contribute.dox 13.9 KB
Newer Older
Francis Dupont's avatar
Francis Dupont committed
// Copyright (C) 2013-2017 Internet Systems Consortium, Inc. ("ISC")
3 4 5
// This Source Code Form is subject to the terms of the Mozilla Public
// License, v. 2.0. If a copy of the MPL was not distributed with this
// file, You can obtain one at
6 7 8


 @page contributorGuide Kea Contributor's Guide

So you found a bug in Kea or plan to develop an extension and want to
send us a patch? Great! This page will explain how to contribute your
changes smoothly.

15 16
@section contributorGuideWritePatch Writing a patch

17 18 19 20
Before you start working on a patch or a new feature, it is a good
idea to discuss it first with Kea developers. You can post your
questions to the \c kea-dev mailing list
( or kea-users
21 22 23
( The kea-users list
is intended for users who are not interested in the internal workings
or development details of Kea: it is OK to ask for feedback regarding new
design or the best proposed solution to a certain problem, but all
25 26 27 28 29 30 31 32 33
the internal details should be discussed on kea-dev and not posted
to kea-users.

The first step in writing the patch or new feature should be to get
the source code from our Git repository. The procedure is very easy and
is explained here:  While it is
possible to provide a patch against the latest stable release, it makes
the review process much easier if it is for latest code from the Git \c
master branch.

OK, so you have written a patch? Great! Before you submit it, make sure
that your code compiles. This may seem obvious, but there's more to
it. You have surely checked that it compiles on your system, but Kea
is portable software. Besides Linux, it is compiled and used on
relatively uncommon systems like OpenBSD. Will your code
Francis Dupont's avatar
Francis Dupont committed
compile and work there? What about endianness? It is likely that you used
a regular x86 architecture machine to write your patch, but the software
is expected to run on many other architectures. You may take a look at
system specific build notes (
For a complete list of systems we build on, you may take a look at the
following build farm report: .

Does your patch conform to Kea coding guidelines
48 49
( You can submit a
patch that does not adhere to them, but that will reduce its chances of
being accepted. If the deviations are minor, one of the Kea engineers who
51 52 53
does the review will likely fix the issues. However, if there are lots
of issues, the reviewer may simply reject the patch and ask you to fix
it before re-submitting.

55 56
@section contributorGuideUnittests Running unit-tests

One of the ground rules in Kea development is that every piece of
58 59
code has to be tested. We now have an extensive set of unit-tests for
almost every line of code. Even if you are fixing something small,
60 61
like a single line fix, you are encouraged to write unit-tests for that
change. That is even more true for new code: if you write a new
function, method or a class, you definitely should write unit-tests
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
for it.

To ensure that everything is tested, ISC uses a development method
called Test Driven Development (TDD).  In TDD, a feature is developed
alongside the tests, with the tests being written first.  In detail,
a test is written for a small piece of functionality and run against
the existing code. (In the case where the test is a unit test for
a function, it would be run against an empty (unimplemented)
function.)  The test should fail.  A minimal amount of code is then
written, just enough to get the test to pass.  Then the process is
repeated for the next small piece of functionality. This continues
until all the functionality has been implemented.

This approach has two advantages:

- By writing a test first and then only enough code to pass the
test, that code is fully tested.  By repeating this process until
the feature is fully implemented, all the code gets test coverage.
You avoid the situation where not enough tests have been written
to check all the code.

- By running the test before the code implementing the function is
written and observing the test fail, you can detect the situation
where a bug in the test code will cause it to pass regardless of
the code being tested.

89 90 91 92 93 94 95 96
Initially, some people unfamiliar with that approach react with "but
my change is simple and I tested that it works". That approach
is both insufficient and short-sighted. It is insufficient, because
manual testing is by definition laborious and can't really be done
on the multitude of systems we run Kea on. It is short-sighted, because
even with your best intentions you will not be able to dedicate
any significant amount of time for repeated testing of your improved
code. The old ISC DHCP has two decades of history behind it and we
97 98
hope to make Kea last similar time span. Over such long periods,
code tends to be refactored several times. The change you
made may be affected by some other change or by the code that
hasn't been written yet.
101 102 103 104 105 106

See @ref unitTests for instructions on how to run unit-tests. If you
happen to touch any database related code, make sure you compile
your code with @c --with-dhcp-mysql, @c --with-dhcp-pgsql and/or
@c --with-cql as needed. For example, if you change something
substantial, make sure the other compilation options still work.

108 109 110
If you happen to add new files or have modified any \c
files, it is also a good idea to check if you haven't broken the
distribution process:

make distcheck
114 115

116 117
There are other useful switches which can be passed to configure. It is
always a good idea to use \c --enable-logger-checks, which does sanity
118 119 120
checks on logger parameters. Use \c --enable-debug to enable various
additional consistency checks that reduce performance but help during
development. If you happen to modify anything in the
documentation, use \c --enable-generate-docs. If you are modifying DHCP
122 123 124 125
code, you are likely to be interested in enabling a non-default database
backends for DHCP. Note that if the backend is not enabled,
the database-specific unit-tests are skipped. To enable the MySQL backend,
use the switch \c --with-dhcp-mysql; for PostgreSQL, use \c --with-dhcp-pgsql.
A complete list of all switches can be obtained with the command:
127 128 129 130 131

 ./configure --help

132 133 134 135 136 137
@section contributorGuideSendPatch Sending the patch

There are several ways how you can send your changes. They are listed
here from most to least preferred.

- Send pull request (PR) on github. This is by far the most convenient way for
everyone. See the excellent documentation on github:
139 for details.  In
140 141 142 143
essence, you need github account (spam/hassle free, takes one minute to set
up). Then you can fork the Kea repository, commit changes to your repo and ask us to pull
your changes into official Kea repository. This has a number of advantages. First, it is
made against a specific code version, which can be easily checked with git log
command. Second, this request pops up instantly on our list of open pull
requests and will stay there. The third benefit is that the pull request mechanism is
146 147
very flexible. Kea engineers (and other users, too) can comment on it, attach
links, mention other users etc.  You as a submitter can augment the patch by
Francis Dupont's avatar
Francis Dupont committed
committing extra changes to your repository. Those extra commits will appear instantly
149 150 151
in the pull request. This is really useful during the review. Finally, ISC engineers can
better assess all open pull requests and add labels to them, such as "enhancement", "bug", or
"unit-tests missing". This makes our life easier. Oh, and your commits will later
152 153
be shown as yours in github history. If you care for that kind of things, once the
patch is merged, you'll be automatically listed as contributor and Kea will be
listed as project you have contributed to.

- Create a ticket in the Kea trac and attach your patch to it. Sending a patch has a
157 158 159 160 161 162
number of disadvantages. First, if you don't specify the base version against
which it was created, one of ISC engineers will have to guess that or go through
a series of trials and errors to find that out. If the code doesn't compile, the
reviewer will not know if the patch is broken or maybe it was applied to
incorrect base code. Another frequent problem is that it may be possible that
the patch didn't include any new files you have added. If we happen to have any
163 164
comments that you as submitter are expected to address (and in the overwhelming majority
of cases, we have), you will be asked to send an updated patch. It is not
165 166
uncommon to see several rounds of such reviews, so this can get very complicated
very quickly. Please make sure your ticket is created in the default milestone
"kea-proposed". ISC engineers review new tickets once a week and assign them to
168 169 170
specific milestones. Please do not add tickets to working milestones directly.
Having a ticket in trac ensures that the patch will never be forgotten and it
will show up on our trac reports. It's not required, but much appreciated if
you send a short note to the kea-dev mailing list explaining what you did with the code
172 173 174
and announce the ticket number.

- Send a patch to the kea-dev list. This is the third preferred method, if you
can't or don't want to use github and trac for whatever reason.  If you
176 177 178 179 180 181
send a patch to a mailing list in a wrong time, e.g. shortly before a release,
ISC engineers may miss it or perhaps they will see it and then forget about
it. Nevertheless, it is still doable and we successfully accepted patches that
way. It just takes more time from everyone involved, so it's a slower process in

- Send a tarball with your modified code. This is really the worst way one
183 184 185 186
can contribute a patch. It has a number of disadvantages. In particular, someone
will need to find out which version the code was based on and generate the
patch. It's not a rocket science, but it may be a very mundane thing to do
if the ISC engineer does not know the version in advance. The mailing list
has a limit on the message size (for good reasons), so you'll likely need to
188 189 190 191 192 193 194
upload it somewhere first. ISC engineers often don't pick up new tickets
instantly, so it may have to wait weeks before the tarball is looked at.
The tarball does not benefit from most of the advantages mentioned for github,
like the ability to easily update the code, have a meaningful discussion
or see what the exact scope of changes are. Nevertheless, if we given a choice
of getting a tarball or not getting a patch at all, we prefer tarballs. Just
keep in mind that processing a tarball is really cumbersome for ISC
Francis Dupont's avatar
Francis Dupont committed
engineers, so it may take significantly longer than other ways.

@section contributorGuideReview Going through a review

199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215
Once you let us submitted a patch using one of the ways above, the action
is on one of the ISC engineers. First, we will need either a trac ticket
or PR on github. We prefer the original submitter fill them as he or she
has the best understanding of the purpose of the change and may have
any extra information, e.g. "this patch fixes compilation issue on
FreeBSD 10.1". If there there is no PR and no trac ticket, we will create
one. Depending on the subjective importance and urgency as perceived by
the ISC engineer, the ticket or PR will be assigned to one of the

Sooner or later, one of ISC engineers will do the review. Here's the tricky
part. One of Kea developers will review your patch, but it may not happen
immediately. Unfortunately, developers are usually working under a tight
schedule, so any extra unplanned review work may take a while sometimes. Having
said that, we value external contributions very much and will do whatever we can
to review patches in a timely manner. Don't get discouraged if your patch is not
accepted after first review. To keep the code quality high, we use the same
review processes for external patches as we do for internal code. It may take
217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233
some cycles of review/updated patch submissions before the code is finally
accepted. The nature of the review process is that it emphasizes areas that need
improvement. If you are not used to the review process, you may get the
impression that the feedback is negative. It is not: even the Kea developers
seldom see reviews that say "All OK please merge".

Once the process is almost complete, the developer will likely ask you how you
would like to be credited. The typical answers are by first and last name, by
nickname, by company name or anonymously. Typically we will add a note to the \c
ChangeLog and also set you as the author of the commit applying the patch and
update the contributors section in the AUTHORS file. If the contributed feature
is big or critical for whatever reason, it may also be mentioned in release

Sadly, we sometimes see patches that are submitted and then the submitter never
responds to our comments or requests for an updated patch. Depending on the
nature of the patch, we may either fix the outstanding issues on our own and get
234 235 236
another ISC engineer to review them or the ticket may end up in our "Outstanding
Tasks" milestone. When a new release is started, we go through the tickets in
Outstanding Tasks, select a small number of them and move them to whatever the
current milestone is. Keep that in mind if you plan to submit a patch and forget
about it. We may accept it eventually, but it's much, much faster process if you
participate in it.

241 242
@section contributorGuideExtra Extra steps

243 244 245 246 247 248 249 250 251
If you are interested in knowing the results of more in-depth testing, you are
welcome to visit the ISC Jenkins page: This is a live
result page with all tests being run on various systems.  Besides basic
unit-tests, we also have reports from valgrind (memory debugger), cppcheck and
clang-analyzer (static code analyzers), Lettuce system tests and more. Although
it is not possible for non ISC employees to run tests on that farm, it is
possible that your contributed patch will end up there sooner or later. We also
have ISC Forge tests running, but currently the test results are not publicly