[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Question about intended behavior of 'insert-for-yank-1'.

From: Eli Zaretskii
Subject: Re: Question about intended behavior of 'insert-for-yank-1'.
Date: Tue, 04 Oct 2016 21:19:06 +0300

> From: Karl Fogel <address@hidden>
> Cc: address@hidden,  address@hidden
> Date: Tue, 04 Oct 2016 12:40:49 -0500
> >You need to understand the logic behind the instructions in
> >CONTRIBUTE: the release branch should only get fixes that are either
> >necessary, or are safe, i.e. cannot possibly break the release.  And
> >doc changes obviously belong to this latter class.
> I understand that logic.  However, do you mean s/only/always/ above?

I don't see the difference.  "Only" means "always" in this case.

> The question is whether it is wrong to commit it to master instead, not 
> whether it would have been right to commit it to emacs-25.

It is "wrong" in the sense that committing to master will prevent the
next release from having the fix, for no good reason, but committing
to the release branch will never prevent master from having it.

> A non-urgent doc fix that lands on master will eventually "get into Emacs", 
> when new release lines are created.  This was good enough for me.

Doc fixes are always "urgent", in the sense that there's no reason to
delay them.  They can never do any harm.

> You seem to be saying that, once the safe doc fix is available, it should 
> always be committed to the current release branch.

Yes.  And not just "safe" doc fix: _any_ doc fix.  Doc fixes are by
definition safe.

> For the committer, this would add the mental overhead of figuring out whether 
> or not a particular release branch is in a freeze state or not.

The freeze state is never relevant to doc fixes.  It is only relevant
to code changes.  The purpose of a freeze is to prevent unsafe
changes, and doc changes are never unsafe.

> This is actually non-trivial to figure out, since it's not maintained as a 
> visible flag in some designated place, but rather as a thread on the mailing 
> list, which I then have to go search for.

If you have trouble with that, just ask before pushing.  I think the
decision is very simple, but if it isn't, John and myself are here to

> So for non-urgent stuff like this, I find it easier to just commit to master, 
> because then I don't have to worry about the state of the current release 
> branch.

Once again, doc fixes are always "urgent".  It may be easier for you
to always commit to master, but that's not our process here, so it
means someone else will have to watch your commits, detect problems
like this, and then cherry-pick for you, or ask you to do that.  It's
a burden we'd like to avoid, and I think the rules are simple enough
to allow anyone to follow them seamlessly.  And if they are not simple
enough, please just ask when in doubt.

> I can start worrying about it, if you think it's important.

It's important, yes.

> (When I want a fix to go into the current release, then I expend the extra 
> effort to determine whether I can put it on the release branch.)

No such considerations are necessary for documentation changes.  the
only reason not to commit a doc fix to the release branch is if the
fix is relevant to behavior that's only available on master.

> Now, *maybe* it is the case that a super-safe change like this is allowed 
> onto a release branch even when that branch is frozen.

Doc fixes are _always_ safe.

> But I don't know that for sure, and CONTRIBUTE doesn't say.  Maybe this is 
> just a doc bug in CONTRIBUTE and we should fix it.

CONTRIBUTE already does say that, I made the change there today.

> IOW, committing to master wasn't an accident; it resulted from a balancing of 
> priorities.

I didn't think it was an accident, I just explained why your decision
was incorrect.  I understand that it was a good-faith mistake, so no

reply via email to

[Prev in Thread] Current Thread [Next in Thread]