freshlybakedca.ke
Git fork
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
35t/Makefile defines a target for each test file, such that you can also use
36shell pattern matching to run a subset of the tests:
37
38 make *checkout*
39
40will run all tests with 'checkout' in their filename.
41
42Since the tests all output TAP (see https://testanything.org) they can
43be run with any TAP harness. Here's an example of parallel testing
44powered by a recent version of prove(1):
45
46 $ prove --timer --jobs 15 ./t[0-9]*.sh
47 [19:17:33] ./t0005-signals.sh ................................... ok 36 ms
48 [19:17:33] ./t0022-crlf-rename.sh ............................... ok 69 ms
49 [19:17:33] ./t0024-crlf-archive.sh .............................. ok 154 ms
50 [19:17:33] ./t0004-unwritable.sh ................................ ok 289 ms
51 [19:17:33] ./t0002-gitfile.sh ................................... ok 480 ms
52 ===( 102;0 25/? 6/? 5/? 16/? 1/? 4/? 2/? 1/? 3/? 1... )===
53
54prove and other harnesses come with a lot of useful options. The
55--state option in particular is very useful:
56
57 # Repeat until no more failures
58 $ prove -j 15 --state=failed,save ./t[0-9]*.sh
59
60You can give DEFAULT_TEST_TARGET=prove on the make command (or define it
61in config.mak) to cause "make test" to run tests under prove.
62GIT_PROVE_OPTS can be used to pass additional options, e.g.
63
64 $ make DEFAULT_TEST_TARGET=prove GIT_PROVE_OPTS='--timer --jobs 16' test
65
66You can also run each test individually from command line, like this:
67
68 $ sh ./t3010-ls-files-killed-modified.sh
69 ok 1 - git update-index --add to add various paths.
70 ok 2 - git ls-files -k to show killed files.
71 ok 3 - validate git ls-files -k output.
72 ok 4 - git ls-files -m to show modified files.
73 ok 5 - validate git ls-files -m output.
74 # passed all 5 test(s)
75 1..5
76
77You can pass --verbose (or -v), --debug (or -d), and --immediate
78(or -i) command line argument to the test, or by setting GIT_TEST_OPTS
79appropriately before running "make". Short options can be bundled, i.e.
80'-d -v' is the same as '-dv'.
81
82-v::
83--verbose::
84 This makes the test more verbose. Specifically, the
85 command being run and their output if any are also
86 output.
87
88--verbose-only=<pattern>::
89 Like --verbose, but the effect is limited to tests with
90 numbers matching <pattern>. The number matched against is
91 simply the running count of the test within the file.
92
93-x::
94 Turn on shell tracing (i.e., `set -x`) during the tests
95 themselves. Implies `--verbose`.
96 Ignored in test scripts that set the variable 'test_untraceable'
97 to a non-empty value, unless it's run with a Bash version
98 supporting BASH_XTRACEFD, i.e. v4.1 or later.
99
100-d::
101--debug::
102 This may help the person who is developing a new test.
103 It causes the command defined with test_debug to run.
104 The "trash" directory (used to store all temporary data
105 during testing) is not deleted even if there are no
106 failed tests so that you can inspect its contents after
107 the test finished.
108
109-i::
110--immediate::
111 This causes the test to immediately exit upon the first
112 failed test. Cleanup commands requested with
113 test_when_finished are not executed if the test failed,
114 in order to keep the state for inspection by the tester
115 to diagnose the bug.
116
117-l::
118--long-tests::
119 This causes additional long-running tests to be run (where
120 available), for more exhaustive testing.
121
122-r::
123--run=<test-selector>::
124 Run only the subset of tests indicated by
125 <test-selector>. See section "Skipping Tests" below for
126 <test-selector> syntax.
127
128--valgrind=<tool>::
129 Execute all Git binaries under valgrind tool <tool> and exit
130 with status 126 on errors (just like regular tests, this will
131 only stop the test script when running under -i).
132
133 Since it makes no sense to run the tests with --valgrind and
134 not see any output, this option implies --verbose. For
135 convenience, it also implies --tee.
136
137 <tool> defaults to 'memcheck', just like valgrind itself.
138 Other particularly useful choices include 'helgrind' and
139 'drd', but you may use any tool recognized by your valgrind
140 installation.
141
142 As a special case, <tool> can be 'memcheck-fast', which uses
143 memcheck but disables --track-origins. Use this if you are
144 running tests in bulk, to see if there are _any_ memory
145 issues.
146
147 Note that memcheck is run with the option --leak-check=no,
148 as the git process is short-lived and some errors are not
149 interesting. In order to run a single command under the same
150 conditions manually, you should set GIT_VALGRIND to point to
151 the 't/valgrind/' directory and use the commands under
152 't/valgrind/bin/'.
153
154--valgrind-only=<pattern>::
155 Like --valgrind, but the effect is limited to tests with
156 numbers matching <pattern>. The number matched against is
157 simply the running count of the test within the file.
158
159--tee::
160 In addition to printing the test output to the terminal,
161 write it to files named 't/test-results/$TEST_NAME.out'.
162 As the names depend on the tests' file names, it is safe to
163 run the tests with this option in parallel.
164
165-V::
166--verbose-log::
167 Write verbose output to the same logfile as `--tee`, but do
168 _not_ write it to stdout. Unlike `--tee --verbose`, this option
169 is safe to use when stdout is being consumed by a TAP parser
170 like `prove`. Implies `--tee` and `--verbose`.
171
172--with-dashes::
173 By default tests are run without dashed forms of
174 commands (like git-commit) in the PATH (it only uses
175 wrappers from ../bin-wrappers). Use this option to include
176 the build directory (..) in the PATH, which contains all
177 the dashed forms of commands. This option is currently
178 implied by other options like --valgrind and
179 GIT_TEST_INSTALLED.
180
181--no-bin-wrappers::
182 By default, the test suite uses the wrappers in
183 `../bin-wrappers/` to execute `git` and friends. With this option,
184 `../git` and friends are run directly. This is not recommended
185 in general, as the wrappers contain safeguards to ensure that no
186 files from an installed Git are used, but can speed up test runs
187 especially on platforms where running shell scripts is expensive
188 (most notably, Windows).
189
190--root=<directory>::
191 Create "trash" directories used to store all temporary data during
192 testing under <directory>, instead of the t/ directory.
193 Using this option with a RAM-based filesystem (such as tmpfs)
194 can massively speed up the test suite.
195
196--chain-lint::
197--no-chain-lint::
198 If --chain-lint is enabled, the test harness will check each
199 test to make sure that it properly "&&-chains" all commands (so
200 that a failure in the middle does not go unnoticed by the final
201 exit code of the test). This check is performed in addition to
202 running the tests themselves. You may also enable or disable
203 this feature by setting the GIT_TEST_CHAIN_LINT environment
204 variable to "1" or "0", respectively.
205
206--stress::
207 Run the test script repeatedly in multiple parallel jobs until
208 one of them fails. Useful for reproducing rare failures in
209 flaky tests. The number of parallel jobs is, in order of
210 precedence: the value of the GIT_TEST_STRESS_LOAD
211 environment variable, or twice the number of available
212 processors (as shown by the 'getconf' utility), or 8.
213 Implies `--verbose -x --immediate` to get the most information
214 about the failure. Note that the verbose output of each test
215 job is saved to 't/test-results/$TEST_NAME.stress-<nr>.out',
216 and only the output of the failed test job is shown on the
217 terminal. The names of the trash directories get a
218 '.stress-<nr>' suffix, and the trash directory of the failed
219 test job is renamed to end with a '.stress-failed' suffix.
220
221--stress-jobs=<N>::
222 Override the number of parallel jobs. Implies `--stress`.
223
224--stress-limit=<N>::
225 When combined with --stress run the test script repeatedly
226 this many times in each of the parallel jobs or until one of
227 them fails, whichever comes first. Implies `--stress`.
228
229You can also set the GIT_TEST_INSTALLED environment variable to
230the bindir of an existing git installation to test that installation.
231You still need to have built this git sandbox, from which various
232test-* support programs, templates, and perl libraries are used.
233If your installed git is incomplete, it will silently test parts of
234your built version instead.
235
236When using GIT_TEST_INSTALLED, you can also set GIT_TEST_EXEC_PATH to
237override the location of the dashed-form subcommands (what
238GIT_EXEC_PATH would be used for during normal operation).
239GIT_TEST_EXEC_PATH defaults to `$GIT_TEST_INSTALLED/git --exec-path`.
240
241
242Skipping Tests
243--------------
244
245In some environments, certain tests have no way of succeeding
246due to platform limitation, such as lack of 'unzip' program, or
247filesystem that do not allow arbitrary sequence of non-NUL bytes
248as pathnames.
249
250You should be able to say something like
251
252 $ GIT_SKIP_TESTS=t9200.8 sh ./t9200-git-cvsexport-commit.sh
253
254and even:
255
256 $ GIT_SKIP_TESTS='t[0-4]??? t91?? t9200.8' make
257
258to omit such tests. The value of the environment variable is a
259SP separated list of patterns that tells which tests to skip,
260and either can match the "t[0-9]{4}" part to skip the whole
261test, or t[0-9]{4} followed by ".$number" to say which
262particular test to skip.
263
264For an individual test suite --run could be used to specify that
265only some tests should be run or that some tests should be
266excluded from a run.
267
268The argument for --run, <test-selector>, is a list of description
269substrings or globs or individual test numbers or ranges with an
270optional negation prefix (of '!') that define what tests in a test
271suite to include (or exclude, if negated) in the run. A range is two
272numbers separated with a dash and specifies an inclusive range of tests
273to run. You may omit the first or the second number to
274mean "from the first test" or "up to the very last test" respectively.
275
276The argument to --run is split on commas into separate strings,
277numbers, and ranges, and picks all tests that match any of the
278individual selection criteria. If the substring of the description
279text that you want to match includes a comma, use the glob character
280'?' instead. For example --run='rebase,merge?cherry-pick' would match
281on all tests that match either the glob *rebase* or the glob
282*merge?cherry-pick*.
283
284If --run starts with an unprefixed number or range, the initial
285set of tests to run is empty. If the first item starts with '!',
286all the tests are added to the initial set. After initial set is
287determined, every test number or range is added or excluded from
288the set one by one, from left to right.
289
290For example, to run only tests up to a specific test (21), one
291could do this:
292
293 $ sh ./t9200-git-cvsexport-commit.sh --run='1-21'
294
295or this:
296
297 $ sh ./t9200-git-cvsexport-commit.sh --run='-21'
298
299Common case is to run several setup tests (1, 2, 3) and then a
300specific test (21) that relies on that setup:
301
302 $ sh ./t9200-git-cvsexport-commit.sh --run='1,2,3,21'
303
304or:
305
306 $ sh ./t9200-git-cvsexport-commit.sh --run=1,2,3,21
307
308or:
309
310 $ sh ./t9200-git-cvsexport-commit.sh --run='-3,21'
311
312As noted above, the test set is built by going through the items
313from left to right, so this:
314
315 $ sh ./t9200-git-cvsexport-commit.sh --run='1-4,!3'
316
317will run tests 1, 2, and 4. Items that come later have higher
318precedence. It means that this:
319
320 $ sh ./t9200-git-cvsexport-commit.sh --run='!3,1-4'
321
322would just run tests from 1 to 4, including 3.
323
324You may use negation with ranges. The following will run all
325test in the test suite except from 7 up to 11:
326
327 $ sh ./t9200-git-cvsexport-commit.sh --run='!7-11'
328
329Sometimes there may be multiple tests with e.g. "setup" in their name
330that are needed and rather than figuring out the number for all of them
331we can just use "setup" as a substring/glob to match against the test
332description:
333
334 $ sh ./t0050-filesystem.sh --run=setup,9-11
335
336or one could select both the setup tests and the rename ones (assuming all
337relevant tests had those words in their descriptions):
338
339 $ sh ./t0050-filesystem.sh --run=setup,rename
340
341Some tests in a test suite rely on the previous tests performing
342certain actions, specifically some tests are designated as
343"setup" test, so you cannot _arbitrarily_ disable one test and
344expect the rest to function correctly.
345
346--run is mostly useful when you want to focus on a specific test
347and know what setup is needed for it. Or when you want to run
348everything up to a certain test.
349
350
351Running tests with special setups
352---------------------------------
353
354The whole test suite could be run to test some special features
355that cannot be easily covered by a few specific test cases. These
356could be enabled by running the test suite with correct GIT_TEST_
357environment set.
358
359GIT_TEST_FAIL_PREREQS=<boolean> fails all prerequisites. This is
360useful for discovering issues with the tests where say a later test
361implicitly depends on an optional earlier test.
362
363There's a "FAIL_PREREQS" prerequisite that can be used to test for
364whether this mode is active, and e.g. skip some tests that are hard to
365refactor to deal with it. The "SYMLINKS" prerequisite is currently
366excluded as so much relies on it, but this might change in the future.
367
368GIT_TEST_SPLIT_INDEX=<boolean> forces split-index mode on the whole
369test suite. Accept any boolean values that are accepted by git-config.
370
371GIT_TEST_PROTOCOL_VERSION=<n>, when set, makes 'protocol.version'
372default to n.
373
374GIT_TEST_FULL_IN_PACK_ARRAY=<boolean> exercises the uncommon
375pack-objects code path where there are more than 1024 packs even if
376the actual number of packs in repository is below this limit. Accept
377any boolean values that are accepted by git-config.
378
379GIT_TEST_OE_SIZE=<n> exercises the uncommon pack-objects code path
380where we do not cache object size in memory and read it from existing
381packs on demand. This normally only happens when the object size is
382over 2GB. This variable forces the code path on any object larger than
383<n> bytes.
384
385GIT_TEST_OE_DELTA_SIZE=<n> exercises the uncommon pack-objects code
386path where deltas larger than this limit require extra memory
387allocation for bookkeeping.
388
389GIT_TEST_VALIDATE_INDEX_CACHE_ENTRIES=<boolean> checks that cache-tree
390records are valid when the index is written out or after a merge. This
391is mostly to catch missing invalidation. Default is true.
392
393GIT_TEST_COMMIT_GRAPH=<boolean>, when true, forces the commit-graph to
394be written after every 'git commit' command, and overrides the
395'core.commitGraph' setting to true.
396
397GIT_TEST_COMMIT_GRAPH_CHANGED_PATHS=<boolean>, when true, forces
398commit-graph write to compute and write changed path Bloom filters for
399every 'git commit-graph write', as if the `--changed-paths` option was
400passed in.
401
402GIT_TEST_FSMONITOR=$PWD/t7519/fsmonitor-all exercises the fsmonitor
403code paths for utilizing a (hook based) file system monitor to speed up
404detecting new or changed files.
405
406GIT_TEST_INDEX_VERSION=<n> exercises the index read/write code path
407for the index version specified. Can be set to any valid version
408(currently 2, 3, or 4).
409
410GIT_TEST_PACK_USE_BITMAP_BOUNDARY_TRAVERSAL=<boolean> if enabled will
411use the boundary-based bitmap traversal algorithm. See the documentation
412of `pack.useBitmapBoundaryTraversal` for more details.
413
414GIT_TEST_PACK_SPARSE=<boolean> if disabled will default the pack-objects
415builtin to use the non-sparse object walk. This can still be overridden by
416the --sparse command-line argument.
417
418GIT_TEST_PACK_PATH_WALK=<boolean> if enabled will default the pack-objects
419builtin to use the path-walk API for the object walk. This can still be
420overridden by the --no-path-walk command-line argument.
421
422GIT_TEST_PRELOAD_INDEX=<boolean> exercises the preload-index code path
423by overriding the minimum number of cache entries required per thread.
424
425GIT_TEST_INDEX_THREADS=<n> enables exercising the multi-threaded loading
426of the index for the whole test suite by bypassing the default number of
427cache entries and thread minimums. Setting this to 1 will make the
428index loading single threaded.
429
430GIT_TEST_MULTI_PACK_INDEX=<boolean>, when true, forces the multi-pack-
431index to be written after every 'git repack' command, and overrides the
432'core.multiPackIndex' setting to true.
433
434GIT_TEST_MULTI_PACK_INDEX_WRITE_INCREMENTAL=<boolean>, when true, sets
435the '--incremental' option on all invocations of 'git multi-pack-index
436write'.
437
438GIT_TEST_SIDEBAND_ALL=<boolean>, when true, overrides the
439'uploadpack.allowSidebandAll' setting to true, and when false, forces
440fetch-pack to not request sideband-all (even if the server advertises
441sideband-all).
442
443GIT_TEST_DISALLOW_ABBREVIATED_OPTIONS=<boolean>, when true (which is
444the default when running tests), errors out when an abbreviated option
445is used.
446
447GIT_TEST_DEFAULT_HASH=<hash-algo> specifies which hash algorithm to
448use in the test scripts. Recognized values for <hash-algo> are "sha1"
449and "sha256".
450
451GIT_TEST_DEFAULT_REF_FORMAT=<format> specifies which ref storage format to use
452in the test scripts. Recognized values for <format> are "files" and
453"reftable".
454
455GIT_TEST_NO_WRITE_REV_INDEX=<boolean>, when true disables the
456'pack.writeReverseIndex' setting.
457
458GIT_TEST_SPARSE_INDEX=<boolean>, when true enables index writes to use the
459sparse-index format by default.
460
461GIT_TEST_CHECKOUT_WORKERS=<n> overrides the 'checkout.workers' setting
462to <n> and 'checkout.thresholdForParallelism' to 0, forcing the
463execution of the parallel-checkout code.
464
465GIT_TEST_FATAL_REGISTER_SUBMODULE_ODB=<boolean>, when true, makes
466registering submodule ODBs as alternates a fatal action. Support for
467this environment variable can be removed once the migration to
468explicitly providing repositories when accessing submodule objects is
469complete or needs to be abandoned for whatever reason (in which case the
470migrated codepaths still retain their performance benefits).
471
472GIT_TEST_REQUIRE_PREREQ=<list> allows specifying a space separated list of
473prereqs that are required to succeed. If a prereq in this list is triggered by
474a test and then fails then the whole test run will abort. This can help to make
475sure the expected tests are executed and not silently skipped when their
476dependency breaks or is simply not present in a new environment.
477
478GIT_TEST_NAME_HASH_VERSION=<int>, when set, causes 'git pack-objects' to
479assume '--name-hash-version=<n>'.
480
481
482Naming Tests
483------------
484
485The test files are named as:
486
487 tNNNN-commandname-details.sh
488
489where N is a decimal digit.
490
491First digit tells the family:
492
493 0 - the absolute basics and global stuff
494 1 - the basic commands concerning database
495 2 - the basic commands concerning the working tree
496 3 - the other basic commands (e.g. ls-files)
497 4 - the diff commands
498 5 - the pull and exporting commands
499 6 - the revision tree commands (even e.g. merge-base)
500 7 - the porcelainish commands concerning the working tree
501 8 - the porcelainish commands concerning forensics
502 9 - the git tools
503
504Second digit tells the particular command we are testing.
505
506Third digit (optionally) tells the particular switch or group of switches
507we are testing.
508
509If you create files under t/ directory (i.e. here) that is not
510the top-level test script, never name the file to match the above
511pattern. The Makefile here considers all such files as the
512top-level test script and tries to run all of them. Care is
513especially needed if you are creating a common test library
514file, similar to test-lib.sh, because such a library file may
515not be suitable for standalone execution.
516
517
518Writing Tests
519-------------
520
521The test script is written as a shell script. It should start
522with the standard "#!/bin/sh", and an
523assignment to variable 'test_description', like this:
524
525 #!/bin/sh
526
527 test_description='xxx test (option --frotz)
528
529 This test registers the following structure in the cache
530 and tries to run git-ls-files with option --frotz.'
531
532
533Source 'test-lib.sh'
534--------------------
535
536After assigning test_description, the test script should source
537test-lib.sh like this:
538
539 . ./test-lib.sh
540
541This test harness library does the following things:
542
543 - If the script is invoked with command line argument --help
544 (or -h), it shows the test_description and exits.
545
546 - Creates an empty test directory with an empty .git/objects database
547 and chdir(2) into it. This directory is 't/trash
548 directory.$test_name_without_dotsh', with t/ subject to change by
549 the --root option documented above, and a '.stress-<N>' suffix
550 appended by the --stress option.
551
552 - Defines standard test helper functions for your scripts to
553 use. These functions are designed to make all scripts behave
554 consistently when command line arguments --verbose (or -v),
555 --debug (or -d), and --immediate (or -i) is given.
556
557Recommended style
558-----------------
559
560 - Keep the test_expect_* function call and test title on
561 the same line.
562
563 For example, with test_expect_success, write it like:
564
565 test_expect_success 'test title' '
566 ... test body ...
567 '
568
569 Instead of:
570
571 test_expect_success \
572 'test title' \
573 '... test body ...'
574
575 - End the line with an opening single quote.
576
577 - Indent here-document bodies, and use "<<-" instead of "<<"
578 to strip leading TABs used for indentation:
579
580 test_expect_success 'test something' '
581 cat >expect <<-\EOF &&
582 one
583 two
584 three
585 EOF
586 test_something > actual &&
587 test_cmp expect actual
588 '
589
590 Instead of:
591
592 test_expect_success 'test something' '
593 cat >expect <<\EOF &&
594 one
595 two
596 three
597 EOF
598 test_something > actual &&
599 test_cmp expect actual
600 '
601
602 - Quote or escape the EOF delimiter that begins a here-document if
603 there is no parameter or other expansion in it, to signal readers
604 that they can skim it more casually:
605
606 cmd <<-\EOF
607 literal here-document text without any expansion
608 EOF
609
610
611Do's & don'ts
612-------------
613
614Here are a few examples of things you probably should and shouldn't do
615when writing tests.
616
617The "do's:"
618
619 - Put all code inside test_expect_success and other assertions.
620
621 Even code that isn't a test per se, but merely some setup code
622 should be inside a test assertion.
623
624 - Chain your test assertions
625
626 Write test code like this:
627
628 git merge foo &&
629 git push bar &&
630 test ...
631
632 Instead of:
633
634 git merge hla
635 git push gh
636 test ...
637
638 That way all of the commands in your tests will succeed or fail. If
639 you must ignore the return value of something, consider using a
640 helper function (e.g. use sane_unset instead of unset, in order
641 to avoid unportable return value for unsetting a variable that was
642 already unset), or prepending the command with test_might_fail or
643 test_must_fail.
644
645 - Check the test coverage for your tests. See the "Test coverage"
646 below.
647
648 Don't blindly follow test coverage metrics; if a new function you added
649 doesn't have any coverage, then you're probably doing something wrong,
650 but having 100% coverage doesn't necessarily mean that you tested
651 everything.
652
653 Tests that are likely to smoke out future regressions are better
654 than tests that just inflate the coverage metrics.
655
656 - When a test checks for an absolute path that a git command generated,
657 construct the expected value using $(pwd) rather than $PWD,
658 $TEST_DIRECTORY, or $TRASH_DIRECTORY. It makes a difference on
659 Windows, where the shell (MSYS bash) mangles absolute path names.
660 For details, see the commit message of 4114156ae9.
661
662 - Remember that inside the <script> part, the standard output and
663 standard error streams are discarded, and the test harness only
664 reports "ok" or "not ok" to the end user running the tests. Under
665 --verbose, they are shown to help debug the tests.
666
667 - Be careful when you loop
668
669 You may need to verify multiple things in a loop, but the
670 following does not work correctly:
671
672 test_expect_success 'test three things' '
673 for i in one two three
674 do
675 test_something "$i"
676 done &&
677 test_something_else
678 '
679
680 Because the status of the loop itself is the exit status of the
681 test_something in the last round, the loop does not fail when
682 "test_something" for "one" or "two" fails. This is not what you
683 want.
684
685 Instead, you can break out of the loop immediately when you see a
686 failure. Because all test_expect_* snippets are executed inside
687 a function, "return 1" can be used to fail the test immediately
688 upon a failure:
689
690 test_expect_success 'test three things' '
691 for i in one two three
692 do
693 test_something "$i" || return 1
694 done &&
695 test_something_else
696 '
697
698 Note that we still &&-chain the loop to propagate failures from
699 earlier commands.
700
701 - Repeat tests with slightly different arguments in a loop.
702
703 In some cases it may make sense to re-run the same set of tests with
704 different options or commands to ensure that the command behaves
705 despite the different parameters. This can be achieved by looping
706 around a specific parameter:
707
708 for arg in '' "--foo"
709 do
710 test_expect_success "test command ${arg:-without arguments}" '
711 command $arg
712 '
713 done
714
715 Note that while the test title uses double quotes ("), the test body
716 should continue to use single quotes (') to avoid breakage in case the
717 values contain e.g. quoting characters. The loop variable will be
718 accessible regardless of the single quotes as the test body is passed
719 to `eval`.
720
721
722And here are the "don'ts:"
723
724 - Don't exit() within a <script> part.
725
726 The harness will catch this as a programming error of the test.
727 Use test_done instead if you need to stop the tests early (see
728 "Skipping tests" below).
729
730 - Don't use '! git cmd' when you want to make sure the git command
731 exits with failure in a controlled way by calling "die()". Instead,
732 use 'test_must_fail git cmd'. This will signal a failure if git
733 dies in an unexpected way (e.g. segfault).
734
735 On the other hand, don't use test_must_fail for running regular
736 platform commands; just use '! cmd'. We are not in the business
737 of verifying that the world given to us sanely works.
738
739 - Don't feed the output of a git command to a pipe, as in:
740
741 git -C repo ls-files |
742 xargs -n 1 basename |
743 grep foo
744
745 which will discard git's exit code and may mask a crash. In the
746 above example, all exit codes are ignored except grep's.
747
748 Instead, write the output of that command to a temporary
749 file with ">" or assign it to a variable with "x=$(git ...)" rather
750 than pipe it.
751
752 - Don't use command substitution in a way that discards git's exit
753 code. When assigning to a variable, the exit code is not discarded,
754 e.g.:
755
756 x=$(git cat-file -p $sha) &&
757 ...
758
759 is OK because a crash in "git cat-file" will cause the "&&" chain
760 to fail, but:
761
762 test "refs/heads/foo" = "$(git symbolic-ref HEAD)"
763
764 is not OK and a crash in git could go undetected.
765
766 - Don't use perl without spelling it as "$PERL_PATH". This is to help
767 our friends on Windows where the platform Perl often adds CR before
768 the end of line, and they bundle Git with a version of Perl that
769 does not do so, whose path is specified with $PERL_PATH. Note that we
770 provide a "perl" function which uses $PERL_PATH under the hood, so
771 you do not need to worry when simply running perl in the test scripts
772 (but you do, for example, on a shebang line or in a sub script
773 created via "write_script").
774
775 - Don't use sh without spelling it as "$SHELL_PATH", when the script
776 can be misinterpreted by broken platform shell (e.g. Solaris).
777
778 - Don't chdir around in tests. It is not sufficient to chdir to
779 somewhere and then chdir back to the original location later in
780 the test, as any intermediate step can fail and abort the test,
781 causing the next test to start in an unexpected directory. Do so
782 inside a subshell if necessary.
783
784 - Don't save and verify the standard error of compound commands, i.e.
785 group commands, subshells, and shell functions (except test helper
786 functions like 'test_must_fail') like this:
787
788 ( cd dir && git cmd ) 2>error &&
789 test_cmp expect error
790
791 When running the test with '-x' tracing, then the trace of commands
792 executed in the compound command will be included in standard error
793 as well, quite possibly throwing off the subsequent checks examining
794 the output. Instead, save only the relevant git command's standard
795 error:
796
797 ( cd dir && git cmd 2>../error ) &&
798 test_cmp expect error
799
800 - Don't break the TAP output
801
802 The raw output from your test may be interpreted by a TAP harness. TAP
803 harnesses will ignore everything they don't know about, but don't step
804 on their toes in these areas:
805
806 - Don't print lines like "$x..$y" where $x and $y are integers.
807
808 - Don't print lines that begin with "ok" or "not ok".
809
810 TAP harnesses expect a line that begins with either "ok" and "not
811 ok" to signal a test passed or failed (and our harness already
812 produces such lines), so your script shouldn't emit such lines to
813 their output.
814
815 You can glean some further possible issues from the TAP grammar
816 (see https://metacpan.org/pod/TAP::Parser::Grammar#TAP-GRAMMAR)
817 but the best indication is to just run the tests with prove(1),
818 it'll complain if anything is amiss.
819
820
821Skipping tests
822--------------
823
824If you need to skip tests you should do so by using the three-arg form
825of the test_expect_* functions (see the "Test harness library" section
826below), e.g.:
827
828 test_expect_success PERL 'I need Perl' '
829 perl -e "hlagh() if unf_unf()"
830 '
831
832The advantage of skipping tests like this is that platforms that don't
833have the PERL and other optional dependencies get an indication of how
834many tests they're missing.
835
836If the test code is too hairy for that (i.e. does a lot of setup work
837outside test assertions) you can also skip all remaining tests by
838setting skip_all and immediately call test_done:
839
840 if ! test_have_prereq PERL
841 then
842 skip_all='skipping perl interface tests, perl not available'
843 test_done
844 fi
845
846The string you give to skip_all will be used as an explanation for why
847the test was skipped.
848
849End with test_done
850------------------
851
852Your script will be a sequence of tests, using helper functions
853from the test harness library. At the end of the script, call
854'test_done'.
855
856
857Test harness library
858--------------------
859
860There are a handful helper functions defined in the test harness
861library for your script to use. Some of them are listed below;
862see test-lib-functions.sh for the full list and their options.
863
864 - test_expect_success [<prereq>] <message> <script>
865
866 Usually takes two strings as parameters, and evaluates the
867 <script>. If it yields success, test is considered
868 successful. <message> should state what it is testing.
869
870 Example:
871
872 test_expect_success \
873 'git-write-tree should be able to write an empty tree.' \
874 'tree=$(git-write-tree)'
875
876 If <script> is `-` (a single dash), then the script to run is read
877 from stdin. This lets you more easily use single quotes within the
878 script by using a here-doc. For example:
879
880 test_expect_success 'output contains expected string' - <<\EOT
881 grep "this string has 'quotes' in it" output
882 EOT
883
884 If you supply three parameters the first will be taken to be a
885 prerequisite; see the test_set_prereq and test_have_prereq
886 documentation below:
887
888 test_expect_success TTY 'git --paginate rev-list uses a pager' \
889 ' ... '
890
891 You can also supply a comma-separated list of prerequisites, in the
892 rare case where your test depends on more than one:
893
894 test_expect_success PERL,PYTHON 'yo dawg' \
895 ' test $(perl -E '\''print eval "1 +" . qx[python -c "print(2)"]'\'') = "4" '
896
897 - test_expect_failure [<prereq>] <message> <script>
898
899 This is NOT the opposite of test_expect_success, but is used
900 to mark a test that demonstrates a known breakage. Unlike
901 the usual test_expect_success tests, which say "ok" on
902 success and "FAIL" on failure, this will say "FIXED" on
903 success and "still broken" on failure. Failures from these
904 tests won't cause -i (immediate) to stop.
905
906 Like test_expect_success this function can optionally use a three
907 argument invocation with a prerequisite as the first argument.
908
909 - test_debug <script>
910
911 This takes a single argument, <script>, and evaluates it only
912 when the test script is started with --debug command line
913 argument. This is primarily meant for use during the
914 development of a new test script.
915
916 - debug [options] <git-command>
917
918 Run a git command inside a debugger. This is primarily meant for
919 use when debugging a failing test script. With '-t', use your
920 original TERM instead of test-lib.sh's "dumb", so that your
921 debugger interface has colors.
922
923 - test_done
924
925 Your test script must have test_done at the end. Its purpose
926 is to summarize successes and failures in the test script and
927 exit with an appropriate error code.
928
929 - test_tick
930
931 Make commit and tag names consistent by setting the author and
932 committer times to defined state. Subsequent calls will
933 advance the times by a fixed amount.
934
935 - test_commit <message> [<filename> [<contents>]]
936
937 Creates a commit with the given message, committing the given
938 file with the given contents (default for both is to reuse the
939 message string), and adds a tag (again reusing the message
940 string as name). Calls test_tick to make the SHA-1s
941 reproducible.
942
943 - test_merge <message> <commit-or-tag>
944
945 Merges the given rev using the given message. Like test_commit,
946 creates a tag and calls test_tick before committing.
947
948 - test_set_prereq <prereq>
949
950 Set a test prerequisite to be used later with test_have_prereq. The
951 test-lib will set some prerequisites for you, see the
952 "Prerequisites" section below for a full list of these.
953
954 Others you can set yourself and use later with either
955 test_have_prereq directly, or the three argument invocation of
956 test_expect_success and test_expect_failure.
957
958 - test_have_prereq <prereq>
959
960 Check if we have a prerequisite previously set with test_set_prereq.
961 The most common way to use this explicitly (as opposed to the
962 implicit use when an argument is passed to test_expect_*) is to skip
963 all the tests at the start of the test script if we don't have some
964 essential prerequisite:
965
966 if ! test_have_prereq PERL
967 then
968 skip_all='skipping perl interface tests, perl not available'
969 test_done
970 fi
971
972 - test_lazy_prereq <prereq> <script>
973
974 Declare the way to determine if a test prerequisite <prereq> is
975 satisified or not, but delay the actual determination until the
976 prerequisite is actually used by "test_have_prereq" or the
977 three-arg form of the test_expect_* functions. For example, this
978 is how the SYMLINKS prerequisite is declared to see if the platform
979 supports symbolic links:
980
981 test_lazy_prereq SYMLINKS '
982 ln -s x y && test -h y
983 '
984
985 The script is lazily invoked when SYMLINKS prerequisite is first
986 queried by either "test_have_prereq SYMLINKS" or "test_expect_*
987 SYMLINKS ...". The script is run in a temporary directory inside
988 a subshell, so you do not have to worry about removing temporary
989 files you create there. When the script exits with status 0, the
990 prerequisite is set. Exiting with non-zero status other than 125
991 makes the prerequisite unsatisified. Exiting the script with 125
992 signals a programming error and is used to mark a prerequisite that
993 should not be used by test scripts.
994
995 - test_expect_code <exit-code> <command>
996
997 Run a command and ensure that it exits with the given exit code.
998 For example:
999
1000 test_expect_success 'Merge with d/f conflicts' '
1001 test_expect_code 1 git merge "merge msg" B master
1002 '
1003
1004 - test_must_fail [<options>] <git-command>
1005
1006 Run a git command and ensure it fails in a controlled way. Use
1007 this instead of "! <git-command>". When git-command dies due to a
1008 segfault, test_must_fail diagnoses it as an error; "! <git-command>"
1009 treats it as just another expected failure, which would let such a
1010 bug go unnoticed.
1011
1012 Accepts the following options:
1013
1014 ok=<signal-name>[,<...>]:
1015 Don't treat an exit caused by the given signal as error.
1016 Multiple signals can be specified as a comma separated list.
1017 Currently recognized signal names are: sigpipe, success.
1018 (Don't use 'success', use 'test_might_fail' instead.)
1019
1020 - test_might_fail [<options>] <git-command>
1021
1022 Similar to test_must_fail, but tolerate success, too. Use this
1023 instead of "<git-command> || :" to catch failures due to segv.
1024
1025 Accepts the same options as test_must_fail.
1026
1027 - test_cmp <expected> <actual>
1028
1029 Check whether the content of the <actual> file matches the
1030 <expected> file. This behaves like "cmp" but produces more
1031 helpful output when the test is run with "-v" option.
1032
1033 - test_cmp_rev <expected> <actual>
1034
1035 Check whether the <expected> rev points to the same commit as the
1036 <actual> rev.
1037
1038 - test_line_count (= | -lt | -ge | ...) <length> <file>
1039
1040 Check whether a file has the length it is expected to.
1041
1042 - test_path_is_file <path>
1043 test_path_is_dir <path>
1044 test_path_is_missing <path>
1045
1046 Check if the named path is a file, if the named path is a
1047 directory, or if the named path does not exist, respectively,
1048 and fail otherwise.
1049
1050 - test_when_finished <script>
1051
1052 Prepend <script> to a list of commands to run to clean up
1053 at the end of the current test. If some clean-up command
1054 fails, the test will not pass.
1055
1056 Example:
1057
1058 test_expect_success 'branch pointing to non-commit' '
1059 git rev-parse HEAD^{tree} >.git/refs/heads/invalid &&
1060 test_when_finished "git update-ref -d refs/heads/invalid" &&
1061 ...
1062 '
1063
1064 - test_atexit <script>
1065
1066 Prepend <script> to a list of commands to run unconditionally to
1067 clean up before the test script exits, e.g. to stop a daemon:
1068
1069 test_expect_success 'test git daemon' '
1070 git daemon &
1071 daemon_pid=$! &&
1072 test_atexit 'kill $daemon_pid' &&
1073 hello world
1074 '
1075
1076 The commands will be executed before the trash directory is removed,
1077 i.e. the atexit commands will still be able to access any pidfiles or
1078 socket files.
1079
1080 Note that these commands will be run even when a test script run
1081 with '--immediate' fails. Be careful with your atexit commands to
1082 minimize any changes to the failed state.
1083
1084 - test_write_lines <lines>
1085
1086 Write <lines> on standard output, one line per argument.
1087 Useful to prepare multi-line files in a compact form.
1088
1089 Example:
1090
1091 test_write_lines a b c d e f g >foo
1092
1093 Is a more compact equivalent of:
1094 cat >foo <<-EOF
1095 a
1096 b
1097 c
1098 d
1099 e
1100 f
1101 g
1102 EOF
1103
1104
1105 - test_pause [options]
1106
1107 This command is useful for writing and debugging tests and must be
1108 removed before submitting. It halts the execution of the test and
1109 spawns a shell in the trash directory. Exit the shell to continue
1110 the test. Example:
1111
1112 test_expect_success 'test' '
1113 git do-something >actual &&
1114 test_pause &&
1115 test_cmp expected actual
1116 '
1117
1118 - test_ln_s_add <path1> <path2>
1119
1120 This function helps systems whose filesystem does not support symbolic
1121 links. Use it to add a symbolic link entry to the index when it is not
1122 important that the file system entry is a symbolic link, i.e., instead
1123 of the sequence
1124
1125 ln -s foo bar &&
1126 git add bar
1127
1128 Sometimes it is possible to split a test in a part that does not need
1129 the symbolic link in the file system and a part that does; then only
1130 the latter part need be protected by a SYMLINKS prerequisite (see below).
1131
1132 - test_path_is_executable
1133
1134 This tests whether a file is executable and prints an error message
1135 if not. This must be used only under the POSIXPERM prerequisite
1136 (see below).
1137
1138 - test_oid_init
1139
1140 This function loads facts and useful object IDs related to the hash
1141 algorithm(s) in use from the files in t/oid-info.
1142
1143 - test_oid_cache
1144
1145 This function reads per-hash algorithm information from standard
1146 input (usually a heredoc) in the format described in
1147 t/oid-info/README. This is useful for test-specific values, such as
1148 object IDs, which must vary based on the hash algorithm.
1149
1150 Certain fixed values, such as hash sizes and common placeholder
1151 object IDs, can be loaded with test_oid_init (described above).
1152
1153 - test_oid <key>
1154
1155 This function looks up a value for the hash algorithm in use, based
1156 on the key given. The value must have been loaded using
1157 test_oid_init or test_oid_cache. Providing an unknown key is an
1158 error.
1159
1160 - yes [<string>]
1161
1162 This is often seen in modern UNIX but some platforms lack it, so
1163 the test harness overrides the platform implementation with a
1164 more limited one. Use this only when feeding a handful lines of
1165 output to the downstream---unlike the real version, it generates
1166 only up to 99 lines.
1167
1168 - test_bool_env <env-variable-name> <default-value>
1169
1170 Given the name of an environment variable with a bool value,
1171 normalize its value to a 0 (true) or 1 (false or empty string)
1172 return code. Return with code corresponding to the given default
1173 value if the variable is unset.
1174 Abort the test script if either the value of the variable or the
1175 default are not valid bool values.
1176
1177
1178Prerequisites
1179-------------
1180
1181These are the prerequisites that the test library predefines with
1182test_have_prereq.
1183
1184See the prereq argument to the test_* functions in the "Test harness
1185library" section above and the "test_have_prereq" function for how to
1186use these, and "test_set_prereq" for how to define your own.
1187
1188 - PYTHON
1189
1190 Git wasn't compiled with NO_PYTHON=YesPlease. Wrap any tests that
1191 need Python with this.
1192
1193 - PERL
1194
1195 Git wasn't compiled with NO_PERL=YesPlease.
1196
1197 Even without the PERL prerequisite, tests can assume there is a
1198 usable perl interpreter at $PERL_PATH, though it need not be
1199 particularly modern.
1200
1201 - POSIXPERM
1202
1203 The filesystem supports POSIX style permission bits.
1204
1205 - BSLASHPSPEC
1206
1207 Backslashes in pathspec are not directory separators. This is not
1208 set on Windows. See 6fd1106a for details.
1209
1210 - EXECKEEPSPID
1211
1212 The process retains the same pid across exec(2). See fb9a2bea for
1213 details.
1214
1215 - PIPE
1216
1217 The filesystem we're on supports creation of FIFOs (named pipes)
1218 via mkfifo(1).
1219
1220 - SYMLINKS
1221
1222 The filesystem we're on supports symbolic links. E.g. a FAT
1223 filesystem doesn't support these. See 704a3143 for details.
1224
1225 - SANITY
1226
1227 Test is not run by root user, and an attempt to write to an
1228 unwritable file is expected to fail correctly.
1229
1230 - PCRE
1231
1232 Git was compiled with support for PCRE. Wrap any tests
1233 that use git-grep --perl-regexp or git-grep -P in these.
1234
1235 - CASE_INSENSITIVE_FS
1236
1237 Test is run on a case insensitive file system.
1238
1239 - UTF8_NFD_TO_NFC
1240
1241 Test is run on a filesystem which converts decomposed utf-8 (nfd)
1242 to precomposed utf-8 (nfc).
1243
1244 - PTHREADS
1245
1246 Git wasn't compiled with NO_PTHREADS=YesPlease.
1247
1248 - REFFILES
1249
1250 Test is specific to packed/loose ref storage, and should be
1251 disabled for other ref storage backends
1252
1253
1254Tips for Writing Tests
1255----------------------
1256
1257As with any programming projects, existing programs are the best
1258source of the information. However, do _not_ emulate
1259t0000-basic.sh when writing your tests. The test is special in
1260that it tries to validate the very core of Git. For example, it
1261knows that there will be 256 subdirectories under .git/objects/,
1262and it knows that the object ID of an empty tree is a certain
126340-byte string. This is deliberately done so in t0000-basic.sh
1264because the things the very basic core test tries to achieve is
1265to serve as a basis for people who are changing the Git internals
1266drastically. For these people, after making certain changes,
1267not seeing failures from the basic test _is_ a failure. Any
1268Git core changes so drastic that they change even these
1269otherwise supposedly stable object IDs should be accompanied by
1270an update to t0000-basic.sh.
1271
1272However, other tests that simply rely on basic parts of the core
1273Git working properly should not have that level of intimate
1274knowledge of the core Git internals. If all the test scripts
1275hardcoded the object IDs like t0000-basic.sh does, that defeats
1276the purpose of t0000-basic.sh, which is to isolate that level of
1277validation in one place. Your test also ends up needing
1278an update whenever the internals change, so do _not_
1279do it and leave the low level of validation to t0000-basic.sh.
1280
1281Test coverage
1282-------------
1283
1284You can use the coverage tests to find code paths that are not being
1285used or properly exercised yet.
1286
1287To do that, run the coverage target at the top-level (not in the t/
1288directory):
1289
1290 make coverage
1291
1292That'll compile Git with GCC's coverage arguments, and generate a test
1293report with gcov after the tests finish. Running the coverage tests
1294can take a while, since running the tests in parallel is incompatible
1295with GCC's coverage mode.
1296
1297After the tests have run you can generate a list of untested
1298functions:
1299
1300 make coverage-untested-functions
1301
1302You can also generate a detailed per-file HTML report using the
1303Devel::Cover module. To install it do:
1304
1305 # On Debian or Ubuntu:
1306 sudo aptitude install libdevel-cover-perl
1307
1308 # From the CPAN with cpanminus
1309 curl -L https://cpanmin.us/ | perl - --sudo --self-upgrade
1310 cpanm --sudo Devel::Cover
1311
1312Then, at the top-level:
1313
1314 make cover_db_html
1315
1316That'll generate a detailed cover report in the "cover_db_html"
1317directory, which you can then copy to a webserver, or inspect locally
1318in a browser.