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         The Windows API follows no standard, it is itself a de facto standard,
   8         and deviations from that standard, even small ones, often cause
   9         applications to crash or misbehave in some way.
  10       </para>
  11       <para>
  12         The question becomes, "How do we ensure compliance with that standard?"
  13         The answer is, "By using the API documentation available to us and
  14         backing that up with conformance tests." Furthermore, a conformance
  15         test suite is the most accurate (if not necessarily the most complete)
  16         form of API documentation and can be used to supplement the Windows
  17         API documentation.
  18       </para>
  19       <para>
  20         Writing a conformance test suite for more than 10000 APIs is no small
  21         undertaking. Fortunately it can prove very useful to the development
  22         of Wine way before it is complete.
  23         <itemizedlist>
  24           <listitem>
  25             <para>
  26               The conformance test suite must run on Windows. This is
  27               necessary to provide a reasonable way to verify its accuracy.
  28               Furthermore the tests must pass successfully on all Windows
  29               platforms (tests not relevant to a given platform should be
  30               skipped).
  31             </para>
  32             <para>
  33               A consequence of this is that the test suite will provide a
  34               great way to detect variations in the API between different
  35               Windows versions. For instance, this can provide insights
  36               into the differences between the, often undocumented, Win9x and
  37               NT Windows families.
  38             </para>
  39             <para>
  40               However, one must remember that the goal of Wine is to run
  41               Windows applications on Linux, not to be a clone of any specific
  42               Windows version. So such variations must only be tested for when
  43               relevant to that goal.
  44             </para>
  45           </listitem>
  46           <listitem>
  47             <para>
  48               Writing conformance tests is also an easy way to discover
  49               bugs in Wine. Of course, before fixing the bugs discovered in
  50               this way, one must first make sure that the new tests do pass
  51               successfully on at least one Windows 9x and one Windows NT
  52               version.
  53             </para>
  54             <para>
  55               Bugs discovered this way should also be easier to fix. Unlike
  56               some mysterious application crashes, when a conformance test
  57               fails, the expected behavior and APIs tested for are known thus
  58               greatly simplifying the diagnosis.
  59             </para>
  60           </listitem>
  61           <listitem>
  62             <para>
  63               To detect regressions. Simply running the test suite regularly
  64               in Wine turns it into a great tool to detect regressions.
  65               When a test fails, one immediately knows what was the expected
  66               behavior and which APIs are involved. Thus regressions caught
  67               this way should be detected earlier, because it is easy to run
  68               all tests on a regular basis, and easier to fix because of the
  69               reduced diagnosis work.
  70             </para>
  71           </listitem>
  72           <listitem>
  73             <para>
  74               Tests written in advance of the Wine development (possibly even
  75               by non Wine developers) can also simplify the work of the
  76               future implementer by making it easier for him to check the
  77               correctness of his code.
  78             </para>
  79           </listitem>
  80           <listitem>
  81             <para>
  82               Conformance tests will also come in handy when testing Wine on
  83               new (or not as widely used) architectures such as FreeBSD,
  84               Solaris x86 or even non-x86 systems. Even when the port does
  85               not involve any significant change in the thread management,
  86               exception handling or other low-level aspects of Wine, new
  87               architectures can expose subtle bugs that can be hard to
  88               diagnose when debugging regular (complex) applications.
  89             </para>
  90           </listitem>
  91         </itemizedlist>
  92       </para>
  93     </sect1>
  94
  95
  96     <sect1 id="testing-what">
  97       <title>What to test for?</title>
  98       <para>
  99         The first thing to test for is the documented behavior of APIs
 100         and such as CreateFile. For instance one can create a file using a
 101         long pathname, check that the behavior is correct when the file
 102         already exists, try to open the file using the corresponding short
 103         pathname, convert the filename to Unicode and try to open it using
 104         CreateFileW, and all other things which are documented and that
 105         applications rely on.
 106       </para>
 107       <para>
 108         While the testing framework is not specifically geared towards this
 109         type of tests, it is also possible to test the behavior of Windows
 110         messages. To do so, create a window, preferably a hidden one so that
 111         it does not steal the focus when running the tests, and send messages
 112         to that window or to controls in that window. Then, in the message
 113         procedure, check that you receive the expected messages and with the
 114         correct parameters.
 115       </para>
 116       <para>
 117         For instance you could create an edit control and use WM_SETTEXT to
 118         set its contents, possibly check length restrictions, and verify the
 119         results using WM_GETTEXT. Similarly one could create a listbox and
 120         check the effect of LB_DELETESTRING on the list's number of items,
 121         selected items list, highlighted item, etc.
 122       </para>
 123       <para>
 124         However, undocumented behavior should not be tested for unless there
 125         is an application that relies on this behavior, and in that case the
 126         test should mention that application, or unless one can strongly
 127         expect applications to rely on this behavior, typically APIs that
 128         return the required buffer size when the buffer pointer is NULL.
 129       </para>
 130     </sect1>
 131
 132
 133     <sect1 id="testing-wine">
 134       <title>Running the tests in Wine</title>
 135       <para>
 136         The simplest way to run the tests in Wine is to type 'make test' in
 137         the Wine sources top level directory. This will run all the Wine
 138         conformance tests.
 139       </para>
 140       <para>
 141         The tests for a specific Wine library are located in a 'tests'
 142         directory in that library's directory. Each test is contained in a
 143         file (e.g. <filename>dlls/kernel/tests/thread.c</>). Each
 144         file itself contains many checks concerning one or more related APIs.
 145       </para>
 146       <para>
 147         So to run all the tests related to a given Wine library, go to the
 148         corresponding 'tests' directory and type 'make test'. This will
 149         compile the tests, run them, and create an '<replaceable>xxx</>.ok'
 150         file for each test that passes successfully. And if you only want to
 151         run the tests contained in the <filename>thread.c</> file of the
 152         kernel library, you would do:
 153 <screen>
 154 <prompt>$ </>cd dlls/kernel/tests
 155 <prompt>$ </>make thread.ok
 156 </screen>
 157       </para>
 158       <para>
 159         Note that if the test has already been run and is up to date (i.e. if
 160         neither the kernel library nor the <filename>thread.c</> file has
 161         changed since the <filename>thread.ok</> file was created), then make
 162         will say so. To force the test to be re-run, delete the
 163         <filename>thread.ok</> file, and run the make command again.
 164       </para>
 165       <para>
 166         You can also run tests manually using a command similar to the
 167         following:
 168 <screen>
 169 <prompt>$ </>../../../tools/runtest -q -M kernel32.dll -p kernel32_test.exe.so thread.c
 170 <prompt>$ </>../../../tools/runtest -p kernel32_test.exe.so thread.c
 171 thread.c: 86 tests executed, 5 marked as todo, 0 failures.
 172 </screen>
 173         The '-P wine' options defines the platform that is currently being
 174         tested. Remove the '-q' option if you want the testing framework
 175         to report statistics about the number of successful and failed tests.
 176         Run <command>runtest -h</> for more details.
 177       </para>
 178     </sect1>
 179
 180
 181     <sect1 id="cross-compiling-tests">
 182       <title>Cross-compiling the tests with MinGW</title>
 183       <sect2>
 184         <title>Setup of the MinGW cross-compiling environment</title>
 185         <para>
 186           Here are some instructions to setup MinGW on different Linux
 187           distributions and *BSD.
 188         </para>
 189         <sect3>
 190           <title>Debian GNU/Linux</title>
 191           <para>
 192             On Debian all you need to do is type <command>apt-get install
 193             mingw32</>.
 194           </para>
 195         </sect3>
 196         <sect3>
 197           <title>Red Hat Linux like rpm systems</title>
 198           <para>
 199             This includes Fedora Core, Red Hat Enterprise Linux, Mandrake,
 200             most probably SuSE Linux too, etc. But this list isn't exhaustive;
 201             the following steps should probably work on any rpm based system.
 202           </para>
 203           <para>
 204             Download and install the latest rpm's from
 205             <ulink url="http://mirzam.it.vu.nl/mingw/">MinGW RPM packages</>.
 206             Alternatively you can follow the instructions on that page and
 207             build your own packages from the source rpm's listed there as well.
 208           </para>
 209         </sect3>
 210         <sect3>
 211           <title>*BSD</title>
 212           <para>
 213             The *BSD systems have in their ports collection a port for the
 214             MinGW cross-compiling environment. Please see the documentation
 215             of your system about how to build and install a port.
 216           </para>
 217         </sect3>
 218       </sect2>
 219       <sect2>
 220       <title>Compiling the tests</title>
 221       <para>
 222         Having the cross-compiling environment set up the generation of the
 223         Windows executables is easy by using the Wine build system.
 224       </para>
 225       <para>
 226         If you had already run <command>configure</>, then delete
 227         <filename>config.cache</> and re-run <command>configure</>.
 228         You can then run <command>make crosstest</>. To sum up:
 229 <screen>
 230 <prompt>$ </><userinput>rm config.cache</>
 231 <prompt>$ </><userinput>./configure</>
 232 <prompt>$ </><userinput>make crosstest</>
 233 </screen>
 234       </para>
 235       </sect2>
 236     </sect1>
 237
 238
 239     <sect1 id="testing-windows">
 240       <title>Building and running the tests on Windows</title>
 241       <sect2>
 242         <title>Using pre-compiled binaries</title>
 243         <para>
 244           Unfortunately there are no pre-compiled binaries yet. However if
 245           send an email to the Wine development list you can probably get
 246           someone to send them to you, and maybe motivate some kind soul to
 247           put in place a mechanism for publishing such binaries on a regular
 248           basis.
 249         </para>
 250       </sect2>
 251       <sect2>
 252         <title>With Visual C++</title>
 253         <screen>
 254         Visual Studio 6 users:
 255         - MSVC headers may not work well, try with Wine headers
 256         - Ensure that you have the "processor pack" from
 257           <ulink url="http://msdn.microsoft.com/vstudio/downloads/tools/ppack/default.aspx">http://msdn.microsoft.com/vstudio/downloads/tools/ppack/default.aspx</>
 258           as well as the latest service packs.  The processor pack fixes <emphasis>"error C2520: conversion from unsigned
 259            __int64 to double not implemented, use signed __int64"</>
 260         </screen>
 261         <itemizedlist>
 262           <listitem><para>
 263             get the Wine sources
 264           </para></listitem>
 265           <listitem><para>
 266             Run msvcmaker to generate Visual C++ project files for the tests.
 267             'msvcmaker' is a perl script so you may be able to run it on
 268             Windows.
 269 <screen>
 270 <prompt>$ </>./tools/winapi/msvcmaker --no-wine
 271 </screen>
 272           </para></listitem>
 273           <listitem><para>
 274             If the previous steps were done on your Linux development
 275             machine, make the Wine sources accessible to the Windows machine
 276             on which you are going to compile them. Typically you would do
 277             this using Samba but copying them altogether would work too.
 278           </para></listitem>
 279           <listitem><para>
 280             On the Windows machine, open the <filename>winetest.dsw</>
 281             workspace. This will load each test's project. For each test there
 282             are two configurations: one compiles the test with the Wine
 283             headers, and the other uses the Visual C++ headers. Some tests
 284             will compile fine with the former, but most will require the
 285             latter.
 286           </para></listitem>
 287           <listitem><para>
 288             Open the <menuchoice><guimenu>Build</> <guimenu>Batch
 289             build...</></> menu and select the tests and build configurations
 290             you want to build. Then click on <guibutton>Build</>.
 291           </para></listitem>
 292           <listitem><para>
 293             To run a specific test from Visual C++, go to
 294             <menuchoice><guimenu>Project</> <guimenu>Settings...</></>. There
 295             select that test's project and build configuration and go to the
 296             <guilabel>Debug</> tab. There type the name of the specific test
 297             to run (e.g. 'thread') in the <guilabel>Program arguments</>
 298             field. Validate your change by clicking on <guibutton>Ok</> and
 299             start the test by clicking the red exclamation mark (or hitting
 300             'F5' or any other usual method).
 301           </para></listitem>
 302           <listitem><para>
 303             You can also run the tests from the command line. You will find
 304             them in either <filename>Output\Win32_Wine_Headers</> or
 305             <filename>Output\Win32_MSVC_Headers</> depending on the build
 306             method. So to run the kernel 'path' tests you would do:
 307 <screen>
 308 <prompt>C:\&gt;</>cd dlls\kernel\tests\Output\Win32_MSVC_Headers
 309 <prompt>C:\dlls\kernel\tests\Output\Win32_MSVC_Headers&gt;</>kernel32_test path
 310 </screen>
 311           </para></listitem>
 312         </itemizedlist>
 313       </sect2>
 314       <sect2>
 315         <title>With MinGW</title>
 316         <para>
 317          Wine's build system already has support for building tests with a MinGW
 318          cross-compiler. See the section above called 'Setup of the MinGW
 319          cross-compiling environment' for instructions on how to set things up.
 320          When you have a MinGW environment installed all you need to do is rerun
 321          configure and it should detect the MinGW compiler and tools. Then run
 322          'make crosstest' to start building the tests.
 323         </para>
 324       </sect2>
 325     </sect1>
 326
 327
 328     <sect1 id="testing-test">
 329       <title>Inside a test</title>
 330
 331       <para>
 332         When writing new checks you can either modify an existing test file or
 333         add a new one. If your tests are related to the tests performed by an
 334         existing file, then add them to that file. Otherwise create a new .c
 335         file in the tests directory and add that file to the
 336         <varname>CTESTS</> variable in <filename>Makefile.in</>.
 337       </para>
 338       <para>
 339         A new test file will look something like the following:
 340 <screen>
 341 #include &lt;wine/test.h&gt;
 342 #include &lt;winbase.h&gt;
 343
 344 /* Maybe auxiliary functions and definitions here */
 345
 346 START_TEST(paths)
 347 {
 348    /* Write your checks there or put them in functions you will call from
 349     * there
 350     */
 351 }
 352 </screen>
 353       </para>
 354       <para>
 355         The test's entry point is the START_TEST section. This is where
 356         execution will start. You can put all your tests in that section but
 357         it may be better to split related checks in functions you will call
 358         from the START_TEST section. The parameter to START_TEST must match
 359         the name of the C file. So in the above example the C file would be
 360         called <filename>paths.c</>.
 361       </para>
 362       <para>
 363         Tests should start by including the <filename>wine/test.h</> header.
 364         This header will provide you access to all the testing framework
 365         functions. You can then include the windows header you need, but make
 366         sure to not include any Unix or Wine specific header: tests must
 367         compile on Windows.
 368       </para>
 369       <para>
 370         You can use <function>trace</> to print informational messages. Note
 371         that these messages will only be printed if 'runtest -v' is being used.
 372 <screen>
 373   trace("testing GlobalAddAtomA");
 374   trace("foo=%d",foo);
 375 </screen>
 376       </para>
 377       <para>
 378         Then just call functions and use <function>ok</> to make sure that
 379         they behaved as expected:
 380 <screen>
 381   ATOM atom = GlobalAddAtomA( "foobar" );
 382   ok( GlobalFindAtomA( "foobar" ) == atom, "could not find atom foobar" );
 383   ok( GlobalFindAtomA( "FOOBAR" ) == atom, "could not find atom FOOBAR" );
 384 </screen>
 385         The first parameter of <function>ok</> is an expression which must
 386         evaluate to true if the test was successful. The next parameter is a
 387         printf-compatible format string which is displayed in case the test
 388         failed, and the following optional parameters depend on the format
 389         string.
 390       </para>
 391     </sect1>
 392
 393     <sect1 id="testing-error-messages">
 394       <title>Writing good error messages</title>
 395       <para>
 396         The message that is printed when a test fails is
 397         <emphasis>extremely</> important.
 398       </para>
 399       <para>
 400         Someone will take your test, run it on a Windows platform that
 401         you don't have access to, and discover that it fails. They will then
 402         post an email with the output of the test, and in particular your
 403         error message. Someone, maybe you, will then have to figure out from
 404         this error message why the test failed.
 405       </para>
 406       <para>
 407         If the error message contains all the relevant information that will
 408         be easy. If not, then it will require modifying the test, finding
 409         someone to compile it on Windows, sending the modified version to the
 410         original tester and waiting for his reply. In other words, it will
 411         be long and painful.
 412       </para>
 413       <para>
 414         So how do you write a good error message? Let's start with an example
 415         of a bad error message:
 416 <screen>
 417     ok(GetThreadPriorityBoost(curthread,&amp;disabled)!=0,
 418        "GetThreadPriorityBoost Failed");
 419 </screen>
 420         This will yield:
 421 <screen>
 422 thread.c:123: Test failed: GetThreadPriorityBoost Failed
 423 </screen>
 424       </para>
 425       <para>
 426         Did you notice how the error message provides no information about
 427         why the test failed? We already know from the line number exactly
 428         which test failed. In fact the error message gives strictly no
 429         information that cannot already be obtained by reading the code. In
 430         other words it provides no more information than an empty string!
 431       </para>
 432       <para>
 433         Let's look at how to rewrite it:
 434 <screen>
 435     BOOL rc;
 436 ...
 437     rc=GetThreadPriorityBoost(curthread,&amp;disabled);
 438     ok(rc!=0 && disabled==0,"rc=%d error=%ld disabled=%d",
 439        rc,GetLastError(),disabled);
 440 </screen>
 441         This will yield:
 442 <screen>
 443 thread.c:123: Test failed: rc=0 error=120 disabled=0
 444 </screen>
 445       </para>
 446       <para>
 447         When receiving such a message, one would check the source, see that
 448         it's a call to GetThreadPriorityBoost, that the test failed not
 449         because the API returned the wrong value, but because it returned an
 450         error code. Furthermore we see that GetLastError() returned 120 which
 451         winerror.h defines as ERROR_CALL_NOT_IMPLEMENTED. So the source of
 452         the problem is obvious: this Windows platform (here Windows 98) does
 453         not support this API and thus the test must be modified to detect
 454         such a condition and skip the test.
 455       </para>
 456       <para>
 457         So a good error message should provide all the information which
 458         cannot be obtained by reading the source, typically the function
 459         return value, error codes, and any function output parameter. Even if
 460         more information is needed to fully understand a problem,
 461         systematically providing the above is easy and will help cut down the
 462         number of iterations required to get to a resolution.
 463       </para>
 464       <para>
 465         It may also be a good idea to dump items that may be hard to retrieve
 466         from the source, like the expected value in a test if it is the
 467         result of an earlier computation, or comes from a large array of test
 468         values (e.g. index 112 of _pTestStrA in vartest.c). In that respect,
 469         for some tests you may want to define a macro such as the following:
 470 <screen>
 471 #define eq(received, expected, label, type) \
 472         ok((received) == (expected), "%s: got " type " instead of " type, (label),(received),(expected))
 473
 474 ...
 475
 476     eq( b, curr_val, "SPI_{GET,SET}BEEP", "%d" );
 477 </screen>
 478        </para>
 479     </sect1>
 480
 481
 482     <sect1 id="testing-platforms">
 483       <title>Handling platform issues</title>
 484       <para>
 485         Some checks may be written before they pass successfully in Wine.
 486         Without some mechanism, such checks would potentially generate
 487         hundred of known failures for months each time the tests are being run.
 488         This would make it hard to detect new failures caused by a regression.
 489         or to detect that a patch fixed a long standing issue.
 490       </para>
 491       <para>
 492         Thus the Wine testing framework has the concept of platforms and
 493         groups of checks can be declared as expected to fail on some of them.
 494         In the most common case, one would declare a group of tests as
 495         expected to fail in Wine. To do so, use the following construct:
 496 <screen>
 497 todo_wine {
 498     SetLastError( 0xdeadbeef );
 499     ok( GlobalAddAtomA(0) == 0 && GetLastError() == 0xdeadbeef, "failed to add atom 0" );
 500 }
 501 </screen>
 502         On Windows the above check would be performed normally, but on Wine it
 503         would be expected to fail, and not cause the failure of the whole
 504         test. However. If that check were to succeed in Wine, it would
 505         cause the test to fail, thus making it easy to detect when something
 506         has changed that fixes a bug. Also note that todo checks are accounted
 507         separately from regular checks so that the testing statistics remain
 508         meaningful. Finally, note that todo sections can be nested so that if
 509         a test only fails on the cygwin and reactos platforms, one would
 510         write:
 511 <screen>
 512 todo("cygwin") {
 513     todo("reactos") {
 514         ...
 515     }
 516 }
 517 </screen>
 518         <!-- FIXME: Would we really have platforms such as reactos, cygwin, freebsd & co? -->
 519         But specific platforms should not be nested inside a todo_wine section
 520         since that would be redundant.
 521       </para>
 522       <para>
 523         When writing tests you will also encounter differences between Windows
 524         9x and Windows NT platforms. Such differences should be treated
 525         differently from the platform issues mentioned above. In particular
 526         you should remember that the goal of Wine is not to be a clone of any
 527         specific Windows version but to run Windows applications on Unix.
 528       </para>
 529       <para>
 530         So, if an API returns a different error code on Windows 9x and
 531         Windows NT, your check should just verify that Wine returns one or
 532         the other:
 533 <screen>
 534 ok ( GetLastError() == WIN9X_ERROR || GetLastError() == NT_ERROR, ...);
 535 </screen>
 536       </para>
 537       <para>
 538         If an API is only present on some Windows platforms, then use
 539         LoadLibrary and GetProcAddress to check if it is implemented and
 540         invoke it. Remember, tests must run on all Windows platforms.
 541         Similarly, conformance tests should nor try to correlate the Windows
 542         version returned by GetVersion with whether given APIs are
 543         implemented or not. Again, the goal of Wine is to run Windows
 544         applications (which do not do such checks), and not be a clone of a
 545         specific Windows version.
 546       </para>
 547       <!--para>
 548         FIXME: What about checks that cause the process to crash due to a bug?
 549       </para-->
 550     </sect1>
 551
 552
 553 <!-- FIXME: Strategies for testing threads, testing network stuff,
 554  file handling, eq macro... -->
 555
 556   </chapter>
 557
 558 <!-- Keep this comment at the end of the file
 559 Local variables:
 560 mode: sgml
 561 sgml-parent-document:("wine-devel.sgml" "set" "book" "part" "chapter" "")
 562 End:
 563 -->