Re: pending-pullups

[Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>]

jnemeth@ wrote:

> On Nov 29,  3:14pm, Izumi Tsutsui wrote:
> } dholland@ wrote:
> } > On Sun, Nov 29, 2015 at 12:58:35PM +0900, Izumi Tsutsui wrote:
> } >  > > (the reason, besides "that's how it's been used since it was added",
> } >  > > is that the pending-pullups state distinguishes "PRs that somebody
> } >  > > needs to work on" from "PRs that are waiting on releng", which is an
> } >  > > important distinction when looking for work to do.)
> } >  > > 
> } >  > > we ought to have a pullups-needed state; maybe we should just add it
> } >  > > to gnats.
> } >  > 
> } >  > Why do you think two independent states are necessary? What's benefit?
> } > 
> } > So that someone searching gnats can tell if a PR they're looking at
> } > requires action or not.
> } > 
> } > Releng doesn't do that, that's what the ticket queues are for. (And
> } > that's why PRs with pending pullups are supposed to have ticket
> } > numbers listed, to make sure we can check from the gnats end whether
> } > things have been handled or not.)
> } > 
> } > However, for people looking for PRs to work on, or people maintaining
> } > the database, or people looking at the PR counts, it's an important
> } > distinction.
> } 
> } The current problem we have is there is no way to remind
> } "commits that should be pulled up to release branches once after
> }  it's confirmed that they won't have any bad side effect on HEAD."
> 
>      In FreeBSD, commits are often marked:
> 
> MFC: <some time period>
> 
> MFC stands for Merge From Current.  We could potentially do something
> like this.  By itself, it doesn't do much other then tag the commit
> and thus could be the target of a search.  We could also have it
> trigger an insertion into some sort of database/queue.  This would
> allow for reminders to the original committer and/or provide a
> central place where somebody looking for something to do can find
> them.

"MFC" is fine for me, though I don't see particular difference
from "pending-pullups without ticket" which requires no system change.

> } Adding a new status is still fine, though I just wonder if
> } the additional states/tasks are appropriate for the benefit,
> } than just adding a new similar meaning to the existing state.
> 
>      This would just be confusing.

Really?  Currently there are only 18 pending-pullups and
the number won't be so large.

Anyway, releng should make amake a decision, IMO.

---
Izumi Tsutsui