summaryrefslogtreecommitdiff
path: root/HACKING
blob: adc305ff24d273b389bb6e23c215ceb120072f53 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
Coreutils Contribution Guidelines


Prerequisites
=============
You will need the "git" version control tools.
On Fedora-based systems, do "yum install git".
On Debian-based ones install the "git-core" package.
Then run "git --version".  If that says it's older than
version 1.4.4, then you'd do well to get a newer version.
At worst, just download the latest stable release from
http://git.or.cz/ and build from source.

For details on building the programs in this package, see
the file, README-hacking.


Use the latest upstream sources
===============================
Base any changes you make on the latest upstream sources.
You can get a copy of the latest with this command:

    git clone git://git.sv.gnu.org/coreutils
    cd coreutils

That downloads the entire repository, including revision control history
dating back to 1991.  The repository (the part you download, and which
resides in coreutils/.git) currently weighs in at about 30MB.  So you
don't want to download it more often than necessary.  Once downloaded,
you can get incremental updates by running one of these commands from
inside your new coreutils/ directory:

If you have made *no* changes:
    git pull

If you *have* made changes and mistakenly committed them to "master",
do the following to put your changes on a private branch, "br", and
to restore master to its unmodified (relative-to-upstream) state:
    git checkout -b br
    git checkout master
    git reset --hard origin

Then "git pull" should work.


*Before* you commit changes
===========================

In this project, we much prefer patches that automatically record
authorship.  That is important not just to give credit where due, but
also from a legal standpoint (see below).  To create author-annotated
patches with git, you must first tell git who you are.  That information
is best recorded in your ~/.gitconfig file.  Edit that file, creating
it if needed, and put your name and email address in place of these
example values:

[user]
  name = Joe X. User
  email = joe.user@example.com


Your first commit: the quick and dirty way
==========================================
First of all, realize that to "commit" a change in git is a purely
local operation.  It affects only the local repository (the .git/ dir)
in your current coreutils/ hierarchy.

To try this out, modify a file or two.  If you create a new file, you'll
need to tell git about it with "git add new-file.c".  Commit all changes
with "git commit -a".  That prompts you for a log message, which should
include a one-line summary, a blank line, and ChangeLog-style entries
for all affected files.  More on that below.

Once your change is committed, you can create a proper patch that includes
a log message and authorship information as well as any permissions
changes.  Use this command to save that single, most-recent change set:

  git format-patch --stdout -1 > DIFF

The trouble with this approach is that you've just checked in a change
(remember, it's only local) on the "master" branch, and that's where new
changes would normally appear when you pull the latest from "upstream".
When you "pull" from a remote repository to get the latest, your local
changes on "master" may well induce conflicts.   For this reason, you
may want to keep "master" free of any local changes, so that you can
use it to track unadulterated upstream sources.

However, if your cloned directory is for a one-shot patch submission and
you're going to remove it right afterwards, then this approach is fine.
Otherwise, for a more sustainable (and more generally useful, IMHO)
process, read on about "topic" branches.


Make your changes on a private "topic" branch
=============================================
So you checked out coreutils like this:

  git clone git://git.sv.gnu.org/coreutils

Now, cd into the coreutils/ directory and run:

  git checkout -b my-topic

That creates the my-topic branch and puts you on it.
To see which branch you're on, type "git branch".
Right after the clone, you were on "master" (aka the trunk).
To get back to the trunk, do this:

  git checkout master

Note 1:
    Be careful to run "git pull" only when on the "master" branch,
    not when on a branch.  With newer versions of git, you can't cause
    trouble if you forget, so this is a good reason to ensure you're
    using 1.5.3.1 or newer.

Note 2:
    It's best not to try to switch from one branch to another if
    you have pending (uncommitted) changes.  Sometimes it works,
    sometimes the checkout will fail, telling you that your local
    modifications conflict with changes required to switch branches.
    However, in any case, you will *not* lose your uncommitted changes.
    Run "git stash" to temporarily hide uncommitted changes in your
    local directory, restoring a clean working directory.

Anyhow, get back onto your just-created branch:

  git checkout my-topic

Now, modify some file and commit it:

  git commit some-file.c

Personally, no matter what package I'm working on, I find it useful to
put the ChangeLog entries *only* in the commit log, initially, unless
I plan to commit/push right away.  Otherwise, I tend to get unnecessary
merge conflicts with each rebase (see below).  In coreutils, I've gone
a step further, and no longer maintain an explicit ChangeLog file in
version control.  Instead, in a git working directory, you can view
ChangeLog information via "git log".  However, each distribution tarball
does include a ChangeLog file that is automatically generated from the
git logs.

So, you've committed a change.  But it's only in your local repository,
and only on your "my-topic" branch.  Let's say you wait a day, and
then see that someone else changed something and pushed it to the
public repository.  Now, you want to update your trunk and "rebase"
your changes on the branch so that they are once again relative to the
tip of the trunk.  Currently, your branch is attached to the trunk at
the next-to-last change set.

First: update the trunk from the public repo:
[you've first made sure that "git diff" produces no output]

  git checkout master
  git pull

Now, return to your branch, and "rebase" relative to trunk (master):

  git checkout my-topic
  git rebase master

If there are no conflicts, this requires no more work from you.
However, let's say there was one in ChangeLog, since you didn't
follow my advice and modified it anyway.
git rebase will tell you there was a conflict and in which
file, and instruct you to resolve it and then resume with
"git rebase --continue" once that's done.

So you resolve as usual, by editing ChangeLog (which has the
usual conflict markers), then type "git rebase --continue".
That will fail, with a diagnostic telling you to mark
the file as "conflict resolved" by doing this:

  git add ChangeLog

Then, finally, you can proceed (possibly onto more conflict resolution,
if there are conflicts in other files):

  git rebase --continue

Once it finishes, your changes on the branch are now relative to
the tip of the trunk.

Now use git format-patch, as above.


Amending the most recent change on your private branch
======================================================
Let's say you've just committed a change on your private
branch, and then realize that something about it is not right.
It's easy to adjust:

  edit your files # this can include running "git add NEW" or "git rm BAD"
  git commit --amend -a
  git format-patch --stdout -1 > your-branch.diff

That replaces the most recent change-set with the revised one.



Coreutils-specific:

No more ChangeLog files
=======================
Do not modify any of the ChangeLog files in coreutils.  Starting in
2008, the policy changed.  Before, we would insert the exact same text
(or worse, sometimes slightly differing) into both the ChangeLog file
and the commit log.  Now we put that information only in the commit log,
and generate the top-level ChangeLog file from logs at "make dist" time.
As such, there are strict requirements on the form of the commit log
messages.


Commit log requirements
=======================
Your commit log should always start with a one-line summary, the second
line should be blank, and the remaining lines are usually ChangeLog-style
entries for all affected files.  However, it's fine -- even recommended --
to write a few lines of prose describing the change, when the summary
and ChangeLog entries don't give enough of the big picture.  Omit the
leading TABs that you're used to seeing in a "real" ChangeLog file, but
keep the maximum line length at 72 or smaller, so that the generated
ChangeLog lines, each with its leading TAB, will not exceed 80 columns.
As for the ChangeLog-style content, please follow these guidelines:

  http://www.gnu.org/software/guile/changelogs/guile-changelogs_3.html

Try to make the summary line fit one of the following forms:

  program_name: change-description
  prog1, prog2: change-description
  doc: change-description
  tests: change-description
  build: change-description
  maint: change-description

If your commit fixes a bug, try to find the commit that introduced that
bug.  If you do that, add a note in your new commit log saying something
like "Introduced by commit v8.12-103-g54cbe6e." and add something like
[bug introduced in coreutils-8.13] in the corresponding NEWS blurb.
Assuming you found the bug in commit 54cbe6e6, "git describe 54cbe6e6"
will print the longer tag-relative string that you'll need.
Note that we used to use an 8-byte SHA1 prefix like "54cbe6e6", because
that was automatically rendered as a clickable link by "gitk", but with
git-1.7.10, the more descriptive version-containing "git describe" format
that we now require is also highlighted.


Curly braces: use judiciously
=============================
Omit the curly braces around an "if", "while", "for" etc. body only when
that body occupies a single line.  In every other case we require the braces.
This ensures that it is trivially easy to identify a single-*statement* loop:
each has only one *line* in its body.

Omitting braces with a single-line body is fine:

     while (expr)
       single_line_stmt ();

However, the moment your loop/if/else body extends onto a second line,
for whatever reason (even if it's just an added comment), then you should
add braces.  Otherwise, it would be too easy to insert a statement just
before that comment (without adding braces), thinking it is already a
multi-statement loop:

     while (true)
       /* comment... */      // BAD: multi-line body without braces
       single_line_stmt ();

Do this instead:

     while (true)
       {  /* Always put braces around a multi-line body.  */
         /* explanation... */
         single_line_stmt ();
       }

There is one exception: when the second body line is not at the same
indentation level as the first body line.

     if (expr)
       error (0, 0, _("a diagnostic that would make this line"
                      " extend past the 80-column limit"));

It is safe to omit the braces in the code above, since the
further-indented second body line makes it obvious that this is still
a single-statement body.

To reiterate, don't do this:

     if (expr)
       while (expr_2)        // BAD: multi-line body without braces
         {
           ...
         }

Do this, instead:

     if (expr)
       {
         while (expr_2)
           {
             ...
           }
       }

However, there is one exception in the other direction, when even a
one-line block should have braces.  That occurs when that one-line,
brace-less block is an "else" block, and the corresponding "then" block
*does* use braces.  In that case, either put braces around the "else"
block, or negate the "if"-condition and swap the bodies, putting the
one-line block first and making the longer, multi-line block be the
"else" block.

    if (expr)
      {
        ...
        ...
      }
    else
      x = y;    // BAD: braceless "else" with braced "then"

This is preferred, especially when the multi-line body is more than a
few lines long, because it is easier to read and grasp the semantics of
an if-then-else block when the simpler block occurs first, rather than
after the more involved block:

    if (!expr)
      x = y;                  /* more readable */
    else
      {
        ...
        ...
      }

If you'd rather not negate the condition, then add braces:

    if (expr)
      {
        ...
        ...
      }
    else
      {
        x = y;
      }


Use SPACE-only indentation in all[*] files
==========================================
We use space-only indentation in nearly all files.
If you use Emacs and your coreutils working directory name matches,
this code enables the right mode:

  ;; In coreutils, indent with spaces everywhere (not TABs).
  ;; Exceptions: Makefile and ChangeLog modes.
  (add-hook 'find-file-hook '(lambda ()
    (if (and buffer-file-name
             (string-match "/coreutils\\>" (buffer-file-name))
             (not (string-equal mode-name "Change Log"))
             (not (string-equal mode-name "Makefile")))
        (setq indent-tabs-mode nil))))

If you use vim (7+ compiled with autocommands), and coreutils working
directory name also matches, add the following in ~/.vimrc:

  " Set GNU style indentation, spaces instead of TABs
  function! CoreutilsIndent()
      " Check if 'coreutils' is part of the current working directory
      if match(getcwd(), "coreutils") > 0
          " The next 3 lines below set the GNU indentation
          setlocal cinoptions=>4,n-2,{2,^-2,:2,=2,g0,h2,p5,t0,+2,(0,u0,w1,m1
          setlocal shiftwidth=2
          setlocal tabstop=8
          " Coreutils specific, expand TABs with spaces
          setlocal expandtab
      endif
  endfunction

  autocmd BufEnter *.c,*.h call CoreutilsIndent()

[*] Makefile and ChangeLog files are exempt, of course.


Send patches to the address listed in --help output
===================================================
Please follow the guidelines in the "Sending your patches." section of
git's own SubmittingPatches:

  http://git.kernel.org/?p=git/git.git;a=blob;f=Documentation/SubmittingPatches


Add documentation
=================
If you add a feature or change some user-visible aspect of a program,
document it.  If you add an option, document it both in --help output
(i.e., in the usage function that generates the --help output) and in
doc/*.texi.  The man pages are generated from --help output, so
you shouldn't need to change anything under man/.  User-visible changes
are usually documented in NEWS, too.

When writing prose (documentation, comments, log entries), use an
active voice, not a passive one.  I.e., say "print the frobnozzle",
not "the frobnozzle will be printed".

Please add comments per the GNU Coding Standard:
  http://www.gnu.org/prep/standards/html_node/Comments.html


Minor syntactic preferences
===========================
[I hesitate to write this one down, because it appears to be an
 acquired taste, at least for native-English speakers.  It seems odd
 (if not truly backwards) to nearly anyone who doesn't have a strong
 mathematics background and perhaps a streak of something odd in their
 character ;-) ]
In writing arithmetic comparisons, use "<" and "<=" rather than
">" and ">=".  For some justification, read this:
  http://thread.gmane.org/gmane.comp.version-control.git/3903/focus=4126

const placement:
Write "Type const *var", not "const Type *var".
FIXME: dig up justification


Be nice to translators
======================
Don't change translatable strings if you can avoid it.
If you must rearrange individual lines (e.g., in multi-line --help
strings), extract and create new strings, rather than extracting
and moving into existing blocks.  This avoids making unnecessary
work for translators.


Add tests
==========
Nearly every significant change must be accompanied by a test suite
addition that exercises it.  If you fix a bug, add at least one test that
fails without the patch, but that succeeds once your patch is applied.
If you add a feature, add tests to exercise as much of the new code
as possible.  If you add a new test file (as opposed to adding a test to
an existing test file) add the new test file to 'tests/local.mk'.
Note to run tests/misc/new-test in isolation you can do:

  make TESTS=tests/misc/new-test SUBDIRS=. VERBOSE=yes

Variables that are significant for tests with their default values are:

  VERBOSE=yes
  RUN_EXPENSIVE_TESTS=no
  RUN_VERY_EXPENSIVE_TESTS=no
  SHELL=/bin/sh
  NON_ROOT_USERNAME=nobody
  NON_ROOT_GROUP=$(id -g $NON_ROOT_USERNAME)
  COREUTILS_GROUPS=$(id -G)

There are hundreds of tests in the tests/ directories.  You can use
tests/sample-test as a template, or one of the various Perl-based ones
in tests/misc.

If writing tests is not your thing, don't worry too much about it,
but do provide scenarios, input/output pairs, or whatever, along with
examples of running the tool to demonstrate the new or changed feature,
and someone else will massage that into a test (writing portable tests
can be a challenge).


Copyright assignment
====================
If your change is significant (i.e., if it adds more than ~10 lines),
then you'll have to have a copyright assignment on file with the FSF.
Since that involves first an email exchange between you and the FSF,
and then the exchange (FSF to you, then back) of an actual sheet of paper
with your signature on it, and finally, some administrative processing
in Boston, the process can take a few weeks.

The forms to choose from are in gnulib's doc/Copyright/ directory.
If you want to assign a single change, you should use the file,
doc/Copyright/request-assign.changes:

    http://www.gnu.org/software/gnulib/Copyright/request-assign.changes

If you would like to assign past and future contributions to a project,
you'd use doc/Copyright/request-assign.future:

    http://www.gnu.org/software/gnulib/Copyright/request-assign.future

You may make assignments for up to four projects at a time.

In case you're wondering why we bother with all of this, read this:

    http://www.gnu.org/licenses/why-assign.html


Run "make syntax-check", or even "make distcheck"
================================================
Making either of those targets runs many integrity and
project-specific policy-conformance tests.  For example, the former
ensures that you add no trailing blanks and no uses of certain deprecated
functions.  The latter performs all "syntax-check" tests, and also
ensures that the build completes with no warnings when using a certain
set of gcc -W... options.  Don't even bother running "make distcheck"
unless you have a reasonably up to date installation including recent
versions of gcc and the linux kernel, and modern GNU tools.


Ensure that your changes are indented properly.
===============================================
Format the code the way GNU indent does.
Filtering most source files through "indent --no-tabs" should
induce no change in indentation.  Try not to add any more.


Avoid trailing white space
==========================
You may notice that the only trailing blanks in coreutils'
version-controlled files are in a single directory: tests/pr,
which contains expected output from various invocations of pr.

Do not add any more trailing blanks anywhere.  While "make syntax-check"
will alert you if you slip up, it's better to nip any problem in the
bud, as you're typing.  A good way to help you adapt to this rule is
to configure your editor to highlight any offending characters in the
files you edit.  If you use Emacs, customize its font-lock mode
or use its WhiteSpace mode:

    http://www.emacswiki.org/emacs/WhiteSpace

If you use vim, add this to ~/.vimrc:

    let c_space_errors=1
    highlight RedundantSpaces ctermbg=red guibg=red
    match RedundantSpaces /\s\+$\| \+\ze\t/


Git can help too, by stopping you from committing any change that would
add trailing blanks.  The example pre-commit hook contains code to check
for trailing whitespace and spaces before tabs; enable it by moving it
to the right place and making sure it is executable:

    mv .git/hooks/pre-commit.sample .git/hooks/pre-commit

With a repository created by git-1.5.6 or older, use this command:

    chmod +x .git/hooks/pre-commit

To manually check for whitespace errors before committing, you can use

    git diff --check

Git also has some settings to enable suitable internal whitespace checks.
See the manpage for git-apply for details.


-------------------------------------------

Miscellaneous useful git commands
=================================

  * gitk: give a graphical view of the revision graph of the current branch
  * gitk --all: same, but display all branches
  * git log: to get most of the same info in text form
  * git log -p: same as above, but with diffs
  * git log -p SOME_FILE: same as above, but limit to SOME_FILE
  * git log -p -2 SOME_FILE: same as above, but print only two deltas
  * git log -p -1: print the most recently committed change set
  * git format-patch --stdout -1 > FILE: output the most recently committed
      change set, in a format suitable to be submitted and/or applied via
      "git am FILE".
  * git reset --soft HEAD^: Commit the delta required to restore
      state to the revision just before HEAD (i.e., next-to-last).
  * git rebase -i master: run this from on a branch, and it gives
      you an interface with which you can reorder and modify arbitrary
      change sets on that branch.

  * if you "misplace" a change set, i.e., via git reset --hard ..., so that
    it's no longer reachable by any branch, you can use "git fsck" to find
    its SHA1 and then tag it or cherry-pick it onto an existing branch.
    For example, run this:
      git fsck --lost-found HEAD && cd .git/lost-found/commit \
        && for i in *; do git show $i|grep SOME_IDENTIFYING_STRING \
        && echo $i; done
    The "git fsck ..." command creates the .git/lost-found/... hierarchy
    listing all unreachable objects.  Then the for loop
    print SHA1s for commits that match via log or patch.
    For example, say that found 556fbb57216b119155cdda824c98dc579b8121c8,
    you could run "git show 556fbb57216b119" to examine the change set,
    or "git checkout -b found 556fbb5721" to give it a branch name.
    Finally, you might run "git checkout master && git cherry-pick 556fbb5721"
    to put that change on the tip of "master".

-------------------------------------------

Finding things to do
====================
If you don't know where to start, check out the TODO file for projects
that look like they're at your skill-/interest-level.  Another good
option is always to improve tests.  You never know what you might
uncover when you improve test coverage, and even if you don't find
any bugs your contribution is sure to be appreciated.

A good way to quickly assess current test coverage is to use "lcov"
to generate HTML coverage reports.  Follow these steps:

  # configure with coverage information
  ./configure CFLAGS="-g -fprofile-arcs -ftest-coverage"
  make
  # run whatever tests you want, i.e.:
  make check
  # run lcov
  lcov -t coreutils -q -d lib -b lib -o lib.lcov -c
  lcov -t coreutils -q -d src -b src -o src.lcov -c
  # generate HTML from the output
  genhtml -p `pwd` -t coreutils -q --output-directory lcov-html *.lcov

Then just open the index.html file (in the generated lcov-html directory)
in your favorite web browser.

========================================================================
Copyright (C) 2009-2012 Free Software Foundation, Inc.

Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.3 or
any later version published by the Free Software Foundation; with no
Invariant Sections, with no Front-Cover Texts, and with no Back-Cover
Texts.  A copy of the license is included in the "GNU Free
Documentation License" file as part of this distribution.