I've worked on projects where tests were included in the final code that "could never be reached". If that part of the code was ever reached, it meant heavy corruption of the source data had caused the code to execute an illegal branch that should never ever be executed.

Therefore the error text was something along the lines of "you f'ed up roally, you useless piece of..."...

Same thing, after I left the company, somebody changed the code and they made a mistake in the logic and the code started being foul-mouthed!

As I've recounted here before, I was once tracking a mainframe bug report (a bit over 25 years ago) and got a message on my terminal stating that "if you reached this point you're fucked".

Although the original coder was no longer working with us, I eventually confirmed his prescient statement.

Any code "that could not be reached" will be.

Just another variant of Murphy's Law

Nearest I came was when a client wanted the source code, and I had to give it a quick once-over for any non-diplomatic comments.

Yes, I've been through and done that on a few projects, also for error messages when they went from test to production. The one above was internal and went into production after I left, and nobody sanitized the error dialogs!

We got a code drop from a supplier, and one of our engineers was looking through it to see what was new. He found the line 'printf("Fuck me, no memory left!\n")' and sent a note to the team saying "maybe we should change this?".

He got an immediate reply from our manager, saying "absolutely, it's not properly internationalized!"

Quite, it should at the very least be printf(ErrorMessages.FuckMeNoMemoryLeft);

"not properly internationalized!"

# include

printf ( _("Fuck me, no memory left!\n"));

Ça suce, plus de mémoire! # LC_MESSAGES=fr_FR

Oh Dear, we appear to have exhausted memory! # LC_MESSAGES=en_US

For the avoidance of this very situation, I have long since perfected the art of highly sarcastic euphemisms in comments, usually in response to the naïveté of one of my predecessors. Every now and then, I see one of the other devs chuckling away to themselves when they read one of these amongst the thousands of lines of long-forgotten code.

This is anecdotal, but knowing the source I believe them :-)

Source code for a new 'build' of telephone exchange code (Running nationwide!) was shared with $CUSTOMER. Their review process found a comment along the lines of 'How the f**k did I get here? Never mind, I'll just call to TRAP). Needless to say, 'trap' was a very bad thing :-)

Nothing sweary involving words (although some expletives were heard in my office), but I once got a weird error from a NAG (Numerical Algorithms Group) library routine for solving ordinary differential equations (ODEs). In this case I was using a routine with the perfectly sensible name of D02BAE (FORTRAN with its limits of 6 characters for identifiers at work). I was running several instances on the 32-processor Cray J932, and got an error message:

IMPOSSIBLE ERROR

Apparently, a negative value was found at some point in the loop where negative values should be impossible. The cause was fairly simple: FORTRAN doesn't deal with the notion of scope well, and data frequently needs to be shared through "common blocks". Apparently, D02BAE used some other elements of the NAG library, and shared data in a common block. However, in this case a single named common block was being shared by 16 different instances of D02BAE and its helper routines. so different instances of the routine were overwriting each other's data. They had forgotten to compile the library on our Cray J932 (long since dead) with the --taskcommon switch, which makes private copies of common blocks. There was no workaround for that (AARGH), so I wrote my own ODE solver.

I had the reverse recently, albeit fairly trivial. I was using an external control chip for some LEDs in a multi-threaded application on a Raspberry Pi. The results were just strange, things not happening in the right order, LEDs being off when they should have been on, etc. It turned out the library supplied didn't store the state of the chip in a common variable, it simply read the state before writing so if two processes ran at the same time (which they typically would at start up) you'd get Read - Read - Write - Write and all sorts of mess ensuing.

The fix was the same, write my own routine that properly handled the multi-threading.

You have a problem so you decide to use threads. you problems.2 have Now

Also not sweary, but a company I worked for acquired another company who had an NT 4 cluster. When it was ripped out and shipped to our office, it didn't work. I was asked to look at it. When I tried to start the cluster service it just popped up an error dialogue with the helpful message 'September'.

In the 90s I managed products with embedded code and the company discipline around comments was lax because the customer never got to see the source. We had a particularly awkward** customer QA, Pete, on one product and he could be a right-royal pain in the arse, which was frequently referenced in comments along the lines of "Pointless trap for an impossible error to keep the Pete-the-f***ing-pedant happy", and so on. What we didn't know is that the contract had a customer code-review option in the contract which Pete called up as part of final acceptance and we handed it over without even checking it ourselves! You can imagine the rest. No one was formally disciplined cos they'd have had to put the whole of SW engineering on a written, but stern words were had and it did result in a change in SW procedures on comments.

** i.e. good.

CAD was very new..... Our designers were tasked with creating a CAD drawing of the entire factory and offices on the new system to 'demonstrate its capabilities'.

Somewhere, in the corner of an office on the second floor at a scale unimaginable, Albert's caricature sits with arms folded in the classic pose as he admonishes a lowly apprentice.

Allegedly.....

...the presence or absence of vituperative error message in my code when something that should never happen occurs, usually due to the input data being so F-ed it deserved its own onlyfans account.

We once had a client ask "who is Chris Waddle and why is he in my network traces?"

It was ... tricky ... to explain. Here goes:

Why? Because Wake-on-LAN packets have a lot of FF in them, there was a bit of extra space in the ethernet frame, and we'd been watching far too much of "The Fast Show", specifically the "Chanel 9" skit.

I was the person who made final sanity checks before the code was released to the great unwashed, and spotted a few 'interesting' comments in the latest pull request, so I replaced them, with an added commit titled "formalised some comments".

I got a somewhat embarrassed email from one of our best developers apologising profusely.

Teachers fill in end of year reports, I proof read them for grammar, spelling, wrong pupil's name or gender used when copy and pasting comments etc.

I usually end up doing this late at night, so sometimes I miss things.

Also clearly working late at night, and sometimes after a few too many adult beverages, are the teachers who fill in these reports.

Usually the odd slip up is fine, but you only need to send one sweary report to a parent who happens to be a bishop for the shit to hit the fan, as I found out the next morning, and the teacher responsible found out an hour later.

I kept my job after stern admonishments, the teacher wasn't so lucky...

Back in the day, I got hired by a new employer. As the new job was a very long way from my previous the company was kind enough to arrange for the car I had ordered to be available before I formally started work so I could find somewhere to live and move my meagre collection of belongings to #NewCity. IT was awash with high performance cars in those days and I took full advantage of this to bag a car that no spotty 20-year old had any normal right to be in charge of.

Dutifully collected the car and drove it home, proud as anything. Took it back out that evening to show off to my mates and my lack of experience shone out: I gave it far too much power coming out of a bend and ran out of road. Upshot: damage to the passenger door and rear quarter thanks to a metal barrier and a smashed-up alloy wheel rim due to violent impact with the curb. 156 miles on the clock and my job didn't start for another two weeks! Fearing getting fired before I'd even started, I had the brother of a friend work his magic: he bashed the door back into shape, filled the dents and resprayed the whole side to make it all blend in. One new wheel from the main dealer and the job's a good 'un. If you looked at it from the rear quarter, it was obvious the door wasn't aligned - but I always parked it nose-out in the company's car park and got away with it. On straight roads, it had a tendency to run towards the middle of the road so I suspect the wheel damage was more serious than we allowed for.

The job didn't really work out and, only a year later, I was safely working elsewhere. My car, still on its three-year lease, got used as a 'pool car' so any number of people borrowed it and its oddities were never really noticed. Thankfully, it got stolen shortly before the lease expired - which was a weight off my mind!

/Anon obviously.