t / READMEon commit Sync with 2.18.1 (1958ad5)
   1Core GIT Tests
   2==============
   3
   4This directory holds many test scripts for core GIT tools.  The
   5first part of this short document describes how to run the tests
   6and read their output.
   7
   8When fixing the tools or adding enhancements, you are strongly
   9encouraged to add tests in this directory to cover what you are
  10trying to fix or enhance.  The later part of this short document
  11describes how your test scripts should be organized.
  12
  13
  14Running Tests
  15-------------
  16
  17The easiest way to run tests is to say "make".  This runs all
  18the tests.
  19
  20    *** t0000-basic.sh ***
  21    ok 1 - .git/objects should be empty after git init in an empty repo.
  22    ok 2 - .git/objects should have 3 subdirectories.
  23    ok 3 - success is reported like this
  24    ...
  25    ok 43 - very long name in the index handled sanely
  26    # fixed 1 known breakage(s)
  27    # still have 1 known breakage(s)
  28    # passed all remaining 42 test(s)
  29    1..43
  30    *** t0001-init.sh ***
  31    ok 1 - plain
  32    ok 2 - plain with GIT_WORK_TREE
  33    ok 3 - plain bare
  34
  35Since the tests all output TAP (see http://testanything.org) they can
  36be run with any TAP harness. Here's an example of parallel testing
  37powered by a recent version of prove(1):
  38
  39    $ prove --timer --jobs 15 ./t[0-9]*.sh
  40    [19:17:33] ./t0005-signals.sh ................................... ok       36 ms
  41    [19:17:33] ./t0022-crlf-rename.sh ............................... ok       69 ms
  42    [19:17:33] ./t0024-crlf-archive.sh .............................. ok      154 ms
  43    [19:17:33] ./t0004-unwritable.sh ................................ ok      289 ms
  44    [19:17:33] ./t0002-gitfile.sh ................................... ok      480 ms
  45    ===(     102;0  25/?  6/?  5/?  16/?  1/?  4/?  2/?  1/?  3/?  1... )===
  46
  47prove and other harnesses come with a lot of useful options. The
  48--state option in particular is very useful:
  49
  50    # Repeat until no more failures
  51    $ prove -j 15 --state=failed,save ./t[0-9]*.sh
  52
  53You can give DEFAULT_TEST_TARGET=prove on the make command (or define it
  54in config.mak) to cause "make test" to run tests under prove.
  55GIT_PROVE_OPTS can be used to pass additional options, e.g.
  56
  57    $ make DEFAULT_TEST_TARGET=prove GIT_PROVE_OPTS='--timer --jobs 16' test
  58
  59You can also run each test individually from command line, like this:
  60
  61    $ sh ./t3010-ls-files-killed-modified.sh
  62    ok 1 - git update-index --add to add various paths.
  63    ok 2 - git ls-files -k to show killed files.
  64    ok 3 - validate git ls-files -k output.
  65    ok 4 - git ls-files -m to show modified files.
  66    ok 5 - validate git ls-files -m output.
  67    # passed all 5 test(s)
  68    1..5
  69
  70You can pass --verbose (or -v), --debug (or -d), and --immediate
  71(or -i) command line argument to the test, or by setting GIT_TEST_OPTS
  72appropriately before running "make".
  73
  74-v::
  75--verbose::
  76        This makes the test more verbose.  Specifically, the
  77        command being run and their output if any are also
  78        output.
  79
  80--verbose-only=<pattern>::
  81        Like --verbose, but the effect is limited to tests with
  82        numbers matching <pattern>.  The number matched against is
  83        simply the running count of the test within the file.
  84
  85-x::
  86        Turn on shell tracing (i.e., `set -x`) during the tests
  87        themselves. Implies `--verbose`.
  88        Ignored in test scripts that set the variable 'test_untraceable'
  89        to a non-empty value, unless it's run with a Bash version
  90        supporting BASH_XTRACEFD, i.e. v4.1 or later.
  91
  92-d::
  93--debug::
  94        This may help the person who is developing a new test.
  95        It causes the command defined with test_debug to run.
  96        The "trash" directory (used to store all temporary data
  97        during testing) is not deleted even if there are no
  98        failed tests so that you can inspect its contents after
  99        the test finished.
 100
 101-i::
 102--immediate::
 103        This causes the test to immediately exit upon the first
 104        failed test. Cleanup commands requested with
 105        test_when_finished are not executed if the test failed,
 106        in order to keep the state for inspection by the tester
 107        to diagnose the bug.
 108
 109-l::
 110--long-tests::
 111        This causes additional long-running tests to be run (where
 112        available), for more exhaustive testing.
 113
 114-r::
 115--run=<test-selector>::
 116        Run only the subset of tests indicated by
 117        <test-selector>.  See section "Skipping Tests" below for
 118        <test-selector> syntax.
 119
 120--valgrind=<tool>::
 121        Execute all Git binaries under valgrind tool <tool> and exit
 122        with status 126 on errors (just like regular tests, this will
 123        only stop the test script when running under -i).
 124
 125        Since it makes no sense to run the tests with --valgrind and
 126        not see any output, this option implies --verbose.  For
 127        convenience, it also implies --tee.
 128
 129        <tool> defaults to 'memcheck', just like valgrind itself.
 130        Other particularly useful choices include 'helgrind' and
 131        'drd', but you may use any tool recognized by your valgrind
 132        installation.
 133
 134        As a special case, <tool> can be 'memcheck-fast', which uses
 135        memcheck but disables --track-origins.  Use this if you are
 136        running tests in bulk, to see if there are _any_ memory
 137        issues.
 138
 139        Note that memcheck is run with the option --leak-check=no,
 140        as the git process is short-lived and some errors are not
 141        interesting. In order to run a single command under the same
 142        conditions manually, you should set GIT_VALGRIND to point to
 143        the 't/valgrind/' directory and use the commands under
 144        't/valgrind/bin/'.
 145
 146--valgrind-only=<pattern>::
 147        Like --valgrind, but the effect is limited to tests with
 148        numbers matching <pattern>.  The number matched against is
 149        simply the running count of the test within the file.
 150
 151--tee::
 152        In addition to printing the test output to the terminal,
 153        write it to files named 't/test-results/$TEST_NAME.out'.
 154        As the names depend on the tests' file names, it is safe to
 155        run the tests with this option in parallel.
 156
 157--verbose-log::
 158        Write verbose output to the same logfile as `--tee`, but do
 159        _not_ write it to stdout. Unlike `--tee --verbose`, this option
 160        is safe to use when stdout is being consumed by a TAP parser
 161        like `prove`. Implies `--tee` and `--verbose`.
 162
 163--with-dashes::
 164        By default tests are run without dashed forms of
 165        commands (like git-commit) in the PATH (it only uses
 166        wrappers from ../bin-wrappers).  Use this option to include
 167        the build directory (..) in the PATH, which contains all
 168        the dashed forms of commands.  This option is currently
 169        implied by other options like --valgrind and
 170        GIT_TEST_INSTALLED.
 171
 172--root=<directory>::
 173        Create "trash" directories used to store all temporary data during
 174        testing under <directory>, instead of the t/ directory.
 175        Using this option with a RAM-based filesystem (such as tmpfs)
 176        can massively speed up the test suite.
 177
 178--chain-lint::
 179--no-chain-lint::
 180        If --chain-lint is enabled, the test harness will check each
 181        test to make sure that it properly "&&-chains" all commands (so
 182        that a failure in the middle does not go unnoticed by the final
 183        exit code of the test). This check is performed in addition to
 184        running the tests themselves. You may also enable or disable
 185        this feature by setting the GIT_TEST_CHAIN_LINT environment
 186        variable to "1" or "0", respectively.
 187
 188You can also set the GIT_TEST_INSTALLED environment variable to
 189the bindir of an existing git installation to test that installation.
 190You still need to have built this git sandbox, from which various
 191test-* support programs, templates, and perl libraries are used.
 192If your installed git is incomplete, it will silently test parts of
 193your built version instead.
 194
 195When using GIT_TEST_INSTALLED, you can also set GIT_TEST_EXEC_PATH to
 196override the location of the dashed-form subcommands (what
 197GIT_EXEC_PATH would be used for during normal operation).
 198GIT_TEST_EXEC_PATH defaults to `$GIT_TEST_INSTALLED/git --exec-path`.
 199
 200
 201Skipping Tests
 202--------------
 203
 204In some environments, certain tests have no way of succeeding
 205due to platform limitation, such as lack of 'unzip' program, or
 206filesystem that do not allow arbitrary sequence of non-NUL bytes
 207as pathnames.
 208
 209You should be able to say something like
 210
 211    $ GIT_SKIP_TESTS=t9200.8 sh ./t9200-git-cvsexport-commit.sh
 212
 213and even:
 214
 215    $ GIT_SKIP_TESTS='t[0-4]??? t91?? t9200.8' make
 216
 217to omit such tests.  The value of the environment variable is a
 218SP separated list of patterns that tells which tests to skip,
 219and either can match the "t[0-9]{4}" part to skip the whole
 220test, or t[0-9]{4} followed by ".$number" to say which
 221particular test to skip.
 222
 223For an individual test suite --run could be used to specify that
 224only some tests should be run or that some tests should be
 225excluded from a run.
 226
 227The argument for --run is a list of individual test numbers or
 228ranges with an optional negation prefix that define what tests in
 229a test suite to include in the run.  A range is two numbers
 230separated with a dash and matches a range of tests with both ends
 231been included.  You may omit the first or the second number to
 232mean "from the first test" or "up to the very last test"
 233respectively.
 234
 235Optional prefix of '!' means that the test or a range of tests
 236should be excluded from the run.
 237
 238If --run starts with an unprefixed number or range the initial
 239set of tests to run is empty. If the first item starts with '!'
 240all the tests are added to the initial set.  After initial set is
 241determined every test number or range is added or excluded from
 242the set one by one, from left to right.
 243
 244Individual numbers or ranges could be separated either by a space
 245or a comma.
 246
 247For example, to run only tests up to a specific test (21), one
 248could do this:
 249
 250    $ sh ./t9200-git-cvsexport-commit.sh --run='1-21'
 251
 252or this:
 253
 254    $ sh ./t9200-git-cvsexport-commit.sh --run='-21'
 255
 256Common case is to run several setup tests (1, 2, 3) and then a
 257specific test (21) that relies on that setup:
 258
 259    $ sh ./t9200-git-cvsexport-commit.sh --run='1 2 3 21'
 260
 261or:
 262
 263    $ sh ./t9200-git-cvsexport-commit.sh --run=1,2,3,21
 264
 265or:
 266
 267    $ sh ./t9200-git-cvsexport-commit.sh --run='-3 21'
 268
 269As noted above, the test set is built by going through the items
 270from left to right, so this:
 271
 272    $ sh ./t9200-git-cvsexport-commit.sh --run='1-4 !3'
 273
 274will run tests 1, 2, and 4.  Items that come later have higher
 275precedence.  It means that this:
 276
 277    $ sh ./t9200-git-cvsexport-commit.sh --run='!3 1-4'
 278
 279would just run tests from 1 to 4, including 3.
 280
 281You may use negation with ranges.  The following will run all
 282test in the test suite except from 7 up to 11:
 283
 284    $ sh ./t9200-git-cvsexport-commit.sh --run='!7-11'
 285
 286Some tests in a test suite rely on the previous tests performing
 287certain actions, specifically some tests are designated as
 288"setup" test, so you cannot _arbitrarily_ disable one test and
 289expect the rest to function correctly.
 290
 291--run is mostly useful when you want to focus on a specific test
 292and know what setup is needed for it.  Or when you want to run
 293everything up to a certain test.
 294
 295
 296Running tests with special setups
 297---------------------------------
 298
 299The whole test suite could be run to test some special features
 300that cannot be easily covered by a few specific test cases. These
 301could be enabled by running the test suite with correct GIT_TEST_
 302environment set.
 303
 304GIT_TEST_SPLIT_INDEX=<boolean> forces split-index mode on the whole
 305test suite. Accept any boolean values that are accepted by git-config.
 306
 307GIT_TEST_FULL_IN_PACK_ARRAY=<boolean> exercises the uncommon
 308pack-objects code path where there are more than 1024 packs even if
 309the actual number of packs in repository is below this limit. Accept
 310any boolean values that are accepted by git-config.
 311
 312GIT_TEST_OE_SIZE=<n> exercises the uncommon pack-objects code path
 313where we do not cache object size in memory and read it from existing
 314packs on demand. This normally only happens when the object size is
 315over 2GB. This variable forces the code path on any object larger than
 316<n> bytes.
 317
 318GIT_TEST_OE_DELTA_SIZE=<n> exercises the uncomon pack-objects code
 319path where deltas larger than this limit require extra memory
 320allocation for bookkeeping.
 321
 322Naming Tests
 323------------
 324
 325The test files are named as:
 326
 327        tNNNN-commandname-details.sh
 328
 329where N is a decimal digit.
 330
 331First digit tells the family:
 332
 333        0 - the absolute basics and global stuff
 334        1 - the basic commands concerning database
 335        2 - the basic commands concerning the working tree
 336        3 - the other basic commands (e.g. ls-files)
 337        4 - the diff commands
 338        5 - the pull and exporting commands
 339        6 - the revision tree commands (even e.g. merge-base)
 340        7 - the porcelainish commands concerning the working tree
 341        8 - the porcelainish commands concerning forensics
 342        9 - the git tools
 343
 344Second digit tells the particular command we are testing.
 345
 346Third digit (optionally) tells the particular switch or group of switches
 347we are testing.
 348
 349If you create files under t/ directory (i.e. here) that is not
 350the top-level test script, never name the file to match the above
 351pattern.  The Makefile here considers all such files as the
 352top-level test script and tries to run all of them.  Care is
 353especially needed if you are creating a common test library
 354file, similar to test-lib.sh, because such a library file may
 355not be suitable for standalone execution.
 356
 357
 358Writing Tests
 359-------------
 360
 361The test script is written as a shell script.  It should start
 362with the standard "#!/bin/sh", and an
 363assignment to variable 'test_description', like this:
 364
 365        #!/bin/sh
 366
 367        test_description='xxx test (option --frotz)
 368
 369        This test registers the following structure in the cache
 370        and tries to run git-ls-files with option --frotz.'
 371
 372
 373Source 'test-lib.sh'
 374--------------------
 375
 376After assigning test_description, the test script should source
 377test-lib.sh like this:
 378
 379        . ./test-lib.sh
 380
 381This test harness library does the following things:
 382
 383 - If the script is invoked with command line argument --help
 384   (or -h), it shows the test_description and exits.
 385
 386 - Creates an empty test directory with an empty .git/objects database
 387   and chdir(2) into it.  This directory is 't/trash
 388   directory.$test_name_without_dotsh', with t/ subject to change by
 389   the --root option documented above.
 390
 391 - Defines standard test helper functions for your scripts to
 392   use.  These functions are designed to make all scripts behave
 393   consistently when command line arguments --verbose (or -v),
 394   --debug (or -d), and --immediate (or -i) is given.
 395
 396Do's, don'ts & things to keep in mind
 397-------------------------------------
 398
 399Here are a few examples of things you probably should and shouldn't do
 400when writing tests.
 401
 402Do:
 403
 404 - Put all code inside test_expect_success and other assertions.
 405
 406   Even code that isn't a test per se, but merely some setup code
 407   should be inside a test assertion.
 408
 409 - Chain your test assertions
 410
 411   Write test code like this:
 412
 413        git merge foo &&
 414        git push bar &&
 415        test ...
 416
 417   Instead of:
 418
 419        git merge hla
 420        git push gh
 421        test ...
 422
 423   That way all of the commands in your tests will succeed or fail. If
 424   you must ignore the return value of something, consider using a
 425   helper function (e.g. use sane_unset instead of unset, in order
 426   to avoid unportable return value for unsetting a variable that was
 427   already unset), or prepending the command with test_might_fail or
 428   test_must_fail.
 429
 430 - Check the test coverage for your tests. See the "Test coverage"
 431   below.
 432
 433   Don't blindly follow test coverage metrics; if a new function you added
 434   doesn't have any coverage, then you're probably doing something wrong,
 435   but having 100% coverage doesn't necessarily mean that you tested
 436   everything.
 437
 438   Tests that are likely to smoke out future regressions are better
 439   than tests that just inflate the coverage metrics.
 440
 441 - When a test checks for an absolute path that a git command generated,
 442   construct the expected value using $(pwd) rather than $PWD,
 443   $TEST_DIRECTORY, or $TRASH_DIRECTORY. It makes a difference on
 444   Windows, where the shell (MSYS bash) mangles absolute path names.
 445   For details, see the commit message of 4114156ae9.
 446
 447Don't:
 448
 449 - exit() within a <script> part.
 450
 451   The harness will catch this as a programming error of the test.
 452   Use test_done instead if you need to stop the tests early (see
 453   "Skipping tests" below).
 454
 455 - use '! git cmd' when you want to make sure the git command exits
 456   with failure in a controlled way by calling "die()".  Instead,
 457   use 'test_must_fail git cmd'.  This will signal a failure if git
 458   dies in an unexpected way (e.g. segfault).
 459
 460   On the other hand, don't use test_must_fail for running regular
 461   platform commands; just use '! cmd'.  We are not in the business
 462   of verifying that the world given to us sanely works.
 463
 464 - use perl without spelling it as "$PERL_PATH". This is to help our
 465   friends on Windows where the platform Perl often adds CR before
 466   the end of line, and they bundle Git with a version of Perl that
 467   does not do so, whose path is specified with $PERL_PATH. Note that we
 468   provide a "perl" function which uses $PERL_PATH under the hood, so
 469   you do not need to worry when simply running perl in the test scripts
 470   (but you do, for example, on a shebang line or in a sub script
 471   created via "write_script").
 472
 473 - use sh without spelling it as "$SHELL_PATH", when the script can
 474   be misinterpreted by broken platform shell (e.g. Solaris).
 475
 476 - chdir around in tests.  It is not sufficient to chdir to
 477   somewhere and then chdir back to the original location later in
 478   the test, as any intermediate step can fail and abort the test,
 479   causing the next test to start in an unexpected directory.  Do so
 480   inside a subshell if necessary.
 481
 482 - save and verify the standard error of compound commands, i.e. group
 483   commands, subshells, and shell functions (except test helper
 484   functions like 'test_must_fail') like this:
 485
 486     ( cd dir && git cmd ) 2>error &&
 487     test_cmp expect error
 488
 489   When running the test with '-x' tracing, then the trace of commands
 490   executed in the compound command will be included in standard error
 491   as well, quite possibly throwing off the subsequent checks examining
 492   the output.  Instead, save only the relevant git command's standard
 493   error:
 494
 495     ( cd dir && git cmd 2>../error ) &&
 496     test_cmp expect error
 497
 498 - Break the TAP output
 499
 500   The raw output from your test may be interpreted by a TAP harness. TAP
 501   harnesses will ignore everything they don't know about, but don't step
 502   on their toes in these areas:
 503
 504   - Don't print lines like "$x..$y" where $x and $y are integers.
 505
 506   - Don't print lines that begin with "ok" or "not ok".
 507
 508   TAP harnesses expect a line that begins with either "ok" and "not
 509   ok" to signal a test passed or failed (and our harness already
 510   produces such lines), so your script shouldn't emit such lines to
 511   their output.
 512
 513   You can glean some further possible issues from the TAP grammar
 514   (see https://metacpan.org/pod/TAP::Parser::Grammar#TAP-GRAMMAR)
 515   but the best indication is to just run the tests with prove(1),
 516   it'll complain if anything is amiss.
 517
 518Keep in mind:
 519
 520 - Inside the <script> part, the standard output and standard error
 521   streams are discarded, and the test harness only reports "ok" or
 522   "not ok" to the end user running the tests. Under --verbose, they
 523   are shown to help debugging the tests.
 524
 525
 526Skipping tests
 527--------------
 528
 529If you need to skip tests you should do so by using the three-arg form
 530of the test_* functions (see the "Test harness library" section
 531below), e.g.:
 532
 533    test_expect_success PERL 'I need Perl' '
 534        perl -e "hlagh() if unf_unf()"
 535    '
 536
 537The advantage of skipping tests like this is that platforms that don't
 538have the PERL and other optional dependencies get an indication of how
 539many tests they're missing.
 540
 541If the test code is too hairy for that (i.e. does a lot of setup work
 542outside test assertions) you can also skip all remaining tests by
 543setting skip_all and immediately call test_done:
 544
 545        if ! test_have_prereq PERL
 546        then
 547            skip_all='skipping perl interface tests, perl not available'
 548            test_done
 549        fi
 550
 551The string you give to skip_all will be used as an explanation for why
 552the test was skipped.
 553
 554End with test_done
 555------------------
 556
 557Your script will be a sequence of tests, using helper functions
 558from the test harness library.  At the end of the script, call
 559'test_done'.
 560
 561
 562Test harness library
 563--------------------
 564
 565There are a handful helper functions defined in the test harness
 566library for your script to use.
 567
 568 - test_expect_success [<prereq>] <message> <script>
 569
 570   Usually takes two strings as parameters, and evaluates the
 571   <script>.  If it yields success, test is considered
 572   successful.  <message> should state what it is testing.
 573
 574   Example:
 575
 576        test_expect_success \
 577            'git-write-tree should be able to write an empty tree.' \
 578            'tree=$(git-write-tree)'
 579
 580   If you supply three parameters the first will be taken to be a
 581   prerequisite; see the test_set_prereq and test_have_prereq
 582   documentation below:
 583
 584        test_expect_success TTY 'git --paginate rev-list uses a pager' \
 585            ' ... '
 586
 587   You can also supply a comma-separated list of prerequisites, in the
 588   rare case where your test depends on more than one:
 589
 590        test_expect_success PERL,PYTHON 'yo dawg' \
 591            ' test $(perl -E 'print eval "1 +" . qx[python -c "print 2"]') == "4" '
 592
 593 - test_expect_failure [<prereq>] <message> <script>
 594
 595   This is NOT the opposite of test_expect_success, but is used
 596   to mark a test that demonstrates a known breakage.  Unlike
 597   the usual test_expect_success tests, which say "ok" on
 598   success and "FAIL" on failure, this will say "FIXED" on
 599   success and "still broken" on failure.  Failures from these
 600   tests won't cause -i (immediate) to stop.
 601
 602   Like test_expect_success this function can optionally use a three
 603   argument invocation with a prerequisite as the first argument.
 604
 605 - test_debug <script>
 606
 607   This takes a single argument, <script>, and evaluates it only
 608   when the test script is started with --debug command line
 609   argument.  This is primarily meant for use during the
 610   development of a new test script.
 611
 612 - debug <git-command>
 613
 614   Run a git command inside a debugger. This is primarily meant for
 615   use when debugging a failing test script.
 616
 617 - test_done
 618
 619   Your test script must have test_done at the end.  Its purpose
 620   is to summarize successes and failures in the test script and
 621   exit with an appropriate error code.
 622
 623 - test_tick
 624
 625   Make commit and tag names consistent by setting the author and
 626   committer times to defined state.  Subsequent calls will
 627   advance the times by a fixed amount.
 628
 629 - test_commit <message> [<filename> [<contents>]]
 630
 631   Creates a commit with the given message, committing the given
 632   file with the given contents (default for both is to reuse the
 633   message string), and adds a tag (again reusing the message
 634   string as name).  Calls test_tick to make the SHA-1s
 635   reproducible.
 636
 637 - test_merge <message> <commit-or-tag>
 638
 639   Merges the given rev using the given message.  Like test_commit,
 640   creates a tag and calls test_tick before committing.
 641
 642 - test_set_prereq <prereq>
 643
 644   Set a test prerequisite to be used later with test_have_prereq. The
 645   test-lib will set some prerequisites for you, see the
 646   "Prerequisites" section below for a full list of these.
 647
 648   Others you can set yourself and use later with either
 649   test_have_prereq directly, or the three argument invocation of
 650   test_expect_success and test_expect_failure.
 651
 652 - test_have_prereq <prereq>
 653
 654   Check if we have a prerequisite previously set with test_set_prereq.
 655   The most common way to use this explicitly (as opposed to the
 656   implicit use when an argument is passed to test_expect_*) is to skip
 657   all the tests at the start of the test script if we don't have some
 658   essential prerequisite:
 659
 660        if ! test_have_prereq PERL
 661        then
 662            skip_all='skipping perl interface tests, perl not available'
 663            test_done
 664        fi
 665
 666 - test_external [<prereq>] <message> <external> <script>
 667
 668   Execute a <script> with an <external> interpreter (like perl). This
 669   was added for tests like t9700-perl-git.sh which do most of their
 670   work in an external test script.
 671
 672        test_external \
 673            'GitwebCache::*FileCache*' \
 674            perl "$TEST_DIRECTORY"/t9503/test_cache_interface.pl
 675
 676   If the test is outputting its own TAP you should set the
 677   test_external_has_tap variable somewhere before calling the first
 678   test_external* function. See t9700-perl-git.sh for an example.
 679
 680        # The external test will outputs its own plan
 681        test_external_has_tap=1
 682
 683 - test_external_without_stderr [<prereq>] <message> <external> <script>
 684
 685   Like test_external but fail if there's any output on stderr,
 686   instead of checking the exit code.
 687
 688        test_external_without_stderr \
 689            'Perl API' \
 690            perl "$TEST_DIRECTORY"/t9700/test.pl
 691
 692 - test_expect_code <exit-code> <command>
 693
 694   Run a command and ensure that it exits with the given exit code.
 695   For example:
 696
 697        test_expect_success 'Merge with d/f conflicts' '
 698                test_expect_code 1 git merge "merge msg" B master
 699        '
 700
 701 - test_must_fail [<options>] <git-command>
 702
 703   Run a git command and ensure it fails in a controlled way.  Use
 704   this instead of "! <git-command>".  When git-command dies due to a
 705   segfault, test_must_fail diagnoses it as an error; "! <git-command>"
 706   treats it as just another expected failure, which would let such a
 707   bug go unnoticed.
 708
 709   Accepts the following options:
 710
 711     ok=<signal-name>[,<...>]:
 712       Don't treat an exit caused by the given signal as error.
 713       Multiple signals can be specified as a comma separated list.
 714       Currently recognized signal names are: sigpipe, success.
 715       (Don't use 'success', use 'test_might_fail' instead.)
 716
 717 - test_might_fail [<options>] <git-command>
 718
 719   Similar to test_must_fail, but tolerate success, too.  Use this
 720   instead of "<git-command> || :" to catch failures due to segv.
 721
 722   Accepts the same options as test_must_fail.
 723
 724 - test_cmp <expected> <actual>
 725
 726   Check whether the content of the <actual> file matches the
 727   <expected> file.  This behaves like "cmp" but produces more
 728   helpful output when the test is run with "-v" option.
 729
 730 - test_cmp_rev <expected> <actual>
 731
 732   Check whether the <expected> rev points to the same commit as the
 733   <actual> rev.
 734
 735 - test_line_count (= | -lt | -ge | ...) <length> <file>
 736
 737   Check whether a file has the length it is expected to.
 738
 739 - test_path_is_file <path> [<diagnosis>]
 740   test_path_is_dir <path> [<diagnosis>]
 741   test_path_is_missing <path> [<diagnosis>]
 742
 743   Check if the named path is a file, if the named path is a
 744   directory, or if the named path does not exist, respectively,
 745   and fail otherwise, showing the <diagnosis> text.
 746
 747 - test_when_finished <script>
 748
 749   Prepend <script> to a list of commands to run to clean up
 750   at the end of the current test.  If some clean-up command
 751   fails, the test will not pass.
 752
 753   Example:
 754
 755        test_expect_success 'branch pointing to non-commit' '
 756                git rev-parse HEAD^{tree} >.git/refs/heads/invalid &&
 757                test_when_finished "git update-ref -d refs/heads/invalid" &&
 758                ...
 759        '
 760
 761 - test_write_lines <lines>
 762
 763   Write <lines> on standard output, one line per argument.
 764   Useful to prepare multi-line files in a compact form.
 765
 766   Example:
 767
 768        test_write_lines a b c d e f g >foo
 769
 770   Is a more compact equivalent of:
 771        cat >foo <<-EOF
 772        a
 773        b
 774        c
 775        d
 776        e
 777        f
 778        g
 779        EOF
 780
 781
 782 - test_pause
 783
 784        This command is useful for writing and debugging tests and must be
 785        removed before submitting. It halts the execution of the test and
 786        spawns a shell in the trash directory. Exit the shell to continue
 787        the test. Example:
 788
 789        test_expect_success 'test' '
 790                git do-something >actual &&
 791                test_pause &&
 792                test_cmp expected actual
 793        '
 794
 795 - test_ln_s_add <path1> <path2>
 796
 797   This function helps systems whose filesystem does not support symbolic
 798   links. Use it to add a symbolic link entry to the index when it is not
 799   important that the file system entry is a symbolic link, i.e., instead
 800   of the sequence
 801
 802        ln -s foo bar &&
 803        git add bar
 804
 805   Sometimes it is possible to split a test in a part that does not need
 806   the symbolic link in the file system and a part that does; then only
 807   the latter part need be protected by a SYMLINKS prerequisite (see below).
 808
 809Prerequisites
 810-------------
 811
 812These are the prerequisites that the test library predefines with
 813test_have_prereq.
 814
 815See the prereq argument to the test_* functions in the "Test harness
 816library" section above and the "test_have_prereq" function for how to
 817use these, and "test_set_prereq" for how to define your own.
 818
 819 - PYTHON
 820
 821   Git wasn't compiled with NO_PYTHON=YesPlease. Wrap any tests that
 822   need Python with this.
 823
 824 - PERL
 825
 826   Git wasn't compiled with NO_PERL=YesPlease.
 827
 828   Even without the PERL prerequisite, tests can assume there is a
 829   usable perl interpreter at $PERL_PATH, though it need not be
 830   particularly modern.
 831
 832 - POSIXPERM
 833
 834   The filesystem supports POSIX style permission bits.
 835
 836 - BSLASHPSPEC
 837
 838   Backslashes in pathspec are not directory separators. This is not
 839   set on Windows. See 6fd1106a for details.
 840
 841 - EXECKEEPSPID
 842
 843   The process retains the same pid across exec(2). See fb9a2bea for
 844   details.
 845
 846 - PIPE
 847
 848   The filesystem we're on supports creation of FIFOs (named pipes)
 849   via mkfifo(1).
 850
 851 - SYMLINKS
 852
 853   The filesystem we're on supports symbolic links. E.g. a FAT
 854   filesystem doesn't support these. See 704a3143 for details.
 855
 856 - SANITY
 857
 858   Test is not run by root user, and an attempt to write to an
 859   unwritable file is expected to fail correctly.
 860
 861 - PCRE
 862
 863   Git was compiled with support for PCRE. Wrap any tests
 864   that use git-grep --perl-regexp or git-grep -P in these.
 865
 866 - LIBPCRE1
 867
 868   Git was compiled with PCRE v1 support via
 869   USE_LIBPCRE1=YesPlease. Wrap any PCRE using tests that for some
 870   reason need v1 of the PCRE library instead of v2 in these.
 871
 872 - LIBPCRE2
 873
 874   Git was compiled with PCRE v2 support via
 875   USE_LIBPCRE2=YesPlease. Wrap any PCRE using tests that for some
 876   reason need v2 of the PCRE library instead of v1 in these.
 877
 878 - CASE_INSENSITIVE_FS
 879
 880   Test is run on a case insensitive file system.
 881
 882 - UTF8_NFD_TO_NFC
 883
 884   Test is run on a filesystem which converts decomposed utf-8 (nfd)
 885   to precomposed utf-8 (nfc).
 886
 887 - PTHREADS
 888
 889   Git wasn't compiled with NO_PTHREADS=YesPlease.
 890
 891Tips for Writing Tests
 892----------------------
 893
 894As with any programming projects, existing programs are the best
 895source of the information.  However, do _not_ emulate
 896t0000-basic.sh when writing your tests.  The test is special in
 897that it tries to validate the very core of GIT.  For example, it
 898knows that there will be 256 subdirectories under .git/objects/,
 899and it knows that the object ID of an empty tree is a certain
 90040-byte string.  This is deliberately done so in t0000-basic.sh
 901because the things the very basic core test tries to achieve is
 902to serve as a basis for people who are changing the GIT internal
 903drastically.  For these people, after making certain changes,
 904not seeing failures from the basic test _is_ a failure.  And
 905such drastic changes to the core GIT that even changes these
 906otherwise supposedly stable object IDs should be accompanied by
 907an update to t0000-basic.sh.
 908
 909However, other tests that simply rely on basic parts of the core
 910GIT working properly should not have that level of intimate
 911knowledge of the core GIT internals.  If all the test scripts
 912hardcoded the object IDs like t0000-basic.sh does, that defeats
 913the purpose of t0000-basic.sh, which is to isolate that level of
 914validation in one place.  Your test also ends up needing
 915updating when such a change to the internal happens, so do _not_
 916do it and leave the low level of validation to t0000-basic.sh.
 917
 918Test coverage
 919-------------
 920
 921You can use the coverage tests to find code paths that are not being
 922used or properly exercised yet.
 923
 924To do that, run the coverage target at the top-level (not in the t/
 925directory):
 926
 927    make coverage
 928
 929That'll compile Git with GCC's coverage arguments, and generate a test
 930report with gcov after the tests finish. Running the coverage tests
 931can take a while, since running the tests in parallel is incompatible
 932with GCC's coverage mode.
 933
 934After the tests have run you can generate a list of untested
 935functions:
 936
 937    make coverage-untested-functions
 938
 939You can also generate a detailed per-file HTML report using the
 940Devel::Cover module. To install it do:
 941
 942   # On Debian or Ubuntu:
 943   sudo aptitude install libdevel-cover-perl
 944
 945   # From the CPAN with cpanminus
 946   curl -L http://cpanmin.us | perl - --sudo --self-upgrade
 947   cpanm --sudo Devel::Cover
 948
 949Then, at the top-level:
 950
 951    make cover_db_html
 952
 953That'll generate a detailed cover report in the "cover_db_html"
 954directory, which you can then copy to a webserver, or inspect locally
 955in a browser.