Starting to develop
Gitlab account
The first step is to make an account on
To get access to the repository you will also need to have an ssh key associated to your
Workflow
If you have never used git before, you should make yourself familiar with it. There are many good online tutorials about how to use git, so this should not be a problem.
The development is done following the oneflow workflow.
Cloning
Once you are part of the octopus-code group, you need to clone the main repository on your desktop
git clone git@gitlab.com:octopus-code/octopus.git
With this you get a complete copy of the Git repository that you can use for your developments.
At this point, you might want to compile and install the version from the develop branch following these instructions.
Working on the code
Before starting to implement some new feature, we recommend you create a new issue on
git checkout develop
git checkout -b myfeature
You can commit your modification in two steps
git add modified_or_new_files
git commit -m "Your commit message."
If you only have modified files, the two steps can be combined into one unique command
git commit -a -m "Your commit message."
Note that the “-m” flags is not mandatory: if you do not give a message, git will open a text editor where you can type your commit message. Make sure that you give a ‘‘‘meaningful description’’’ of your changes to help other developers understand what you did. It might be a good idea to have a look at the changes that other developers made to see some examples of their messages.
Do not be afraid to change things in Octopus, you can always revert them back to the original version that you got in the beginning. This is one of the best features of using a version control system like git. You will also have the opportunity to get feedback from some of the other developers before your changes are included in the main development branch.
To share your work with the other developers, you need to send your changes to the main repository. This is done with the push
command. The first time you push your branch, you need to tell git where to send your changes. In this case you want to send them to the origin
, which is the default name of the original repository you cloned
git push --set-upstream origin myfeature
Git will remember this, so the next time you push your branch you will only need to do
git push
In order to prevent the testsuite from failing on code-formatting issues, it is highly recommended to run a style check before
pushing any changes to a merge request. You can do that by running octopus-findent
, which can be found in
this repository. Instructions on how to compile and install the findent
tool can be found there as well.
In most cases, your changes will be pushed to the repository without problems. However, if you have been developing in your branch for some time, it is very likely that the develop
branch will have evolved as well. You need to make sure that you are keeping up to date with those changes. Otherwise it might not be possible to accept a merge request based on your branch, either because of conflicts or because the testsuite fails (more about merge requests and the testsuite bellow). So from time to time you should incorporate new changes from the develop
branch into you branch by doing a git merge
(another option is to use git rebase
, but this requires some extra care). First switch to the develop
branch and make sure it is up to date
git checkout develop
git pull
Then switch back to your branch and merge develop into it
git checkout myfeature
git merge develop
There is the possibility for conflicts between your changes and the ones in the develop
branch, so read the output of the merge
command carefully. If the changes in the develop
branch are unrelated to your changes the merge will just go through and you can continue to make changes in your branch. If you did the merge because of a failed git push, you can try again and it should be accepted now.
In case of conflicts, the merge output it will tell you in which files the conflicts appear. Open one of these files in your favorite text editor and search for HEAD
, this will mark the start of the conflicting lines. You will first see the lines from your branch and then the corresponding lines from the develop
branch. You will need to make a choice which lines you keep. Depending on the situation different solutions can be appropriate, so select carefully. You then need to delete the lines that mark the conflict, like the HEAD
that you initially searched for. Solve all the conflicts in the file and save your changes. Then let git know that you have resolved the conflict in this file by using
git add filename
If there is more than one file having conflicts, move on to the next one until you have resolved all conflicts in all files. Commit the new version of your branch with
git commit
It will automatically create a message saying that this commit is for merging develop into myfeature and list the files that had conflicts. If you started the whole procedure because git push failed, you can try again and it should be accepted.
Testsuite
Octopus comes with a whole set of tests in various categories. You can find them in the testsuite/ folder in your Octopus directory. You should make sure that your changes do not lead to a failure in this testsuite. To run the complete set of tests run
make check
For greater convenience, the tests are also divided into two subsets: “long” and “short” tests. You can run these by using
make check-short
and
make check-long
The first should take about 20 minutes and the second about 40 minutes (depending on your computer, of course). If you start the tests before going for lunch, they should be done easily by the time you return, so no excuses for not running them.
If you are implementing a new feature, make sure that you also provide a test inside the testsuite so that other people do not break your code.
Any changes to existing tests ‘‘‘must’’’ be discussed with the other developers first.
For more details on the testsuite, you might want to check the [[media:Strubbe_Octopus_testsuite.pptx.pdf | talk by David Strubbe]] and the testsuite/README file in the source code.
Buildbot
Octopus has a set of computers with different architectures set up that run the tests with different compilers etc. Whenever a branch gets pushed to the main repository on
In case the tests appear as failed, you will probably want to know more about what went wrong. To do this, go to the pipelines page and select the pipeline corresponding to your push. Next click on the button that says “Octopus Buildbot”. This should send you to the Octopus buildbot page. You need to be logged in to see its contents. The authentication is done using your
Due to the complexity of available architectures (e.g. parallelism, use of accelerator cards) results can slightly differ from one to another computer. In order to account for those variations, the test files contain precision goals for each number. These should be set as tight as possible. We have a tool, which can analyze the different results and visualize the spread of the values, which can help to identify whether there is some real problem, giving rise to outliers on some computers. The tool can be accesses through the web app. There also exists a tutorial video.
Merge requests
Once a new feature is ready, it should be merged into the develop
branch. This should be done through a merge request. The simplest way to do this is by using the
You can also create a merge request in case you want to get some feedback about your work from the other developers before a new feature is ready to be merged. In that case create a merge request as explained above, but start the merge request title with WIP:
.
Bug fixes
Sooner or later new bugs are found in the code that need to be fixed. If a bug is found, the first thing to do is to create a new issue explaining what the problem is. The exact workflow to fix the bug will depend whether the bug is found in the master branch or not. In any case, once the bug has been fixed, we recommend that you add a regression test, or modify an existing test, such that it returns a failure in the event that the same bug is reintroduced in the code. For more details on how to handle bug fixes and more unusual situations, please see the [[Developers:Workflow]] page.
Bugs present in develop branch, but ‘‘‘not’’’ in master
In this case follow the same workflow as above to introduce new features: create a new branch from develop
, fix the bug, commit your changes, push them to
Bugs present in master
If the bug is present in the master
branch, the workflow to be followed is slightly different, because we want to make the bug fix available to users ASAP through a bug fix release.
To do this, first create a new branch from the master
branch. The name of this new branch should start with hotfix-
to make its purpose clear. Then fix the bug, bump the Octopus release number in the configure.ac
file (see the [[FAQ-How_do_the_version_numbers_in_Octopus_work.3F|FAQ]] to learn how version numbers in Octopus work), commit your changes, and push everything to
Using a fork
If you do not have write permissions to the main git repository, you first need to fork it. This is done using the
Note that you can always do your developments using a fork if you think this is more convenient, even if you have write permissions to the main git repository.
Cloning
Forking will create a new git repository that lives in your
git clone git@gitlab.com:my-username/octopus.git
Note the ‘‘my-username’’ in the git url. This should be replaced by the appropriate name.
Staying synchronized with the main repository
You will want to regularly get the changes made by others and accepted in merge requests. For this you need to set a pointer, called a remote, to the main repository of Octopus:
git remote add upstream git@gitlab.com:octopus-code/octopus.git
In this case we have named this pointer upstream . This is a widely used convention, but you can name it differently if you prefer. To check that the remote was set properly, do
>git remote -v
origin git@gitlab.com:my-username/octopus.git (fetch)
origin git@gitlab.com:my-username/octopus.git (push)
upstream git@gitlab.com:octopus-code/octopus.git (fetch)
upstream git@gitlab.com:octopus-code/octopus.git (push)}}
Once this is set, you can get the latest changes from the main repository
git fetch upstream
The fetch
command will get the changes, but will not merge them into your local branches. Usually the local branch that you want to keep synchronized with the upstream is the develop
branch. This is because you typically branch off from develop
to create a feature branch (as explained above). To merge the upstream develop
into the local one, do
git checkout develop
git merge upstream/develop
Since you are not supposed to directly change your local develop branch, this should not raise any conflicts.
Merge requests
Merge requests work pretty much like before, except that you need to specify that the source branch lives in your forked repository instead of the main repository.