Hacker News Reader

I had the joy of re-engineering a terrible in-house web server because of that bug.

It had used the system sleep to pause execution between the fake threads in their awful program. Suddenly everything was hitting race problems, instead of about 10%. Not something I really understood why it was going so bad, so I just dumped the inhouse implementation of threads, and took up the Solaris one, which was a joy by comparison.

View on HN

Sleep() returns early if a signal is delivered to the program, so doesn't your code need to handle this case anyway? Of course, this is super rare, so incorrect code could survive for a long time on a correctly implemented system.

this is the point of that terrible 'threading' strategy, actually, if i guess what they are doing correctly.

You should always handle your errors, but people trust in exceptions and "what works" often enough that we have a culture that doesn't.

http://geocar.sdf1.org/close.html

Reply on news.ycombinator.com