]> git.ipfire.org Git - thirdparty/git.git/blob - t/README
clone: allow "--bare" with "-o"
[thirdparty/git.git] / t / README
1 Core Git Tests
2 ==============
3
4 This directory holds many test scripts for core Git tools. The
5 first part of this short document describes how to run the tests
6 and read their output.
7
8 When fixing the tools or adding enhancements, you are strongly
9 encouraged to add tests in this directory to cover what you are
10 trying to fix or enhance. The later part of this short document
11 describes how your test scripts should be organized.
12
13
14 Running Tests
15 -------------
16
17 The easiest way to run tests is to say "make". This runs all
18 the 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
35 Since the tests all output TAP (see http://testanything.org) they can
36 be run with any TAP harness. Here's an example of parallel testing
37 powered 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
47 prove 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
53 You can give DEFAULT_TEST_TARGET=prove on the make command (or define it
54 in config.mak) to cause "make test" to run tests under prove.
55 GIT_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
59 You 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
70 You 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
72 appropriately before running "make". Short options can be bundled, i.e.
73 '-d -v' is the same as '-dv'.
74
75 -v::
76 --verbose::
77 This makes the test more verbose. Specifically, the
78 command being run and their output if any are also
79 output.
80
81 --verbose-only=<pattern>::
82 Like --verbose, but the effect is limited to tests with
83 numbers matching <pattern>. The number matched against is
84 simply the running count of the test within the file.
85
86 -x::
87 Turn on shell tracing (i.e., `set -x`) during the tests
88 themselves. Implies `--verbose`.
89 Ignored in test scripts that set the variable 'test_untraceable'
90 to a non-empty value, unless it's run with a Bash version
91 supporting BASH_XTRACEFD, i.e. v4.1 or later.
92
93 -d::
94 --debug::
95 This may help the person who is developing a new test.
96 It causes the command defined with test_debug to run.
97 The "trash" directory (used to store all temporary data
98 during testing) is not deleted even if there are no
99 failed tests so that you can inspect its contents after
100 the test finished.
101
102 -i::
103 --immediate::
104 This causes the test to immediately exit upon the first
105 failed test. Cleanup commands requested with
106 test_when_finished are not executed if the test failed,
107 in order to keep the state for inspection by the tester
108 to diagnose the bug.
109
110 -l::
111 --long-tests::
112 This causes additional long-running tests to be run (where
113 available), for more exhaustive testing.
114
115 -r::
116 --run=<test-selector>::
117 Run only the subset of tests indicated by
118 <test-selector>. See section "Skipping Tests" below for
119 <test-selector> syntax.
120
121 --valgrind=<tool>::
122 Execute all Git binaries under valgrind tool <tool> and exit
123 with status 126 on errors (just like regular tests, this will
124 only stop the test script when running under -i).
125
126 Since it makes no sense to run the tests with --valgrind and
127 not see any output, this option implies --verbose. For
128 convenience, it also implies --tee.
129
130 <tool> defaults to 'memcheck', just like valgrind itself.
131 Other particularly useful choices include 'helgrind' and
132 'drd', but you may use any tool recognized by your valgrind
133 installation.
134
135 As a special case, <tool> can be 'memcheck-fast', which uses
136 memcheck but disables --track-origins. Use this if you are
137 running tests in bulk, to see if there are _any_ memory
138 issues.
139
140 Note that memcheck is run with the option --leak-check=no,
141 as the git process is short-lived and some errors are not
142 interesting. In order to run a single command under the same
143 conditions manually, you should set GIT_VALGRIND to point to
144 the 't/valgrind/' directory and use the commands under
145 't/valgrind/bin/'.
146
147 --valgrind-only=<pattern>::
148 Like --valgrind, but the effect is limited to tests with
149 numbers matching <pattern>. The number matched against is
150 simply the running count of the test within the file.
151
152 --tee::
153 In addition to printing the test output to the terminal,
154 write it to files named 't/test-results/$TEST_NAME.out'.
155 As the names depend on the tests' file names, it is safe to
156 run the tests with this option in parallel.
157
158 -V::
159 --verbose-log::
160 Write verbose output to the same logfile as `--tee`, but do
161 _not_ write it to stdout. Unlike `--tee --verbose`, this option
162 is safe to use when stdout is being consumed by a TAP parser
163 like `prove`. Implies `--tee` and `--verbose`.
164
165 --with-dashes::
166 By default tests are run without dashed forms of
167 commands (like git-commit) in the PATH (it only uses
168 wrappers from ../bin-wrappers). Use this option to include
169 the build directory (..) in the PATH, which contains all
170 the dashed forms of commands. This option is currently
171 implied by other options like --valgrind and
172 GIT_TEST_INSTALLED.
173
174 --no-bin-wrappers::
175 By default, the test suite uses the wrappers in
176 `../bin-wrappers/` to execute `git` and friends. With this option,
177 `../git` and friends are run directly. This is not recommended
178 in general, as the wrappers contain safeguards to ensure that no
179 files from an installed Git are used, but can speed up test runs
180 especially on platforms where running shell scripts is expensive
181 (most notably, Windows).
182
183 --root=<directory>::
184 Create "trash" directories used to store all temporary data during
185 testing under <directory>, instead of the t/ directory.
186 Using this option with a RAM-based filesystem (such as tmpfs)
187 can massively speed up the test suite.
188
189 --chain-lint::
190 --no-chain-lint::
191 If --chain-lint is enabled, the test harness will check each
192 test to make sure that it properly "&&-chains" all commands (so
193 that a failure in the middle does not go unnoticed by the final
194 exit code of the test). This check is performed in addition to
195 running the tests themselves. You may also enable or disable
196 this feature by setting the GIT_TEST_CHAIN_LINT environment
197 variable to "1" or "0", respectively.
198
199 A few test scripts disable some of the more advanced
200 chain-linting detection in the name of efficiency. You can
201 override this by setting the GIT_TEST_CHAIN_LINT_HARDER
202 environment variable to "1".
203
204 --stress::
205 Run the test script repeatedly in multiple parallel jobs until
206 one of them fails. Useful for reproducing rare failures in
207 flaky tests. The number of parallel jobs is, in order of
208 precedence: the value of the GIT_TEST_STRESS_LOAD
209 environment variable, or twice the number of available
210 processors (as shown by the 'getconf' utility), or 8.
211 Implies `--verbose -x --immediate` to get the most information
212 about the failure. Note that the verbose output of each test
213 job is saved to 't/test-results/$TEST_NAME.stress-<nr>.out',
214 and only the output of the failed test job is shown on the
215 terminal. The names of the trash directories get a
216 '.stress-<nr>' suffix, and the trash directory of the failed
217 test job is renamed to end with a '.stress-failed' suffix.
218
219 --stress-jobs=<N>::
220 Override the number of parallel jobs. Implies `--stress`.
221
222 --stress-limit=<N>::
223 When combined with --stress run the test script repeatedly
224 this many times in each of the parallel jobs or until one of
225 them fails, whichever comes first. Implies `--stress`.
226
227 You can also set the GIT_TEST_INSTALLED environment variable to
228 the bindir of an existing git installation to test that installation.
229 You still need to have built this git sandbox, from which various
230 test-* support programs, templates, and perl libraries are used.
231 If your installed git is incomplete, it will silently test parts of
232 your built version instead.
233
234 When using GIT_TEST_INSTALLED, you can also set GIT_TEST_EXEC_PATH to
235 override the location of the dashed-form subcommands (what
236 GIT_EXEC_PATH would be used for during normal operation).
237 GIT_TEST_EXEC_PATH defaults to `$GIT_TEST_INSTALLED/git --exec-path`.
238
239
240 Skipping Tests
241 --------------
242
243 In some environments, certain tests have no way of succeeding
244 due to platform limitation, such as lack of 'unzip' program, or
245 filesystem that do not allow arbitrary sequence of non-NUL bytes
246 as pathnames.
247
248 You should be able to say something like
249
250 $ GIT_SKIP_TESTS=t9200.8 sh ./t9200-git-cvsexport-commit.sh
251
252 and even:
253
254 $ GIT_SKIP_TESTS='t[0-4]??? t91?? t9200.8' make
255
256 to omit such tests. The value of the environment variable is a
257 SP separated list of patterns that tells which tests to skip,
258 and either can match the "t[0-9]{4}" part to skip the whole
259 test, or t[0-9]{4} followed by ".$number" to say which
260 particular test to skip.
261
262 For an individual test suite --run could be used to specify that
263 only some tests should be run or that some tests should be
264 excluded from a run.
265
266 The argument for --run, <test-selector>, is a list of description
267 substrings or globs or individual test numbers or ranges with an
268 optional negation prefix (of '!') that define what tests in a test
269 suite to include (or exclude, if negated) in the run. A range is two
270 numbers separated with a dash and matches a range of tests with both
271 ends been included. You may omit the first or the second number to
272 mean "from the first test" or "up to the very last test" respectively.
273
274 The argument to --run is split on commas into separate strings,
275 numbers, and ranges, and picks all tests that match any of the
276 individual selection criteria. If the substring of the description
277 text that you want to match includes a comma, use the glob character
278 '?' instead. For example --run='rebase,merge?cherry-pick' would match
279 on all tests that match either the glob *rebase* or the glob
280 *merge?cherry-pick*.
281
282 If --run starts with an unprefixed number or range the initial
283 set of tests to run is empty. If the first item starts with '!'
284 all the tests are added to the initial set. After initial set is
285 determined every test number or range is added or excluded from
286 the set one by one, from left to right.
287
288 For example, to run only tests up to a specific test (21), one
289 could do this:
290
291 $ sh ./t9200-git-cvsexport-commit.sh --run='1-21'
292
293 or this:
294
295 $ sh ./t9200-git-cvsexport-commit.sh --run='-21'
296
297 Common case is to run several setup tests (1, 2, 3) and then a
298 specific test (21) that relies on that setup:
299
300 $ sh ./t9200-git-cvsexport-commit.sh --run='1,2,3,21'
301
302 or:
303
304 $ sh ./t9200-git-cvsexport-commit.sh --run=1,2,3,21
305
306 or:
307
308 $ sh ./t9200-git-cvsexport-commit.sh --run='-3,21'
309
310 As noted above, the test set is built by going through the items
311 from left to right, so this:
312
313 $ sh ./t9200-git-cvsexport-commit.sh --run='1-4,!3'
314
315 will run tests 1, 2, and 4. Items that come later have higher
316 precedence. It means that this:
317
318 $ sh ./t9200-git-cvsexport-commit.sh --run='!3,1-4'
319
320 would just run tests from 1 to 4, including 3.
321
322 You may use negation with ranges. The following will run all
323 test in the test suite except from 7 up to 11:
324
325 $ sh ./t9200-git-cvsexport-commit.sh --run='!7-11'
326
327 Sometimes there may be multiple tests with e.g. "setup" in their name
328 that are needed and rather than figuring out the number for all of them
329 we can just use "setup" as a substring/glob to match against the test
330 description:
331
332 $ sh ./t0050-filesystem.sh --run=setup,9-11
333
334 or one could select both the setup tests and the rename ones (assuming all
335 relevant tests had those words in their descriptions):
336
337 $ sh ./t0050-filesystem.sh --run=setup,rename
338
339 Some tests in a test suite rely on the previous tests performing
340 certain actions, specifically some tests are designated as
341 "setup" test, so you cannot _arbitrarily_ disable one test and
342 expect the rest to function correctly.
343
344 --run is mostly useful when you want to focus on a specific test
345 and know what setup is needed for it. Or when you want to run
346 everything up to a certain test.
347
348
349 Running tests with special setups
350 ---------------------------------
351
352 The whole test suite could be run to test some special features
353 that cannot be easily covered by a few specific test cases. These
354 could be enabled by running the test suite with correct GIT_TEST_
355 environment set.
356
357 GIT_TEST_FAIL_PREREQS=<boolean> fails all prerequisites. This is
358 useful for discovering issues with the tests where say a later test
359 implicitly depends on an optional earlier test.
360
361 There's a "FAIL_PREREQS" prerequisite that can be used to test for
362 whether this mode is active, and e.g. skip some tests that are hard to
363 refactor to deal with it. The "SYMLINKS" prerequisite is currently
364 excluded as so much relies on it, but this might change in the future.
365
366 GIT_TEST_SPLIT_INDEX=<boolean> forces split-index mode on the whole
367 test suite. Accept any boolean values that are accepted by git-config.
368
369 GIT_TEST_PASSING_SANITIZE_LEAK=<boolean> when compiled with
370 SANITIZE=leak will run only those tests that have whitelisted
371 themselves as passing with no memory leaks. Tests can be whitelisted
372 by setting "TEST_PASSES_SANITIZE_LEAK=true" before sourcing
373 "test-lib.sh" itself at the top of the test script. This test mode is
374 used by the "linux-leaks" CI target.
375
376 GIT_TEST_PROTOCOL_VERSION=<n>, when set, makes 'protocol.version'
377 default to n.
378
379 GIT_TEST_FULL_IN_PACK_ARRAY=<boolean> exercises the uncommon
380 pack-objects code path where there are more than 1024 packs even if
381 the actual number of packs in repository is below this limit. Accept
382 any boolean values that are accepted by git-config.
383
384 GIT_TEST_OE_SIZE=<n> exercises the uncommon pack-objects code path
385 where we do not cache object size in memory and read it from existing
386 packs on demand. This normally only happens when the object size is
387 over 2GB. This variable forces the code path on any object larger than
388 <n> bytes.
389
390 GIT_TEST_OE_DELTA_SIZE=<n> exercises the uncommon pack-objects code
391 path where deltas larger than this limit require extra memory
392 allocation for bookkeeping.
393
394 GIT_TEST_VALIDATE_INDEX_CACHE_ENTRIES=<boolean> checks that cache-tree
395 records are valid when the index is written out or after a merge. This
396 is mostly to catch missing invalidation. Default is true.
397
398 GIT_TEST_COMMIT_GRAPH=<boolean>, when true, forces the commit-graph to
399 be written after every 'git commit' command, and overrides the
400 'core.commitGraph' setting to true.
401
402 GIT_TEST_COMMIT_GRAPH_CHANGED_PATHS=<boolean>, when true, forces
403 commit-graph write to compute and write changed path Bloom filters for
404 every 'git commit-graph write', as if the `--changed-paths` option was
405 passed in.
406
407 GIT_TEST_FSMONITOR=$PWD/t7519/fsmonitor-all exercises the fsmonitor
408 code paths for utilizing a (hook based) file system monitor to speed up
409 detecting new or changed files.
410
411 GIT_TEST_INDEX_VERSION=<n> exercises the index read/write code path
412 for the index version specified. Can be set to any valid version
413 (currently 2, 3, or 4).
414
415 GIT_TEST_PACK_SPARSE=<boolean> if disabled will default the pack-objects
416 builtin to use the non-sparse object walk. This can still be overridden by
417 the --sparse command-line argument.
418
419 GIT_TEST_PRELOAD_INDEX=<boolean> exercises the preload-index code path
420 by overriding the minimum number of cache entries required per thread.
421
422 GIT_TEST_ADD_I_USE_BUILTIN=<boolean>, when false, disables the
423 built-in version of git add -i. See 'add.interactive.useBuiltin' in
424 git-config(1).
425
426 GIT_TEST_INDEX_THREADS=<n> enables exercising the multi-threaded loading
427 of the index for the whole test suite by bypassing the default number of
428 cache entries and thread minimums. Setting this to 1 will make the
429 index loading single threaded.
430
431 GIT_TEST_MULTI_PACK_INDEX=<boolean>, when true, forces the multi-pack-
432 index to be written after every 'git repack' command, and overrides the
433 'core.multiPackIndex' setting to true.
434
435 GIT_TEST_MULTI_PACK_INDEX_WRITE_BITMAP=<boolean>, when true, sets the
436 '--bitmap' option on all invocations of 'git multi-pack-index write',
437 and ignores pack-objects' '--write-bitmap-index'.
438
439 GIT_TEST_SIDEBAND_ALL=<boolean>, when true, overrides the
440 'uploadpack.allowSidebandAll' setting to true, and when false, forces
441 fetch-pack to not request sideband-all (even if the server advertises
442 sideband-all).
443
444 GIT_TEST_DISALLOW_ABBREVIATED_OPTIONS=<boolean>, when true (which is
445 the default when running tests), errors out when an abbreviated option
446 is used.
447
448 GIT_TEST_DEFAULT_HASH=<hash-algo> specifies which hash algorithm to
449 use in the test scripts. Recognized values for <hash-algo> are "sha1"
450 and "sha256".
451
452 GIT_TEST_WRITE_REV_INDEX=<boolean>, when true enables the
453 'pack.writeReverseIndex' setting.
454
455 GIT_TEST_SPARSE_INDEX=<boolean>, when true enables index writes to use the
456 sparse-index format by default.
457
458 GIT_TEST_CHECKOUT_WORKERS=<n> overrides the 'checkout.workers' setting
459 to <n> and 'checkout.thresholdForParallelism' to 0, forcing the
460 execution of the parallel-checkout code.
461
462 GIT_TEST_FATAL_REGISTER_SUBMODULE_ODB=<boolean>, when true, makes
463 registering submodule ODBs as alternates a fatal action. Support for
464 this environment variable can be removed once the migration to
465 explicitly providing repositories when accessing submodule objects is
466 complete or needs to be abandoned for whatever reason (in which case the
467 migrated codepaths still retain their performance benefits).
468
469 GIT_TEST_REQUIRE_PREREQ=<list> allows specifying a space separated list of
470 prereqs that are required to succeed. If a prereq in this list is triggered by
471 a test and then fails then the whole test run will abort. This can help to make
472 sure the expected tests are executed and not silently skipped when their
473 dependency breaks or is simply not present in a new environment.
474
475 Naming Tests
476 ------------
477
478 The test files are named as:
479
480 tNNNN-commandname-details.sh
481
482 where N is a decimal digit.
483
484 First digit tells the family:
485
486 0 - the absolute basics and global stuff
487 1 - the basic commands concerning database
488 2 - the basic commands concerning the working tree
489 3 - the other basic commands (e.g. ls-files)
490 4 - the diff commands
491 5 - the pull and exporting commands
492 6 - the revision tree commands (even e.g. merge-base)
493 7 - the porcelainish commands concerning the working tree
494 8 - the porcelainish commands concerning forensics
495 9 - the git tools
496
497 Second digit tells the particular command we are testing.
498
499 Third digit (optionally) tells the particular switch or group of switches
500 we are testing.
501
502 If you create files under t/ directory (i.e. here) that is not
503 the top-level test script, never name the file to match the above
504 pattern. The Makefile here considers all such files as the
505 top-level test script and tries to run all of them. Care is
506 especially needed if you are creating a common test library
507 file, similar to test-lib.sh, because such a library file may
508 not be suitable for standalone execution.
509
510
511 Writing Tests
512 -------------
513
514 The test script is written as a shell script. It should start
515 with the standard "#!/bin/sh", and an
516 assignment to variable 'test_description', like this:
517
518 #!/bin/sh
519
520 test_description='xxx test (option --frotz)
521
522 This test registers the following structure in the cache
523 and tries to run git-ls-files with option --frotz.'
524
525
526 Source 'test-lib.sh'
527 --------------------
528
529 After assigning test_description, the test script should source
530 test-lib.sh like this:
531
532 . ./test-lib.sh
533
534 This test harness library does the following things:
535
536 - If the script is invoked with command line argument --help
537 (or -h), it shows the test_description and exits.
538
539 - Creates an empty test directory with an empty .git/objects database
540 and chdir(2) into it. This directory is 't/trash
541 directory.$test_name_without_dotsh', with t/ subject to change by
542 the --root option documented above, and a '.stress-<N>' suffix
543 appended by the --stress option.
544
545 - Defines standard test helper functions for your scripts to
546 use. These functions are designed to make all scripts behave
547 consistently when command line arguments --verbose (or -v),
548 --debug (or -d), and --immediate (or -i) is given.
549
550 Recommended style
551 -----------------
552 Here are some recommented styles when writing test case.
553
554 - Keep test title the same line with test helper function itself.
555
556 Take test_expect_success helper for example, write it like:
557
558 test_expect_success 'test title' '
559 ... test body ...
560 '
561
562 Instead of:
563
564 test_expect_success \
565 'test title' \
566 '... test body ...'
567
568
569 - End the line with a single quote.
570
571 - Indent the body of here-document, and use "<<-" instead of "<<"
572 to strip leading TABs used for indentation:
573
574 test_expect_success 'test something' '
575 cat >expect <<-\EOF &&
576 one
577 two
578 three
579 EOF
580 test_something > actual &&
581 test_cmp expect actual
582 '
583
584 Instead of:
585
586 test_expect_success 'test something' '
587 cat >expect <<\EOF &&
588 one
589 two
590 three
591 EOF
592 test_something > actual &&
593 test_cmp expect actual
594 '
595
596 - Quote or escape the EOF delimiter that begins a here-document if
597 there is no parameter and other expansion in it, to signal readers
598 that they can skim it more casually:
599
600 cmd <<-\EOF
601 literal here-document text without any expansion
602 EOF
603
604
605 Do's & don'ts
606 -------------
607
608 Here are a few examples of things you probably should and shouldn't do
609 when writing tests.
610
611 Here are the "do's:"
612
613 - Put all code inside test_expect_success and other assertions.
614
615 Even code that isn't a test per se, but merely some setup code
616 should be inside a test assertion.
617
618 - Chain your test assertions
619
620 Write test code like this:
621
622 git merge foo &&
623 git push bar &&
624 test ...
625
626 Instead of:
627
628 git merge hla
629 git push gh
630 test ...
631
632 That way all of the commands in your tests will succeed or fail. If
633 you must ignore the return value of something, consider using a
634 helper function (e.g. use sane_unset instead of unset, in order
635 to avoid unportable return value for unsetting a variable that was
636 already unset), or prepending the command with test_might_fail or
637 test_must_fail.
638
639 - Check the test coverage for your tests. See the "Test coverage"
640 below.
641
642 Don't blindly follow test coverage metrics; if a new function you added
643 doesn't have any coverage, then you're probably doing something wrong,
644 but having 100% coverage doesn't necessarily mean that you tested
645 everything.
646
647 Tests that are likely to smoke out future regressions are better
648 than tests that just inflate the coverage metrics.
649
650 - When a test checks for an absolute path that a git command generated,
651 construct the expected value using $(pwd) rather than $PWD,
652 $TEST_DIRECTORY, or $TRASH_DIRECTORY. It makes a difference on
653 Windows, where the shell (MSYS bash) mangles absolute path names.
654 For details, see the commit message of 4114156ae9.
655
656 - Remember that inside the <script> part, the standard output and
657 standard error streams are discarded, and the test harness only
658 reports "ok" or "not ok" to the end user running the tests. Under
659 --verbose, they are shown to help debug the tests.
660
661 - Be careful when you loop
662
663 You may need to verify multiple things in a loop, but the
664 following does not work correctly:
665
666 test_expect_success 'test three things' '
667 for i in one two three
668 do
669 test_something "$i"
670 done &&
671 test_something_else
672 '
673
674 Because the status of the loop itself is the exit status of the
675 test_something in the last round, the loop does not fail when
676 "test_something" for "one" or "two" fails. This is not what you
677 want.
678
679 Instead, you can break out of the loop immediately when you see a
680 failure. Because all test_expect_* snippets are executed inside
681 a function, "return 1" can be used to fail the test immediately
682 upon a failure:
683
684 test_expect_success 'test three things' '
685 for i in one two three
686 do
687 test_something "$i" || return 1
688 done &&
689 test_something_else
690 '
691
692 Note that we still &&-chain the loop to propagate failures from
693 earlier commands.
694
695
696 And here are the "don'ts:"
697
698 - Don't exit() within a <script> part.
699
700 The harness will catch this as a programming error of the test.
701 Use test_done instead if you need to stop the tests early (see
702 "Skipping tests" below).
703
704 - Don't use '! git cmd' when you want to make sure the git command
705 exits with failure in a controlled way by calling "die()". Instead,
706 use 'test_must_fail git cmd'. This will signal a failure if git
707 dies in an unexpected way (e.g. segfault).
708
709 On the other hand, don't use test_must_fail for running regular
710 platform commands; just use '! cmd'. We are not in the business
711 of verifying that the world given to us sanely works.
712
713 - Don't feed the output of a git command to a pipe, as in:
714
715 git -C repo ls-files |
716 xargs -n 1 basename |
717 grep foo
718
719 which will discard git's exit code and may mask a crash. In the
720 above example, all exit codes are ignored except grep's.
721
722 Instead, write the output of that command to a temporary
723 file with ">" or assign it to a variable with "x=$(git ...)" rather
724 than pipe it.
725
726 - Don't use command substitution in a way that discards git's exit
727 code. When assigning to a variable, the exit code is not discarded,
728 e.g.:
729
730 x=$(git cat-file -p $sha) &&
731 ...
732
733 is OK because a crash in "git cat-file" will cause the "&&" chain
734 to fail, but:
735
736 test "refs/heads/foo" = "$(git symbolic-ref HEAD)"
737
738 is not OK and a crash in git could go undetected.
739
740 - Don't use perl without spelling it as "$PERL_PATH". This is to help
741 our friends on Windows where the platform Perl often adds CR before
742 the end of line, and they bundle Git with a version of Perl that
743 does not do so, whose path is specified with $PERL_PATH. Note that we
744 provide a "perl" function which uses $PERL_PATH under the hood, so
745 you do not need to worry when simply running perl in the test scripts
746 (but you do, for example, on a shebang line or in a sub script
747 created via "write_script").
748
749 - Don't use sh without spelling it as "$SHELL_PATH", when the script
750 can be misinterpreted by broken platform shell (e.g. Solaris).
751
752 - Don't chdir around in tests. It is not sufficient to chdir to
753 somewhere and then chdir back to the original location later in
754 the test, as any intermediate step can fail and abort the test,
755 causing the next test to start in an unexpected directory. Do so
756 inside a subshell if necessary.
757
758 - Don't save and verify the standard error of compound commands, i.e.
759 group commands, subshells, and shell functions (except test helper
760 functions like 'test_must_fail') like this:
761
762 ( cd dir && git cmd ) 2>error &&
763 test_cmp expect error
764
765 When running the test with '-x' tracing, then the trace of commands
766 executed in the compound command will be included in standard error
767 as well, quite possibly throwing off the subsequent checks examining
768 the output. Instead, save only the relevant git command's standard
769 error:
770
771 ( cd dir && git cmd 2>../error ) &&
772 test_cmp expect error
773
774 - Don't break the TAP output
775
776 The raw output from your test may be interpreted by a TAP harness. TAP
777 harnesses will ignore everything they don't know about, but don't step
778 on their toes in these areas:
779
780 - Don't print lines like "$x..$y" where $x and $y are integers.
781
782 - Don't print lines that begin with "ok" or "not ok".
783
784 TAP harnesses expect a line that begins with either "ok" and "not
785 ok" to signal a test passed or failed (and our harness already
786 produces such lines), so your script shouldn't emit such lines to
787 their output.
788
789 You can glean some further possible issues from the TAP grammar
790 (see https://metacpan.org/pod/TAP::Parser::Grammar#TAP-GRAMMAR)
791 but the best indication is to just run the tests with prove(1),
792 it'll complain if anything is amiss.
793
794
795 Skipping tests
796 --------------
797
798 If you need to skip tests you should do so by using the three-arg form
799 of the test_* functions (see the "Test harness library" section
800 below), e.g.:
801
802 test_expect_success PERL 'I need Perl' '
803 perl -e "hlagh() if unf_unf()"
804 '
805
806 The advantage of skipping tests like this is that platforms that don't
807 have the PERL and other optional dependencies get an indication of how
808 many tests they're missing.
809
810 If the test code is too hairy for that (i.e. does a lot of setup work
811 outside test assertions) you can also skip all remaining tests by
812 setting skip_all and immediately call test_done:
813
814 if ! test_have_prereq PERL
815 then
816 skip_all='skipping perl interface tests, perl not available'
817 test_done
818 fi
819
820 The string you give to skip_all will be used as an explanation for why
821 the test was skipped.
822
823 End with test_done
824 ------------------
825
826 Your script will be a sequence of tests, using helper functions
827 from the test harness library. At the end of the script, call
828 'test_done'.
829
830
831 Test harness library
832 --------------------
833
834 There are a handful helper functions defined in the test harness
835 library for your script to use. Some of them are listed below;
836 see test-lib-functions.sh for the full list and their options.
837
838 - test_expect_success [<prereq>] <message> <script>
839
840 Usually takes two strings as parameters, and evaluates the
841 <script>. If it yields success, test is considered
842 successful. <message> should state what it is testing.
843
844 Example:
845
846 test_expect_success \
847 'git-write-tree should be able to write an empty tree.' \
848 'tree=$(git-write-tree)'
849
850 If you supply three parameters the first will be taken to be a
851 prerequisite; see the test_set_prereq and test_have_prereq
852 documentation below:
853
854 test_expect_success TTY 'git --paginate rev-list uses a pager' \
855 ' ... '
856
857 You can also supply a comma-separated list of prerequisites, in the
858 rare case where your test depends on more than one:
859
860 test_expect_success PERL,PYTHON 'yo dawg' \
861 ' test $(perl -E 'print eval "1 +" . qx[python -c "print 2"]') == "4" '
862
863 - test_expect_failure [<prereq>] <message> <script>
864
865 This is NOT the opposite of test_expect_success, but is used
866 to mark a test that demonstrates a known breakage. Unlike
867 the usual test_expect_success tests, which say "ok" on
868 success and "FAIL" on failure, this will say "FIXED" on
869 success and "still broken" on failure. Failures from these
870 tests won't cause -i (immediate) to stop.
871
872 Like test_expect_success this function can optionally use a three
873 argument invocation with a prerequisite as the first argument.
874
875 - test_debug <script>
876
877 This takes a single argument, <script>, and evaluates it only
878 when the test script is started with --debug command line
879 argument. This is primarily meant for use during the
880 development of a new test script.
881
882 - debug [options] <git-command>
883
884 Run a git command inside a debugger. This is primarily meant for
885 use when debugging a failing test script. With '-t', use your
886 original TERM instead of test-lib.sh's "dumb", so that your
887 debugger interface has colors.
888
889 - test_done
890
891 Your test script must have test_done at the end. Its purpose
892 is to summarize successes and failures in the test script and
893 exit with an appropriate error code.
894
895 - test_tick
896
897 Make commit and tag names consistent by setting the author and
898 committer times to defined state. Subsequent calls will
899 advance the times by a fixed amount.
900
901 - test_commit <message> [<filename> [<contents>]]
902
903 Creates a commit with the given message, committing the given
904 file with the given contents (default for both is to reuse the
905 message string), and adds a tag (again reusing the message
906 string as name). Calls test_tick to make the SHA-1s
907 reproducible.
908
909 - test_merge <message> <commit-or-tag>
910
911 Merges the given rev using the given message. Like test_commit,
912 creates a tag and calls test_tick before committing.
913
914 - test_set_prereq <prereq>
915
916 Set a test prerequisite to be used later with test_have_prereq. The
917 test-lib will set some prerequisites for you, see the
918 "Prerequisites" section below for a full list of these.
919
920 Others you can set yourself and use later with either
921 test_have_prereq directly, or the three argument invocation of
922 test_expect_success and test_expect_failure.
923
924 - test_have_prereq <prereq>
925
926 Check if we have a prerequisite previously set with test_set_prereq.
927 The most common way to use this explicitly (as opposed to the
928 implicit use when an argument is passed to test_expect_*) is to skip
929 all the tests at the start of the test script if we don't have some
930 essential prerequisite:
931
932 if ! test_have_prereq PERL
933 then
934 skip_all='skipping perl interface tests, perl not available'
935 test_done
936 fi
937
938 - test_external [<prereq>] <message> <external> <script>
939
940 Execute a <script> with an <external> interpreter (like perl). This
941 was added for tests like t9700-perl-git.sh which do most of their
942 work in an external test script.
943
944 test_external \
945 'GitwebCache::*FileCache*' \
946 perl "$TEST_DIRECTORY"/t9503/test_cache_interface.pl
947
948 If the test is outputting its own TAP you should set the
949 test_external_has_tap variable somewhere before calling the first
950 test_external* function. See t9700-perl-git.sh for an example.
951
952 # The external test will outputs its own plan
953 test_external_has_tap=1
954
955 - test_external_without_stderr [<prereq>] <message> <external> <script>
956
957 Like test_external but fail if there's any output on stderr,
958 instead of checking the exit code.
959
960 test_external_without_stderr \
961 'Perl API' \
962 perl "$TEST_DIRECTORY"/t9700/test.pl
963
964 - test_expect_code <exit-code> <command>
965
966 Run a command and ensure that it exits with the given exit code.
967 For example:
968
969 test_expect_success 'Merge with d/f conflicts' '
970 test_expect_code 1 git merge "merge msg" B master
971 '
972
973 - test_must_fail [<options>] <git-command>
974
975 Run a git command and ensure it fails in a controlled way. Use
976 this instead of "! <git-command>". When git-command dies due to a
977 segfault, test_must_fail diagnoses it as an error; "! <git-command>"
978 treats it as just another expected failure, which would let such a
979 bug go unnoticed.
980
981 Accepts the following options:
982
983 ok=<signal-name>[,<...>]:
984 Don't treat an exit caused by the given signal as error.
985 Multiple signals can be specified as a comma separated list.
986 Currently recognized signal names are: sigpipe, success.
987 (Don't use 'success', use 'test_might_fail' instead.)
988
989 - test_might_fail [<options>] <git-command>
990
991 Similar to test_must_fail, but tolerate success, too. Use this
992 instead of "<git-command> || :" to catch failures due to segv.
993
994 Accepts the same options as test_must_fail.
995
996 - test_cmp <expected> <actual>
997
998 Check whether the content of the <actual> file matches the
999 <expected> file. This behaves like "cmp" but produces more
1000 helpful output when the test is run with "-v" option.
1001
1002 - test_cmp_rev <expected> <actual>
1003
1004 Check whether the <expected> rev points to the same commit as the
1005 <actual> rev.
1006
1007 - test_line_count (= | -lt | -ge | ...) <length> <file>
1008
1009 Check whether a file has the length it is expected to.
1010
1011 - test_path_is_file <path>
1012 test_path_is_dir <path>
1013 test_path_is_missing <path>
1014
1015 Check if the named path is a file, if the named path is a
1016 directory, or if the named path does not exist, respectively,
1017 and fail otherwise.
1018
1019 - test_when_finished <script>
1020
1021 Prepend <script> to a list of commands to run to clean up
1022 at the end of the current test. If some clean-up command
1023 fails, the test will not pass.
1024
1025 Example:
1026
1027 test_expect_success 'branch pointing to non-commit' '
1028 git rev-parse HEAD^{tree} >.git/refs/heads/invalid &&
1029 test_when_finished "git update-ref -d refs/heads/invalid" &&
1030 ...
1031 '
1032
1033 - test_atexit <script>
1034
1035 Prepend <script> to a list of commands to run unconditionally to
1036 clean up before the test script exits, e.g. to stop a daemon:
1037
1038 test_expect_success 'test git daemon' '
1039 git daemon &
1040 daemon_pid=$! &&
1041 test_atexit 'kill $daemon_pid' &&
1042 hello world
1043 '
1044
1045 The commands will be executed before the trash directory is removed,
1046 i.e. the atexit commands will still be able to access any pidfiles or
1047 socket files.
1048
1049 Note that these commands will be run even when a test script run
1050 with '--immediate' fails. Be careful with your atexit commands to
1051 minimize any changes to the failed state.
1052
1053 - test_write_lines <lines>
1054
1055 Write <lines> on standard output, one line per argument.
1056 Useful to prepare multi-line files in a compact form.
1057
1058 Example:
1059
1060 test_write_lines a b c d e f g >foo
1061
1062 Is a more compact equivalent of:
1063 cat >foo <<-EOF
1064 a
1065 b
1066 c
1067 d
1068 e
1069 f
1070 g
1071 EOF
1072
1073
1074 - test_pause [options]
1075
1076 This command is useful for writing and debugging tests and must be
1077 removed before submitting. It halts the execution of the test and
1078 spawns a shell in the trash directory. Exit the shell to continue
1079 the test. Example:
1080
1081 test_expect_success 'test' '
1082 git do-something >actual &&
1083 test_pause &&
1084 test_cmp expected actual
1085 '
1086
1087 - test_ln_s_add <path1> <path2>
1088
1089 This function helps systems whose filesystem does not support symbolic
1090 links. Use it to add a symbolic link entry to the index when it is not
1091 important that the file system entry is a symbolic link, i.e., instead
1092 of the sequence
1093
1094 ln -s foo bar &&
1095 git add bar
1096
1097 Sometimes it is possible to split a test in a part that does not need
1098 the symbolic link in the file system and a part that does; then only
1099 the latter part need be protected by a SYMLINKS prerequisite (see below).
1100
1101 - test_oid_init
1102
1103 This function loads facts and useful object IDs related to the hash
1104 algorithm(s) in use from the files in t/oid-info.
1105
1106 - test_oid_cache
1107
1108 This function reads per-hash algorithm information from standard
1109 input (usually a heredoc) in the format described in
1110 t/oid-info/README. This is useful for test-specific values, such as
1111 object IDs, which must vary based on the hash algorithm.
1112
1113 Certain fixed values, such as hash sizes and common placeholder
1114 object IDs, can be loaded with test_oid_init (described above).
1115
1116 - test_oid <key>
1117
1118 This function looks up a value for the hash algorithm in use, based
1119 on the key given. The value must have been loaded using
1120 test_oid_init or test_oid_cache. Providing an unknown key is an
1121 error.
1122
1123 - yes [<string>]
1124
1125 This is often seen in modern UNIX but some platforms lack it, so
1126 the test harness overrides the platform implementation with a
1127 more limited one. Use this only when feeding a handful lines of
1128 output to the downstream---unlike the real version, it generates
1129 only up to 99 lines.
1130
1131 - test_bool_env <env-variable-name> <default-value>
1132
1133 Given the name of an environment variable with a bool value,
1134 normalize its value to a 0 (true) or 1 (false or empty string)
1135 return code. Return with code corresponding to the given default
1136 value if the variable is unset.
1137 Abort the test script if either the value of the variable or the
1138 default are not valid bool values.
1139
1140
1141 Prerequisites
1142 -------------
1143
1144 These are the prerequisites that the test library predefines with
1145 test_have_prereq.
1146
1147 See the prereq argument to the test_* functions in the "Test harness
1148 library" section above and the "test_have_prereq" function for how to
1149 use these, and "test_set_prereq" for how to define your own.
1150
1151 - PYTHON
1152
1153 Git wasn't compiled with NO_PYTHON=YesPlease. Wrap any tests that
1154 need Python with this.
1155
1156 - PERL
1157
1158 Git wasn't compiled with NO_PERL=YesPlease.
1159
1160 Even without the PERL prerequisite, tests can assume there is a
1161 usable perl interpreter at $PERL_PATH, though it need not be
1162 particularly modern.
1163
1164 - POSIXPERM
1165
1166 The filesystem supports POSIX style permission bits.
1167
1168 - BSLASHPSPEC
1169
1170 Backslashes in pathspec are not directory separators. This is not
1171 set on Windows. See 6fd1106a for details.
1172
1173 - EXECKEEPSPID
1174
1175 The process retains the same pid across exec(2). See fb9a2bea for
1176 details.
1177
1178 - PIPE
1179
1180 The filesystem we're on supports creation of FIFOs (named pipes)
1181 via mkfifo(1).
1182
1183 - SYMLINKS
1184
1185 The filesystem we're on supports symbolic links. E.g. a FAT
1186 filesystem doesn't support these. See 704a3143 for details.
1187
1188 - SANITY
1189
1190 Test is not run by root user, and an attempt to write to an
1191 unwritable file is expected to fail correctly.
1192
1193 - PCRE
1194
1195 Git was compiled with support for PCRE. Wrap any tests
1196 that use git-grep --perl-regexp or git-grep -P in these.
1197
1198 - CASE_INSENSITIVE_FS
1199
1200 Test is run on a case insensitive file system.
1201
1202 - UTF8_NFD_TO_NFC
1203
1204 Test is run on a filesystem which converts decomposed utf-8 (nfd)
1205 to precomposed utf-8 (nfc).
1206
1207 - PTHREADS
1208
1209 Git wasn't compiled with NO_PTHREADS=YesPlease.
1210
1211 - REFFILES
1212
1213 Test is specific to packed/loose ref storage, and should be
1214 disabled for other ref storage backends
1215
1216
1217 Tips for Writing Tests
1218 ----------------------
1219
1220 As with any programming projects, existing programs are the best
1221 source of the information. However, do _not_ emulate
1222 t0000-basic.sh when writing your tests. The test is special in
1223 that it tries to validate the very core of Git. For example, it
1224 knows that there will be 256 subdirectories under .git/objects/,
1225 and it knows that the object ID of an empty tree is a certain
1226 40-byte string. This is deliberately done so in t0000-basic.sh
1227 because the things the very basic core test tries to achieve is
1228 to serve as a basis for people who are changing the Git internals
1229 drastically. For these people, after making certain changes,
1230 not seeing failures from the basic test _is_ a failure. And
1231 such drastic changes to the core Git that even changes these
1232 otherwise supposedly stable object IDs should be accompanied by
1233 an update to t0000-basic.sh.
1234
1235 However, other tests that simply rely on basic parts of the core
1236 Git working properly should not have that level of intimate
1237 knowledge of the core Git internals. If all the test scripts
1238 hardcoded the object IDs like t0000-basic.sh does, that defeats
1239 the purpose of t0000-basic.sh, which is to isolate that level of
1240 validation in one place. Your test also ends up needing
1241 updating when such a change to the internal happens, so do _not_
1242 do it and leave the low level of validation to t0000-basic.sh.
1243
1244 Test coverage
1245 -------------
1246
1247 You can use the coverage tests to find code paths that are not being
1248 used or properly exercised yet.
1249
1250 To do that, run the coverage target at the top-level (not in the t/
1251 directory):
1252
1253 make coverage
1254
1255 That'll compile Git with GCC's coverage arguments, and generate a test
1256 report with gcov after the tests finish. Running the coverage tests
1257 can take a while, since running the tests in parallel is incompatible
1258 with GCC's coverage mode.
1259
1260 After the tests have run you can generate a list of untested
1261 functions:
1262
1263 make coverage-untested-functions
1264
1265 You can also generate a detailed per-file HTML report using the
1266 Devel::Cover module. To install it do:
1267
1268 # On Debian or Ubuntu:
1269 sudo aptitude install libdevel-cover-perl
1270
1271 # From the CPAN with cpanminus
1272 curl -L http://cpanmin.us | perl - --sudo --self-upgrade
1273 cpanm --sudo Devel::Cover
1274
1275 Then, at the top-level:
1276
1277 make cover_db_html
1278
1279 That'll generate a detailed cover report in the "cover_db_html"
1280 directory, which you can then copy to a webserver, or inspect locally
1281 in a browser.