Post

Early in my career I was called on to figure out what the hell was wrong with a FORTRAN program which had worked just fine for years, but when a contractor updated some formulas, it stopped working.

They reverted to a backup and it worked, but whenever they tried to update the maths -- which were also full of weird, redundant calculations -- it stopped working again. Nobody could figure it out. It made literally zero sense. They'd change a long formula to a shorter one and everything went nuts.

What this means is that just like you can have a variable named x in C++, you can have a variable named 7 in FORTRAN. Or, in this particular case, 3. Apparently, the original programmer got tired of typing pi. But instead of defining a variable named "pi" -- well. You can guess, can't you?

No, this programmer assigned the value of pi to the integer 3. So at various points in the code he would have (1 + 2) or (4 - 1) in the maths, because when he used 3 it was pi. And nobody knew this, so they went "what the fuck is this" and corrected the weird redundant shit and the whole thing broke

Who on earth thought it was a good idea to allow the use of an integer literal as a variable name!? And how the **** did the original programmer not think twice when they had to use (4 - 1)!?

It's weirder than you think. In older versions of Fortran, it was pure pass by reference, so to make that work all numeric constants were compiled as variables that contained themselves as the value. Which means if you assigned to such a constant, you changed the meaning of that constant everywhere

Honest question: do you have a source on this? I started programming I'm fortran IV in 1975, and AFAIK it was never allowed...

Yeah I recall you needed to start with a letter... except i, j, k, l, m and n. Those were, uh, numbers.