documentation/testing.sgml

   1   <chapter id="testing">
   2     <title>Writing Conformance tests</title>
   3
   4     <sect1 id="testing-intro">
   5       <title>Introduction</title>
   6       <para>
   7         With more The Windows API follows no standard, it is itself a de facto
   8         standard, and deviations from that standard, even small ones, often
   9         cause applications to crash or misbehave in some way. Furthermore
  10         a conformance test suite is the most accurate (if not necessarily
  11         the most complete) form of API documentation and can be used to
  12         supplement the Windows API documentation.
  13       </para>
  14       <para>
  15         Writing a conformance test suite for more than 10000 APIs is no small
  16         undertaking. Fortunately it can prove very useful to the development
  17         of Wine way before it is complete.
  18         <itemizedlist>
  19           <listitem>
  20             <para>
  21               The conformance test suite must run on Windows. This is
  22               necessary to provide a reasonable way to verify its accuracy.
  23               Furthermore the tests must pass successfully on all Windows
  24               platforms (tests not relevant to a given platform should be
  25               skipped).
  26             </para>
  27             <para>
  28               A consequence of this is that the test suite will provide a
  29               great way to detect variations in the API between different
  30               Windows versions. For instance, this can provide insights
  31               into the differences between the, often undocumented, Win9x and
  32               NT Windows families.
  33             </para>
  34             <para>
  35               However, one must remember that the goal of Wine is to run
  36               Windows applications on Linux, not to be a clone of any specific
  37               Windows version. So such variations must only be tested for when
  38               relevant to that goal.
  39             </para>
  40           </listitem>
  41           <listitem>
  42             <para>
  43               Writing conformance tests is also an easy way to discover
  44               bugs in Wine. Of course, before fixing the bugs discovered in
  45               this way, one must first make sure that the new tests do pass
  46               successfully on at least one Windows 9x and one Windows NT
  47               version.
  48             </para>
  49             <para>
  50               Bugs discovered this way should also be easier to fix. Unlike
  51               some mysterious application crashes, when a conformance test
  52               fails, the expected behavior and APIs tested for are known thus
  53               greatly simplifying the diagnosis.
  54             </para>
  55           </listitem>
  56           <listitem>
  57             <para>
  58               To detect regressions. Simply running the test suite regularly
  59               in Wine turns it into a great tool to detect regressions.
  60               When a test fails, one immediately knows what was the expected
  61               behavior and which APIs are involved. Thus regressions caught
  62               this way should be detected earlier, because it is easy to run
  63               all tests on a regular basis, and easier to fix because of the
  64               reduced diagnosis work.
  65             </para>
  66           </listitem>
  67           <listitem>
  68             <para>
  69               Tests written in advance of the Wine development (possibly even
  70               by non Wine developers) can also simplify the work of the
  71               future implementer by making it easier for him to check the
  72               correctness of his code.
  73             </para>
  74           </listitem>
  75           <listitem>
  76             <para>
  77               Conformance tests will also come in handy when testing Wine on
  78               new (or not as widely used) architectures such as FreeBSD,
  79               Solaris x86 or even non-x86 systems. Even when the port does
  80               not involve any significant change in the thread management,
  81               exception handling or other low-level aspects of Wine, new
  82               architectures can expose subtle bugs that can be hard to
  83               diagnose when debugging regular (complex) applications.
  84             </para>
  85           </listitem>
  86         </itemizedlist>
  87       </para>
  88     </sect1>
  89
  90
  91     <sect1 id="testing-what">
  92       <title>What to test for?</title>
  93       <para>
  94         The first thing to test for is the documented behavior of APIs
  95         and such as CreateFile. For instance one can create a file using a
  96         long pathname, check that the behavior is correct when the file
  97         already exists, try to open the file using the corresponding short
  98         pathname, convert the filename to Unicode and try to open it using
  99         CreateFileW, and all other things which are documented and that
 100         applications rely on.
 101       </para>
 102       <para>
 103         While the testing framework is not specifically geared towards this
 104         type of tests, it is also possible to test the behavior of Windows
 105         messages. To do so, create a window, preferably a hidden one so that
 106         it does not steal the focus when running the tests, and send messages
 107         to that window or to controls in that window. Then, in the message
 108         procedure, check that you receive the expected messages and with the
 109         correct parameters.
 110       </para>
 111       <para>
 112         For instance you could create an edit control and use WM_SETTEXT to
 113         set its contents, possibly check length restrictions, and verify the
 114         results using WM_GETTEXT. Similarly one could create a listbox and
 115         check the effect of LB_DELETESTRING on the list's number of items,
 116         selected items list, highlighted item, etc.
 117       </para>
 118       <para>
 119         However, undocumented behavior should not be tested for unless there
 120         is an application that relies on this behavior, and in that case the
 121         test should mention that application, or unless one can strongly
 122         expect applications to rely on this behavior, typically APIs that
 123         return the required buffer size when the buffer pointer is NULL.
 124       </para>
 125     </sect1>
 126
 127
 128     <sect1 id="testing-wine">
 129       <title>Running the tests in Wine</title>
 130       <para>
 131         The simplest way to run the tests in Wine is to type 'make test' in
 132         the Wine sources top level directory. This will run all the Wine
 133         conformance tests.
 134       </para>
 135       <para>
 136         The tests for a specific Wine library are located in a 'tests'
 137         directory in that library's directory. Each test is contained in a
 138         file (e.g. <filename>dlls/kernel/tests/thread.c</>). Each
 139         file itself contains many checks concerning one or more related APIs.
 140       </para>
 141       <para>
 142         So to run all the tests related to a given Wine library, go to the
 143         corresponding 'tests' directory and type 'make test'. This will
 144         compile the tests, run them, and create an '<replaceable>xxx</>.ok'
 145         file for each test that passes successfully. And if you only want to
 146         run the tests contained in the <filename>thread.c</> file of the
 147         kernel library, you would do:
 148 <screen>
 149 <prompt>$ </>cd dlls/kernel/tests
 150 <prompt>$ </>make thread.ok
 151 </screen>
 152       </para>
 153       <para>
 154         Note that if the test has already been run and is up to date (i.e. if
 155         neither the kernel library nor the <filename>thread.c</> file has
 156         changed since the <filename>thread.ok</> file was created), then make
 157         will say so. To force the test to be re-run, delete the
 158         <filename>thread.ok</> file, and run the make command again.
 159       </para>
 160       <para>
 161         You can also run tests manually using a command similar to the
 162         following:
 163 <screen>
 164 <prompt>$ </>../../../tools/runtest -q -M kernel32.dll -p kernel32_test.exe.so thread.c
 165 <prompt>$ </>../../../tools/runtest -p kernel32_test.exe.so thread.c
 166 thread.c: 86 tests executed, 5 marked as todo, 0 failures.
 167 </screen>
 168         The '-P wine' options defines the platform that is currently being
 169         tested. Remove the '-q' option if you want the testing framework
 170         to report statistics about the number of successful and failed tests.
 171         Run <command>runtest -h</> for more details.
 172       </para>
 173     </sect1>
 174
 175
 176     <sect1 id="cross-compiling-tests">
 177       <title>Cross-compiling the tests with MinGW</title>
 178       <sect2>
 179         <title>Setup of the MinGW cross-compiling environment</title>
 180         <para>
 181           Here are some instructions to setup MinGW on different Linux
 182           distributions and *BSD.
 183         </para>
 184         <sect3>
 185           <title>Debian GNU/Linux</title>
 186           <para>
 187             On Debian all you need to do is type <command>apt-get install
 188             mingw32</>.
 189           </para>
 190         </sect3>
 191         <sect3>
 192           <title>Red Hat Linux like rpm systems</title>
 193           <para>
 194             This includes Fedora Core, Red Hat Enterprise Linux, Mandrake,
 195             most probably SuSE Linux too, etc. But this list isn't exhaustive;
 196             the following steps should probably work on any rpm based system.
 197           </para>
 198           <para>
 199             Download and install the latest rpm's from
 200             <ulink url="http://mirzam.it.vu.nl/mingw/">MinGW RPM packages</>.
 201             Alternatively you can follow the instructions on that page and
 202             build your own packages from the source rpm's listed there as well.
 203           </para>
 204         </sect3>
 205         <sect3>
 206           <title>*BSD</title>
 207           <para>
 208             The *BSD systems have in their ports collection a port for the
 209             MinGW cross-compiling environment. Please see the documentation
 210             of your system about how to build and install a port.
 211           </para>
 212         </sect3>
 213       </sect2>
 214       <sect2>
 215       <title>Compiling the tests</title>
 216       <para>
 217         Having the cross-compiling environment set up the generation of the
 218         Windows executables is easy by using the Wine build system.
 219       </para>
 220       <para>
 221         If you had already run <command>configure</>, then delete
 222         <filename>config.cache</> and re-run <command>configure</>.
 223         You can then run <command>make crosstest</>. To sum up:
 224 <screen>
 225 <prompt>$ </><userinput>rm config.cache</>
 226 <prompt>$ </><userinput>./configure</>
 227 <prompt>$ </><userinput>make crosstest</>
 228 </screen>
 229       </para>
 230       </sect2>
 231     </sect1>
 232
 233
 234     <sect1 id="testing-windows">
 235       <title>Building and running the tests on Windows</title>
 236       <sect2>
 237         <title>Using pre-compiled binaries</title>
 238         <para>
 239           Unfortunately there are no pre-compiled binaries yet. However if
 240           send an email to the Wine development list you can probably get
 241           someone to send them to you, and maybe motivate some kind soul to
 242           put in place a mechanism for publishing such binaries on a regular
 243           basis.
 244         </para>
 245       </sect2>
 246       <sect2>
 247         <title>With Visual C++</title>
 248         <screen>
 249         Visual Studio 6 users:
 250         - MSVC headers may not work well, try with Wine headers
 251         - Ensure that you have the "processor pack" from
 252           <ulink url="http://msdn.microsoft.com/vstudio/downloads/tools/ppack/default.aspx">http://msdn.microsoft.com/vstudio/downloads/tools/ppack/default.aspx</>
 253           as well as the latest service packs.  The processor pack fixes <emphasis>"error C2520: conversion from unsigned
 254            __int64 to double not implemented, use signed __int64"</>
 255         </screen>
 256         <itemizedlist>
 257           <listitem><para>
 258             get the Wine sources
 259           </para></listitem>
 260           <listitem><para>
 261             Run msvcmaker to generate Visual C++ project files for the tests.
 262             'msvcmaker' is a perl script so you may be able to run it on
 263             Windows.
 264 <screen>
 265 <prompt>$ </>./tools/winapi/msvcmaker --no-wine
 266 </screen>
 267           </para></listitem>
 268           <listitem><para>
 269             If the previous steps were done on your Linux development
 270             machine, make the Wine sources accessible to the Windows machine
 271             on which you are going to compile them. Typically you would do
 272             this using Samba but copying them altogether would work too.
 273           </para></listitem>
 274           <listitem><para>
 275             On the Windows machine, open the <filename>winetest.dsw</>
 276             workspace. This will load each test's project. For each test there
 277             are two configurations: one compiles the test with the Wine
 278             headers, and the other uses the Visual C++ headers. Some tests
 279             will compile fine with the former, but most will require the
 280             latter.
 281           </para></listitem>
 282           <listitem><para>
 283             Open the <menuchoice><guimenu>Build</> <guimenu>Batch
 284             build...</></> menu and select the tests and build configurations
 285             you want to build. Then click on <guibutton>Build</>.
 286           </para></listitem>
 287           <listitem><para>
 288             To run a specific test from Visual C++, go to
 289             <menuchoice><guimenu>Project</> <guimenu>Settings...</></>. There
 290             select that test's project and build configuration and go to the
 291             <guilabel>Debug</> tab. There type the name of the specific test
 292             to run (e.g. 'thread') in the <guilabel>Program arguments</>
 293             field. Validate your change by clicking on <guibutton>Ok</> and
 294             start the test by clicking the red exclamation mark (or hitting
 295             'F5' or any other usual method).
 296           </para></listitem>
 297           <listitem><para>
 298             You can also run the tests from the command line. You will find
 299             them in either <filename>Output\Win32_Wine_Headers</> or
 300             <filename>Output\Win32_MSVC_Headers</> depending on the build
 301             method. So to run the kernel 'path' tests you would do:
 302 <screen>
 303 <prompt>C:\&gt;</>cd dlls\kernel\tests\Output\Win32_MSVC_Headers
 304 <prompt>C:\dlls\kernel\tests\Output\Win32_MSVC_Headers&gt;</>kernel32_test path
 305 </screen>
 306           </para></listitem>
 307         </itemizedlist>
 308       </sect2>
 309       <sect2>
 310         <title>With MinGW</title>
 311         <para>
 312          Wine's build system already has support for building tests with a MinGW
 313          cross-compiler. See the section above called 'Setup of the MinGW
 314          cross-compiling environment' for instructions on how to set things up.
 315          When you have a MinGW environment installed all you need to do is rerun
 316          configure and it should detect the MinGW compiler and tools. Then run
 317          'make crosstest' to start building the tests.
 318         </para>
 319       </sect2>
 320     </sect1>
 321
 322
 323     <sect1 id="testing-test">
 324       <title>Inside a test</title>
 325
 326       <para>
 327         When writing new checks you can either modify an existing test file or
 328         add a new one. If your tests are related to the tests performed by an
 329         existing file, then add them to that file. Otherwise create a new .c
 330         file in the tests directory and add that file to the
 331         <varname>CTESTS</> variable in <filename>Makefile.in</>.
 332       </para>
 333       <para>
 334         A new test file will look something like the following:
 335 <screen>
 336 #include &lt;wine/test.h&gt;
 337 #include &lt;winbase.h&gt;
 338
 339 /* Maybe auxiliary functions and definitions here */
 340
 341 START_TEST(paths)
 342 {
 343    /* Write your checks there or put them in functions you will call from
 344     * there
 345     */
 346 }
 347 </screen>
 348       </para>
 349       <para>
 350         The test's entry point is the START_TEST section. This is where
 351         execution will start. You can put all your tests in that section but
 352         it may be better to split related checks in functions you will call
 353         from the START_TEST section. The parameter to START_TEST must match
 354         the name of the C file. So in the above example the C file would be
 355         called <filename>paths.c</>.
 356       </para>
 357       <para>
 358         Tests should start by including the <filename>wine/test.h</> header.
 359         This header will provide you access to all the testing framework
 360         functions. You can then include the windows header you need, but make
 361         sure to not include any Unix or Wine specific header: tests must
 362         compile on Windows.
 363       </para>
 364       <para>
 365         You can use <function>trace</> to print informational messages. Note
 366         that these messages will only be printed if 'runtest -v' is being used.
 367 <screen>
 368   trace("testing GlobalAddAtomA");
 369   trace("foo=%d",foo);
 370 </screen>
 371       </para>
 372       <para>
 373         Then just call functions and use <function>ok</> to make sure that
 374         they behaved as expected:
 375 <screen>
 376   ATOM atom = GlobalAddAtomA( "foobar" );
 377   ok( GlobalFindAtomA( "foobar" ) == atom, "could not find atom foobar" );
 378   ok( GlobalFindAtomA( "FOOBAR" ) == atom, "could not find atom FOOBAR" );
 379 </screen>
 380         The first parameter of <function>ok</> is an expression which must
 381         evaluate to true if the test was successful. The next parameter is a
 382         printf-compatible format string which is displayed in case the test
 383         failed, and the following optional parameters depend on the format
 384         string.
 385       </para>
 386     </sect1>
 387
 388     <sect1 id="testing-error-messages">
 389       <title>Writing good error messages</title>
 390       <para>
 391         The message that is printed when a test fails is
 392         <emphasis>extremely</> important.
 393       </para>
 394       <para>
 395         Someone will take your test, run it on a Windows platform that
 396         you don't have access to, and discover that it fails. They will then
 397         post an email with the output of the test, and in particular your
 398         error message. Someone, maybe you, will then have to figure out from
 399         this error message why the test failed.
 400       </para>
 401       <para>
 402         If the error message contains all the relevant information that will
 403         be easy. If not, then it will require modifying the test, finding
 404         someone to compile it on Windows, sending the modified version to the
 405         original tester and waiting for his reply. In other words, it will
 406         be long and painful.
 407       </para>
 408       <para>
 409         So how do you write a good error message? Let's start with an example
 410         of a bad error message:
 411 <screen>
 412     ok(GetThreadPriorityBoost(curthread,&amp;disabled)!=0,
 413        "GetThreadPriorityBoost Failed");
 414 </screen>
 415         This will yield:
 416 <screen>
 417 thread.c:123: Test failed: GetThreadPriorityBoost Failed
 418 </screen>
 419       </para>
 420       <para>
 421         Did you notice how the error message provides no information about
 422         why the test failed? We already know from the line number exactly
 423         which test failed. In fact the error message gives strictly no
 424         information that cannot already be obtained by reading the code. In
 425         other words it provides no more information than an empty string!
 426       </para>
 427       <para>
 428         Let's look at how to rewrite it:
 429 <screen>
 430     BOOL rc;
 431 ...
 432     rc=GetThreadPriorityBoost(curthread,&amp;disabled);
 433     ok(rc!=0 && disabled==0,"rc=%d error=%ld disabled=%d",
 434        rc,GetLastError(),disabled);
 435 </screen>
 436         This will yield:
 437 <screen>
 438 thread.c:123: Test failed: rc=0 error=120 disabled=0
 439 </screen>
 440       </para>
 441       <para>
 442         When receiving such a message, one would check the source, see that
 443         it's a call to GetThreadPriorityBoost, that the test failed not
 444         because the API returned the wrong value, but because it returned an
 445         error code. Furthermore we see that GetLastError() returned 120 which
 446         winerror.h defines as ERROR_CALL_NOT_IMPLEMENTED. So the source of
 447         the problem is obvious: this Windows platform (here Windows 98) does
 448         not support this API and thus the test must be modified to detect
 449         such a condition and skip the test.
 450       </para>
 451       <para>
 452         So a good error message should provide all the information which
 453         cannot be obtained by reading the source, typically the function
 454         return value, error codes, and any function output parameter. Even if
 455         more information is needed to fully understand a problem,
 456         systematically providing the above is easy and will help cut down the
 457         number of iterations required to get to a resolution.
 458       </para>
 459       <para>
 460         It may also be a good idea to dump items that may be hard to retrieve
 461         from the source, like the expected value in a test if it is the
 462         result of an earlier computation, or comes from a large array of test
 463         values (e.g. index 112 of _pTestStrA in vartest.c). In that respect,
 464         for some tests you may want to define a macro such as the following:
 465 <screen>
 466 #define eq(received, expected, label, type) \
 467         ok((received) == (expected), "%s: got " type " instead of " type, (label),(received),(expected))
 468
 469 ...
 470
 471     eq( b, curr_val, "SPI_{GET,SET}BEEP", "%d" );
 472 </screen>
 473        </para>
 474     </sect1>
 475
 476
 477     <sect1 id="testing-platforms">
 478       <title>Handling platform issues</title>
 479       <para>
 480         Some checks may be written before they pass successfully in Wine.
 481         Without some mechanism, such checks would potentially generate
 482         hundred of known failures for months each time the tests are being run.
 483         This would make it hard to detect new failures caused by a regression.
 484         or to detect that a patch fixed a long standing issue.
 485       </para>
 486       <para>
 487         Thus the Wine testing framework has the concept of platforms and
 488         groups of checks can be declared as expected to fail on some of them.
 489         In the most common case, one would declare a group of tests as
 490         expected to fail in Wine. To do so, use the following construct:
 491 <screen>
 492 todo_wine {
 493     SetLastError( 0xdeadbeef );
 494     ok( GlobalAddAtomA(0) == 0 && GetLastError() == 0xdeadbeef, "failed to add atom 0" );
 495 }
 496 </screen>
 497         On Windows the above check would be performed normally, but on Wine it
 498         would be expected to fail, and not cause the failure of the whole
 499         test. However. If that check were to succeed in Wine, it would
 500         cause the test to fail, thus making it easy to detect when something
 501         has changed that fixes a bug. Also note that todo checks are accounted
 502         separately from regular checks so that the testing statistics remain
 503         meaningful. Finally, note that todo sections can be nested so that if
 504         a test only fails on the cygwin and reactos platforms, one would
 505         write:
 506 <screen>
 507 todo("cygwin") {
 508     todo("reactos") {
 509         ...
 510     }
 511 }
 512 </screen>
 513         <!-- FIXME: Would we really have platforms such as reactos, cygwin, freebsd & co? -->
 514         But specific platforms should not be nested inside a todo_wine section
 515         since that would be redundant.
 516       </para>
 517       <para>
 518         When writing tests you will also encounter differences between Windows
 519         9x and Windows NT platforms. Such differences should be treated
 520         differently from the platform issues mentioned above. In particular
 521         you should remember that the goal of Wine is not to be a clone of any
 522         specific Windows version but to run Windows applications on Unix.
 523       </para>
 524       <para>
 525         So, if an API returns a different error code on Windows 9x and
 526         Windows NT, your check should just verify that Wine returns one or
 527         the other:
 528 <screen>
 529 ok ( GetLastError() == WIN9X_ERROR || GetLastError() == NT_ERROR, ...);
 530 </screen>
 531       </para>
 532       <para>
 533         If an API is only present on some Windows platforms, then use
 534         LoadLibrary and GetProcAddress to check if it is implemented and
 535         invoke it. Remember, tests must run on all Windows platforms.
 536         Similarly, conformance tests should nor try to correlate the Windows
 537         version returned by GetVersion with whether given APIs are
 538         implemented or not. Again, the goal of Wine is to run Windows
 539         applications (which do not do such checks), and not be a clone of a
 540         specific Windows version.
 541       </para>
 542       <!--para>
 543         FIXME: What about checks that cause the process to crash due to a bug?
 544       </para-->
 545     </sect1>
 546
 547
 548 <!-- FIXME: Strategies for testing threads, testing network stuff,
 549  file handling, eq macro... -->
 550
 551   </chapter>
 552
 553 <!-- Keep this comment at the end of the file
 554 Local variables:
 555 mode: sgml
 556 sgml-parent-document:("wine-devel.sgml" "set" "book" "part" "chapter" "")
 557 End:
 558 -->