At one of my old jobs, we had a suite of browser tests that would run on PR. It’d stand up the application, open headless chrome, and click through stuff. This was the final end-to-end test suite to make sure that yes, you can still log in and everything plays nicely together.
Developers were constantly pinging slack about “why is this test broken??”. Most of the time, the error message would be like “Never found an element matching css selector #whatever” or “Element with css selector #loading-spinner never went away”. There’d be screenshots and logs, and usually when you’d look you’d see like the loading spinner was stuck, and the client had gotten a 400 back from the server because someone broke something.
We put a giant red box on the CI/CD page explaining what to do. Where to read the traces, reminding them there’s a screenshot, etc. Still got questions.
I put a giant ascii cat in the test output, right before the error trace, with instructions in a word bubble. People would ping me, “why is this test broken?”. I’d say “What did the cat say?” They’d say “What cat?” And I’d know they hadn’t even looked at the error message.
There’s a kind of learned helplessness with some developers and tests. It’s weird.
On the context of a node package, I’m pretty sure that “solution” is utterly worthless and doesn’t come even close to targeting the same functionality the old code had.
But odds are the one place the library author used that function can be replaced by a completely different functionality that happens to use the suggestion.
My #1 pet peeve is when someone comes to me with a problem, and the solution is in the fucking console output or error message.
On a bad day, if I had unilateral power, I would fire those people on the spot.
At one of my old jobs, we had a suite of browser tests that would run on PR. It’d stand up the application, open headless chrome, and click through stuff. This was the final end-to-end test suite to make sure that yes, you can still log in and everything plays nicely together.
Developers were constantly pinging slack about “why is this test broken??”. Most of the time, the error message would be like “Never found an element matching css selector #whatever” or “Element with css selector #loading-spinner never went away”. There’d be screenshots and logs, and usually when you’d look you’d see like the loading spinner was stuck, and the client had gotten a 400 back from the server because someone broke something.
We put a giant red box on the CI/CD page explaining what to do. Where to read the traces, reminding them there’s a screenshot, etc. Still got questions.
I put a giant ascii cat in the test output, right before the error trace, with instructions in a word bubble. People would ping me, “why is this test broken?”. I’d say “What did the cat say?” They’d say “What cat?” And I’d know they hadn’t even looked at the error message.
There’s a kind of learned helplessness with some developers and tests. It’s weird.
i think that guy works for us now :D
the best bit is he pings multiple senior devs in slack separately.
so we are all wasting our time doing the same shit in parallel for the same muppet.
Another reason why I don’t want more copilot and chatgpt to beginners
Same here. For some fucking reason reading is so damn difficult
Error: pull your head out of your assHey I got this error what do
to be fair you couldnt read the error in that situation
Error message: “you must manually run ‘sudo dpkg --configure -a’ to fix this”
Junior dev: 😵
Sometimes I’ll copy paste the error message back to them. Apparently it works better when it’s in a text message.
On the context of a node package, I’m pretty sure that “solution” is utterly worthless and doesn’t come even close to targeting the same functionality the old code had.
But odds are the one place the library author used that function can be replaced by a completely different functionality that happens to use the suggestion.