Collaborative Development with Git-Trac¶
Sometimes you will only want to work on local changes to Sage, for your own private needs. However, typically it is beneficial to share code and ideas with others; the manner in which the Sage project does this (as well as fixing bugs and upgrading components) is in a very collaborative and public setting on the Sage Trac server (the Sage bug and enhancement tracker).
One can use git the hard way for this,
but this section explains how to use the helper git trac command, which
simplifies many of the most common actions in collaboration on Sage. Some
of the tutorials we suggest may be helpful
in navigating what they are for.
Most of the commands in the following section will not work unless you have an account on Trac. If you want to contribute to Sage, it is a good idea to get an account now (see Obtaining an Account).
Installing the Git-Trac Command¶
Git is a separate project from trac, and the two do not know how to
talk to each other. To simplify the development, we have a special
git trac subcommand for the git suite. Note that this really is
only to simplify interaction with our trac issue management, you can
perform every development task with just git and a web browser. See
Git the Hard Way instead if you prefer to do everything by
hand:
[user@localhost]$ git clone https://github.com/sagemath/git-trac-command.git
Cloning into 'git-trac-command'...
[...]
Checking connectivity... done.
[user@localhost]$ source git-trac-command/enable.sh
Prepending the git-trac command to your search PATH
This creates a directory git-trac-command.
Sourcing the enable.sh script in there is just a quick and dirty
way to enable it temporarily. For a more permanent installation on
your system later, make sure to put the git-trac command in your
PATH. Assuming that ~/bin is already in your PATH, you can
do this by symlinking:
[user@localhost]$ echo $PATH
/home/user/bin:/usr/local/bin:/usr/bin:/bin:/usr/local/sbin:/usr/sbin
[user@localhost]$ cd git-trac-command
[user@localhost git-trac-command]$ ln -s `pwd`/git-trac ~/bin/
See the git-trac README for
more details. At this point you leave git-trac-command subdirectory, and only go
there whenever you need to update the git-trac command.
Git and Trac Configuration¶
Note
- trac uses username/password for authentication.
- Our git repository server uses SSH public key authentication for write access.
You need to set up both authentication mechanisms to be able to upload
your changes with “git trac”. For read-only access neither
authentication mechanism is needed. To set up git trac, first go
to the Sage directory and tell git trac about your trac account:
[user@localhost sage]$ git trac config --user USERNAME --pass 'PASSWORD'
Trac xmlrpc URL:
http://trac.sagemath.org/xmlrpc (anonymous)
http://trac.sagemath.org/login/xmlrpc (authenticated)
realm sage.math.washington.edu
Username: USERNAME
Password: PASSWORD
Retrieving SSH keys...
1024 ab:1b:7c:c9:9b:48:fe:dd:59:56:1e:9d:a4:a6:51:9d My SSH Key
where you have to replace USERNAME with your trac user name and
PASSWORD with your trac password. If you don’t have a trac account,
use git trac config without any arguments. The single quotes in
'PASSWORD' escape special characters that you might have in your
password. The password is stored in plain-text in .git/config, so
make sure that it is not readable by other users on your system. For
example, by running chmod 0600 .git/config if your home directory
is not already private.
Instead of a username and password you may also configure authentication via
a generated token by passing --token=<token> instead of --pass:
[user@localhost sage]$ git trac config --user=<username> --token=<token>
This is required if you authenticate to Trac with your GitHub account, as you do not have a Trac password. Logged in users can find their token under the token tab in preferences on the trac site .
If both a token and a username/password are configured, the token-based authentication takes precedence.
If you do not want to store your trac username/password/token on disk you
can temporarily override it with the environment variables
TRAC_USERNAME, TRAC_PASSWORD, and TRAC_TOKEN respectively.
These take precedence over any other configuration.
If there is no SSH key listed then you haven’t uploaded your SSH public key to the trac server. You should do that now following the instructions to Linking your Public Key to your Trac Account, if you want to upload any changes. You may have to add your private key to your authentication agent:
[user@localhost sage]$ ssh-add
Note
The git trac config command will automatically add a trac
remote git repository to your list of remotes if necessary.
If you followed the above instructions then you will have two remote repositories set up:
[user@localhost sage]$ git remote -v
origin git://github.com/sagemath/sage.git (fetch)
origin git://github.com/sagemath/sage.git (push)
trac git://trac.sagemath.org/sage.git (fetch)
trac git@trac.sagemath.org:sage.git (push)
The git@... part of the push url means that write access is
secured with SSH keys, which you must have set up as in
Linking your Public Key to your Trac Account. Read-only access happens through the
fetch url and does not require SSH.
Finally, if you do not want to use the git trac subcommand at all
then you can set up the remote by hand as described in the section on
The Trac Server.
Trac Tickets and Git Branches¶
Now let’s start adding code to Sage!
Create a Ticket¶
Suppose you have written an algorithm for calculating the last twin prime, and want to add it to Sage. You would first open a ticket for that:
[user@localhost sage]$ git trac create 'Last Twin Prime'
Remote branch: u/user/last_twin_prime
Newly-created ticket number: 12345
Ticket URL: http://trac.sagemath.org/12345
Local branch: t/12345/last_twin_prime
This will create a new trac ticket titled “Last Twin Prime” with a
remote branch u/user/last_twin_prime attached to it. The remote
branch name is automatically derived from the ticket title; If you
don’t like this then you can use the -b switch to specify it
explicitly. See git trac create -h for details. This new branch is
automatically checked out for you with the local branch name
t/12345/last_twin_prime.
Note
Only some trac fields are filled in automatically. See The Ticket Fields for what trac fields are available and how we use them.
Check out an Existing Ticket¶
Alternatively, you can use the web interface to the Sage trac development server to open a new ticket. Just log in and click on “Create Ticket”.
Or maybe somebody else already opened a ticket. Then, to get a suitable local branch to make your edits, you would just run:
[user@localhost sage]$ git trac checkout 12345
Loading ticket #12345...
Checking out Trac #13744 remote branch u/user/last_twin_prime -> local branch t/12345/last_twin_prime...
The git trac checkout command downloads an existing branch (as
specified in the “Branch:” field on the trac ticket) or creates a new
one if there is none yet. Just like the create command, you can
specify the remote branch name explicitly using the -b switch if
you want.
Note on Branch Names¶
The “Branch:” field of a trac ticket (see The Ticket Fields) indicates the git branch containing its code. Our git server implements the following access restrictions for remote branch names:
- You can read/write/create a branch named
u/your_username/whatever_you_like. Everybody else can read. - Everybody can read/write/create a branch named
public/whatever_you_like.
Depending on your style of collaboration, you can use one or the
other. The git trac subcommands defaults to the former.
As a convention, the git trac subcommand uses local branch
names of the form t/12345/description, where the number is the
trac ticket number. The script uses this number to figure out the
ticket from the local branch name. You can rename the local branches
if you want, but if they don’t contain the ticket number then you will
have to specify the ticket number manually when you are uploading your
changes.
Making Changes¶
Once you have checked out a ticket, edit the appropriate files and commit your changes to the branch as described in Editing the Source Code and Commits (Snapshots).
Uploading Changes to Trac¶
Automatic Push¶
At some point, you may wish to share your changes with the rest of us: maybe it is ready for review, or maybe you are collaborating with someone and want to share your changes “up until now”. This is simply done by:
[user@localhost sage]$ git trac push
Pushing to Trac #12345...
Guessed remote branch: u/user/last_twin_prime
To git@trac.sagemath.org:sage.git
* [new branch] HEAD -> u/user/last_twin_prime
Changing the trac "Branch:" field...
This uploads your changes to a remote branch on the Sage git server. The git trac command uses
the following logic to find out the remote branch name:
- By default, the remote branch name will be whatever is already on the trac ticket.
- If there is no remote branch yet, the branch will be called
u/user/description(u/user/last_twin_primein the example). - You can use the
--branchoption to specify the remote branch name explicitly, but it needs to follow the naming convention from Note on Branch Names for you to have write permission.
Specifying the Ticket Number¶
You can upload any local branch to an existing ticket, whether or not
you created the local branch with git trac. This works exactly
like in the case where you started with a ticket, except that you have
to specify the ticket number (since there is no way to tell which
ticket you have in mind). That is:
[user@localhost sage]$ git trac push TICKETNUM
where you have to replace TICKETNUM with the number of the trac
ticket.
Finishing It Up¶
It is common to go through a few iterations of commits before you upload, and you will probably also have pushed your changes a few times before your changes are ready for review.
Once you are happy with the changes you uploaded, they must be
reviewed by somebody else before they can be included in the next
version of Sage. To mark your ticket as ready for review, you should
set it to needs_review on the trac server. Also, add yourself as
the (or one of the) author(s) for that ticket by inserting the
following as the first line:
Authors: Your Real Name
Downloading Changes from Trac¶
If somebody else worked on a ticket, or if you just switched computers, you’ll want to get the latest version of the branch from a ticket into your local branch. This is done with:
[user@localhost sage]$ git trac pull
Technically, this does a merge (just like the standard git pull)
command. See Merging and Rebasing for more background information.
Merging¶
As soon as you are working on a bigger project that spans multiple tickets you will want to base your work on branches that have not been merged into Sage yet. This is natural in collaborative development, and in fact you are very much encouraged to split your work into logically different parts. Ideally, each part that is useful on its own and can be reviewed independently should be a different ticket instead of a huge patch bomb.
For this purpose, you can incorporate branches from other tickets (or just other local branches) into your current branch. This is called merging, and all it does is include commits from other branches into your current branch. In particular, this is done when a new Sage release is made: the finished tickets are merged with the Sage master and the result is the next Sage version. Git is smart enough to not merge commits twice. In particular, it is possible to merge two branches, one of which had already merged the other branch. The syntax for merging is easy:
[user@localhost sage]$ git merge other_branch
This creates a new “merge” commit, joining your current branch and
other_branch.
Warning
You should avoid merging branches both ways. Once A merged B and B merged A, there is no way to distinguish commits that were originally made in A or B. Effectively, merging both ways combines the branches and makes individual review impossible.
In practice, you should only merge when one of the following holds:
- Either two tickets conflict, then you have to merge one into the other in order to resolve the merge conflict.
- Or you definitely need a feature that has been developed as part of another branch.
A special case of merging is merging in the develop branch. This
brings your local branch up to date with the newest Sage version. The
above warning against unnecessary merges still applies, though. Try to
do all of your development with the Sage version that you originally
started with. The only reason for merging in the develop branch is if
you need a new feature or if your branch conflicts. See
Update Branch to Latest SageMath Version (and Minimizing Recompilation Time) for details.
Collaboration and conflict resolution¶
Exchanging Branches¶
It is very easy to collaborate by just going through the above steps any number of times. For example, Alice starts a ticket and adds some initial code:
[alice@laptop sage]$ git trac create "A and B Ticket"
... EDIT EDIT ...
[alice@laptop sage]$ git add .
[alice@laptop sage]$ git commit
[alice@laptop sage]$ git trac push
The trac ticket now has “Branch:” set to
u/alice/a_and_b_ticket. Bob downloads the branch and works some
more on it:
[bob@home sage]$ git trac checkout TICKET_NUMBER
... EDIT EDIT ...
[bob@home sage]$ git add .
[bob@home sage]$ git commit
[bob@home sage]$ git trac push
The trac ticket now has “Branch:” set to u/bob/a_and_b_ticket,
since Bob cannot write to u/alice/.... Now the two authors just
pull/push in their collaboration:
[alice@laptop sage]$ git trac pull
... EDIT EDIT ...
[alice@laptop sage]$ git add .
[alice@laptop sage]$ git commit
[alice@laptop sage]$ git trac push
[bob@home sage]$ git trac pull
... EDIT EDIT ...
[bob@home sage]$ git add .
[bob@home sage]$ git commit
[bob@home sage]$ git trac push
Alice and Bob need not alternate, they can also add further commits on top of their own remote branch. As long as their changes do not conflict (edit the same lines simultaneously), this is fine.
Conflict Resolution¶
Merge conflicts happen if there are overlapping edits, and they are an unavoidable consequence of distributed development. Fortunately, resolving them is common and easy with git. As a hypothetical example, consider the following code snippet:
def fibonacci(i):
"""
Return the `i`-th Fibonacci number
"""
return fibonacci(i-1) * fibonacci(i-2)
This is clearly wrong; Two developers, namely Alice and Bob, decide to fix it. First, in a cabin in the woods far away from any internet connection, Alice corrects the seed values:
def fibonacci(i):
"""
Return the `i`-th Fibonacci number
"""
if i > 1:
return fibonacci(i-1) * fibonacci(i-2)
return [0, 1][i]
and turns those changes into a new commit:
[alice@laptop sage]$ git add fibonacci.py
[alice@laptop sage]$ git commit -m 'return correct seed values'
However, not having an internet connection, she cannot immediately send her changes to the trac server. Meanwhile, Bob changes the multiplication to an addition since that is the correct recursion formula:
def fibonacci(i):
"""
Return the `i`-th Fibonacci number
"""
return fibonacci(i-1) + fibonacci(i-2)
and immediately uploads his change:
[bob@home sage]$ git add fibonacci.py
[bob@home sage]$ git commit -m 'corrected recursion formula, must be + instead of *'
[bob@home sage]$ git trac push
Eventually, Alice returns to civilization. In her mailbox, she finds a trac notification email that Bob has uploaded further changes to their joint project. Hence, she starts out by getting his changes into her own local branch:
[alice@laptop sage]$ git trac pull
...
CONFLICT (content): Merge conflict in fibonacci.py
Automatic merge failed; fix conflicts and then commit the result.
The file now looks like this:
def fibonacci(i):
"""
Return the `i`-th Fibonacci number
"""
<<<<<<< HEAD
if i > 1:
return fibonacci(i-1) * fibonacci(i-2)
return [0, 1][i]
=======
return fibonacci(i-1) + fibonacci(i-2)
>>>>>>> 41675dfaedbfb89dcff0a47e520be4aa2b6c5d1b
The conflict is shown between the conflict markers <<<<<<< and
>>>>>>>. The first half (up to the ======= marker) is Alice’s
current version, the second half is Bob’s version. The 40-digit hex
number after the second conflict marker is the SHA1 hash of the most
recent common parent of both.
It is now Alice’s job to resolve the conflict by reconciling their changes, for example by editing the file. Her result is:
def fibonacci(i):
"""
Return the `i`-th Fibonacci number
"""
if i > 1:
return fibonacci(i-1) + fibonacci(i-2)
return [0, 1][i]
And then upload both her original change and her merge commit to trac:
[alice@laptop sage]$ git add fibonacci.py
[alice@laptop sage]$ git commit -m "merged Bob's changes with mine"
The resulting commit graph now has a loop:
[alice@laptop sage]$ git log --graph --oneline
* 6316447 merged Bob's changes with mine
|\
| * 41675df corrected recursion formula, must be + instead of *
* | 14ae1d3 return correct seed values
|/
* 14afe53 initial commit
If Bob decides to do further work on the ticket then he will have to pull Alice’s changes. However, this time there is no conflict on his end: git downloads both Alice’s conflicting commit and her resolution.
Reviewing¶
For an explanation of what should be checked by the reviewer, see The reviewer’s check list.
If you go to the web interface to the Sage trac development server then you can click on the “Branch:” field and see the code that is added by combining all commits of the ticket. This is what needs to be reviewed.
The git trac command gives you two commands that might be handy
(replace 12345 with the actual ticket number) if you do not want
to use the web interface:
git trac print 12345displays the trac ticket directly in your terminal.git trac review 12345downloads the branch from the ticket and shows you what is being added, analogous to clicking on the “Branch:” field.
To review tickets with minimal recompiling, start by building the “develop” branch, that is, the latest beta. Just checking out an older ticket would most likely reset the Sage tree to an older version, so you would have to compile older versions of packages to make it work. Instead, you can create an anonymous (“detached HEAD”) merge of the ticket and the develop branch using
$ git trac try 12345
This will only touch files that are really modified by the ticket. In particular,
if only Python files are changed by the ticket (which is true for most tickets)
then you just have to run sage -b to rebuild the Sage library. If files other
than Python have been changed, you must run make. When you are finished
reviewing, just check out a named branch, for example
$ git checkout develop
If you want to edit the ticket branch (that is, add additional commits) you cannot
use git trac try. You must Check out an Existing Ticket to get the actual ticket
branch as a starting point.