Merge branch 'maint-1.6.0' into maint-1.6.1
[git/git.git] / Documentation / git-bisect.txt
1 git-bisect(1)
2 =============
3
4 NAME
5 ----
6 git-bisect - Find the change that introduced a bug by binary search
7
8
9 SYNOPSIS
10 --------
11 'git bisect' <subcommand> <options>
12
13 DESCRIPTION
14 -----------
15 The command takes various subcommands, and different options depending
16 on the subcommand:
17
18 git bisect help
19 git bisect start [<bad> [<good>...]] [--] [<paths>...]
20 git bisect bad [<rev>]
21 git bisect good [<rev>...]
22 git bisect skip [(<rev>|<range>)...]
23 git bisect reset [<branch>]
24 git bisect visualize
25 git bisect replay <logfile>
26 git bisect log
27 git bisect run <cmd>...
28
29 This command uses 'git rev-list --bisect' to help drive the
30 binary search process to find which change introduced a bug, given an
31 old "good" commit object name and a later "bad" commit object name.
32
33 Getting help
34 ~~~~~~~~~~~~
35
36 Use "git bisect" to get a short usage description, and "git bisect
37 help" or "git bisect -h" to get a long usage description.
38
39 Basic bisect commands: start, bad, good
40 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
41
42 The way you use it is:
43
44 ------------------------------------------------
45 $ git bisect start
46 $ git bisect bad # Current version is bad
47 $ git bisect good v2.6.13-rc2 # v2.6.13-rc2 was the last version
48 # tested that was good
49 ------------------------------------------------
50
51 When you give at least one bad and one good versions, it will bisect
52 the revision tree and say something like:
53
54 ------------------------------------------------
55 Bisecting: 675 revisions left to test after this
56 ------------------------------------------------
57
58 and check out the state in the middle. Now, compile that kernel, and
59 boot it. Now, let's say that this booted kernel works fine, then just
60 do
61
62 ------------------------------------------------
63 $ git bisect good # this one is good
64 ------------------------------------------------
65
66 which will now say
67
68 ------------------------------------------------
69 Bisecting: 337 revisions left to test after this
70 ------------------------------------------------
71
72 and you continue along, compiling that one, testing it, and depending
73 on whether it is good or bad, you say "git bisect good" or "git bisect
74 bad", and ask for the next bisection.
75
76 Until you have no more left, and you'll have been left with the first
77 bad kernel rev in "refs/bisect/bad".
78
79 Bisect reset
80 ~~~~~~~~~~~~
81
82 Oh, and then after you want to reset to the original head, do a
83
84 ------------------------------------------------
85 $ git bisect reset
86 ------------------------------------------------
87
88 to get back to the original branch, instead of being on the bisection
89 commit ("git bisect start" will do that for you too, actually: it will
90 reset the bisection state).
91
92 Bisect visualize
93 ~~~~~~~~~~~~~~~~
94
95 During the bisection process, you can say
96
97 ------------
98 $ git bisect visualize
99 ------------
100
101 to see the currently remaining suspects in 'gitk'. `visualize` is a bit
102 too long to type and `view` is provided as a synonym.
103
104 If 'DISPLAY' environment variable is not set, 'git log' is used
105 instead. You can even give command line options such as `-p` and
106 `--stat`.
107
108 ------------
109 $ git bisect view --stat
110 ------------
111
112 Bisect log and bisect replay
113 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
114
115 The good/bad input is logged, and
116
117 ------------
118 $ git bisect log
119 ------------
120
121 shows what you have done so far. You can truncate its output somewhere
122 and save it in a file, and run
123
124 ------------
125 $ git bisect replay that-file
126 ------------
127
128 if you find later you made a mistake telling good/bad about a
129 revision.
130
131 Avoiding to test a commit
132 ~~~~~~~~~~~~~~~~~~~~~~~~~
133
134 If in a middle of bisect session, you know what the bisect suggested
135 to try next is not a good one to test (e.g. the change the commit
136 introduces is known not to work in your environment and you know it
137 does not have anything to do with the bug you are chasing), you may
138 want to find a near-by commit and try that instead.
139
140 It goes something like this:
141
142 ------------
143 $ git bisect good/bad # previous round was good/bad.
144 Bisecting: 337 revisions left to test after this
145 $ git bisect visualize # oops, that is uninteresting.
146 $ git reset --hard HEAD~3 # try 3 revs before what
147 # was suggested
148 ------------
149
150 Then compile and test the one you chose to try. After that, tell
151 bisect what the result was as usual.
152
153 Bisect skip
154 ~~~~~~~~~~~~
155
156 Instead of choosing by yourself a nearby commit, you may just want git
157 to do it for you using:
158
159 ------------
160 $ git bisect skip # Current version cannot be tested
161 ------------
162
163 But computing the commit to test may be slower afterwards and git may
164 eventually not be able to tell the first bad among a bad and one or
165 more "skip"ped commits.
166
167 You can even skip a range of commits, instead of just one commit,
168 using the "'<commit1>'..'<commit2>'" notation. For example:
169
170 ------------
171 $ git bisect skip v2.5..v2.6
172 ------------
173
174 would mean that no commit between `v2.5` excluded and `v2.6` included
175 can be tested.
176
177 Note that if you want to also skip the first commit of a range you can
178 use something like:
179
180 ------------
181 $ git bisect skip v2.5 v2.5..v2.6
182 ------------
183
184 and the commit pointed to by `v2.5` will be skipped too.
185
186 Cutting down bisection by giving more parameters to bisect start
187 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
188
189 You can further cut down the number of trials if you know what part of
190 the tree is involved in the problem you are tracking down, by giving
191 paths parameters when you say `bisect start`, like this:
192
193 ------------
194 $ git bisect start -- arch/i386 include/asm-i386
195 ------------
196
197 If you know beforehand more than one good commits, you can narrow the
198 bisect space down without doing the whole tree checkout every time you
199 give good commits. You give the bad revision immediately after `start`
200 and then you give all the good revisions you have:
201
202 ------------
203 $ git bisect start v2.6.20-rc6 v2.6.20-rc4 v2.6.20-rc1 --
204 # v2.6.20-rc6 is bad
205 # v2.6.20-rc4 and v2.6.20-rc1 are good
206 ------------
207
208 Bisect run
209 ~~~~~~~~~~
210
211 If you have a script that can tell if the current source code is good
212 or bad, you can automatically bisect using:
213
214 ------------
215 $ git bisect run my_script
216 ------------
217
218 Note that the "run" script (`my_script` in the above example) should
219 exit with code 0 in case the current source code is good. Exit with a
220 code between 1 and 127 (inclusive), except 125, if the current
221 source code is bad.
222
223 Any other exit code will abort the automatic bisect process. (A
224 program that does "exit(-1)" leaves $? = 255, see exit(3) manual page,
225 the value is chopped with "& 0377".)
226
227 The special exit code 125 should be used when the current source code
228 cannot be tested. If the "run" script exits with this code, the current
229 revision will be skipped, see `git bisect skip` above.
230
231 You may often find that during bisect you want to have near-constant
232 tweaks (e.g., s/#define DEBUG 0/#define DEBUG 1/ in a header file, or
233 "revision that does not have this commit needs this patch applied to
234 work around other problem this bisection is not interested in")
235 applied to the revision being tested.
236
237 To cope with such a situation, after the inner 'git bisect' finds the
238 next revision to test, with the "run" script, you can apply that tweak
239 before compiling, run the real test, and after the test decides if the
240 revision (possibly with the needed tweaks) passed the test, rewind the
241 tree to the pristine state. Finally the "run" script can exit with
242 the status of the real test to let the "git bisect run" command loop to
243 determine the outcome.
244
245 EXAMPLES
246 --------
247
248 * Automatically bisect a broken build between v1.2 and HEAD:
249 +
250 ------------
251 $ git bisect start HEAD v1.2 -- # HEAD is bad, v1.2 is good
252 $ git bisect run make # "make" builds the app
253 ------------
254
255 * Automatically bisect a broken test suite:
256 +
257 ------------
258 $ cat ~/test.sh
259 #!/bin/sh
260 make || exit 125 # this "skip"s broken builds
261 make test # "make test" runs the test suite
262 $ git bisect start v1.3 v1.1 -- # v1.3 is bad, v1.1 is good
263 $ git bisect run ~/test.sh
264 ------------
265 +
266 Here we use a "test.sh" custom script. In this script, if "make"
267 fails, we "skip" the current commit.
268 +
269 It's safer to use a custom script outside the repo to prevent
270 interactions between the bisect, make and test processes and the
271 script.
272 +
273 And "make test" should "exit 0", if the test suite passes, and
274 "exit 1" (for example) otherwise.
275
276 * Automatically bisect a broken test case:
277 +
278 ------------
279 $ cat ~/test.sh
280 #!/bin/sh
281 make || exit 125 # this "skip"s broken builds
282 ~/check_test_case.sh # does the test case passes ?
283 $ git bisect start HEAD HEAD~10 -- # culprit is among the last 10
284 $ git bisect run ~/test.sh
285 ------------
286 +
287 Here "check_test_case.sh" should "exit 0", if the test case passes,
288 and "exit 1" (for example) otherwise.
289 +
290 It's safer if both "test.sh" and "check_test_case.sh" scripts are
291 outside the repo to prevent interactions between the bisect, make and
292 test processes and the scripts.
293
294 Author
295 ------
296 Written by Linus Torvalds <torvalds@osdl.org>
297
298 Documentation
299 -------------
300 Documentation by Junio C Hamano and the git-list <git@vger.kernel.org>.
301
302 GIT
303 ---
304 Part of the linkgit:git[1] suite