hey @pnkfelix , re this, do you want me to extract the doctests and bisect their code to see the commits which do change resolution wrt extern crate and the likes ?
the TLDR of this issue is: there were some rustc bugs which seem fixed, there are invalid tests, there were some rustdoc changes which started detecting these invalid tests, and there might still be a miscommunication between rustdoc and rustc wrt the soft_unstable lint
since they're now fixed I don't think we really need to, but I'll extract the soft_unstable piece of code and post it on the tracking issue
deny by default, but the error message is
error: use of unstable library feature 'test': `bench` is a part of custom test frameworks which are unstable --> src/lib.rs:384:7 | 9 | #[bench] | ^^^^^ | = warning: this was previously accepted by the compiler but is being phased out; it will become a hard error in a future release! = note: for more information, see issue #64266 <https://github.com/rust-lang/rust/issues/64266>
so maybe it's not a miscommunication, but just the message confusing me
let me know if we need to do more testing
maybe I misunderstood what happened here
I had thought the issue was now hypothesized to be: There was no actual changes to resolution (we think) -- instead, a change to the doc-comment parser used by rustdoc caused it to start picking up tests that it wasn't picking up before,
and some of those tests were buggy
but those buggy tests were skipped (and thus silently accepted) in the past by crater
which means I don't think there's any follow-up work for the rustc team to do. The only questions I could imagine are whether the rustdoc team wants/needs to do anything to "address" this
(I suspect the answer at this point is "there's no value in going back to the old parser, or even trying to warning-cycle this change" -- but its not my decision to make, I think.)
there are many crates in that issue and that describes most of them yes, however some of the code which was ignored and unignored seemed to showcase rustc issues, which seemed either known, or have since been fixed
the other issue which could be a duplicate, and mentions the new test detection behaviour also mentions what you say, unlikely to rollback as it fixed other issues, but I don't know
If it seems to you that all of the rustc issues that have been identified this way are all already known and/or fixed
then I still don't think there's any follow-up work for rustc team?
Or is your point that you want to know how we can determine
if we need to look more deeply or not?
I _could_ bisect those if you want, but since they seem fixed I'm not sure we need to
yeah I don't think we need to do that at this point
Now I understand your Q
that is, I expect bisection to point to these known fixed issues
the only new information I had was discovering a deny lint had a lint-like message and that's it
Right. The main benefit I could see from bisection would be to double check if the problems being fixed here are actually covered by whatever tests were added when the relevant bugs were fixed.
(as in, there might be E-needs-test cases here)
but I don't think its worth the effort
we have got a lot of other bugs that I think deserve attention
alright, I agree :)
I think we can close this one then, the "rustdoc changing behaviour" problem is probably for the better and "tracked" in another T-rustdoc issue #61562
sounds good to me