During the last couple of years, internally we’ve moved over to using Git as our Revision Control System (RCS). It’s been an interesting exercise, especially where, like me, you’ve come from a traditional model (such as subversion or even back to good old SCCS). I’m sure you’ve all got your own “top 5” and I don’t necessarily expect you to agree with me, but here’s my key learning points:
#1 “Branch always, branch often”
At the outset this was certainly the biggest mindset change*. In many RCSs, branching is particularly “heavyweight” and merging branches (and branches of branches, and branches… you get the gist) can be, to say the least, challenging. So you could end up being counter-productive by trying to avoid branching unless absolutely necessary.
It couldn’t be more different using Git, and once I’d “got it”, then it is an instinct to always work on a branch. I guess it all fits in with the aim to keep work cycles too much shorter timeframes, so branching and merging must be trivial. There is an excellent tutorial to help learn branching on Git.
#2 “Don’t ignore your .gitignore”
You know you’re going to need it, so do it before you start any real work. It’ll just save you a whole load of tidying up later on (ok so maybe this is an age thing? don’t procrastinate!). There are even a language specific .gitignore templates at https://github.com/github/gitignore, so no excuses.
It’s also worth getting to know the basics of the pattern matching that can be used in the .gitignore file, they’re well documented.
#3 “GitHub is not Git”
Maybe it’s my onset of senility, but initially it wasn’t exactly clear where Git ended and GitHub started. Most of the early stuff I looked at to help understand and learn Git almost always used GitHub to host your project. In hindsight it’s all obvious (well of course it’s just a web-based hosting service for software development projects that use Git), but at the time I can say my understanding was a little “fuzzy” around the edges.
Two things really helped; first the excellent book Pro Git written by Scott Chacon which you can find online at http://git-scm.com/book. I found it so useful to I also bought the eBook (Kindle) version.
Second, moving to BitBucket away from GitHub. In fact we use both, GitHub for our publicly hosted projects and BitBucket for internal work. This just made it plainly obvious where Git ended and GitHub/BitBucket started.
#4 “Quickly learn how to undo”
There are many ways to mess it up, if you really try. Committing without adding, modifying files without branching (see #1), committing unwanted files (see #2), not pushing changes (see #3) and the list goes on (and I’ve done them all).
Be assured, you can always fix it, from a simple commit-amend to, probably one of the coolest things, being able to create a branch from a stash. But I’d highly recommend spending a bit of time understanding unstaging as opposed to unmodifying a file.
Using Git is very much about understanding workflow and typically messing up means you’ve not quite got your workflow model running smoothly. The nice folks at Atlassian have put together a set of guides on workflow.
#5 “Initially do all your work on the command line”
For Linux people (and hopefully most programmers on a Mac), this doesn’t need saying as why would you do it any other way? However for all those Windows-based people (and in the embedded world this is the majority) then there is a tendency to use graphical tools. Both GitHub and BitBucket offer graphical frontends to managing Git local/remote repo’s but without understanding the underlying Git model (which you only really get from using it at the command line) then it’s easy to get lost.
I didn’t find the GitHub Windows client tool especially friendly, whereas SourceTree from Alassian I found a more intuitive client. Nevertheless I’d still rather work at the shell.
So that’s my top 5; What have I missed (remember you’ve got to drop one to add one)?
What would I do differently? certainly I’d read “Pro Git” cover-to-cover much sooner than I did. Otherwise, as usual, just throw yourself in and have a go.
Now what was that someone said about Mercurial being better…
* Actually the initial challenge was getting used to say “Git” outside of its usual context in the UK.
Niall has been designing and programming embedded systems for over 30 years. He has worked in different sectors, including aerospace, telecomms, government and banking.
His current interest lie in IoT Security and Agile for Embedded Systems.
Latest posts by Niall Cooling (see all)
- Updated: Developing a Generic Hard Fault handler for ARM Cortex-M3/Cortex-M4 using GCC - September 21, 2018
- Code Quality – Cyclomatic Complexity - July 27, 2018
- Bugs do matter…(unsurprisingly) - January 18, 2018