regression: std no longer in toolchain dir

[Mark-Simulacrum]

https://crater-reports.s3.amazonaws.com/beta-1.83-3/beta-2024-10-19/reg/bevy_dexterous_developer_library-0.4.0-alpha.3/log.txt

root is in prefer-dynamic crate, probably just needs a compat note

[apiraino]

I quickly skimmed at the most recent merges with relnotes and didn't find anything from {t-libs,-api} about that. Very likely have missed that, just wanted to check if it's tagged for the changelog

[jieyouxu]

I quickly skimmed at the most recent merges with relnotes and didn't find anything from {t-libs,-api} about that. Very likely have missed that, just wanted to check if it's tagged for the changelog

As discussed in https://rust-lang.zulipchat.com/#narrow/channel/245100-t-compiler.2Fwg-prioritization.2Falerts/topic/.23132202.20regression.3A.20std.20no.20longer.20in.20toolchain.20dir, this is probably T-bootstrap (and maybe weakly T-compiler) change from #131188. But yes, we should probably add a compat note.

cc @Kobzol as you may know more about this, but I suspect the prefer-dynamic crate was relying on undocumented unguaranteed behavior and that this isn't considered a "breakage" in the strict sense.

EDIT: nevermind @bjorn3 already commented on the prefer-dynamic crate WilliamVenner/prefer-dynamic#2, oops.

[Kobzol]

Yeah I think that if anyone depended on libstd.so within the lib directory, it was an instance of Hyrum's Law.

[cuviper]

This will be documented as a compat note via #133352.

[jieyouxu]

I think this can be closed since 1.83.0 is already released containing said compat notes?

[apiraino]

I think so, so I will go and close this one.

This thing of closing regressions accepted for one reason or another was mentioned on Zulip but I'd want to document this reasoning in some place to make it more "official" and discoverable.