| 1 |
This is a guide to submitting patches to the concordance project. |
| 2 |
|
| 3 |
TABLE OF CONTENTS |
| 4 |
1. Preparing a change |
| 5 |
2. Generating patches |
| 6 |
3. Preparing and submitting your patch |
| 7 |
4. Additional notes for major changes |
| 8 |
5. A word on comments |
| 9 |
|
| 10 |
|
| 11 |
PREPARING A CHANGE |
| 12 |
|
| 13 |
If you would like to make a change to the codebase, you need to decide if this |
| 14 |
is a major change or a minor change. Major changes include anything that |
| 15 |
affects the API, ABI, anything that affects lots of files, adds/removes |
| 16 |
classes, etc. |
| 17 |
|
| 18 |
If your plan is to make a major change, then you should send a description of |
| 19 |
the changes you plan to make to the concordance-devel list FIRST. By doing |
| 20 |
this you give the developers a chance to comment on your plan and provide |
| 21 |
suggestions and guidance before you invest the time. This will save you time |
| 22 |
and effort and provide a much higher chance of your patch being accepted. Also |
| 23 |
see "ADDITIONAL NOTES FOR MAJOR CHANGES" below. |
| 24 |
|
| 25 |
If your plan is to make a minor change, you are still welcome to send an email |
| 26 |
before you start to the list, but it's not as big of a deal. |
| 27 |
|
| 28 |
|
| 29 |
GENERATING PATCHES |
| 30 |
|
| 31 |
After making your changes, you should always make sure you have done a 'cvs |
| 32 |
update' to ensure you'll be generating your patch against the latest CVS |
| 33 |
sources. |
| 34 |
|
| 35 |
Patches should always be generated via CVS's diff, and from top-level CVS |
| 36 |
directory. Additionally, they should always be in unified diff format ("-u"), |
| 37 |
should show new/removed files ("-N"), be recursive ("-R"), and use the -p flag |
| 38 |
to make sure the C function name is included for all changes. For example: |
| 39 |
|
| 40 |
CVSTREE=~/cvs/concordance |
| 41 |
cd $CVSTREE |
| 42 |
vi libconcord/remote.cpp |
| 43 |
cvs diff -NRup >/tmp/libconcord_remote.patch |
| 44 |
|
| 45 |
You should ALWAYS read your patch through after generating it to make sure it |
| 46 |
contains the changes you expect and ONLY the changes you expect. |
| 47 |
|
| 48 |
You should ensure the following: |
| 49 |
1. You've maintained the same coding style. See the CodingStyle document |
| 50 |
in this directory. |
| 51 |
2. You should make sure that your code is properly commented. |
| 52 |
|
| 53 |
|
| 54 |
PREPARING AND SUBMITTING YOUR PATCH |
| 55 |
|
| 56 |
Your patch should be sent in as an attachement. Mail clients often wrap emails |
| 57 |
(and rightly so!) which can break a patch. Attach your patch to your email. |
| 58 |
|
| 59 |
The body of your email should have a good descriptiong of your patch. The more |
| 60 |
far-reaching your patch is, the more detailed your description should be. |
| 61 |
|
| 62 |
A good description includes justifying any ABI/API changes, an explanation of |
| 63 |
the goal and outcome of the patch, and a list of changes by file, where |
| 64 |
appropriate. |
| 65 |
|
| 66 |
The subject of an email SHOULD start with "PATCH: " and include a _useful_ |
| 67 |
description of your patch. Where possible, include the component affected |
| 68 |
(concordance vs. libconcord) GOOD examples include: |
| 69 |
|
| 70 |
PATCH: concordance: fix double-free bug |
| 71 |
PATCH: libconcord: Add remote_info table entry for 4847 byte location |
| 72 |
|
| 73 |
BAD examples include: |
| 74 |
|
| 75 |
PATCH: update |
| 76 |
PATCH: libconcord fix |
| 77 |
PATCH: new remote |
| 78 |
|
| 79 |
NOTE: PATCHES SHOULD ALWAYS BE SENT TO THE -devel LIST! Even if you are |
| 80 |
attaching a patch to a bug, you should still send it to the devel list if you |
| 81 |
want it committed. |
| 82 |
|
| 83 |
|
| 84 |
PATCHSET SUBMISSIONS |
| 85 |
|
| 86 |
Sometimes it makes sense to break a given change into several pieces, but you |
| 87 |
still want to submit them together. In this case, you may do this in two ways. |
| 88 |
You could just you attach all patches to a single email with useful filenames |
| 89 |
like 1_update_foo_API.patch 2_update_bar_API and a list of the patches and |
| 90 |
their details in the email. |
| 91 |
|
| 92 |
Alternatively, if you prefer multiple emails, you can send a group of emails |
| 93 |
as follows. |
| 94 |
|
| 95 |
An initial email with a subject that starts with "[PATCH 0/X]" (where X is the |
| 96 |
total number of patches) and a short description should be sent. This email |
| 97 |
should contain all the information about the patchset as a whole, as well as a |
| 98 |
list of patch-containing-emails that will follow. This list should look like |
| 99 |
this: |
| 100 |
|
| 101 |
1/2 - libconcord: update foo() API to accept new parameter |
| 102 |
2/2 - concordance: use new foo() paremeter to do bar |
| 103 |
|
| 104 |
Assuming here that X is 2. Then these patches should be submitted with |
| 105 |
subjects like: "[PATCH 1/2] libconcord: update foo() API to accept new param." |
| 106 |
|
| 107 |
Note that if there's a possiblity of merging some parts of your patchset in |
| 108 |
stages, then they should definitely be split into multiple emails. |
| 109 |
|
| 110 |
|
| 111 |
ADDITIONAL NOTES FOR MAJOR CHANGES |
| 112 |
|
| 113 |
As previously mentioned, all major changes should be mentioned on the list. |
| 114 |
Note, however, that changes to the ABI or API need to be justified. The API, |
| 115 |
in particular, should change as little as possible. |
| 116 |
|
| 117 |
If a change to the API is needed, remember that the API is *C* API even though |
| 118 |
the library itself is written in C++. Any changes to the API must ensure it's |
| 119 |
still C-compatible, and that all of the API is exported properly for C |
| 120 |
clients. |
| 121 |
|
| 122 |
In addition, all API changes must be added to all bindings we support as well. |
| 123 |
If you don't have experience with some of the languages, we can help. |
| 124 |
|
| 125 |
A WORD ON COMMENTS |
| 126 |
Comments should be immediately above the block of code they are explaining, |
| 127 |
NOT next to it. For example, we prefer this: |
| 128 |
|
| 129 |
/* |
| 130 |
* This special algorith for XXX does A, then B, then C, then D. |
| 131 |
* It is necessary because of bar. |
| 132 |
*/ |
| 133 |
if (foo) { |
| 134 |
... |
| 135 |
} |
| 136 |
|
| 137 |
Not this: |
| 138 |
|
| 139 |
if (foo) { |
| 140 |
a = *(b+8); // Some comment here |
| 141 |
|
| 142 |
In addition, comments should explain code, not just narrate it. A comments |
| 143 |
that says "skip the first byte" isn't useful, but a comment that says "the |
| 144 |
first byte is all 0 for Windows, so we skip it when not in Windows" is very |
| 145 |
useful. |
| 146 |
|
| 147 |
|
| 148 |
|
| 149 |
|
| 150 |
|
| 151 |
|
| 152 |
|
| 153 |
|
| 154 |
|
| 155 |
# for vim to text-wrap |
| 156 |
vim:textwidth=78: |
| 157 |
|