Taken from: https://discordapp.com/channels/442252698964721669/443151225160990732/555416685260046337
[4:47 PM] pnkfelix: hey @centril , you've been interested in trying to get discussion started about how we schedule turning future-incompatibility lints into hard errors, right?
[4:47 PM] centril: @pnkfelix yep
[4:49 PM] pnkfelix: how broad do you think such a conversation needs to be? I suspect it may be quite a cross-cutting concern
[4:50 PM] centril: @pnkfelix broad in the sense of the personnel involved? I think T-Lang + T-Compiler members are a sufficient set of people probably
[4:50 PM] centril: or a subset of those
[4:51 PM] pnkfelix: I need to review how it interacts with cargo and its use of --cap-lints
[4:52 PM] centril: @pnkfelix oh about that... can you #![allow(..)] a future compat lint?
[4:52 PM] centril: to make the warning go away?
[4:52 PM] pnkfelix: or at least @SimonSapin (ping me on IRC) raised a concern to me about whether upstream crates would get broken while their clients wouldn't know due to --cap-lints silencing the lint
[4:52 PM] pnkfelix: I believe today you can #![allow(..)] a future compat lint.
[4:52 PM] pnkfelix: at least, I know I landed a test that did so
[4:52 PM] pnkfelix: for nested_impl_trait
[4:52 PM] centril: @pnkfelix I think maybe that's not so good
[4:52 PM] pnkfelix: I agree
[4:53 PM] centril: its not like any other warning
[4:53 PM] pnkfelix: agree again
[4:53 PM] pnkfelix: so that's one reason why this may be a cross-cutting concern
[4:53 PM] centril: right; it is probably
[4:53 PM] centril: but cutting across what teams? :wink:
[4:54 PM] pnkfelix: everyone
[4:54 PM] centril: oh... that's a large group of people ^^
[4:54 PM] pnkfelix: (cue clip from "Leon" aka "The Professional")
[4:55 PM] centril: (saved in a tab)
[4:55 PM] centril: My view on scheduling: we should have a standard set of dates for warning => deny => error based on N where N is the number of crater reported regressions
[4:56 PM] centril: those dates may be adjusted based on what happens but they are a good baseline/guideline
[4:57 PM] centril: @pnkfelix realistically I think to have anything actionable here we must narrow everyone to a group of people or if we need broader consultation an RFC under the purview of some teams...
[4:58 PM] pnkfelix: an RFC might be prudent
[4:59 PM] pnkfelix: only question is whether to try to form an actual WG dedicated to the question, or just try to press on without that overhead
[4:59 PM] centril: @pnkfelix maybe wg-meta?
[4:59 PM] centril: probably not
[4:59 PM] centril: an rfc sgtm; I think t-lang & t-compiler are the affected teams so those suffice
[5:00 PM] pnkfelix: I think cargo might be affected. Not sure.
[5:00 PM] pnkfelix: I suppose it will depend on how we attack the --cap-lints part of it
[5:01 PM] centril: @pnkfelix so I think there are 2 independent questions: 1. how to tackle allow => warning; 2. how to tackle scheduling
[5:01 PM] pnkfelix: if we manage to deploy a change to how future-incompat lints are handled, such that they simply cannot be silenced, then maybe that won't require cargo changes.
[5:01 PM] centril: I think T-cargo is not relevant for 2. but it is possibly for 1.
[5:01 PM] pnkfelix: Ah true, its good to tease apart the problems here.
[5:03 PM] centril: @pnkfelix you know how I don't like Zulip? Well... I'm starting to come around to your way of viewing it... It's still ugly but I was going to suggest we move this into a topic there :blush: ^^
[5:03 PM] pnkfelix: I'm happy to move this to Zulip
[5:03 PM] pnkfelix: maybe under #wg-meta
[5:03 PM] centril: sgtm
[5:03 PM] pnkfelix: since I agree they seem most relevant.
[5:04 PM] centril: @pnkfelix are you summarizing this discussion? I can otherwise
[5:04 PM] centril: ah great