And it’s been nine days since we learned 0.19.1 didn’t fix the problem. There’s a lesson to be learned here: don’t push a massive update close to the holidays because, ya know, you might break federation and go on vacation for a week+.
And it’s been nine days since we learned 0.19.1 didn’t fix the problem. There’s a lesson to be learned here: don’t push a massive update close to the holidays because, ya know, you might break federation and go on vacation for a week+.
There is also a lesson in implementing proper tests. During these holidays I started to play a bit more with Rust and went on to look at Lemmy’s backend code. Not a single unit test in sight…
We all await your commit containing all those simple unit tests with bated breath.
I want to work an issue that is open since 2020, but I can only justify dropping all my other work for that if I have enough paying customers interested in some new feature. So, help me get 50 customers to my “all in one” hosting service and I will dedicate a week to it, which should be more than enough time for even a Rust newbie like me to submit a proper PR to that issue. Ok?
There is already an open PR for Oauth, though Im not sure whats the difference to LDAP. Regarding tests you must have looked in the wrong place, we have a couple hundred already. Docs are here.
Unless there is some thing inherent to how Rust runs tests, the test script seems be mostly integration tests exercising the API. There does not seem to be unit tests which would be the first thing you’d need to have to test like the function that seems to be the source of the bug. (Hint, if your test suite needs to have a running db, you are not unit testing.)
As for OAuth vs LDAP: both could be used as part of a Single Sign-On solution, but the actual use cases are completely different. Having LDAP authentication would allow, e.g., to let users authenticate to Lemmy using an intranet account or (in the case of Communick) customers could use the same credentials for their Mastodon or Matrix accounts.
Sounds like you are looking at the api tests then. There are also unit tests which you can run with
./scripts/test.sh. The database is a fundamental part of Lemmy, so its obviously required for many of the unit tests. It may be possible to mock the db, but that would be a lot of effort for very little benefit, and might even introduce additional bugs.Yeah and plz send 2000€ and I’ll work on it too.
deleted by creator
To be fair, I’ve yet had a job that actually pulls off unit testing. Most either don’t bother or just go for the grunt work bare minimum to force pass tests. Most friends in my field have had pretty much the same experience. Unit tests can be just a chore with little to no real benefit. Maybe an opensource project that actually cares about its code can pull it off, but I wouldn’t bat an eye if they never get to it.
If you go through the comments, you will see that the devs talk about an issue with the logic in the for loop, which “may be stopping before it should”. Writing a couple of test cases that check whether this is true or not should be trivial.
I’d expect at the very least some type of regression tests to be implemented for every bug that makes into production, to avoid cases like this one where the developers spend weeks figuring out whether their patches even fix the bug in the first place.
I’m surprised that’s your experience. I’ve worked 4 software engineering jobs now, and every single project has had well done unit test suites.
What kind of job do you have, making websites?
Usually a React dev, have been some other stuff, but generally yeah, websites. Anything from resort chain websites to complex internal applications. Unit tests were optional at best in most jobs I’ve been at. I’ve heard of jobs where they’re pulled off, but from what I’ve seen, those are the exception and not the rule.
Edit: given the downvotes on my other comment, I should add that this is both anecdotal and unopinionated from my behalf. My opinion on unit testing is “meh”, if I’m asked to do tests, I’ll do tests, if not, I won’t. I wouldn’t go out of my way to implement them, say, on a personal project or most work projects, but if I was tasked to lead certain project and that project could clearly benefit from them (i.e. Fintech, data security, high availability operation-critical tools), I wouldn’t think twice about it. Most of what I’ve worked on, however, has not been that operation-critical. What few things were critical in my work experience, would sometimes be the only code being unit tested.
That explains why you haven’t seen many unit tests then.
I guess you haven’t really worked on complex software just yet too, like a 2GB sourscode 700 projects all in one where tests saves the day all the time.
So maybe you’re the lucky one :-D
Cheers
Same. Every team talks a good game about unit testing but in nearly 20 years I’ve yet to see a single worthwhile test in a company’s code base. I eventually stopped asking about it when doing M&A research.