups, Dir, Suite, Groups, {cases,Cases}}.
 {groups, NodeRefs, Dir, Suite, Groups, {cases,Cases}}.

 {cases, Dir, Suite, Cases}.                           
 {cases, NodeRefs, Dir, Suite, Cases}.

 {skip_suites, Dir, Suites, Comment}.
 {skip_suites, NodeRefs, Dir, Suites, Comment}.

 {skip_groups, Dir, Suite, GroupNames, Comment}.
 {skip_groups, NodeRefs, Dir, Suite, GroupNames, Comment}.

 {skip_cases, Dir, Suite, Cases, Comment}.
 {skip_cases, NodeRefs, Dir, Suite, Cases, Comment}.

Types:

 Bool            = true | false
 Constant        = atom()
 Value           = term()
 InclSpecsOption = join | separate
 TestSpecs       = string() | [string()]
 NodeAlias       = atom()
 Node            = node()
 NodeRef         = NodeAlias | Node | master
 NodeRefs        = all_nodes | [NodeRef] | NodeRef
 InitOptions     = term()
 Label           = atom() | string()
 VerbosityLevels = integer() | [{Category,integer()}]
 Category        = atom()
 CSSFile         = string()
 ConnTypes       = all | [atom()]
 N               = integer()
 CoverSpecFile   = string()
 IncludeDirs     = string() | [string()]
 ConfigFiles     = string() | [string()]
 ConfigDir       = string()
 ConfigBaseNames = string() | [string()]
 CallbackModule  = atom()
 ConfigStrings   = string() | [string()]
 LogDir          = string()
 LogOpts         = [term()]
 PrivDirOption   = auto_per_run | auto_per_tc | manual_per_tc
 EventHandlers   = atom() | [atom()]
 InitArgs        = [term()]
 CTHModules      = [CTHModule |
		    {CTHModule, CTHInitArgs} |
		    {CTHModule, CTHInitArgs, CTHPriority}]
 CTHModule       = atom()
 CTHInitArgs     = term()
 Dir             = string()
 Suites          = atom() | [atom()] | all
 Suite           = atom()
 Groups          = GroupPath | GroupSpec | [GroupSpec] | all
 GroupPath       = [[GroupSpec]]
 GroupSpec       = GroupName | {GroupName,Properties} | {GroupName,Properties,[GroupSpec]}
 GroupName       = atom()
 GroupNames      = GroupName | [GroupName]
 Cases           = atom() | [atom()] | all
 Comment         = string() | ""

The difference between the config terms above is that with ConfigDir, ConfigBaseNames is a list of base names, that is, without directory paths. ConfigFiles must be full names, including paths. For example, the following two terms have the same meaning:

 {config, ["/home/testuser/tests/config/nodeA.cfg",
           "/home/testuser/tests/config/nodeB.cfg"]}.

 {config, "/home/testuser/tests/config", ["nodeA.cfg","nodeB.cfg"]}.
Note

Any relative paths, specified in the test specification, are relative to the directory containing the test specification file if ct_run -spec TestSpecFile ... or ct:run:test([{spec,TestSpecFile},...]) executes the test.

The path is relative to the top-level log directory if ct:run:testspec(TestSpec) executes the test.

The term define introduces a constant that is used to replace the name Constant with Value, wherever it is found in the test specification. This replacement occurs during an initial iteration through the test specification. Constants can be used anywhere in the test specification, for example, in any lists and tuples, and even in strings and inside the value part of other constant definitions. A constant can also be part of a node name, but that is the only place where a constant can be part of an atom.

Note

For the sake of readability, the name of the constant must always begin with an uppercase letter, or a $, ?, or _. This means that it must always be single quoted (as the constant name is an atom, not text).

The main benefit of constants is that they can be used to reduce the size (and avoid repetition) of long strings, such as file paths.

Examples:

 %% 1a. no constant
 {config, "/home/testuser/tests/config", ["nodeA.cfg","nodeB.cfg"]}.
 {suites, "/home/testuser/tests/suites", all}.

 %% 1b. with constant
 {define, 'TESTDIR', "/home/testuser/tests"}.
 {config, "'TESTDIR'/config", ["nodeA.cfg","nodeB.cfg"]}.
 {suites, "'TESTDIR'/suites", all}.

 %% 2a. no constants
 {config, [testnode@host1, testnode@host2], "../config", ["nodeA.cfg","nodeB.cfg"]}.
 {suites, [testnode@host1, testnode@host2], "../suites", [x_SUITE, y_SUITE]}.

 %% 2b. with constants
 {define, 'NODE', testnode}.
 {define, 'NODES', ['NODE'@host1, 'NODE'@host2]}.
 {config, 'NODES', "../config", ["nodeA.cfg","nodeB.cfg"]}.
 {suites, 'NODES', "../suites", [x_SUITE, y_SUITE]}.

Constants make the test specification term alias, in previous versions of Common Test, redundant. This term is deprecated but remains supported in upcoming Common Test releases. Replacing alias terms with define is strongly recommended though. An example of such replacement follows:

 %% using the old alias term
 {config, "/home/testuser/tests/config/nodeA.cfg"}.
 {alias, suite_dir, "/home/testuser/tests/suites"}.
 {groups, suite_dir, x_SUITE, group1}.

 %% replacing with constants
 {define, 'TestDir', "/home/testuser/tests"}.
 {define, 'CfgDir', "'TestDir'/config"}.
 {define, 'SuiteDir', "'TestDir'/suites"}.
 {config, 'CfgDir', "nodeA.cfg"}.
 {groups, 'SuiteDir', x_SUITE, group1}.

Constants can well replace term node also, but this still has a declarative value, mainly when used in combination with NodeRefs == all_nodes (see Types).

Here follows a simple test specification example:

 {define, 'Top', "/home/test"}.
 {define, 'T1', "'Top'/t1"}.
 {define, 'T2', "'Top'/t2"}.
 {define, 'T3', "'Top'/t3"}.
 {define, 'CfgFile', "config.cfg"}.

 {logdir, "'Top'/logs"}.

 {config, ["'T1'/'CfgFile'", "'T2'/'CfgFile'", "'T3'/'CfgFile'"]}.

 {suites, 'T1', all}.
 {skip_suites, 'T1', [t1B_SUITE,t1D_SUITE], "Not implemented"}.
 {skip_cases, 'T1', t1A_SUITE, [test3,test4], "Irrelevant"}.
 {skip_cases, 'T1', t1C_SUITE, [test1], "Ignore"}.

 {suites, 'T2', [t2B_SUITE,t2C_SUITE]}.
 {cases, 'T2', t2A_SUITE, [test4,test1,test7]}.

 {skip_suites, 'T3', all, "Not implemented"}.

The example specifies the following:

  • The specified logdir directory is used for storing the HTML log files (in subdirectories tagged with node name, date, and time).
  • The variables in the specified test system configuration files are imported for the test.
  • The first test to run includes all suites for system t1. Suites t1B and t1D are excluded from the test. Test cases test3 and test4 in t1A and test1 case in t1C are also excluded from the test.
  • The second test to run is for system t2. The included suites are t2B and t2C. Test cases test4, test1, and test7 in suite t2A are also included. The test cases are executed in the specified order.
  • The last test to run is for system t3. Here, all suites are skipped and this is explicitly noted in the log files.

With term init it is possible to specify initialization options for nodes defined in the test specification. There are options to start the node and to evaluate any function on the node. For details, see section Automatic Startup of Test Target Nodes in section Using Common Test for Large Scale Testing.

The user can provide a test specification including (for Common Test) unrecognizable terms. If this is desired, use flag -allow_user_terms when starting tests with ct_run. This forces Common Test to ignore unrecognizable terms. In this mode, Common Test is not able to check the specification for errors as efficiently as if the scanner runs in default mode. If ct:run_test/1 is used for starting the tests, the relaxed scanner mode is enabled by tuple {allow_user_terms,true}.

Terms in the current test specification (that is, the specification that has been used to configure and run the current test) can be looked up. The function get_testspec_terms() returns a list of all test specification terms (both configuration terms and test terms), and get_testspec_terms(Tags) returns the term (or a list of terms) matching the tag (or tags) in Tags.

For example, in the test specification:

 ...
 {label, my_server_smoke_test}.
 {config, "../../my_server_setup.cfg"}.
 {config, "../../my_server_interface.cfg"}.
 ...

And in, for example, a test suite or a Common Test Hook function:

 ...
 [{label,[{_Node,TestType}]}, {config,CfgFiles}] =
     ct:get_testspec_terms([label,config]),

 [verify_my_server_cfg(TestType, CfgFile) || {Node,CfgFile} <- CfgFiles,
					     Node == node()];
 ...

As the execution of the test suites proceed, events are logged in the following four different ways:

  • Text to the operator console.
  • Suite-related information is sent to the major log file.
  • Case-related information is sent to the minor log file.
  • The HTML overview log file is updated with test results.
  • A link to all runs executed from a certain directory is written in the log named all_runs.html and direct links to all tests (the latest results) are written to the top-level index.html.

Typically the operator, possibly running hundreds or thousands of test cases, does not want to fill the console with details about, or printouts from, specific test cases. By default, the operator only sees the following:

  • A confirmation that the test has started and information about how many test cases are executed in total.
  • A small note about each failed test case.
  • A summary of all the run test cases.
  • A confirmation when the test run is complete.
  • Some special information, such as error reports, progress reports, and printouts written with erlang:display/1, or io:format/3 specifically addressed to a receiver other than standard_io (for example, the default group leader process user).

To dig deeper into the general results, or the result of a specific test case, the operator can do so by following the links in the HTML presentation and read the major or minor log files. The "all_runs.html" page is a good starting point. It is located in logdir and contains a link to each test run, including a quick overview (with date and time, node name, number of tests, test names, and test result totals).

An "index.html" page is written for each test run (that is, stored in the ct_run directory tagged with node name, date, and time). This file provides an overview of all individual tests performed in the same test run. The test names follow the following convention:

  • TopLevelDir.TestDir (all suites in TestDir executed)
  • TopLevelDir.TestDir:suites (specific suites executed)
  • TopLevelDir.TestDir.Suite (all cases in Suite executed)
  • TopLevelDir.TestDir.Suite:cases (specific test cases executed)
  • TopLevelDir.TestDir.Suite.Case (only Case executed)

The "test run index" page includes a link to the Common Test Framework Log file in which information about imported configuration data and general test progress is written. This log file is useful to get snapshot information about the test run during execution. It can also be helpful when analyzing test results or debugging test suites.

The "test run index" page indicates if a test has missing suites (that is, suites that Common Test failed to compile). Names of the missing suites can be found in the Common Test Framework Log file.

The major log file shows a detailed report of the test run. It includes test suite and test case names, execution time, the exact reason for failures, and so on. The information is available in both a file with textual and with HTML representation. The HTML file shows a summary that gives a good overview of the test run. It also has links to each individual test case log file for quick viewing with an HTML browser.

The minor log files contain full details of every single test case, each in a separate file. This way, it is straightforward to compare the latest results to that of previous test runs, even if the set of test cases changes. If application SASL is running, its logs are also printed to the current minor log file by the cth_log_redirect built-in hook.

The full name of the minor log file (that is, the name of the file including the absolute directory path) can be read during execution of the test case. It comes as value in tuple {tc_logfile,LogFileName} in the Config list (which means it can also be read by a pre- or post Common Test Hook function). Also, at the start of a test case, this data is sent with an event to any installed event handler. For details, see section Event Handling.

The log files are written continuously during a test run and links are always created initially when a test starts. Thevtest progress can therefore be followed simply by refreshing pages in the HTML browser. Statistics totals are not presented until a test is complete however.

With start flag logopts options that modify some aspects of the logging behavior can be specified. The following options are available:

The HTML version of the test suite source code is not generated during the test run (and is consequently not available in the log file system).

Common Test does not add a newline character (\n) to the end of an output string that it receives from a call to, for example, io:format/2, and which it prints to the test case log.

For example, if a test is started with:

$ ct_run -suite my_SUITE -logopts no_nl

then printouts during the test made by successive calls to io:format("x"), appears in the test case log as:

xxx

instead of each x printed on a new line, which is the default behavior.

By clicking the name in the column header of any table (for example, "Ok", "Case", "Time", and so on), the table rows are sorted in whatever order makes sense for the type of value (for example, numerical for "Ok" or "Time", and alphabetical for "Case"). The sorting is performed through JavaScript code, automatically inserted into the HTML log files. Common Test uses the jQuery library and the tablesorter plugin, with customized sorting functions, for this implementation.

The test suites overview page includes a link to the Unexpected I/O Log. In this log, Common Test saves printouts made with ct:log/1,2,3,4,5 and ct:pal/1,2,3,4,5, as well as captured system error- and progress reports, which cannot be associated with particular test cases and therefore cannot be written to individual test case log files. This occurs, for example, if a log printout is made from an external process (not a test case process), or if an error- or progress report comes in, during a short interval while Common Test is not executing a test case or configuration function, or while Common Test is currently executing a parallel test case group.

The Pre- and Post Test I/O Log

The Common Test Framework Log page includes links to the Pre- and Post Test I/O Log. In this log, Common Test saves printouts made with ct:log/1,2,3,4,5 and ct:pal/1,2,3,4,5, as well as captured system error- and progress reports, which take place before, and after, the test run. Examples of this are printouts from a CT hook init- or terminate function, or progress reports generated when an OTP application is started from a CT hook init function. Another example is an error report generated because of a failure when an external application is stopped from a CT hook terminate function. All information in these examples ends up in the Pre- and Post Test I/O Log. For more information on how to synchronize test runs with external user applications, see section Synchronizing in section Common Test Hooks.

Note

Logging to file with ct:log/1,2,3,4,5 or ct:pal/1,2,3,4,5 only works when Common Test is running. Printouts with ct:pal/1,2,3,4,5 are however always displayed on screen.

Common Test can automatically delete old log. This is specified with the keep_logs option. The default value for this option is all, which means that no logs are deleted. If the value is set to an integer, N, Common Test deletes all ct_run.<timestamp> directories, except the N newest.

Common Test uses an HTML Style Sheet (CSS file) to control the look of the HTML log files generated during test runs. If the log files are not displayed correctly in the browser of your choice, or you prefer a more primitive ("pre Common Test v1.6") look of the logs, use the start flag/option:

 basic_html

This disables the use of style sheets and JavaScripts (see Sorting HTML Table Columns).

Common Test includes an optional feature to allow user HTML style sheets for customizing printouts. The functions in ct that print to a test case HTML log file (log/3,4,5 and pal/3,4,5) accept Category as first argument. With this argument a category can be specified that can be mapped to a named div selector in a CSS rule-set. This is useful, especially for coloring text differently depending on the type of (or reason for) the printout. Say you want one particular background color for test system configuration information, a different one for test system state information, and finally one for errors detected by the test case functions. The corresponding style sheet can look as follows:

 div.sys_config  { background:blue }
 div.sys_state   { background:yellow }
 div.error       { background:red }

Common Test prints the text from ct:log/3,4,5 or ct:pal/3,4,5 inside a pre element nested under the named div element. Since the pre selector has a predefined CSS rule (in file ct_default.css) for the attributes color, font-family and font-size, if a user wants to change any of the predefined attribute settings, a new rule for pre must be added to the user stylesheet. Example:

div.error pre { color:white }

Here, white text is used instead of the default black for div.error printouts (and no other attribute settings for pre are affected).

To install the CSS file (Common Test inlines the definition in the HTML code), the file name can be provided when executing ct_run.

Example:

 $ ct_run -dir $TEST/prog -stylesheet $TEST/styles/test_categories.css

Categories in a CSS file installed with flag -stylesheet are on a global test level in the sense that they can be used in any suite that is part of the test run.

Style sheets can also be installed on a per suite and per test case basis.

Example:

 -module(my_SUITE).
 ...
 suite() -> [..., {stylesheet,"suite_categories.css"}, ...].
 ...
 my_testcase(_) ->
     ...
     ct:log(sys_config, "Test node version: ~p", [VersionInfo]),
     ...
     ct:log(sys_state, "Connections: ~p", [ConnectionInfo]),
     ...
     ct:pal(error, "Error ~p detected! Info: ~p", [SomeFault,ErrorInfo]),
     ct:fail(SomeFault).

If the style sheet is installed as in this example, the categories are private to the suite in question. They can be used by all test cases in the suite, but cannot be used by other suites. A suite private style sheet, if specified, is used in favor of a global style sheet (one specified with flag -stylesheet). A stylesheet tuple (as returned by suite/0 above) can also be returned from a test case information function. In this case the categories specified in the style sheet can only be used in that particular test case. A test case private style sheet is used in favor of a suite or global level style sheet.

In a tuple {stylesheet,CSSFile}, if CSSFile is specified with a path, for example, "$TEST/styles/categories.css", this full name is used to locate the file. However, if only the file name is specified, for example, categories.css, the CSS file is assumed to be located in the data directory, data_dir, of the suite. The latter use is recommended, as it is portable compared to hard coding path names in the suite.

Argument Category in the previous example can have the value (atom) sys_config (blue background), sys_state (yellow background), or error (white text on red background).

You can order Common Test to repeat the tests you specify. You can choose to repeat tests a number of times, repeat tests for a specific period of time, or repeat tests until a particular stop time is reached. If repetition is controlled by time, an action for Common Test to take upon time-out can be specified. Either Common Test performs all tests in the current run before stopping, or it stops when the current test job is finished. Repetition can be activated by ct_run start flags, or tuples in the ct:run:test/1 option list argument. The flags (options in parentheses) are the following:

  • -repeat N ({repeat,N}), where N is a positive integer
  • -duration DurTime ({duration,DurTime}), where DurTime is the duration
  • -until StopTime ({until,StopTime}), where StopTime is finish time
  • -force_stop ({force_stop,true})
  • -force_stop skip_rest ({force_stop,skip_rest})

The duration time is specified as HHMMSS, for example, -duration 012030 or {duration,"012030"}

, which means that the tests are executed and (if time allows) repeated until time-out occurs after 1 hour, 20 minutes, and 30 seconds.

The finish time can be specified as HHMMSS and is then interpreted as a time today (or possibly tomorrow), but can also be specified as YYMoMoDDHHMMSS, for example, -until 071001120000 or {until,"071001120000"}. This means that the tests are executed and (if time allows) repeated, until 12 o'clock on the 1st of October 2007.

When time-out occurs, Common Test never aborts the ongoing test case, as this can leave the SUT in an undefined, and possibly bad, state. Instead Common Test, by default, finishes the current test run before stopping. If flag force_stop is specified, Common Test stops when the current test job is finished. If flag force_stop is specified with skip_rest, Common Test only completes the current test case and skips the remaining tests in the test job.

Note

As Common Test always finishes at least the current test case, the time specified with duration or until is never definitive.

Log files from every repeated test run is saved in normal Common Test fashion (described earlier).

Common Test might later support an optional feature to only store the last (and possibly the first) set of logs of repeated test runs, but for now the user must be careful not to run out of disk space if tests are repeated during long periods of time.

For each test run that is part of a repeated session, information about the particular test run is printed in the Common Test Framework Log. The information includes the repetition number, remaining time, and so on.

Example 1:

 $ ct_run -dir $TEST_ROOT/to1 $TEST_ROOT/to2 -duration 001000 -force_stop

Here, the suites in test directory to1, followed by the suites in to2, are executed in one test run. A time-out event occurs after 10 minutes. As long as there is time left, Common Test repeats the test run (that is, starting over with test to1). After time-out, Common Test stops when the current job is finished (because of flag force_stop). As a result, the specified test run can be aborted after test to1 and before test to2.

Example 2:

 $ ct_run -dir $TEST_ROOT/to1 $TEST_ROOT/to2 -duration 001000 -forces_stop skip_rest

Here, the same tests as in Example 1 are run, but with flag force_stop set to skip_rest. If time-out occurs while executing tests in directory to1, the remaining test cases in to1 are skipped and the test is aborted without running the tests in to2 another time. If time-out occurs while executing tests in directory to2, the remaining test cases in to2 are skipped and the test is aborted.

Example 3:

 $ date
 Fri Sep 28 15:00:00 MEST 2007

 $ ct_run -dir $TEST_ROOT/to1 $TEST_ROOT/to2 -until 160000

Here, the same test run as in the previous examples are executed (and possibly repeated). However, when the time-out occurs, after 1 hour, Common Test finishes the entire test run before stopping (that is, both to1 and to2 are always executed in the same test run).

Example 4:

 $ ct_run -dir $TEST_ROOT/to1 $TEST_ROOT/to2 -repeat 5

Here, the test run, including both the to1 and the to2 test, is repeated five times.

Note

Do not confuse this feature with the repeat property of a test case group. The options described here are used to repeat execution of entire test runs, while the repeat property of a test case group makes it possible to repeat execution of sets of test cases within a suite. For more information about the latter, see section Test Case Groups in section Writing Test Suites.

The protocol handling processes in Common Test, implemented by ct_telnet, ct_ssh, ct_ftp, and so on, do verbose printing to the test case logs. This can be switched off with flag -silent_connections:

 ct_run -silent_connections [conn_types]

Here, conn_types specifies SSH, Telnet, FTP, RPC, and/or SNMP.

Example 1:

 ct_run ... -silent_connections ssh telnet

This switches off logging for SSH and Telnet connections.

Example 2:

 ct_run ... -silent_connections

This switches off logging for all connection types.

Fatal communication error and reconnection attempts are always printed, even if logging has been suppressed for the connection type in question. However, operations such as sending and receiving data are performed silently.

silent_connections can also be specified in a test suite. This is accomplished by returning a tuple, {silent_connections,ConnTypes}, in the suite/0 or test case information list. If ConnTypes is a list of atoms (SSH, Telnet, FTP, RPC and/or SNMP), output for any corresponding connections are suppressed. Full logging is by default enabled for any connection of type not specified in ConnTypes. Hence, if ConnTypes is the empty list, logging is enabled for all connections.

Example 3:

 -module(my_SUITE).

 suite() -> [..., {silent_connections,[telnet,ssh]}, ...].

 ...

 my_testcase1() ->
     [{silent_connections,[ssh]}].

 my_testcase1(_) ->
     ...

 my_testcase2(_) ->
     ...

In this example, suite/0 tells Common Test to suppress printouts from Telnet and SSH connections. This is valid for all test cases. However, my_testcase1/0 specifies that for this test case, only SSH is to be silent. The result is that my_testcase1 gets Telnet information (if any) printed in the log, but not SSH information. my_testcase2 gets no information from either connection printed.

silent_connections can also be specified with a term in a test specification (see section Test Specifications in section Running Tests and Analyzing Results). Connections provided with start flag/option silent_connections are merged with any connections listed in the test specification.

Start flag/option silent_connections and the test specification term override any settings made by the information functions inside the test suite.

Note

In the current Common Test version, the silent_connections feature only works for Telnet and SSH connections. Support for other connection types can be added in future Common Test versions.