4 ms·
I once spent nearly 2 weeks as a team of three chasing a intermittent bug in some multithreaded Windows NT codebase (around 1999). We were sure it was some subt
by amichal 5y ago
I once spent nearly 2 weeks as a team of three chasing a intermittent bug in some multithreaded Windows NT codebase (around 1999). We were sure it was some subtle issue with multiple CPUs since it only happened on the one quad cpu machine we had and took hours of 100% cpu between failures. Feeling increasingly foolish and relatively new to multithreaded code I was sure I was missing something obvious. The guy with a EE/hardware background comes in super early on a Monday (like 3am) and gets the longest run in yet without a crash and then notices that the crash happens increasing often throughout the day. He opens the case of our rented beast of a machine none of us could afford to risk damage to and reseats the CPUs. Test loop ran at full 100% CPU load for 72 hours straight.
- herpderperator 5y agoWhat is the significance of the 3am early morning and later realisation that it happens more frequently throughout the day? How did that lead to him wanting to reseat the CPUs?
- ninjanomnom 5y agoI can at least answer why reseating the cpu was an early response. Heat is one of the few things that would persist across full reboots and gradual heat buildup could easily cause these symptoms. There are a lot of places to check still if that's all you have to go on but the cpu is a decent first choice as it's not uncommon for the thermal past to be applied poorly.
- jfrunyon 5y agoIf you run into an issue that takes longer to occur after a cold start, and gets more frequent with use, odds are very good it's heat-related.