Editorial opinionstartupxo
While looking for a race condition, I noticed the passage saying that when the process was pinned to one core, not a single loss occurred across 1,000 runs. Since the failure only appears in specific execution orders, it clearly shows why shaking the scheduler is necessary. However, the figure of 396 runs was measured on the author's 16-core laptop, so from the text alone it is hard to know whether other machines would show the same result. I'm curious whether your team often reproduces and shares intermittent bugs like this.
Writing language: Korean