gh-90562: Mention slots pitfall in dataclass docs

[thejcannon]

Doesn't fix the issue, but does call it out so developers know that the issue doesn't lie with them (per-se) and gives them more info on where to go.

Will need to be backported to 3.10 and 3.11.

• Issue: RFE: @dataclasses.dataclass(slots=True) doesn't support methods using zero-arg super() #90562

---

📚 Documentation preview 📚: https://cpython-previews--107391.org.readthedocs.build/

[ericvsmith]

I can't say I've seen something like this documented before. I'd like to get some more opinions on it.

[thejcannon]

Yeah. If the exception was more clear, maybe it wouldn't be warranted. But it took me longer than I want to admit to put two-and-two together. And then, because I'm super nosy I was able to find the linked issue.

So I feel for the little guy who has to piece it all together. (Speaking of, I should edit to suggest the workaround, the "old-style" super call)

[AA-Turner]

We should try to capture everything relevant in the documentation -- what is relevant to the reader from the linked issue that isn't captured in your summary?

A

[thejcannon]

I think these two points are it. What to look for and how to workaround.

[carljm]

I feel like early in my CPython contribution I remember someone saying "we don't document bugs, we fix them."

But I'm also sympathetic to the point of view that this is a significant limitation that people are likely to hit and with an extremely confusing symptom, and there's no clear fix for it on the horizon, and it would be kinder to our users to surface it in the docs.

So while it bugs me slightly, on the whole I think I'm in favor of this. I'll leave the final say to our documentation experts (and @ericvsmith as dataclasses expert.)