derive structs (inert attribute?) regression 1.39

[Mark-Simulacrum]

https://crater-reports.s3.amazonaws.com/beta-1.39-1/beta-2019-09-28/reg/derive_less-0.2.0/log.txt
https://crater-reports.s3.amazonaws.com/beta-1.39-1/beta-2019-09-28/reg/rs621-0.5.0-alpha2/log.txt
https://crater-reports.s3.amazonaws.com/beta-1.39-1/beta-2019-09-28/gh/andrei-cacio.distil-wasm/log.txt
https://crater-reports.s3.amazonaws.com/beta-1.39-rustdoc-1/beta-2019-09-28/gh/nasso.get621/log.txt

possibly cc #63221?

cc @petrochenkov

cc #64962

[petrochenkov]

This is a consequence of #63468.
Attributes on fields, variants and other places are now properly passed through name resolution.
This means we now detect attempts to apply derives to them, while previously they were just ignored.

The new behavior is correct, but we can make a deprecation lint for this if necessary.

[Centril]

The new behavior is correct, but we can make a deprecation lint for this if necessary.

The regressions do not seem numerous or serious enough to warrant that work imo.

[pnkfelix]

triage: P-high to decide whether to close as known-bug with no deprecation lint. Nominating for discussion at T-compiler meeting to decide above.

• To be clear: my own inclination/vote is that I'm with @Centril here: I think we can close this as "bugfix" with no deprecation lint. But I want the team to have a chance to discuss.

[nikomatsakis]

We discussed at the @rust-lang/compiler meeting and decided that the scope of this fallout doesn't merit a future compatibility warning. Therefore, I am closing this issue.