About 20 years ago when I was still in school I remember a bug in C++ that took me a whole week to find, where I was accidentally returning a pointer to a local variable. The embarrassing thing was that the compiler warned me about it, but I was so laser focussed on finding why my program was crashing that I decided to leave fixing the compiler warnings til later.
About 10 years ago I spent 3 months on and off trying to track down a memory leak in a nodejs server process. I can’t remember what the fix was - I’m sure it was something stupid. But along the way I learned to connect the chrome dev tools to nodejs and I figured out how to use heap snapshot diffing, and what roots are, and how V8 thinks about memory. I still use that knowledge to this day.
I could tell lots of stories like this. That time I left a debugging statement in some performance critical code and for 3 months it ran 20x slower than it should have. How a diffing function actually needed to be a 3 way diff to function correctly - else some weird corner cases would cause problems. And so on.
Do you seriously not have interesting stories? Are you just not curious when things go wrong, why they go wrong? This sounds like a great interview question to weed out people who don’t have any passion to become excellent engineers.
Your brain remembers things that you think about a lot, or that had a strong emotional experience. That’s how memory works. People who think a lot about engineering become better engineers.
As for strong emotional experiences, it seems obvious to me that people will have a wide range of different emotional reactions to their work. None of these reactions is inherently better or more correct than any other. Here I think you are falling into the common interviewer trap of judging people negatively for not being exactly like you.
It seems like some people have this sequential narrative of their experience in their head, where events are strictly ordered—but for me it's more like a pool, or an unordered set. I learn things Just In Time, they go in the pool, and eventually they fade away if no longer relevant (goodbye, VB6). That doesn't mean I don't care about the craft, it just means I'm focused on the future rather than the past.
One I didn't work on, but was the sort of stuff legends are made of. A very specific revision of Blink that was never shipped as a Chrome release but got shipped as BlackberryOS's default browser had a bug around one of the bitshift operators (can't recall which), which would give a wrong result around once in a million operations. The desktop build of that particular revision would only trigger that bug semi-reliably, but the Blackberry build would trigger it deterministically. Bitshifts are not particularly common in most JS codebases, but this was a bitcoin wallet, and bitshifts are used by the thousands as part of e.g. key derivations. "Key derivation" is exactly where this bug triggered for a user, and he lost a whole bunch of BTC to that bug. We were almost 100% sure the guy was trying to defraud us with that bug report but something made us itchy and we had somebody keep digging until we found the bug. The good news is that, because the Blackberry version was deterministic, we managed to reproduce the bug to the point of recreating the badly-generated keys and successfully recovered the guy's BTC.
One I worked on myself, diagnosing it involved attaching both IntelliJ's debugger and GDB to a running Java application, so I could set breakpoints both in Java land and in the C++/JNI component of a JDBC driver. My colleague tracked it down to an NPE at that boundary, and was stuck on how to proceed from there.
A fun one from a couple of years involved a script that ran fine on Firefox but failed on Chrome. This was around the time when fat arrow notation was introduced for JS functions. Introducing that involved changes to the parser, of course, and a vendor's minified scripts triggered a bug in the Chrome's JS parser such that it was seeing a fat arrow where there wasn't one. We solved that one by re-minifying the vendor's minified script with a different minifier that wouldn't generate the offending pattern.
I suspect this comes down to biological or neurological differences. As someone who remembers lots of trivial facts but cares not for the details of bugs, it might also just be a matter of whether your passion is your work.
Yeah I agree. But in that case, people with a passion for their work make better employees. A good interview process is actively looking to select people like that.
I'll never forget the bug where a function took by copy a mutex (instead of passing it by reference) so locking the mutex didn't work which triggered threading issues, it took me two weeks to fix it, threading issue like this are evil: debuggers are useless, logs are useless..
If you’re the candidate, practice reframing any question you get asked into the variant of the question that brings you the most alive. Your interviewer, and your career, will thank you.
and honestly, it's not about the specific answer: it's about creating enough conversation that you can understand the person. i usually use these two with senior+ engineers, but if i'm interviewing someone more junior, i will ask something different (like, what are you learning right now? what do you wanna understand over the next 12 months? is there anything that you currently do not like in tech?).
The bugs that matter aren't potholes, they're trails with no markers, cut by someone you don't know who has left the organization years ago.
This is more like "What was your harder part of the Rubicon trail?"
I spent 7 months debugging a single issue which happened once a billion requests across 100k web servers or so.
For those 7 months, most machines did a hard restart every 4 hours and the bug never reproduces, so I didn't even know if I fixed it until I had enough confidence to turn that restart loop off.
Finding the issue had a fork of valgrind in the middle, found a CPU errata with AMD hardware when 2 locks placed in the same cache line (but not all failing machines had that CPU version), found a "bug" in UFS FreeBSD where deleting a file + recreating it uses the same inode id (not a bug, but not how linux worked)
The fix was 2 lines.
Your answer is not the wrong answer, but the bugs that you encounter in the wild is often like a tiger hunt in a jungle, alone.
A few bugs are really interesting. I can still describe my most interesting bug, in detail, and it was about 13 years ago.
I'd be hard pressed to choose a "favorite" bug, let alone remember the details years later. I'm trying to build working software, and bugs are just obstacles on my path. I fix the bug and then move on to what I actually care about.
That would be like asking a driver, "what's your favorite pothole you've ever driven over?"