I’m pretty good about backing things up and it took about 30 minutes to burn a new SD Card, update it, install VS Code and the C/C++ extension, then copy my asteroids version over. I use WinSCP so had to enable SSH on the PI. It’s disabled by default but just tick a checkbox on the interface tab of the Preferences->Configuration menu.

Even then it didn’t compile. Of course I had to reinstall the dev versions of SDL, SDL_image, SDL_mixer and SDL_ttf and the clang compiler. Still it didn’t compile. I had created a Projects folder and created an Asteroids folder underneath that. I also had the Vs Code JSON files that you need for compiling C/C++. the main one of which is tasks.json. Those were in a folder .vscode which I had backed up but I’d copied it over into the wrong location. You want it located inside your VS Code folder.

This makes sense. If you have say five different projects then you are going to have a different build, link stuff per project. So you’ll have a unique .vscode in each folder. When you want to switch projects, you just close the Folder in the VS Code File menu and open it in the folder for the project that you next want to work on.

Mind you it still wouldn’t compile. It turned out my tasks.json has clang-6 in it. When I did a clang –version on it, it told me it was clang 7.0.1. So I upped it to clang-7 in tasks.json and that fixed it. It all compiled and ran.

Once you’ve done this a time or two it becomes 2nd nature but I can understand novices frustration; there are a lot of moving parts that all have to be right before you can even write and run C code. It’s not like other programming languages are really any better though. If you have setup virtual environments in Python and installed Python modules, you’ll know what it can be like.

So, if you are new to programming Compiling and linking can seem like magic. The compiler transforms C source code files into obj files. The actual format is different depending on the CPU and Operating system.  Linking is even more magic because it takes a bunch of obj files and produces a single executable.

There’s actual more magic going on than my simplistic explanation. The compiler might be doing optimisation (generally in release mode but not debug). That makes debug compiles faster which is more desirable.

Another blog called Hack the developer has published an article with a lot more detail about compiling and linking. If you want to get a better understanding then this is a petty good piece. It uses Linux, GCC and Clang as examples. It also goes into more depth on the layout of obj files.