#debugging
- The capacity check that let all nine throughNine writers hit one limited slot at the same moment and all nine got on. The capacity check ran for every one of them, the lock was the right lock, and a one-user test passes it every time.
- The diagrams carried the argument, and screen readers got none of itNearly every post on this site opens with a diagram that does the explaining. For as long as the site has existed, not one of those diagrams had alt text, and the theme I'd picked had left nowhere to put it.
- My deploy ran my tests, and my tests ran my deployMy desktop ran out of memory five times in two days, once killing 29,082 processes at a stroke. Two files that were each sensible on their own had made a loop with no way out.
- A memory limit made Grafana hammer the diskGrafana stalled for minutes while the host had more than 4 GB of memory available. The clue that it was starving for memory was the NVMe reading 1.7 GB a second.
- A third of my request was the garbage collectorA slow grid in a Flask app I maintain had an obvious suspect. The biggest cost was Python re-checking 134,000 objects that would never be freed.
- The date that was yesterday until 3amA one-line date bug that only existed between midnight and 03:00, in an app whose heaviest users work nights. Then my fix for it turned out to be wrong in two of the three places I applied it.
- The badge that said 21 for a count of 1I added a small shield badge showing how many months someone had won a challenge. On my own profile, with one win, it said 21. The sample render I had checked it against said 1.
- Safari could not download the fileA download button in an app I maintain failed on a Mac with WebKitBlobResource error 1. The fix was to delete code: the server had been able to do the job by itself all along.
- The scanner that was never brokenThe scan half of my HP OfficeJet had never worked from this machine, and I'd filed that under driver problems. No driver was needed. There was no scanner software installed at all, and the real mistakes started after I installed some.
- The 144 Hz monitor that preferred 60My 144 Hz panel ran at 60 from the day the machine was built, and again after a graphics card swap. Different cause each time, and nothing reported a problem.
- I wrote the lesson down. I broke it two days later.My project keeps a lessons-learned file. Going back through my notes, I found the same pattern five times: the lesson gets written down, and the repeat comes anyway.
- Hugo's module.mounts quietly replaced my static directoryEvery deploy was green, every page worked, and one file returned 404 in production. The cause was a Hugo module mount behaviour that is documented but easy to miss.
- The particle fade that never fadedI wrote fade-out code for explosion particles in a Pygame project, and it looked fine to me for months. It never faded once, and the reason is a quiet rule about how Pygame draws.