Post #1417022
2026-04-17 17:54 UTC
@mhoye @cks every failing test begins with a full investigation and that work can cross a dozen different team jurisdictions and take three months... or it could conclude in ten minutes when a dev realizes "oh shit, I accidentally included some WIP code in that commit which breaks everywhere but on my specific machine".
(That this kind of human error is even possible, let alone that its blast radius can include everyone who depends on the test suite, is a problem but a different one.)
Replies (1)
-
@cks@mastodon.social 2026-04-17 17:57
@gnomon @mhoye I can actually see a potential use for 'we know this test fails consistently and we're investigating so you don't need to keep re-running it all the time', but even if you want to do that, it should be driven by your team, not the infrastructure team, because they don't know if you need to keep re-verifying the failure (... because maybe the rack PDU got replaced again and suddenly it will start working).