On the other hand, Upstart makes heavy use of libnih (it's like glib, but nowhere near as heavy) and is actually very easy to follow. Compared to systemd, it certainly is, at least. Many of the comments are actually annotations for the automated documentation generator, from what I recall.

OpenRC makes use of source code comments in a more conventional way, but they actually do tend to help with adding insight into what was going on in the author's head, and the code is still understandable (though not as clean as Upstart's), regardless. Its small size obviously also helps.

Or that the author understands something about architecture. From an internal code base that I can't cut and paste, there's a nice discussion in comment form that paraphrasing is something like "you'd really like to implement an AVL tree here because we do lots of lookups and AVLs are faster than RBs at lookups, but actual performance testing on the test cluster shows that under our real world workload, the slightly better writing performance of the RB outweighs the slightly faster reads, or TLDR do not blindly change this to a AVL thinking you'll win, because you won't".

(Oh and obviously business logic. I find nothing inherently technologically wrong with the concept of say, floats for currency, perhaps in a hard sci fi novel, although as a business logic concept that is widely considered horrifically wrong, and there's nothing wrong with documenting business logic in comments although float currency is a ridiculous extreme example... how bout, and this GIGO filter forbids negative interest rates or something, which makes math sense and computer sense but from a business logic perspective might be quite icky)

I agree with you WRT comments for regex looking like line noise or programming 101 style syntax comments "x = 1 // this step sets the variable x to contain the integer number 1"

No it isn't. Software developers have been saying for 30-40 years "the code is readable as is", yet we keep coming up with easier to read languages, and comments are still in the spec of all these languages.

Comments are important, it's not worth arguing if the code is readable or not. It's a waste of time that could be spent better documenting what ever you consider readable. Because your co-workers, or co-contributors may have a different definition of what readable is.

The same could be said about comments. You can always say they are not enough. As I said - comments have their use - in my opinion for the cases where language is not flexible enough to express intent in a readable way.

Use code for describing how your program does what it does, function-level comments for what it does in cases where you cannot put that in the name of a type or identifier (getFoo typically does not need a comment, but strtok does) and sprinkle in comments why your program does what it does in the way it does where that may not be clear to the expected reader of your code (as in the example of the AVL tree in a sister comment)

I have, on occasion, seen code that seemed horribly convoluted or inefficient, only to find out the hard way that there was a good reason for it being that way, often having to do with idiosyncrasies or bugs in the OS or external libraries. In those cases, I found it extremely helpful to write a short comment explaining what mess you are trying to work around. Not only will this keep others (or yourself, a year from now) from trying to mess with that code, it will also encourage people to fix it if the reasons you state in your comment no longer apply (maybe the OS/library/compiler has been fixed, maybe you're on a different platform, ...).

If, for some reason, I find myself writing a longer-than-recommended function/method, I to insert comments to make the structure of the code clear, so if I find out later that there is a bug in the code that, say, creates the database connection, I can find that code more easily. Of course, factoring that code into several functions/methods and call them from a short top-level one would be better, but just having to read through a couple of screens of code without any sign posts is worse.

(Much of the same applies to documentation, of course. I you need to explain what a function/method does, because its name does not tell you, you might want to rename it. But if there are any caveats you should be aware of - e.g. "Don't ever pass a NULL pointer to this function, because it does not check for those and will segfault the program" - the docs should point them out to you.)

(I am, of course, writing all this from a more general point of view, I have never looked at the systemd sources, and I'm not sure I want to.)

Comments are used to explain strange things in your code. A small number of comments in a good codebase is an indicator of strange things being sparse - a good thing.