Re: master 00149f18ea9: Support modifying VC change comments for Git

Re: master 00149f18ea9: Support modifying VC change comments for Git

[Eli Zaretskii]

The values of the new option vc-allow-rewriting-published-history
could have their mnemonic value improved (and also become more
consistent with our practices elsewhere), if they were changed as
follows:

  nil (default): don't allow
  ask: ask whether to allow
  t: allow without asking

The current values, where t means "ask for confirmation" and 'no-ask'
allows without asking is IMO sub-optimal, because t usually means in
Emacs "do something unconditionally".

So I suggest to change the values as described above.

Thanks.

Re: master 00149f18ea9: Support modifying VC change comments for Git

[Sean Whitton]

Hello,

On Wed 23 Oct 2024 at 12:17pm +03, Eli Zaretskii wrote:

> The values of the new option vc-allow-rewriting-published-history
> could have their mnemonic value improved (and also become more
> consistent with our practices elsewhere), if they were changed as
> follows:
>
>   nil (default): don't allow
>   ask: ask whether to allow
>   t: allow without asking
>
> The current values, where t means "ask for confirmation" and 'no-ask'
> allows without asking is IMO sub-optimal, because t usually means in
> Emacs "do something unconditionally".
>
> So I suggest to change the values as described above.

`ask' is nicer than `no-ask', indeed.  I am however keen to steer people
towards `ask'.  In addition to t often meaning "do something
conditionally", it often means "turn this on in the most common way"
even if the most common way is not the most permissive way.

I would be grateful for more opinions on how this option should look.
CCing Dmitry.

-- 
Sean Whitton

Re: master 00149f18ea9: Support modifying VC change comments for Git

[Dmitry Gutov]

Hi!

On 23/10/2024 12:47, Sean Whitton wrote:
> On Wed 23 Oct 2024 at 12:17pm +03, Eli Zaretskii wrote:
> 
>> The values of the new option vc-allow-rewriting-published-history
>> could have their mnemonic value improved (and also become more
>> consistent with our practices elsewhere), if they were changed as
>> follows:
>>
>>    nil (default): don't allow
>>    ask: ask whether to allow
>>    t: allow without asking
>>
>> The current values, where t means "ask for confirmation" and 'no-ask'
>> allows without asking is IMO sub-optimal, because t usually means in
>> Emacs "do something unconditionally".
>>
>> So I suggest to change the values as described above.
> `ask' is nicer than `no-ask', indeed.  I am however keen to steer people
> towards `ask'.  In addition to t often meaning "do something
> conditionally", it often means "turn this on in the most common way"
> even if the most common way is not the most permissive way.
> 
> I would be grateful for more opinions on how this option should look.
> CCing Dmitry.

I think I prefer Eli's suggestion, simply because it is more 
straightforward and has precedent. The more exotic schemes we have 
somewhere are probably due to history, where a new possibility had been 
added to an existing option, and the decision was to retain the current 
behavior by default.

If we want to steer towards "ask", changing the default would probably 
be the most efficient approach.

Re: master 00149f18ea9: Support modifying VC change comments for Git

[Sean Whitton]

Hello,

Okay, done, thanks both.

I think the default should remain nil, and I've added some more
docstring text recommending `ask'.

-- 
Sean Whitton

Re: master 00149f18ea9: Support modifying VC change comments for Git

[Eli Zaretskii]

> Date: Wed, 23 Oct 2024 22:40:46 +0300
> Cc: emacs-devel@gnu.org
> From: Dmitry Gutov <dgutov@yandex.ru>
> 
> If we want to steer towards "ask", changing the default would probably 
> be the most efficient approach.

But then we will need to override it in .dir-locals.el of the Emacs
repository, I guess?  Because annoying users with a request for
confirmation for something that will be refused by upstream anyway is
pointless?

Re: master 00149f18ea9: Support modifying VC change comments for Git

[Sean Whitton]

Hello,

On Thu 24 Oct 2024 at 08:50am +03, Eli Zaretskii wrote:

>> Date: Wed, 23 Oct 2024 22:40:46 +0300
>> Cc: emacs-devel@gnu.org
>> From: Dmitry Gutov <dgutov@yandex.ru>
>>
>> If we want to steer towards "ask", changing the default would probably
>> be the most efficient approach.
>
> But then we will need to override it in .dir-locals.el of the Emacs
> repository, I guess?  Because annoying users with a request for
> confirmation for something that will be refused by upstream anyway is
> pointless?

This'll be the situation for most projects, so I think this is another
good reason to have the default be nil.  Thanks for pointing it out.

-- 
Sean Whitton

Re: master 00149f18ea9: Support modifying VC change comments for Git

[Dmitry Gutov]

On 24/10/2024 08:50, Eli Zaretskii wrote:
>> Date: Wed, 23 Oct 2024 22:40:46 +0300
>> Cc:emacs-devel@gnu.org
>> From: Dmitry Gutov<dgutov@yandex.ru>
>>
>> If we want to steer towards "ask", changing the default would probably
>> be the most efficient approach.
> But then we will need to override it in .dir-locals.el of the Emacs
> repository, I guess?  Because annoying users with a request for
> confirmation for something that will be refused by upstream anyway is
> pointless?

Not exactly: after all we also have the scratch/ branches and a solution 
to rewriting their history (delete and recreate the branch). So editing 
messages in such could still be useful.

But it's a more complex workflow, of course, not for every contributor.