Diff for /wikisrc/tutorials/atf.mdwn between versions 1.6 and 1.7

version 1.6, 2010/09/03 15:04:51 version 1.7, 2010/09/03 15:47:34
Line 1 Line 1
 [[!meta title="Creating atf-based tests for NetBSD src"]]  [[!meta title="Creating atf-based tests for NetBSD src"]]
 [[!toc ]]  [[!toc ]]
   
 This quick tutorial is an attempt to workaround the lack of proper documentation  This quick tutorial provides a guideline on how to start creating new test
 in atf.  The tutorial provides a guideline on how to start creating new test  
 programs and/or test cases, how these tests are tied to the NetBSD source tree  programs and/or test cases, how these tests are tied to the NetBSD source tree
 and a short reference of the most commonly used functions.  and includes a short reference of the most commonly used functions.
   
 You should start by reading the  You should start by reading the
 [tests(7)](http://netbsd.gw.com/cgi-bin/man-cgi?tests++NetBSD-current) manual  [tests(7)](http://netbsd.gw.com/cgi-bin/man-cgi?tests++NetBSD-current) manual
 page, which is probably the only sane document in the whole documentation.  Any  page, which provides a user-level overview on how to run the tests included in
 other attempts at reading the atf-* manual pages are probably doomed unless you  NetBSD.  While reading this tutorial, you may also want to refer to these pages
 are already familiar with atf itself and its internals.  Still, you may be able  on a need-to-know basis:
 to get some useful information out of  
 [atf-run(1)](http://netbsd.gw.com/cgi-bin/man-cgi?atf-run++NetBSD-current),  [atf-run(1)](http://netbsd.gw.com/cgi-bin/man-cgi?atf-run++NetBSD-current),
 [atf-report(1)](http://netbsd.gw.com/cgi-bin/man-cgi?atf-report++NetBSD-current),  [atf-report(1)](http://netbsd.gw.com/cgi-bin/man-cgi?atf-report++NetBSD-current),
 [atf-test-program(1)](http://netbsd.gw.com/cgi-bin/man-cgi?atf-test-program++NetBSD-current),  [atf-test-program(1)](http://netbsd.gw.com/cgi-bin/man-cgi?atf-test-program++NetBSD-current),
 [atf-c-api(3)](http://netbsd.gw.com/cgi-bin/man-cgi?atf-c-api++NetBSD-current)  [atf-c-api(3)](http://netbsd.gw.com/cgi-bin/man-cgi?atf-c-api++NetBSD-current),
   [atf-sh-api(3)](http://netbsd.gw.com/cgi-bin/man-cgi?atf-sh-api++NetBSD-current)
 and  and
 [atf-sh-api(3)](http://netbsd.gw.com/cgi-bin/man-cgi?atf-sh-api++NetBSD-current).  [atf-check(1)](http://netbsd.gw.com/cgi-bin/man-cgi?atf-check++NetBSD-current).
   
 **IMPORTANT: Do not take anything for granted, SPECIALLY if you have previously  **IMPORTANT: Do not take anything for granted, SPECIALLY if you have previously
 worked with and/or have seen src/regress/.  Your assumptions are most likely  worked with and/or have seen src/regress/.  Your assumptions are most likely
Line 71  require comments to explain what they in Line 70  require comments to explain what they in
 # Test case parts  # Test case parts
   
 A test case is composed by three parts: the *head*, the *body* and the  A test case is composed by three parts: the *head*, the *body* and the
 *cleanup*.  Only the body is required; the other two routines are optional.  *cleanup*.  Only the body is required; the other two routines are optional, but
   you are strongly encouraged to provide the *head*, if only to describe the
   purpose of your tests by setting the *descr* property.
   
 ## The head  ## The head
   
Line 83  The following properties are commonly us Line 84  The following properties are commonly us
   
 * descr: A textual description of the purpose of the test case.  * descr: A textual description of the purpose of the test case.
   
 * require.user: Set to 'root' to mark the test case as root-specific.  It is  * require.user: Set to 'root' to tell the atf runtime that this test requires
   nice not to abuse this; see puffs and rump.    root privileges.  The test will later be skipped if you are running atf as
     non-root, and the test will be executed otherwise.
   
 * use.fs: Set to 'true' if the test case creates temporary files in the "current  * use.fs: Set to 'true' if the test case creates temporary files in the "current
   directory".  Otherwise the atf runtime will isolate the test case in such a    directory".  If set to false, the atf runtime will set the "current directory"
   way to forbid this, which will misteriously make your test to fail.    to an unwritable directory, which will disallow the creation of the temporary
     files and will make your test misteriously fail.
   
 ## The body  ## The body
   
Line 101  every test case to prevent side-effects  Line 104  every test case to prevent side-effects 
 in-memory data corruption, etc.).  In particular:  in-memory data corruption, etc.).  In particular:
   
 * A test case is **always executed as a subprocess** that is separate from the  * A test case is **always executed as a subprocess** that is separate from the
   head and the cleanup.    head and the cleanup.  This implies that you cannot pass any in-memory state
     between the three parts.  The only exception is that the body and the cleanup
     share the same temporary directory, so you can use the file system to pass
     information between them.
   
 * The current working directory of a test case is changed to a temporary  * The current working directory of a test case is changed to a temporary
   location that gets cleaned up later on automatically.  (Set the use.fs    location that gets cleaned up later on automatically.  (Set the use.fs
Line 111  in-memory data corruption, etc.).  In pa Line 117  in-memory data corruption, etc.).  In pa
 * The environment of the test case is "sanitized" to get rid of variables that  * The environment of the test case is "sanitized" to get rid of variables that
   can cause side-effects; e.g. LC_ALL, TZ, etc.    can cause side-effects; e.g. LC_ALL, TZ, etc.
   
 # Installation of test programs: the why and the where  # Running the test programs
   
   Do:
   
       $ cd /usr/tests/
       $ atf-run | atf-report
   
   Why?
   
 Test programs get installed into the /usr/tests/ hierarchy.  The main reason for  Test programs get installed into the /usr/tests/ hierarchy.  The main reason for
 doing that is to allow *any* user to test his system and to be able to convince  doing that is to allow *any* user to test his system and to be able to convince
Line 133  This Makefile in src/tests/bin/ls/ will  Line 146  This Makefile in src/tests/bin/ls/ will 
 ui_test binary.  The Makefile will also generate an Atffile.  Both files (the  ui_test binary.  The Makefile will also generate an Atffile.  Both files (the
 ui_test binary and the Atffile) will later be installed to /usr/tests/bin/ls/  ui_test binary and the Atffile) will later be installed to /usr/tests/bin/ls/
   
   ## Executing a single test
   
   In general, you **do not want to run a test program by hand**.  If you do so,
   you do not take advantage of any of the isolation provided by the atf runtime.
   This means that the test program will probably leave some temporary files behind
   or will raise some false negatives.
   
   To run a test, use atf-run.  In general:
   
       $ atf-run | atf-report  # To run all the test programs in a directory.
       $ atf-run some_test | atf-report  # To run only the some_test program.
   
   The only "legitimate" case in which you should be running test cases by hand is
   to debug them:
   
       $ gdb --args ./some_test the_broken_test_case
   
   ... but be sure to clean up any leftover files if you do that.
   
   ## Executing tests during development
   
   When you are in a subdirectory of src/tests/, you can generally run "make test"
   to execute the tests of that particular subdirectory.  This assumes that the
   tests have been installed into the destdir.
   
   Please note that this is only provided for convenience but it is completely
   unsupported.  Tests run this way may fail misteriously, and that is perfectly
   fine as long as they work from their canonical locations in /usr/tests.
   
 # Adding a new test  # Adding a new test
   
 To add a new *test case* to the source tree, look for any test program in  To add a new *test case* to the source tree, look for any test program in
Line 172  If the subdirectory does not exist: Line 214  If the subdirectory does not exist:
   
 ## Makefile template  ## Makefile template
   
     # $NetBSD: atf.mdwn,v 1.5 2010/09/03 15:04:05 jmmv Exp $  Follow this template to create your Makefile:
   
     .include <bsd.own.mk>      .include <bsd.own.mk>
   
     # This must always be defined.      # This must always be defined.
     TESTSDIR= ${TESTSBASE}/bin/ls      TESTSDIR= ${TESTSBASE}/bin/ls
   
     # Define only the variables you actually need for the directory.      # These correspond to the test programs you have in the directory.
     TESTS_C+= c1_test c2_test  # Correspond to c1_test.c and c2_test.c.      TESTS_C+= c1_test c2_test  # Correspond to c1_test.c and c2_test.c.
     TESTS_SH+= sh1_test sh2_test  # Correspond to sh1_test.c and sh2_test.c      TESTS_SH+= sh1_test sh2_test  # Correspond to sh1_test.c and sh2_test.c
   
Line 191  If the subdirectory does not exist: Line 233  If the subdirectory does not exist:
   
 ## Atffile template  ## Atffile template
   
   *Atffiles are automatically generated by bsd.test.mk, so in general you will not
   have to deal with them.*
   
 What is an Atffile?  An Atffile is the atf-run counterpart of a "Makefile".  What is an Atffile?  An Atffile is the atf-run counterpart of a "Makefile".
 Given that atf tests *do not rely on a toolchain*, they cannot use make(1) to  Given that atf tests *do not rely on a toolchain*, they cannot use make(1) to
 script their execution as the old tests in src/regress/ did.  script their execution as the old tests in src/regress/ did.
Line 198  script their execution as the old tests  Line 243  script their execution as the old tests 
 The Atffiles, in general, just provide a list of test programs in a particular  The Atffiles, in general, just provide a list of test programs in a particular
 directory and the list of the subdirectories to descend into.  directory and the list of the subdirectories to descend into.
   
 Atffiles are automatically generated by bsd.test.mk, so in general you will not  If you have to provide an Atffile explicitly because the automatic generation
 have to deal with them.  However, if you have to provide one explicitly, they  does not suit your needs, follow this format:
 follow the following format:  
   
     Content-Type: application/X-atf-atffile; version="1"      Content-Type: application/X-atf-atffile; version="1"
   
Line 216  follow the following format: Line 260  follow the following format:
   
 ## Template  ## Template
   
 The following code snippet provides a C test program with two test cases:  The following code snippet provides a C test program with two test cases.  The
   specific details as to how this works follow later:
   
     #include <atf-c.h>      #include <atf-c.h>
   
Line 227  The following code snippet provides a C  Line 272  The following code snippet provides a C 
     }      }
     ATF_TC_BODY(tc, my_test_case)      ATF_TC_BODY(tc, my_test_case)
     {      {
         ATF_CHECK(returns_a_boolean()); /* Non-fatal test. */          ATF_CHECK(true); /* Success; continue execution. */
         ATF_REQUIRE(returns_a_boolean()); /* Non-fatal test. */          ATF_CHECK(false); /* Failure; continue execution. */
   
         ATF_CHECK_EQ(4, 2 + 2); /* Non-fatal test. */          ATF_CHECK_EQ(5, 2 + 2); /* Failure; continue execution. */
         ATF_REQUIRE_EQ(4, 2 + 2); /* Fatal test. */          ATF_REQUIRE_EQ(5, 2 + 2); /* Failure; abort execution. */
   
         if (!condition)          if (!condition)
             atf_tc_fail("Condition not met!"); /* Explicit failure. */              atf_tc_fail("Condition not met!"); /* Abort execution. */
     }      }
   
     ATF_TC_WITHOUT_HEAD(tc, another_test_case);      ATF_TC_WITHOUT_HEAD(tc, another_test_case);
Line 249  The following code snippet provides a C  Line 294  The following code snippet provides a C 
         ATF_TP_ADD_TC(tp, another_test_case);          ATF_TP_ADD_TC(tp, another_test_case);
     }      }
   
 This program needs to be linked against libatf-c as described below.  Once  This program needs to be built with the Makefile shown below.  Once built, the
 linked, the program automatically gains a main() method that provides a  program automatically gains a main() method that provides a consistent user
 consistent user interface to all test programs.  You are simply not inteded to  interface to all test programs.  You are simply not inteded to provide your own
 provide your own main method, nor to deal with the command-line of the  main method, nor to deal with the command-line of the invocation.
 invocation.  
   
 ## How to build  ## How to build
   
Line 323  The following functions are commonly use Line 367  The following functions are commonly use
   
 ## Template  ## Template
   
 The following code snippet provides a shell test program with two test cases:  The following code snippet provides a shell test program with two test cases.
   The details on how this works are provided later:
   
     atf_test_case my_test_case      atf_test_case my_test_case
     my_test_case_head() {      my_test_case_head() {
Line 336  The following code snippet provides a sh Line 381  The following code snippet provides a sh
     file1      file1
     file2      file2
     EOF      EOF
           # The following call validates that the 'ls' command returns an
           # exit code of 0, that its stdout matches exactly the contents
           # previously stored in the 'expout' file and that its stderr is
           # completely empty.  See atf-check(1) for details, which is the
           # auxiliary tool invoked by the atf_check wrapper function.
         atf_check -s eq:0 -o file:expout -e empty 'ls'          atf_check -s eq:0 -o file:expout -e empty 'ls'
   
         atf_check_equal 4 $((2 + 2))          atf_check_equal 4 $((2 + 2))
Line 355  The following code snippet provides a sh Line 405  The following code snippet provides a sh
         atf_add_test_case another_test_case          atf_add_test_case another_test_case
     }      }
   
 This program needs to be be executed with the atf-sh(1) interpreter as described  This program needs to be built with the Makefile shown below.  The program
 below.  The program automatically gains an entry point that provides a  automatically gains an entry point that provides a consistent user interface to
 consistent user interface to all test programs.  You are simply not inteded to  all test programs.  You are simply not inteded to provide your own "main
 provide your own "main method", nor to deal with the command-line of the  method", nor to deal with the command-line of the invocation.
 invocation.  
   
 ## How to build  ## How to build
   
Line 452  keep this in mind: Line 501  keep this in mind:
   integer formatter, one for the string formatter, etc.    integer formatter, one for the string formatter, etc.
   
 * Use the ATF_CHECK* and/or atf_tc_fail functions to do the comparisons and  * Use the ATF_CHECK* and/or atf_tc_fail functions to do the comparisons and
   report errors.  errx should not be used.    report errors.  Neither errx nor any other error reporting and program
     termination functions (read: err, errx, warn, warnx, exit, abor) are to be
     used at all.
   
 The result would look like:  The result would look like:
   
Line 499  The result would look like: Line 550  The result would look like:
   
 Which can later be invoked as any of:  Which can later be invoked as any of:
   
     $ ./snprintf_test integer_formatter      $ atf-run snprintf_test | atf-report  # Normal execution method.
     $ ./snprintf_test string_formatter      $ ./snprintf_test integer_formatter  # For DEBUGGING only.
     $ atf-run snprintf_test | atf-report      $ ./snprintf_test string_formatter  # For DEBUGGING only.
   
 ## How do I write a test case for an unfixed PR?  ## How do I write a test case for an unfixed PR?
   
 Use the "expectations" mechanism to define part of the test case as faulty,  Use the "expectations" mechanism to define part of the test case as faulty,
 crashy, etc.  For example, suppose we have PR 1 that reports a condition in  crashy, etc.  This is for two reasons:
 which snprintf() does the wrong formatting when using %s, and PR 2 that mentions  
   * As long as the bug still exists, he test case will be reported as an "expected
     failure".  Such expected failures do not count towards the success or failure
     of the whole test suite.
   
   * When the bug gets fixed, the bug will not trigger any more in the test case,
     and thus the expectation of failure will not be met any more.  At this point
     the test case will start raising a regular failure, which is usually addressed
     by either just removing the expect_* calls.
   
   For example, suppose we have PR lib/1 that reports a condition in which
   snprintf() does the wrong formatting when using %s, and PR lib/2 that mentions
 that another snprintf() call using %d with number 5 causes a segfault.  We could  that another snprintf() call using %d with number 5 causes a segfault.  We could
 do:  do:
   
Line 523  do: Line 585  do:
         snprintf(buf, sizeof(buf), "Hello %d\n", 1);          snprintf(buf, sizeof(buf), "Hello %d\n", 1);
         ATF_CHECK_STREQ("Hello 1", buf);          ATF_CHECK_STREQ("Hello 1", buf);
   
         atf_tc_expect_signal(SIGSEGV, "PR 2: %%d with 5 causes a crash");          atf_tc_expect_signal(SIGSEGV, "PR lib/2: %%d with 5 causes a crash");
         snprintf(buf, sizeof(buf), "Hello %d\n", 5);          snprintf(buf, sizeof(buf), "Hello %d\n", 5);
         atf_tc_expect_pass();          atf_tc_expect_pass();
         ATF_CHECK_STREQ("Hello 5", buf);          ATF_CHECK_STREQ("Hello 5", buf);
Line 535  do: Line 597  do:
         char buf[1024];          char buf[1024];
   
         snprintf(buf, sizeof(buf), "Hello %s\n", "world!");          snprintf(buf, sizeof(buf), "Hello %s\n", "world!");
         atf_tc_expect_failure("PR 1: %%s does not work");          atf_tc_expect_failure("PR lib/1: %%s does not work");
         ATF_CHECK_STREQ("Hello world!", buf);          ATF_CHECK_STREQ("Hello world!", buf);
         atf_tc_expect_pass();          atf_tc_expect_pass();
     }      }
Line 550  do: Line 612  do:
   
 No.  atf-run does this automatically for you, because it runs every test program  No.  atf-run does this automatically for you, because it runs every test program
 in its own temporary subdirectory.  in its own temporary subdirectory.
   
   ## When do I use ATF_CHECK and when ATF_REQUIRE?
   
   ATF_CHECK logs errors but does not abort the execution of the test program.
   ATF_REQUIRE logs errors in a similar way but immediately terminates the
   execution.
   
   You can use this distinction in the following way: use ATF_REQUIRE to check the
   code that "prepares" your test case.  Use ATF_CHECK to do the actual
   functionality tests once all the set up has been performed.  For example:
   
       ATF_TC_BODY(getline) {
           FILE *f;
           char buf[1024];
   
           /* Opening the file is not part of the functionality under test, but it
            * must succeed before we actually test the relevant code. */
           ATF_REQUIRE((f = fopen("foo")) != NULL);
   
           ATF_CHECK(getline(f, buf, sizeof(buf)) > 0);
           ATF_CHECK_STREQ("line 1", buf);
   
           ATF_CHECK(getline(f, buf, sizeof(buf)) > 0);
           ATF_CHECK_STREQ("line 2", buf);
       }

Removed from v.1.6  
changed lines
  Added in v.1.7


CVSweb for NetBSD wikisrc <wikimaster@NetBSD.org> software: FreeBSD-CVSweb