aboutsummaryrefslogblamecommitdiffstats
path: root/lib/common_test/doc/src/event_handler_chapter.xml
blob: a5508108504760b0093c27e9d39c5e2dd1a4f638 (plain) (tree)





















































































































































































































































































































                                                                                                        
<?xml version="1.0" encoding="latin1" ?>
<!DOCTYPE chapter SYSTEM "chapter.dtd">

<chapter>
  <header>
    <copyright>
      <year>2006</year><year>2009</year>
      <holder>Ericsson AB. All Rights Reserved.</holder>
    </copyright>
    <legalnotice>
      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.
    
    </legalnotice>

    <title>Event Handling</title>
    <prepared>Peter Andersson</prepared>
    <docno></docno>
    <date></date>
    <rev></rev>
    <file>event_handler_chapter.xml</file>
  </header>

  <section>
    <marker id="event_handling"></marker>
    <title>General</title>
    <p>It is possible for the operator of a Common Test system to receive
    event notifications continously during a test run. It is reported e.g. 
    when a test case starts and stops, what the current count of successful, 
    failed and skipped cases is, etc. This information can be used for 
    different purposes such as logging progress and results on 
    other format than HTML, saving statistics to a database for report
    generation and test system supervision.</p>

    <p>Common Test has a framework for event handling which is based on
    the OTP event manager concept and gen_event behaviour. When the Common Test
    server starts, it spawns an event manager. During test execution the
    manager gets a notification from the server every time something
    of potential interest happens. Any event handler plugged into the
    event manager can match on events of interest, take action, or maybe
    simply pass the information on. Event handlers are Erlang modules
    implemented by the Common Test user according to the gen_event 
    behaviour (see the OTP User's Guide and Reference Manual for more 
    information).</p>

    <p>As already described, a Common Test server always starts an event manager. 
    The server also plugs in a default event handler which has as its only
    purpose to relay notifications to a globally registered CT Master 
    event manager (if a CT Master server is running in the system). 
    The CT Master also spawns an event manager at startup.
    Event handlers plugged into this manager will receive the events from 
    all the test nodes as well as information from the CT Master server
    itself.</p>
  </section>
  <section>
    <title>Usage</title>
    <p>Event handlers may be installed by means of an <c>event_handler</c>
    start flag (<c>run_test</c>) or option (<c>ct:run_test/1</c>), where the
    argument specifies the names of one or more event handler modules.
    Example:</p>
    <p><c>$ run_test -suite test/my_SUITE -event_handler handlers/my_evh1 
    handlers/my_evh2 -pa $PWD/handlers</c></p>
    <p>All event handler modules must have gen_event behaviour. Note also that 
    these modules must be precompiled, and that their locations must be
    added explicitly to the Erlang code server search path (like in the
    example).</p>

    <p>It is not possible to specify start arguments to the event handlers when
    using the <c>run_test</c> script. You may however pass along start arguments
    if you use the <c>ct:run_test/1</c> function. An event_handler tuple in the argument 
    <c>Opts</c> has the following definition (see also <c>ct:run_test/1</c> in the 
    reference manual):</p>

    <pre>
    {event_handler,EventHandlers}

    EventHandlers = EH | [EH]
    EH = atom() | {atom(),InitArgs} | {[atom()],InitArgs}
    InitArgs = [term()] 
    </pre>

    <p>Example:</p>
    <pre>
    1> ct:run_test([{suite,"test/my_SUITE"},{event_handler,[my_evh1,{my_evh2,[node()]}]}]).
    </pre>
    <p>This will install two event handlers for the <c>my_SUITE</c> test. Event handler
    <c>my_evh1</c> is started with <c>[]</c> as argument to the init function. Event handler
    <c>my_evh2</c> is started with the name of the current node in the init argument list.</p>

    <p>Event handlers can also be plugged in by means of
    <seealso marker="run_test_chapter#test_specifications">test specification</seealso> 
    terms:</p>

    <p><c>{event_handler, EventHandlers}</c>, or</p>
    <p><c>{event_handler, EventHandlers, InitArgs}</c>, or</p>
    <p><c>{event_handler, NodeRefs, EventHandlers}</c>, or</p>
    <p><c>{event_handler, NodeRefs, EventHandlers, InitArgs}</c></p>
  
    <p><c>EventHandlers</c> is a list of module names. Before a test 
    session starts, the init function of each plugged in event handler 
    is called (with the InitArgs list as argument or [] if
    no start arguments are given).</p>

    <p>To plug a handler into the CT Master event manager, specify 
    <c>master</c> as the node in <c>NodeRefs</c>.</p>

    <p>For an event handler to be able to match on events, the module must
    include the header file <c>ct_event.hrl</c>. An event is a record with the
    following definition:</p>

    <p><c>#event{name, node, data}</c></p>

    <p><c>name</c> is the label (type) of the event. <c>node</c> is the name of the
    node the event has originated from (only relevant for CT Master event handlers).
    <c>data</c> is specific for the particular event.</p>

    <p><em>General events:</em></p>

    <list>
      <item><c>#event{name = start_logging, data = LogDir}</c>
        <p><c>LogDir = string()</c>, top level log directory for the test run.</p>
        <p>Indicates that the logging process of Common Test
	has started successfully and is ready to receive IO
	messages.</p></item>

      <item><c>#event{name = stop_logging, data = []}</c>
        <p>Indicates that the logging process of Common Test
	has been shut down at the end of the test run.
	</p></item>

      <item><c>#event{name = test_start, data = {StartTime,LogDir}}</c>
        <p><c>StartTime = {date(),time()}</c>, test run start date and time.</p>
        <p><c>LogDir = string()</c>, top level log directory for the test run.</p>
        <p>This event indicates that Common Test has finished initial preparations
	and will begin executing test cases.
	</p></item>

      <item><c>#event{name = test_done, data = EndTime}</c>
        <p><c>EndTime = {date(),time()}</c>, date and time the test run finished.</p>
        <p>This indicates that the last test case has been executed and 
	Common Test is shutting down.	
	</p></item>

      <item><c>#event{name = start_info, data = {Tests,Suites,Cases}}</c>
        <p><c>Tests = integer()</c>, the number of tests.</p>
        <p><c>Suites = integer()</c>, the total number of suites.</p>
        <p><c>Cases = integer() | unknown</c>, the total number of test cases.</p>
        <p>Initial test run information that can be interpreted as: "This test
	run will execute <c>Tests</c> separate tests, in total containing
	<c>Cases</c> number of test cases, in <c>Suites</c> number of suites".
	Note that if a test case group with a repeat property exists in any test, 
	the total number of test cases can not be calculated (unknown).
	</p></item>

      <item><c>#event{name = tc_start, data = {Suite,FuncOrGroup}}</c>
        <p><c>Suite = atom()</c>, name of the test suite.</p>
        <p><c>FuncOrGroup = Func | {Conf,GroupName,GroupProperties}</c></p>
        <p><c>Func = atom()</c>, name of test case or configuration function.</p>
        <p><c>Conf = init_per_group | end_per_group</c>, group configuration function.</p>
        <p><c>GroupName = atom()</c>, name of the group.</p>
        <p><c>GroupProperties = list()</c>, list of execution properties for the group.</p>
        <p>This event informs about the start of a test case, or a group configuration
	function. The event is sent also for <c>init_per_suite</c> and <c>end_per_suite</c>,
	but not for <c>init_per_testcase</c> and <c>end_per_testcase</c>. If a group
	configuration function is starting, the group name and execution properties
	are also given.
	</p></item>

      <item><c>#event{name = tc_done, data = {Suite,FuncOrGroup,Result}}</c>
        <p><c>Suite = atom()</c>, name of the suite.</p>
        <p><c>FuncOrGroup = Func | {Conf,GroupName,GroupProperties}</c></p>
        <p><c>Func = atom()</c>, name of test case or configuration function.</p>
        <p><c>Conf = init_per_group | end_per_group</c>, group configuration function.</p>
        <p><c>GroupName = unknown | atom()</c>, name of the group 
	      (unknown if init- or end function times out).</p>
        <p><c>GroupProperties = list()</c>, list of execution properties for the group.</p>
        <p><c>Result = ok | {skipped,SkipReason} | {failed,FailReason}</c>, the result.</p>
        <p><c>SkipReason = {require_failed,RequireInfo} | 
	                   {require_failed_in_suite0,RequireInfo} | 
	                   {failed,{Suite,init_per_testcase,FailInfo}} | 
                           UserTerm</c>, 
	      the reason why the case has been skipped.</p>
	<p><c>FailReason = {error,FailInfo} | 
	                   {error,{RunTimeError,StackTrace}} | 
                           {timetrap_timeout,integer()} | 
			   {failed,{Suite,end_per_testcase,FailInfo}}</c>, reason for failure.</p>	
        <p><c>RequireInfo = {not_available,atom()}</c>, why require has failed.</p>
        <p><c>FailInfo = {timetrap_timeout,integer()} | 
	                 {RunTimeError,StackTrace} | 
			 UserTerm</c>, 
	      detailed information about an error.</p>
        <p><c>RunTimeError = term()</c>, a run-time error, e.g. badmatch, undef, etc.</p>
        <p><c>StackTrace = list()</c>, list of function calls preceeding a run-time error.</p>
        <p><c>UserTerm = term()</c>, arbitrary data specified by user, or <c>exit/1</c> info.</p>
        <p>This event informs about the end of a test case or a configuration function (see the 
	<c>tc_start</c> event for details on the FuncOrGroup element). With this event comes the 
	final result of the function in question. It is possible to determine on the top level
	of <c>Result</c> if the function was successful, skipped (by the user), or if it failed.
	It is of course possible to dig deeper and also perform pattern matching on the various
	reasons for skipped or failed. Note that <c>{'EXIT',Reason}</c> tuples have been translated into
	<c>{error,Reason}</c>. Note also that if a <c>{failed,{Suite,end_per_testcase,FailInfo}</c>
	result is received, it actually means the test case was successful, but that 
	<c>end_per_testcase</c> for the case failed.
	</p></item>

      <item><c>#event{name = tc_auto_skip, data = {Suite,Func,Reason}}</c>
        <p><c>Suite = atom()</c>, the name of the suite.</p>
        <p><c>Func = atom()</c>, the name of the test case or configuration function.</p>
        <p><c>Reason = {failed,FailReason} |
	               {require_failed_in_suite0,RequireInfo}</c>, 
	      reason for auto skipping <c>Func</c>.</p>
        <p><c>FailReason = {Suite,ConfigFunc,FailInfo}} | 
	                   {Suite,FailedCaseInSequence}</c>, reason for failure.</p>	
        <p><c>RequireInfo = {not_available,atom()}</c>, why require has failed.</p>
        <p><c>ConfigFunc = init_per_suite | init_per_group</c></p>
        <p><c>FailInfo = {timetrap_timeout,integer()} | 
	                 {RunTimeError,StackTrace} |
			 bad_return | UserTerm</c>, 
	      detailed information about an error.</p>
	<p><c>FailedCaseInSequence = atom()</c>, name of a case that has failed in a sequence.</p>
        <p><c>RunTimeError = term()</c>, a run-time error, e.g. badmatch, undef, etc.</p>
        <p><c>StackTrace = list()</c>, list of function calls preceeding a run-time error.</p>
        <p><c>UserTerm = term()</c>, arbitrary data specified by user, or <c>exit/1</c> info.</p>
        <p>This event gets sent for every test case or configuration function that Common Test
	has skipped automatically because of either a failed <c>init_per_suite</c> or 
	<c>init_per_group</c>, a failed <c>require</c> in <c>suite/0</c>, or a failed test case
	in a sequence. Note that this event is never received as a result of a test case getting
	skipped because of <c>init_per_testcase</c> failing, since that information is carried with
	the <c>tc_done</c> event.
	</p></item>

      <item><c>#event{name = tc_user_skip, data = {Suite,TestCase,Comment}}</c>
        <p><c>Suite = atom()</c>, name of the suite.</p>
        <p><c>TestCase = atom()</c>, name of the test case.</p>
        <p><c>Comment = string()</c>, reason for skipping the test case.</p>
        <p>This event specifies that a test case has been skipped by the user. 
	It is only ever received if the skip was declared in a test specification. 
	Otherwise, user skip information is received as a <c>{skipped,SkipReason}</c> 
	result in the <c>tc_done</c> event for the test case.
	</p></item>

      <item><c>#event{name = test_stats, data = {Ok,Failed,Skipped}}</c>
        <p><c>Ok = integer()</c>, the current number of successful test cases.</p>
        <p><c>Failed = integer()</c>, the current number of failed test cases.</p>
        <p><c>Skipped = {UserSkipped,AutoSkipped}</c></p>
        <p><c>UserSkipped = integer()</c>, the current number of user skipped test cases.</p>
        <p><c>AutoSkipped = integer()</c>, the current number of auto skipped test cases.</p>
        <p>This is a statistics event with the current count of successful, skipped 
	and failed test cases so far. This event gets sent after the end of each test case,
	immediately following the <c>tc_done</c> event.
	</p></item>
    </list>

     <p><em>Internal events:</em></p>

     <list>
      <item><c>#event{name = start_make, data = Dir}</c>
        <p><c>Dir = string()</c>, running make in this directory.</p>
        <p>An internal event saying that Common Test will start compiling
	modules in directory <c>Dir</c>.
	</p></item>

     <item><c>#event{name = finished_make, data = Dir}</c>
        <p><c>Dir = string()</c>, finished running make in this directory.</p>
        <p>An internal event saying that Common Test is finished compiling
	modules in directory <c>Dir</c>.
	</p></item>

      <item><c>#event{name = start_write_file, data = FullNameFile}</c>
        <p><c>FullNameFile = string(), full name of the file.</c></p>
        <p>An internal event used by the Common Test Master process to 
	synchronize particular file operations.
	</p></item>

      <item><c>#event{name = finished_write_file, data = FullNameFile}</c>
        <p><c>FullNameFile = string(), full name of the file.</c></p>
        <p>An internal event used by the Common Test Master process to 
	synchronize particular file operations.
	</p></item>

    </list>

    <p>The events are also documented in <c>ct_event.erl</c>. This module
    may serve as an example of what an event handler for the CT event 
    manager can look like.</p>

    <note><p>To ensure that printouts to standard out (or printouts made with
    <c>ct:log/2/3</c> or <c>ct:pal/2/3</c>) get written to the test case log
    file, and not to the Common Test framework log, you can syncronize
    with the Common Test server by matching on the <c>tc_start</c> and <c>tc_done</c>
    events. In the period between these events, all IO gets directed to the
    test case log file. These events are sent synchronously to avoid potential
    timing problems (e.g. that the test case log file gets closed just before
    an IO message from an external process gets through). Knowing this, you
    need to be careful that your <c>handle_event/2</c> callback function doesn't
    stall the test execution, possibly causing unexpected behaviour as a result.</p></note>
</section>
</chapter>