I spent nearly six hours tracking down a bug that was caused by not blocking signals correctly. As a result, signal processing could occur in the middle of a critical section. To make matters worse, the crash would only occur much later when the desync from the earlier overlapping execution would result in use-after-free and crash. The crash would not occur every time since it was timing-dependent - the signal would need to hit a relatively small window to trigger the issue.
Figuring out the root cause was a lot of manual elimination of potential causes and old-fashioned mental gymnastics. When everything else was ruled out, the only explanation was that two supposedly synchronised calls really were overlapping, despite this being supposedly "impossible". Once this fact was established, I could turn into verifying signal blocking, and the root cause was found rather quickly.
Is it weird that I loved spending six hours of my Sunday on stuff like this?
