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 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 committed them to "master", do this: git fetch git rebase origin *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 --signoff HEAD~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. 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 -e -a git format-patch --stdout --signoff HEAD~1 > your-branch.diff That replaces the most recent commit 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. Omit the leading TABs that you're used to seeing in a "real" ChangeLog file. Use SPACE-only indentation in new files. ======================================== In any new file, eliminate all leading TABs (e.g., via running GNU indent with --no-tabs) and put these lines at the end of the file: [FIXME: suggest vim syntax to do same thing, if it can be done safely. Most distros now "set nomodeline" by default for a good reason. ] /* * Local variables: * indent-tabs-mode: nil * End: */ Do not change TABs to spaces or vice versa in any existing file. Send patches to bug-coreutils@gnu.org ===================================== 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/coreutils.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. 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. 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://git.sv.gnu.org/gitweb/?p=gnulib.git;a=blob;f=doc/Copyright/request-assign.changes;hb=HEAD If you would like to assign past and future coreutils work, you'd use doc/Copyright/request-assign.future: http://git.sv.gnu.org/gitweb/?p=gnulib.git;a=blob;f=doc/Copyright/request-assign.future;hb=HEAD Run "make syntax-check", or even "make distcheck" ================================================ Making either of those targets runs many integrity and coreutils-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 very 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. In a file with the "indent-tabs-mode: nil" directive at the end, running "indent --no-tabs" should induce no change. With other files, there will be some existing differences. Try not to add any more. ------------------------------------------- Miscellaneous useful git commands ================================= * gitk: give a graphical view of the revision graph * 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 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.