Git: An Unexpected Journey
This summer of ‘18, I’ll be working in Git. Git is an insanely awesome software. I still remember the day I learned about Git via Github and instantly got hooked. Git isn’t just an everyday software you encounter, just look at the documentation and observe closely how it works and you will understand the rock solid design. Enough with Git introduction already. There are too much posts describing Git already.
Sending Patches
Rather than going to the theory of Git, lets learn how to send patch to Git. Sending patches to Git is different than sending patches to other JS Frameworks, Game Engines, etc. You don’t use Github to send patches to Git. Documentation/SubmittingPatches already contains a great deal of information for anybody to learn how to send patch. But, let me summarize it.
Ok, you’ve modified some code. You check it and believe it doesn’t break
anything else. You run tests on it. It looks good. You follow the
standard C writing guideline. So, you decide to send it.
First you must do is fire up terminal and run git diff --check
.
If you don’t see anything there, it’s good you have not introduced
whitespaces in your code. But, if you do see codes there fix
those whitespaces. Git community is picky about whitespaces.
Now, is the right time to run some tests. Always remember to test your changes. Running test suite provided in Git is the most preferred but if your change introduces something new, you can write a new test. There are two kinds of tests you can do:
-
Functional tests: These are the core Git tests. These test scripts are mandatory to be run for changes and test must pass to introduce new patch. The easiest way to run all the tests is navigate to
/t
directory and runmake
. Another way is to run with any TAP harness. You can parallel test by using prove.$ prove --timer --jobs 15 ./t[0-9]*.sh
You can also run single test by running the shell script with a name liketNNNN-descriptor.sh
. If new functionality is added by you then unit tests need to be created by creating helper commands which have limited action. Keep these int/helpers
. To add helper add a line tot/Makefile
and to.gitignore
for the binary file you added. The Git community prefers functional tests using the full git executable, so try to exercise your new code using git commands before creating a test helper. To find out why a test failed, repeat the test with the -x -v -d -i options and then navigate to the appropriate “trash” directory to see the data shape that was used for the test failed step. More information can be found in t/README -
Performance tests: If your changes improve performance or if your patch can affect performance, you need to run the tests in
t/perf
. To check the change in performance uset/perf/run
script. More information about performance tests can be found in this awesome guideline written in Git for Windows Contributing document which is applicable for both Linux and Windows.
Then run git log <file you modified>
and see all the changes you made.
Now, think of good commit message. This is very important as it
describes the maintainers what you are fixing. Write the first line
of the commit message with a short description of about 50 characters.
Also, prefix is needed for the first line like:
test: avoid pipes in git related commands for test
Now, in after the short description, you should write a lengthier
commit message describing the changes you made. The commit message
should be in imperative mood. e.g. “avoid using pipes …”.
This should be very descriptive and very clear of what you’ve done.
Also, remember to sign-off your patch. Run git commit <files> -s
.
Now, time to generate your patch. Git has a tool for generating patches.
Cool, right? Run git format-patch HEAD~
to generate a single patch
for your changes. If your change is of only a single type i.e. change
in typo in different files then, only generate a single file. But, if
your changes are in different files of different nature then, run
git format-patch HEAD~<number of commits to convert to patches>
.
Remember you’ll need multiple commits if you’ve changed multiple files
differently e.g. you changed a test file for typo and git-rebase.sh
to add some extra functionalities, then you have to write two commit
messages describing both changes.
Now, a patch file or multiple patch files will be generated depending
on your changes. You can apply this patchfile by running git am <patchfile>
.
Now, time to send to patch to the mailing list. Git has a tool for this
too. Run git send-email <your-patch-file>
. Now, to use send-email, you need
to set up your mailing id and format message with the mail. This should be
in plain text. Follow git send-email documentation
closely. This explains everything in much more detail.
Your patch now, goes through lot of review process. Follow closely,
discuss with the community, understand and acknowledge their comments.
Write quality codes rather than large quantity.
GSoC ‘18
So, in this GSoC timeline. I’ll be working to make git rebase
a builtin.
i.e. git rebase
is previously written in shell scripts. My job is to convert
it into C.
Why is this necessary one might ask.
Well, Many components of Git are still in the form of shell and Perl scripts.
This has certain advantages of being extensible but causes problems in production
code on multiple platforms like Windows. I’m going to rewrite a couple of shell
and perl scripts into portable and performant C code, making them built-ins.
The major advantage of doing this is improvement in efficiency and performance.
So, what does rebase do? There’s a cool beginner friendly introduction to rebase here Or, you can follow, git-rebase documentation. I’m not going to explain again here as it is already done in a much better way in both the git documentation and the post. Now, this is just a beginning post. I will be updating and keeping a series of what I learn from Git. I will be writing every week. This will be a great journey.
Ah, nearly forgot to mention. If all of our patches in GSoC ‘18 gets merged into Git. Git will be fast. How fast you might ask? Time will tell.