Commit comment rules - and a metaproblem

[Eric S. Raymond]

Paul Eggert <address@hidden>:
> On 12/03/2014 11:27 AM, Eric S. Raymond wrote:
> >You want it even when the ChangeLog part is a trivial
> >repetition of the summary line?
> 
> It's not needed for one-liners.  For commit
> e820f16c06a5a6be4bc87910b349c7c3c6eca0f4, for example, your
> ChangeLog entry was "* files.el (file-tree-walk): Lisp translation
> of ANSI ftw(3).", and that one-liner should have been the git commit
> message, too.
> 
> For longer messages, please use the same text for both the ChangeLog
> file as for the commit message, except omit the 2nd-blank-line in
> the former and omit the leading tabs in the latter.  For commit
> b1a765b3a8586cd53c21579982c8fbc0ce534336, in contrast, it would have
> been better to use the following text (removing indentation) as the
> commit message:
> 
>     In vc, abolish the dir-status method.
> 
>     *  vc.el, all backends: API simplification: Abolish dir-status.
>     It's replaced by dir-status-files.
> 
> and to use the same text (with indentation, but without the blank
> line) as the ChangeLog entry.
> 
> It is a pain to have to have essentially two copies of the same
> text, one in the ChangeLog file and the other in the commit message,
> but we haven't yet had strong consensus on fixing this.

A larger problem is that these rules are both fussy and poorly documented -
in fact I don't know that they're documented at all.

And then there was talk of revoking commit access for violation of fussy
and poorly documented rules.

Emacs has very few contributors for its size.  It desperately needs new
blood, but has chronic trouble attracting any.

Connect the dots.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>