Nobody Expects... | Hacker News Reader

Nobody Expects...

os2museum.com

65 pointskencausey22 comments

What a pleasant surprise, there are people still debugging code from 1990.

I occasionally reverse-engineer code from that era for fun and it's still very manageable. Imagine what it will be like reverse-engineering our current code in 30 years. Everything is just so much more complex.

I spent today patching a binary which was compiled in the 80s - an infocom interpreter of all things - so that I could play hitchhikers's guide to the galaxy on a single-board computer.

So yes, people will probably keep doing things like this long after you think programs are obsolete.

Your comment makes me wonder what the largest elapsed time duration has been so far between the previous release of a popular software project and a confirmed, accepted bugfix.

(naturally it'd be easier to determine that for free/open source projects, although only a subset of the overall space)

Edit: add the word 'popular', to potentially make the answer more widely-accessible by filtering out tiny/niche projects

Self-reply: perhaps an easier way to express this question would be:

"What is the most reliable popular software product in use today, determined by the absence of reported bugs?"

Probably just one example: Mozilla fixed a 14-year-old bug in Firefox https://news.ycombinator.com/item?id=10274245

Impressive - tagged for mozilla 0.91 in the bugtracker twenty (!) years ago.

I didn't express the criteria I had in mind very clearly; I'd been wondering about longest elapsed time without any other bugreports occurring (ruling out Mozilla, and probably many other large codebases).

Essentially trying to determine ways to find the best-known examples of stable and reliable widely-used software.

TeX and METAFont just had bug fix releases after 7 years or something.

Thanks - good examples. That also led me to https://www-cs-faculty.stanford.edu/~knuth/boss.html which is a unique curiosity.

Do you know if that was a 7-year duration from previous release to bug discovery, or a 7-year duration between releases?

It’s the latter (and not sure if it is 7 years) - but Knuth only looks at big reports periodically (and he has people who scrub them for him) and has declared that upon his death any remaining bugs are features and the code will cease changing.

The bugs are super obscure edge cases now - I suppose if someone found an extremely severe bug Knuth would release earlier than the next planned interval.

http://www.tug.org/TUGboat/tb42-1/tb130knuth-tuneup21.pdf has further details. It’s from a different time where programs could be considered to have a “final form” and be completed. True software engineering.

The BSDs had a bug in the seekdir() function that was only discovered after 25 years: https://msys.ch/fixing_seekdir

Incredible, and what a great write-up. I think that's the front-runner at the moment :)

Edit: addendum:

To correct my own comment: I guess technically this doesn't pass the "without any other bugreports" criteria, so it's in the same category as the Mozilla issue report, and perhaps the sort algorithm is the front-runner.

Perhaps it'd be possible to add some kind of focusing criteria around the 'size' of the module/component - since any large enough codebase is - at least using today's development methodologies - likely to include bugs.

14? Hah.... a mere youngster. https://bugzilla.mozilla.org/show_bug.cgi?id=92176 was close to 20 years old when it recently got fixed.

(Pretty sure I've seen even older issues eventually get fixed, but don't recall details offhand.)

Jon Bentley's Programming Pearls had a latent bug in its binary search program for 20 years.

This is so humbling. Not only did the master miss this, despite proving it correct, but also the Java and C++ library authors.

https://ai.googleblog.com/2006/06/extra-extra-read-all-about...

That website is borked, half the content is off screen on mobile.

What in the world means "borked"? Is it some kind of hybrid of "botched" and "broken"?

Ha yes, exactly. There are a few different meanings based on context. But it's primarily just one of those fun sounding nonsense words that I also like to say. Not sure where I picked it up from. It is usable in any sentence as a drop in like a G-rated "fucked".

There's the nonsensical swedish chef muppet catch phrase "bork bork bork", but he usually says that before everything goes wrong.

There was also a judge named "Bork" who was "borked" when his nomination to the supreme court was blocked.

I feel like "borked" is probably of a vintage like "fubared" which is a WW2 thing.

Funny. I just watched this video from 8-Bit Show and Tell last night: https://youtu.be/dG_oZV6s0Zg . This bug appears to be from 1978. Of course, we have no idea if this time was the first time the bug was discovered.

I programmed MASM in the 1990s and found myself deciding exactly where to put the 90H bytes in order to make the code work. It is good to see this stuff.

https://www.youtube.com/watch?v=T2ncJ6ciGyM

At least the code can be easily patched. Back in those days there was no driver signing, but some code would validate checksum/CRC (which is pretty easy to overcome). I doubt that there was any checking at all in the NE2000 driver.

From last week, another NE2000 article from the same site: http://www.os2museum.com/wp/was-the-ne2000-really-that-bad/

I was thinking this was going to be an article on the Spanish Inquisition.

(Sorry)

It is not the Spanish Inquisition if you expect it.

Reply on news.ycombinator.com