Skip to content

update yank section#36

Open
rafaqz wants to merge 2 commits into
SciML:masterfrom
rafaqz:patch-1
Open

update yank section#36
rafaqz wants to merge 2 commits into
SciML:masterfrom
rafaqz:patch-1

Conversation

@rafaqz

@rafaqz rafaqz commented May 6, 2024

Copy link
Copy Markdown

yanked = true is rarely merged in the general registry these days, so we should edit the recomendations to point towards editing Compat.toml instead.

Possible a concrete example will help here?

@rafaqz rafaqz changed the title update yank section WIP update yank section May 6, 2024
@rafaqz

rafaqz commented May 6, 2024

Copy link
Copy Markdown
Author

Some comments from @StefanKarpinski on slack regarding the current permissible context for yanking:

  • If a version is actually dangerous to use, it should be yanked
  • Like it has a security vulnerability or someone accidentally committed an rm -rf / call
  • “Oops, I released a version with a bug” is not a reason

And in response to: "But the readmes for general registry and ColPrac still both say to yank if you register a version with too wide compat bounds"

  • No, that’s not right. You should just fix the compat bounds in the registry
  • And ideally, also make a release that has the right bounds in the project file
  • Largely just because it’s confusing when those don’t match

Comment thread README.md Outdated
@rafaqz rafaqz changed the title WIP update yank section update yank section May 6, 2024
@LilithHafner

Copy link
Copy Markdown
Contributor

See also:

Add guidance recommending yanking in General part 1: JuliaRegistries/General#102636 part 2: JuliaRegistries/General#104800

Ongoing difficulties of retroactive compat bounds adjustment: JuliaRegistries/General#104849

Parallel PR to General: JuliaRegistries/General#106278

I recommend that this PR follows the course of JuliaRegistries/General#106278, whatever that happens to be.

@rafaqz

rafaqz commented May 7, 2024

Copy link
Copy Markdown
Author

But in practice compat bounds are preferred, and yanks are rejected:
JuliaRegistries/General#102850
JuliaRegistries/General#104235

These guidelines should match practice?

@giordano

@ChrisRackauckas

Copy link
Copy Markdown
Member

That is inconsistent with what's normally done. Normally the compat bounds are not manually updated. As @LilithHafner points out, it's recommended that devs don't manually muck with compat bounds like that because there's many issues that can crop up and no tools. That's just weird. That should've been a yank.

@rafaqz

rafaqz commented May 7, 2024

Copy link
Copy Markdown
Author

See comments by @StefanKarpinsky above too... he said compat bounds! This is all very confusing.

@ChrisRackauckas

Copy link
Copy Markdown
Member

That's not how it's generally done in General though? Compat bounds changes to General are not merged because they aren't even adequately tested: you have to manually test whether the registry version you get will be installable. If you do it wrong then CI can pass but you can then crash all installations. That's why it stopped being recommended.

@rafaqz

rafaqz commented May 7, 2024

Copy link
Copy Markdown
Author

The compat changes above were first reccomended and then merged by @giordano

The yank was closed. I had another yank merged only after seriously pestering people on slack.

(I prefer the yank too Im just trying to resolve guidelines with practical experience)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants