I found the following comment inside the source code for TeX in the preface to a function creating Roman numeral representations:
Readers who like puzzles might enjoy trying to figure out how this tricky code works; therefore no explanation will be given.
Even with Knuth does not leave a comment, he leaves a comment!
In the days when you got the source code with your operating system I was looking through a file called LITMACS.SR on a Data General Nova/RDOS system. This was an extension to MAC, the macro assembler, that let you write things like
LDA 0, =45
to load 45 into accumulator 0 (the Nova minicomputers didn’t have immediate addressing). Every so often in your code you’d place a .LPOOL directive, and the literals used so far would be created at that spot, and the code patched up to use PC relative addressing.
One of the comments said something like:
Don’t alter this code unless you understand it properly, and who does?