On Tue, Oct 27, 2026 at 09:42:55AM +0100, Daniel Engberg wrote: > On 2026-10-27 07:30, Mathieu Arnold wrote: > > Daniel, > > > > That is true. That patch should keep the BROKEN line. This is not a > > problem. Anyone can see that and keep the line. > > > > Your personal preferences on how ports go away are only that. They are > > not policy. > > > > The long-standing policy is that ports that do not build get marked > > BROKEN. This has two effects. First, the build clusters stop trying to > > build ports that cannot finish. Second, the mark shows that the port is > > broken. People who care about the port can then look at it. > > > > After a while, a port marked BROKEN gets a DEPRECATED reason and an > > EXPIRATION_DATE. The date is usually a few months in the future. The > > change also needs a PORTREVISION bump. All of this must be merged to the > > quarterly branch. > > > > This shows that the port will be removed in the future, and why. The > > reason is in DEPRECATED. For ports that build, users see a notice when > > they upgrade. The notice says the port is deprecated and will go away. > > > > This is important. The ports tree is not a playground for committers. It > > exists so that users can use software. > > > > Your commit removed six ports. Only one of them had a BROKEN tag. You > > should have added DEPRECATED and EXPIRATION_DATE to the broken port. You > > should also have set those variables on all ports that depend on that > > broken port. Then all of them would be marked for deletion. > > > > In the future, please follow that procedure. > > > > It's not documented in either Porters Handbook or in the new policies as > such and there are conflicting messages, logical and practical shortcomings. > Since there's a discrepancy which several have recognized there's already a > discussion taking place and hopefully document it in the end.
Daniel, I wrote the policy above. If you do not understand it, please tell me which part you do not understand. If you refuse to follow it, please do not remove ports. Kind regards, -- Mathieu Arnold
signature.asc
Description: PGP signature
