I only have production code

Hi,
I'm Nena and I am developing and maintaining my open source software open3A. It's an invoicing application with CRM, staff- and inventory management. I am working on open3A since 2007 by myself and It has gotten really big. Since 2021 I have an employee but I still am the only one coding the entire application.

As a single developer I have to make some compromises because I just don't have the time to do all the fancy stuff like unit tests, CI/CD, younameit. I need to produce working code because that is what I am paid for.

I have one principle for developing new features:

I only develop something if I have someone who will use it directly afterwards. Even if that is myself 😉. This is great because I don't waste time on features no one will use and it has another upside: I don't need to write any tests for the functionality. Because a real person will be using it. In a real use-case. In their own environment. And they usually use it in ways I never considered.

For this to work I need to be able to fix a bug really quickly because there is only code that is in production. I use PHP as language of choice and my software is open source so I can just send my users/customers the new version of a file and tell them where to copy it and everything is fine.

Of course if multiple files are affected I need a process for that, too.
I have two versions of my software. The last stable one and my developer version which I call "nightly". The last stable version is just the nightly version with a feature freeze at some point. After the freeze I only backport some important bug fixes to the stable version. Soooo, a user reports a bug, what do I do? I fix it in nightly and maybe backport it to stable. I then tell the user to make a full update with the last stable version OR with the current nightly version. Yep. With the nightly. For production. I know, I am crazy 😜

But let me take you full circle. I wrote earlier, that I only develop a new feature if there is someone who will use it directly afterwards. This means there is always someone who is already using the nightly version. Because new features are only in the nightly version. I use my users to continuously test the software. This way I can concentrate on production code, make sure everything works as expected and even when (there is no if) something gets broken, people only tell me "wow, you are really fast!". They don't know that I use them for testing. They don't care because they are used to things not working. It is my fast response saving the day.

This also works for the extensions I write for customers. I get a lot of requests to build this and that into the software and often I don't extend my main version for it but instead I write a single file which contains the desired functionality. I have a whole extension system in place but it simply is impossible to check every extension I ever wrote. There are hundreds which I created over the years. Sooo, a user with a custom extension - which I've already forgotten existed ☺️ - makes an update and boom 💥, it breaks because I made a change two versions ago and she didn't update since then. No biggie, I just fix the extension to work with the current version and send it, along with instructions where to copy the file.

Of course this only works because I don't do complete rewrites and making any extension compatible again only includes some code changes. I have this base version of my software which I won't rewrite completely. Ever. It's just too big by now and I couldn't pull it off anyway. I don't do revolution. I only do evolution. Sure enough I clean up technical debt (I have to) when I am working on some code again after months or years. But I never touch stuff that has no reported problems. I accept that there are pieces of my application that might not be working right now. But that is fine. If no one notices, then no one uses it and it might as well be broken.