Re: [PATCH v3 2/2] doc: revert: add discussion

From: Linus Arver <linusa@google.com>
Date: Sep 7, 2023, 21:29 UTC
Message-ID: <owlytts5llje.fsf@fine.c.googlers.com>
In-Reply-To: <ZNcyhUL89WVXOv3F@ugly>

First, I apologize for the long delay in my response. I only work on Git 20% of the time, and that 20% can become 0% due to factors outside my control.

Oswald Buddenhagen <oswald.buddenhagen@gmx.de> writes:

Show 15 quoted lines

> On Fri, Aug 11, 2023 at 04:00:53PM -0700, Linus Arver wrote:
>>Oswald Buddenhagen <oswald.buddenhagen@gmx.de> writes:
>>
>>> On Thu, Aug 10, 2023 at 02:50:59PM -0700, Linus Arver wrote:
>>>>Nit: the "doc: revert: add discussion" subject line should probably be more
>>>>like "revert doc: suggest adding the 'why' behind reverts".
>>>>
>>> this is counter to the prevalent "big endian" prefix style, and is in 
>>> this case really easy to misread.
>>
>>I also learned recently that there should just be one colon ":" in the
>>subject, which is why I suggested "revert doc" as the prefix instead of
>>"doc: revert: ...".
>>
> in what context was this preference expressed?

IIRC, it was from a conversation off-list with the folks at Google's Git-core team.

> because here, it's rather counter-productive: most commands are verbs 
> for obvious reasons, so using that style sets the reader up for 
> misparsing the subject on first try.

I think the convention for commit titles is

    <prefix>: <action>

so the phrase "revert doc: add discussion", where the <prefix> is "revert doc" does not parse any worse than "doc: revert: add discussion". That is, the <prefix> is never confused with the <action> (they are separated by the colon).

> this could be avoided by quoting 
> the command, but that looks noisy in the subject.
> so rather, i'd follow another precedent, 'git-revert.txt: ', which is 
> unambiguous.

SGTM.

Show 13 quoted lines

>>> i also intentionally kept the subject generic, because the content 
>>> covers two matters (the reasoning and the subjects, which is also the 
>>> reason why this is a separate patch to start with).
>>
>>I think the phrase "add discussion" in "doc: revert: add discussion"
>>doesn't add much value, because your patch's diff is very easy to read
>>(in that it adds a new DISCUSSION section). I just wanted to replace it
>>with something more useful that gives more information than
>
>>just repeat
>>(somewhat redundantly) what is obvious by looking at the patch.
>>
> but ... that's exactly what a subject is supposed to do!

I think the rule of thumb is to explain the goodness of what a commit brings, rather than focus on what is literally happening. This is because the former is more valuable. So instead of

    "git-revert.txt: add discussion"

you could say

    "git-revert.txt: advise against default commit message"

and now you don't have to look at the patch to see (roughly) what kind of discussion was added.

Show 12 quoted lines

>>>>Please consider rewording such
>>>>    subject lines to reflect the reason why the original commit is being
>>>>    reapplied again.
>>>>
>>> the reasoning most likely wouldn't fit into the subject.
>>
>>Hence the language "to _reflect_ the reason", because the "reason"
>>should belong in the commit message body text.
>>
> i don't think that's how most people would actually read this.
> and i still don't see how that instruction could be meaningfully 
> followed.

OK, you may be right.

Show 10 quoted lines

>>> also, the original request to explain the reasoning applies 
>>> transitively, so i don't think it's really necessary to point it out 
>>> explicitly.
>>
>>It may be that a user will think only giving the revert reason in the
>>body text is enough, while leaving the subject line as is. I wanted to
>>break this line of thinking by providing additional instructions.
>>
> yes, that's the whole intention of this patch. but i don't see how 
> making it more convoluted than my proposal helps in any way.

Well, even if a review makes something more convoluted, it may generate discussion and drive consensus on the better way(s) of doing something. I see value in that course of events.

Of course you are free to reject review comments that you truly believe are inferior to the approach you've already taken.

But overall, when I see a reviewer's comment on this mailing list, I assume they are trying to make my patch better. Similarly when I reviewed your patch my intent was to provide actionable feedback to try to make it better. I'm sorry if I did not come across that way.

Show 6 quoted lines

>>This is definitely better. But others in this thread have already
>>commented that my version looks good (after seeing your version also,
>>presumably).
>>
> well, i'm also an "others" when it comes to your proposal, and i find it 
> confusing.

I think you did the right thing by responding to my comments, and pointing to things you found confusing.