Re: Script to generate ChangeLogs automatically

[Simon Marchi]

On 2018-11-30 16:08, Alfred M. Szmidt wrote:
> I think this is not really what one (or at least, I) is after.  Here
> is one typical excursion, picking a random file and function in Emacs:
> pr-menu-set-txt-title in lisp/printing.el.
>
> I was unable to follow the history of pr-menu-set-txt-title using git
> log, git blame, or any other such command.  Infact, I get the wrong
> information from git blame, mainly because of things getting wrapped
> with a eval-when:
>
> ebe4c71027c (Vinicius Jose Latorre   2007-10-27 00:25:43 +0000 1564)
> (defun pr-menu-set-txt-title (value &optional item entry index)

This is indeed the last time this code has changed.

> If we look at the ChangeLog, we learn that this function was infact
> moved with in the file in 2008-02-01, and an alias was created for it
> on the same date.

Actually, commit

  aca7a3d201 ("Merge ChangeLog.unicode files into ChangeLogs")

changed the date of that ChangeLog entry from 2007-10-26 to 2008-02-02 
while moving the content, for some reason.  The info that git blame gave 
you is accurate.

> Looking at that particular diff is a mess, it is
> over 2000 lines.

Well, you can search in a diff for the symbol you are interested in.  
You will see it removed at some point and added at another.

But I agree that just with the diff alone, it's hard to tell that the 
function has been moved, but there has been no functional changes.  So 
most importantly, a sensible commit message (which is, I hope, expected 
to have by today's standards) here would have said something along the 
lines of "Moving a bunch of functions for reason X, no function changes 
intended".

When we send patches on mailing lists (at least in the GDB community), 
we need to describe our changes in prose to convince our fellow 
developers why it's needed, what it does, how it does it, etc.  So 
writing the ChangeLog always feels like I'm repeating what I have 
already said in my commit log and what is illustrated by the diff.

Simon