Interesting Links [0x07]

For the Hallowe’en Edition(tm), we’re going to read about more scary bugs! And some of these bugs are interesting indeed.

• Tracking Down A Mysterious Windows Crash (in Chromium)
  Stack corruption caused by some (thing) hijacking a hook to a Windows API call. Typically this is an indication of a malware, but there are legitimate things that does that too, say like access restriction software.

• Hangs from thread holding lock expires, other threads stuck when the signal handler fires
  I can so relate to this one that it isn’t even funny :-/

• Time, technology and leaping seconds
  Who knew that a little thing like a leap-second can cause so much problems?

• The case of the 500-mile email
  This would be the most interesting bug stories, where the email can be sent only to a maximum of 500-miles away.

• What is the hardest bug you have ever debugged?
  Someone else’s hardest bug. Mine? Off-by-one error when storing some values required for stack unwinding at the red-zone boundary - everything will run perfectly fine, until once in a rare while where the signal handler gets fired, trampling over that saved memory location. Because it’s for stack unwinding, the cause of random crashes and execution dispatches to illogical places share no locality in time. If not for that time that I’ve read the x86_64 ABI Manual, and the epiphany came to me - could have never proved that I’ve fixed the issue, but I have never seen that crash again for 5 years now since I’ve adjusted the offset…

• When a crash happens on ‘mov ebx, eax’, there’s not many obvious explanations…
  I can so relate to this one - How in the hell can a segmentation fault occur when the CPU is doing no memory operations?! Explains why I had found and bookmarked Raymond Chen’s page after loads of frantic Googling. (Do read his blog, it’s got plenty of fun stuffs about Windows and all its quirks.)

• Whose bug is this anyway?
  You know, that one time, that you are so sure that it’s a compiler bug? Or maybe even the bug is in the CPU itself? Well, most times, it is PBKAC. But sometimes it isn’t.

• Epic programmer finds CPU bug
  Matthew Dillon (not the actor, but of the DragonFlyBSD fame) found a bug in the AMD CPU. I have to say I’ve got much likes for the OS itself. Their IRC channel is also probably the greatest among the Open Source communities; never been in one that you could learn so much stuff. Which largely explains why I haven’t stopped lurking there since.