Unix warned readers about code its own authors later had to replace
A famous source comment became a warning about depending on one compiler’s habits.

The Unix comment that sounds like a programmer sneering at the next reader had a gentler intended meaning. Dennis Ritchie said it was closer to telling a student that a difficult passage would not appear on the exam.
Then he and Steve Johnson had to move the kernel to another machine. Surprise exam.
In his account of odd Unix comments, Ritchie reproduces the surrounding Sixth Edition code and its explanation of a process switch. He describes a mechanism that returned through a stack frame saved by a different procedure. The PDP-11 compiler’s consistent return conventions let that arrangement work.
The Interdata compiler did not make the same choices about saved registers. Ritchie says the port took about a week of agonizing before they concluded that the control-passing mechanism itself needed replacing. They redid the primitives, and the famous comment disappeared with that code.
He puts the admission plainly: “we didn’t understand what was going on either.”
That is a better story than a trophy for incomprehensible programming. The problem was an assumption that had become invisible while the system stayed on familiar hardware. The code had an explanation. What it lacked was a valid guarantee across the next environment.
You can recognize the modern version without touching a vintage computer. A program succeeds under one toolchain, one operating system, or one convenient deployment setup. That success quietly becomes evidence for a broader promise the program never actually established.
When the next environment exposes the gap, adding more detail around the symptom can feel productive. Another compiler setting. Another special case. Another afternoon trying to make the new machine resemble the old one. Ritchie’s account is useful because the eventual repair changed the mechanism that owned the assumption.
For a reader learning systems programming, the surrounding explanation matters as much as the famous sentence. Quoting only the dismissive-sounding line turns a record of uncertainty into an attitude. Reading the passage restores the actual problem: how control moved between processes and what the compiler had to do for that movement to work.
For somebody maintaining software today, a practical question follows. If this works because two parts happen to behave alike, where is that agreement written down? If the answer is a compiler habit rather than a defined interface, a successful run has not settled the portability question.
None of that makes the original work foolish. It makes the retrospective valuable. Ritchie preserved the part where a working system depended on a mistaken model and the authors had to revise their understanding. I would keep the explanation beside the famous line, especially for the next person asked to make the system work somewhere its first authors never had to consider.
Open-book exam, please.