20032013 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. Running Tests and Analyzing Results Peter Andersson, Kenneth Lundin run_test_chapter.xml
Using the Common Test Framework

The Common Test Framework provides a high level operator interface for testing. It adds the following features to the Erlang/OTP Test Server:

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

When Common Test starts, it will automatically attempt to compile any suites included in the specified tests. If particular suites have been specified, only those suites will be compiled. If a particular test object directory has been specified (meaning all suites in this directory should be part of the test), Common Test runs make:all/1 in the directory to compile the suites.

If compilation should fail for one or more suites, the compilation errors are printed to tty and the operator is asked if the test run should proceed without the missing suites, or be aborted. If the operator chooses to proceed, it is noted in the HTML log which tests have missing suites.

Any help module (i.e. 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, will also be automatically compiled. A help module will not be mistaken for a test suite (unless it has a "_SUITE" name of course). 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, you may specify these include directories by means of the flag with , or the option with . In addition to this, an include path may be specified with an OS environment variable; . Example (bash):

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

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

It is also possible to specify include directories in test specifications (see below).

If the user wants to run all test suites for a test object (or OTP application) by specifying only the top directory (e.g. with the dir start flag/option), Common Test will primarily look for test suite modules in a subdirectory named test. If this subdirectory doesn't exist, the specified top directory is assumed to be the actual test directory, and test suites will be read from there instead.

It is possible to disable the automatic compilation feature by using the flag with , or the option with . 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 can not be loaded from the local file system during startup of Common Test, the user needs to pre-load the modules before starting the test. Common Test will only verify that the specified test suites exist (i.e. that they are, or can be, loaded). This is useful e.g. if the test suites are transferred and loaded as binaries via RPC from a remote node.

Running tests from the OS command line

The ct_run program can be used for running tests from the OS command line, e.g.

-dir ]]> -suite ]]> -suite ]]> -suite -group -case ]]>

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

It is also possible to combine the dir, suite and group/case flags. E.g, to run x_SUITE and y_SUITE in directory testdir:

$ ct_run -dir ./testdir -suite x_SUITE y_SUITE

This has the same effect as calling:

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

For more details on test case group execution, please see below.

Other flags that may be used with ct_run:

]]>, 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. -refresh_logs, refreshes the top level HTML index files. -vts, start web based GUI (see below). -shell, start interactive shell mode (see below). -step [step_opts], step through test cases using the Erlang Debugger (see below). ]]>, use test specification as input (see below). -allow_user_terms, allows user specific terms in a test specification (see below). -silent_connections [conn_types], tells Common Test to suppress printouts for specified connections (see below). ]]>, points out a user HTML style sheet (see below). ]]>, to perform code coverage test (see Code Coverage Analysis). ]]>, to specify if the cover tool shall 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/disable Built-in Common Test Hooks. Default is true. , specifies include directories (see above). , disables the automatic test suite compilation feature (see above). ]]>, extends timetrap timeout values. ]]>, enables automatic timetrap timeout scaling. ]]>, tells Common Test to repeat the tests n times (see below). ]]>, tells Common Test to repeat the tests for duration of time (see below). ]]>, tells Common Test to repeat the tests until stop_time (see below). -force_stop [skip_rest], on timeout, the test run will be aborted when current test job is finished. If skip_rest is provided the rest of the test cases in the current test job will be skipped (see below). ]]>, 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 might not be compatible with older browsers. ]]>, makes it possible to modify aspects of the logging behaviour, see Log options below. ]]>, sets verbosity levels for printouts.

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

Arbitrary start flags to the Erlang Runtime System may also be passed as parameters to ct_run. It is, for example, useful to be able to pass directories that should be added to the Erlang code server search path with the -pa or -pz flag. 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 preferrably added to the code path this way. Example:

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

Note how in this example, the absolute path of the chat_server/ebin directory is passed to the code server. This is essential since relative paths are stored by the code server as relative, and Common Test changes the current working directory of the Erlang Runtime System 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, i.e. one 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 e.g. compilation errors, an illegal return value from an info function, etc.

If auto-skipped test cases should not affect the exit status, you may change the default behaviour using start flag:

-exit_status ignore_config

For more information about the ct_run program, see the Reference Manual and the Installation chapter.

Running tests from the Erlang shell or from an Erlang program

Common Test provides an Erlang API for running tests. The main (and most flexible) function for specifying and executing tests is called ct:run_test/1. This function takes the same start parameters as the ct_run program described above, only the flags are instead given as options in a list of key-value tuples. E.g. a test specified with ct_run like:

$ 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.

Releasing the Erlang shell

During execution of tests, started with ct:run_test/1, the Erlang shell process, controlling stdin, will remain the top level process of the Common Test system of processes. The result is that the Erlang shell is not available for interaction during the test run. If this is not desirable, maybe because the shell is needed for debugging purposes or for interaction with the SUT during test execution, you may set the release_shell start option to true (in the call to ct:run_test/1 or by using the corresponding test specification term, see below). This will make 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, and 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 that in order to use the ct:break/1/2 and ct:continue/0/1 functions, release_shell must be set to true.

For detailed documentation about ct:run_test/1, please see the ct manual page.

Test case group execution

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 is as follows (on the command line):

       [-case ]]]>

or (in the Erlang shell):

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

The group_names_or_paths parameter specifies either one or more group names and/or one or more group paths. At start up, Common Test will search for matching groups in the group definitions tree (i.e. the list returned from Suite:groups/0, please see the Test case groups chapter for details). Given a group name, say g, Common Test will search for all paths that lead to g. By path here we mean a sequence of nested groups, all of which have to be followed in order to get from the top level group to g. Actually, what Common Test needs to do in order to execute the test cases in group g, is to call the init_per_group/2 function for each group in the path to g, as well as all corresponding end_per_group/2 functions afterwards. The obvious reason for this is that 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, etc, all the way up to the top level group.

As you may have already realized, 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 will be performed. Common Test actually interprets a group specification that consists of a single name this way:

"Search and find all paths in the group definitions tree that lead to the specified group and, for each path, create a test which (1) executes all configuration functions in the path to the specified group, then (2) executes all - or all matching - test cases in this group, as well as (3) all - or all matching - test cases in all sub groups of the group".

It is also possible for the user to specify a specific group path with the group_names_or_paths parameter. With this type of specification it's possible to avoid execution of unwanted groups (in otherwise matching paths), and/or the execution of sub groups. The syntax of the group path is a list of group names in the path, e.g. on the command line:

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

or similarly in the Erlang shell (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 will be the terminating group in the test, i.e. no sub groups following this group will be executed. In the example above, g4 is the terminating group, hence Common Test will execute a test that calls all init configuration functions in the path to g4, i.e. g1..g3..g4. It will then call test cases tc1 and tc5 in g4 and finally all end configuration functions in order g4..g3..g1.

Note that the group path specification doesn't necessarily have to include all groups in the path to the terminating group. Common Test will search for all matching paths if given an incomplete group path.

Note also that it's possible to combine group names and group paths with the group_names_or_paths parameter. Each element is treated as an individual specification in combination with the cases parameter. See examples below.

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]}].
    


$ ct_run -suite "x_SUITE" -group all

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

Two tests will be executed, one for all cases and all sub groups under top1, and one for all under top2. (We would get the same result with -group top1 top2, or {group,[top1,top2]}.



$ ct_run -suite "x_SUITE" -group top1

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

This will execute one test for all cases and sub groups under top1.



$ ct_run -suite "x_SUITE" -group top1 -case tc12

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

This will run a test that executes tc12 in top1 and any sub group 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]}]).

This will execute tc12 only in group top1.



$ ct_run -suite "x_SUITE" -group top1 -case tc16

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

This will search top1 and all its sub groups for tc16 and the result will be that this test case executes in group sub121. (The specific path: -group [sub121] or {group,[[sub121]]}, would have given us the same result in this example).



$ ct_run -suite "x_SUITE" -group sub12 [sub12]

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

This will execute two tests, one that includes all cases and sub groups under sub12, and one with only the test cases in sub12.



$ ct_run -suite "x_SUITE" -group sub2X2

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

In this example, Common Test will find and execute two tests, one for the path from top2 to sub2X2 via sub21, and one from top2 to sub2X2 via sub22.



$ ct_run -suite "x_SUITE" -group [sub21,sub2X2]

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

Here, by specifying the unique path: top2 -> sub21 -> sub2X2, only one test is executed. The second possible path from top2 to sub2X2 (above) will be discarded.



$ ct_run -suite "x_SUITE" -group [sub22] -case tc22 tc21

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

In this example only the test cases for sub22 will be executed, and in reverse order compared to the group definition.



If a test case that belongs to a group (according to the group definition), is executed without a group specification, i.e. simply by means of (command line):

$ ct_run -suite "my_SUITE" -case my_tc

or (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, exactly as it has been presented in this section, can also be used in Test Specifications (with some extra features added). Please see below.

Running the interactive shell mode

You can start Common Test in an interactive shell mode where no automatic testing is performed. Instead, in this mode, 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 e.g. for debugging test suites, for analysing and debugging the SUT during "simulated" test case execution, and for trying out various operations during test suite development.

To invoke the interactive shell mode, you can 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 may start the Erlang shell and Common Test in the same go by using the -shell and, optionally, the -config and/or -userconfig flag. Examples:

ct_run -shell

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

If any functions using "required config data" (e.g. ct_telnet or ct_ftp functions) are to be called from the erlang shell, config data must first be required with ct:require/1/2. This is equivalent to a require statement in the Test Suite Info Function or in the Test Case Info 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, will in the interactive mode be written to a log named ctlog.html in the ]]> directory. A link to this file will be available in the file named last_interactive.html in the directory from which you executed ct_run. Currently, specifying a different root directory for the logs than the current working directory, is not supported.

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

Step by step execution of test cases with the Erlang Debugger

By means of ct_run -step [opts], or by passing the {step,Opts} option to ct:run_test/1, it is possible to get the Erlang Debugger started automatically and use its graphical interface to investigate the state of the current test case and to execute it step by step and/or set execution breakpoints.

If no extra options are given with the step flag/option, breakpoints will be set automatically on the test cases that are to be executed by Common Test, and those functions only. If the step option config is specified, breakpoints will also be initially set on the configuration functions in the suite, i.e. 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 will automatically pop up. (This is because each test case executes on a dedicated Erlang process). Whenever a new test case starts, Common Test will attempt to close the inactive trace window of the previous test case. However, if you prefer that Common Test leaves inactive trace windows, use the keep_inactive option.

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

Test Specifications
General description

The most flexible way to specify what to test, is to use a so called test specification. A test specification is a sequence of Erlang terms. The terms are normally declared in one or more text files (see ct:run_test/1), but may 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 possible to e.g. label the test run (similar to ct_run -label), evaluate arbitrary 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 that should 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), and more.

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

With test specification terms it is possible to state exactly which tests should 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.

An arbitrary number of test terms may be declared in sequence. Common Test will by default compile the terms into one or more tests to be performed in one resulting test run. Note that a term that specifies a set of test cases will "swallow" one that only specifies a subset of these cases. E.g. the result of merging one term that specifies that all cases in suite S should 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 behaviour, i.e. to instead perform each test sequentially in a "script-like" manner, the term merge_tests can be set 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.

Using multiple test specification files

When multiple test specification files are given at startup (either with ct_run -spec file1 file2 ... or ct:run_test([{spec, [File1,File2,...]}])), Common Test will either execute one test run per specification file, or join the files and perform all tests within one single test run. The first behaviour is the default one. The latter requires that the start flag/option join_suites is provided, e.g. run_test -spec ./my_tests1.ts ./my_tests2.ts -join_suites.

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

Test specification file inclusion

With the specs term (see syntax below), it's possible to have a test specification include other specifications. An included specification may either be joined with the source specification, or used to produce a separate test run (like with the join_specs start flag/option 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" will be joined with the terms in the source specification "a.spec" (if any). The inclusion of specifications "d.spec" and "e.spec" will result in two separate, and independent, test runs (i.e. one for each included specification).

Note that the join option does not imply that the test terms will be merged (see merge_tests above), 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 configuration that can not 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 the separate option, do not share configuration settings with the source specification. This is useful e.g. if there are clashing configuration settings in included specifications, making it impossible to join them.

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

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

The specs term may of course be used to nest specifications, i.e. have one specification include other specifications, which in turn include others, etc.

Test case groups

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

The test specification utilizes the same mechanism for specifying test case groups by means of names and paths, as explained in the Group Execution section above, with the addition of the GroupSpec element described next.

The GroupSpec element makes it possible to specify group execution properties that will override those in the group definition (i.e. in groups/0). Execution properties for sub-groups may be overridden as well. This feature makes it possible to change properties of groups at the time of execution, without even having to edit the test suite. The very same feature is available for group elements in the Suite:all/0 list. Therefore, more detailed documentation, and examples, can be found in the Test case groups chapter.

Test specification syntax

Below is the test specification syntax. 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 the init term are only relevant in the latter (see the Large Scale Testing chapter for information). For more information about the various terms, please see the corresponding sections in the User's Guide, such as e.g. the ct_run program for an overview of available start flags (since most flags have a corresponding configuration term), and more detailed explanation of e.g. Logging (for the verbosity, stylesheet and basic_html terms), External Configuration Data (for the config and userconfig terms), Event Handling (for the event_handler term), Common Test Hooks (for the ct_hooks term), etc.

Config 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},
	
	{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}.
	
        {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 | [GroupPath] | GroupSpec | [GroupSpec] | all
	GroupPath       = [GroupName]
	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, i.e. without directory paths. ConfigFiles must be full names, including paths. E.g, these 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"]}.

Any relative paths specified in the test specification, will be relative to the directory which contains the test specification file, if ct_run -spec TestSpecFile ... or ct:run:test([{spec,TestSpecFile},...]) executes the test. The path will be relative to the top level log directory, if ct:run:testspec(TestSpec) executes the test.

Constants

The define term introduces a constant, which is used to replace the name Constant with Value, wherever it's found in the test specification. This replacement happens during an initial iteration through the test specification. Constants may be used anywhere in the test specification, e.g. in arbitrary 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.

For the sake of readability, the name of the constant must always begin with an upper case letter, or a $, ?, or _. This also means that it must always be single quoted (obviously, since the constant name is actually 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. Compare these terms:

	    %% 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 has been deprecated but will remain supported in upcoming Common Test releases. Replacing alias terms with define is strongly recommended though! Here's an example of such a replacement:

	      %% 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}.

Actually, constants could well replace the node term too, but this still has declarative value, mainly when used in combination with NodeRefs == all_nodes (see types above).

Example

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 will be used for storing the HTML log files (in subdirectories tagged with node name, date and time). The variables in the specified test system config files will be imported for the test. The first test to run includes all suites for system t1. Excluded from the test are however the t1B and t1D suites. Also test cases test3 and test4 in t1A as well as the test1 case in t1C are excluded from the test. Secondly, the test for system t2 should run. The included suites are t2B and t2C. Included are also test cases test4, test1 and test7 in suite t2A. Note that the test cases will be executed in the specified order. Lastly, all suites for systems t3 are to be completely skipped and this should be explicitly noted in the log files.
The init term

With the init term it's possible to specify initialization options for nodes defined in the test specification. Currently, there are options to start the node and/or to evaluate any function on the node. See the Automatic startup of the test target nodes chapter for details.

User specific terms

It is possible for the user to provide a test specification that includes (for Common Test) unrecognizable terms. If this is desired, the -allow_user_terms flag should be used when starting tests with ct_run. This forces Common Test to ignore unrecognizable terms. Note that 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 means of the tuple: {allow_user_terms,true}

Running tests from the Web based GUI

The web based GUI, VTS, is started with the ct_run program. From the GUI you can load config files, and select directories, suites and cases to run. You can also state the config files, directories, suites and cases on the command line when starting the web based GUI.

ct_run -vts ]]> -suite -case ]]>

From the GUI you can run tests and view the result and the logs.

Note that ct_run -vts will try to open the Common Test start page in an existing web browser window or start the browser if it is not running. Which browser should be started may be specified with the browser start command option:

]]>

Example:

Note that the browser must run as a separate OS process or VTS will hang!

If no specific browser start command is specified, Firefox will be the default browser on Unix platforms and Internet Explorer on Windows. If Common Test fails to start a browser automatically, or 'none' is specified as the value for -browser (i.e. -browser none), start your favourite browser manually and type in the URL that Common Test displays in the shell.

Log files

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

Text to the operator's 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 gets 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, who may run hundreds or thousands of test cases, doesn't want to fill the console with details about, or printouts from, the specific test cases. By default, the operator will only see:

A confirmation that the test has started and information about how many test cases will be executed totally. A small note about each failed test case. A summary of all the run test cases. A confirmation that the test run is complete. Some special information like error reports and progress reports, printouts written with erlang:display/1, or io:format/3 specifically addressed to a receiver other than standard_io (e.g. the default group leader process 'user').

If/when the operator wants to dig deeper into the general results, or the result of a specific test case, he should do so by following the links in the HTML presentation and take a look in the major or minor log files. The "all_runs.html" page is a practical starting point usually. It's located in logdir and contains a link to each test run including a quick overview (date and time, node name, number of tests, test names and test result totals).

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

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

On the test run index page there is 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 very helpful when analyzing test results or debugging test suites.

On the test run index page it is noted if a test has missing suites (i.e. suites that Common Test has 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 etc. The information is available in both a file with textual and with HTML representation. The HTML file shows a summary which 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 one in a separate file. This way, it should be straightforward to compare the latest results to that of previous test runs, even if the set of test cases changes. If SASL is running, its logs will also be printed to the current minor log file by the cth_log_redirect built-in hook.

The full name of the minor log file (i.e. the name of the file including the absolute directory path) can be read during execution of the test case. It comes as value in the 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. Please see the Event Handling chapter for details.

Which information goes where is user configurable via the test server controller. Three threshold values determine what comes out on screen, and in the major or minor log files. See the OTP Test Server manual for information. The contents that goes to the HTML log file is fixed however and cannot be altered.

The log files are written continously during a test run and links are always created initially when a test starts. This makes it possible to follow test progress simply by refreshing pages in the HTML browser. Statistics totals are not presented until a test is complete however.

Log options

With the logopts start flag, it's possible to specify options that modify some aspects of the logging behaviour. Currently, the following options are available:

no_src no_nl

With no_src, the html version of the test suite source code will not be generated during the test run (and consequently not be available in the log file system).

With no_nl, Common Test will not add a newline character (\n) to the end of an output string that it receives from a call to e.g. 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_src

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

xxx

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

Sorting HTML table columns

By clicking the name in the column header of any table (e.g. "Ok", "Case", "Time", etc), the table rows are sorted in whatever order makes sense for the type of value (e.g. numerical for "Ok" or "Time", and alphabetical for "Case"). The sorting is performed by means of 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 Unexpected I/O Log

On the test suites overview page you find a link to the Unexpected I/O Log. In this log, Common Test saves printouts made with ct:log/2 and ct:pal/2, as well as captured system error- and progress reports, that cannot be associated with particular test cases and therefore cannot be written to individual test case log files. This happens e.g. 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 parallell test case group.

The Pre- and Post Test I/O Log

On the Common Test Framework Log page you find links to the so called Pre- and Post Test I/O Log. In this log, Common Test saves printouts made with ct:log/2 and ct:pal/2, as well as captured system error- and progress reports, that take place before - and after - the actual 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 due to 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, please see the Synchronizing section in the Common Test Hooks chapter.

Note that logging to file with ct:log/2 or ct:pal/2 only works when Common Test is running. Printouts with ct:pal/2 are however always displayed on screen.

HTML Style Sheets

Common Test uses an HTML Style Sheet (CSS file) to control the look of the HTML log files generated during test runs. If, for some reason, 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, as well as JavaScripts (see table sorting above).

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 and pal/3) accept Category as first argument. With this argument it's possible to specify a category that can be mapped to a selector in a CSS definition. This is useful especially for coloring text differently depending on the type of (or reason for) the printout. Say you want one 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 may look like this:

	  div.sys_config  { background:blue; color:white }
	  div.sys_state   { background:yellow; color:black }
	  div.error       { background:red; color:white }

To install the CSS file (Common Test inlines the definition in the HTML code), the name may 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 the -stylesheet flag are on a global test level in the sense that they can be used in any suite which is part of the test run.

It is also possible to install style sheets 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 can not be used by other suites. A suite private style sheet, if specified, will be used in favour of a global style sheet (one specified with the -stylesheet flag). A stylesheet tuple (as returned by suite/0 above) can also be returned from a test case info 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 favour of a suite or global level style sheet.

In a tuple {stylesheet,CSSFile}, if CSSFile is specified with a path, e.g. "$TEST/styles/categories.css", this full name will be used to locate the file. If only the file name is specified however, e.g. "categories.css", then the CSS file is assumed to be located in the data directory, data_dir, of the suite. The latter usage is recommended since it is portable compared to hard coding path names in the suite!

The Category argument in the example above may have the value (atom) sys_config (white on blue), sys_state (black on yellow) or error (white on red).

Repeating tests

You can order Common Test to repeat the tests you specify. You can choose to repeat tests a certain 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 means of time, it is also possible to specify what action Common Test should take upon timeout. Either Common Test performs all tests in the current run before stopping, or it stops as soon as the current test job is finished. Repetition can be activated by means of ct_run start flags, or tuples in the ct:run:test/1 option list argument. The flags (options in parenthesis) are:

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

The duration time, DurTime, is specified as HHMMSS. Example: -duration 012030 or {duration,"012030"}, means the tests will be executed and (if time allows) repeated, until timeout occurs after 1 h, 20 min and 30 secs. StopTime can be specified as HHMMSS and is then interpreted as a time today (or possibly tomorrow). StopTime can also be specified as YYMoMoDDHHMMSS. Example: -until 071001120000 or {until,"071001120000"}, which means the tests will be executed and (if time allows) repeated, until 12 o'clock on the 1st of Oct 2007.

When timeout occurs, Common Test will never abort the ongoing test case, since this might leave the system under test in an undefined, and possibly bad, state. Instead Common Test will by default finish the current test run before stopping. If the force_stop flag is given, Common Test will stop as soon as the current test job is finished, and if the force_stop flag is given with skip_rest Common Test will only complete the current test case and skip the rest of the tests in the test job. Note that since Common Test always finishes off at least the current test case, the time specified with duration or until is never definitive!

Log files from every single repeated test run is saved in normal Common Test fashion (see above). Common Test may 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.

Note that 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. There you can read the repetition number, remaining time, etc.

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, will be executed in one test run. A timeout event will occur after 10 minutes. As long as there is time left, Common Test will repeat the test run (i.e. starting over with the to1 test). When the timeout occurs, Common Test will stop as soon as the current job is finished (because of the force_stop flag). As a result, the specified test run might be aborted after the to1 test and before the to2 test.

Example 2:

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

Here the same test run as in Example 1, but with the force_stop flag set to skip_rest. If the timeout occurs while executing tests in directory to1, the rest of the test cases in to1 will be skipped and then the test will be aborted without running the tests in to2 another time. If the timeout occurs while executing tests in directory to2, then the rest of the test cases in to2 will be skipped and then the test will be 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 example above will be executed (and possibly repeated). In this example, however, the timeout will occur after 1 hour and when that happens, Common Test will finish the entire test run before stopping (i.e. the to1 and to2 test will always both be 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, will be repeated 5 times.

This feature should not be confused 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 the Writing Test Suites chapter.

Silent Connections

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

	ct_run -silent_connections [conn_types]
      

where conn_types specifies ssh, telnet, ftp, rpc and/or snmp.

Example:

	ct_run ... -silent_connections ssh telnet

switches off logging for ssh and telnet connections.

	ct_run ... -silent_connections

switches off logging for all connection types.

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

It is possible to also specify silent_connections in a test suite. This is accomplished by returning a tuple, {silent_connections,ConnTypes}, in the suite/0 or test case info list. If ConnTypes is a list of atoms (ssh, telnet, ftp, rpc and/or snmp), output for any corresponding connections will be suppressed. Full logging is per 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:

	
	-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 should be silent. The result is that my_testcase1 will get telnet info (if any) printed in the log, but not ssh info. my_testcase2 will get no info from either connection printed.

silent_connections may also be specified with a term in a test specification (see Test Specifications). Connections provided with the silent_connections start flag/option, will be merged with any connections listed in the test specification.

The silent_connections start flag/option and test specification term, overrides any settings made by the info functions inside the test suite.

Note that in the current Common Test version, the silent_connections feature only works for telnet and ssh connections! Support for other connection types will be added in future Common Test versions.