#wtf
-
Tell me the story about the weirdest bug you’ve ever found that left you wondering how computers even worked in the first place.
There are some fantastic gems in that list and don't want to spoil them all, but here's a couple to tempt you:
1993 SuperSPARC bug where if you depended on the bit that was physically on the corner of the CPU on chip cache array it would sometimes fail. Gcc broke it, SunC compiler didn’t. Took a while to figure out it was a hardware problem. Disabled associativity to fix it.
Found a bug in Valgrind that prevented you from running Firefox under Wine under Valgrind, had to debug using Firefox under Wine under Valgrind under GDB
-
-
-
@pie_flavor said in #wtf:
Not too many people even had enough room for a 1.6 GB outgoing mail spool in 1996. And those who did didn't accept anything above a few megabytes …
-
@Tsaukpaetra said in #wtf:
Firefox under Wine
:butwhy.gif:
Just guessing: debugging Wine's "IE browser control" which was actually running Firefox underneath?
-
@thegoryone Maybe the software was impressed with the blazing 56k speed
-
Not too many people even had enough room for a 1.6 GB outgoing mail spool in 1996. And those who did didn't accept anything above a few megabytes …IIRC it maximum message size was often set to a bit less than 1.44 MB, to prevent warez images of 3.5 inch disks being sent through the mail.
-
@Tsaukpaetra said in #wtf:
Firefox under Wine
:butwhy.gif:
You left out the clue: Valgrind. Sounds like he was trying to find out
Firefox was doing to leak memory.
-
@svieira My favourite of recent years was finding that with gcc newer than a certain version,
f(-g(x) * y)would blow up with an internal compiler bug (the function names are unimportant; the fact that this was using fixed point arithmetic was crucial). Where all the bits and pieces could be in totally different functions and only be brought together by the optimizer. And the internal compiler bug would be reported against the function where the optimizer had actually dumped the code, which was totally different to where the problem pieces were (in a different file, no less).Second favourite: discovering that the supercomputer also functions as a large scale radioactivity detector, just like a physics experiment. Just fire the thing up and wait for those bit errors to occur in your code…
-
Nobody posted the 500-mile email yet? It's a classic, but somebody may not know about it.
The weirdest one I encountered, successfully analyzed and didn't erase from my memory by drinking was me running into GCC bug 66139, a very nasty and relatively easy-to-hit bug that breaks all of C++'s safety guarantees. I don't remember the details, but it actually made a unit test fail somehow. Because it's never the compiler until you've accepted it's never the compiler and no longer expect it to break correct programs.
Debugging ODR violations that randomly fail during linking is also a lot of fun. Especially if your C/C++ code is generated by scripts (don't ask). I don't remember if it was the static or dynamic linker, but one of them basically threw a coin to decide whether the "correct" or "wrong" function would be called. I banged my head against the desk a lot that day.
x.com
