Re: Is it time to drop ChangeLogs?

[Karl Fogel]

Dmitry Gutov <address@hidden> writes:
>On 03/07/2016 10:46 PM, Nikolaus Rath wrote:
>> How can it be an increased burden if reviewers have to review just one
>> thing (the commit message) instead of two (commit message and
>> ChangeLog)?
>
>Eli was stating something more general, but to get into specifics:
>often, I only need to read the patch's introduction (high-level
>description), and its ChangeLog entry, to understand it well enough.
>
>A diff contains the same information, but it's usually longer, could
>be harder to read, and often you don't see right away which function a
>given change is being changed, especially if the function is long, and
>the change is right in the middle of it (though that can be alleviated
>with language-specific diff options).
>
>And then, I can look through the diff, compare it against the
>ChangeLog, and see if there are any discrepancies. So the odds of
>getting some unrelated changes (or missing some related ones) is
>lower.

So much about this conversation is baffling to me :-).

Let's please take it for granted that everyone who's suggesting to get rid of 
ChangeLog files is also assuming that we would, of course, use the same 
conventions for writing git commit messages that we would use for writing 
ChangeLog entries.  (As far as I can tell, everyone who wants to get rid of 
ChangeLog files has either expressed that assumption explicitly or implied it 
pretty clearly.  )

Of course, one can view the diff any time, and a good log entry should prepare 
the reader's mind for comprehending the diff.  Whether that entry's text is 
stored in just the VC system commit logs, or in the commit log *and* in a 
ChangeLog entry, shouldn't affect one's ability to use the text to understand 
the change.  Dmitry, you seem to be saying that a "patch's introduction 
(high-level description), and its ChangeLog entry" are two different things.  
That's confusing, to me at least.  The commit message should be the 
introduction to the patch *and* should be the same as the ChangeLog entry (if 
one is keeping separate ChangeLog files).  At least, that's what those of us on 
the no-ChangeLogs side are saying.

Again, if the majority of people like Eli who do tons of work around here would 
prefer to keep ChangeLogs, then I'm in favor of doing so.  But it would be nice 
to be de-confused about the arguments for it.  Many of those arguments seem to 
be based on the idea that the ChangeLog entry is somehow *different* from the 
VC system commit message -- which no one is proposing, as far as I can tell.

Best regards,
-Karl