Our Flaky Test Was Actually a Daylight Saving Time Bug Hiding for Three Years
It failed twice a year, always on a Monday, and we kept clicking re-run.
Backend engineer in Columbus, Ohio. I like boring code, short meetings, and deleting things nobody uses anymore.
It failed twice a year, always on a Monday, and we kept clicking re-run.
A routine scan found it running firmware from 2016, an open admin page, and a copy of every document scanned for two years.
It is the cheapest design review I know, and it has killed three of my own bad ideas.
It cost $410, compiles our service in 40 seconds more than the new one, and taught me what I actually need.
Nobody asked me to remove the dead code, which is exactly why it had been sitting there for six years.
It does five things, the battery lasts eight days, and she has never once asked me how to use it.
Our team of eight ran the experiment for 13 weeks, kept a log, and the only casualty was a few unnecessary late-night replies.
We had 312 flags. Forty of them had been on or off for so long that nobody remembered what the other branch did.
"Why didn't you just..." felt helpful when I typed it and landed like a slap every single time.