On 08/02/2017 10:26 AM, Richard Stallman wrote:
> But it is unreliable.  You can't rely on the user's having done 'make'
> in the repository before doing the git operations that use these
> settings.

You can pretty much rely on users running either Make or ./autogen.sh; 
otherwise they can't easily build or run Emacs. So the main problem here 
is ./autogen.sh.

Although last year I arranged for ./autogen.sh (when invoked with no 
arguments) to do Git settings like Make does, some other developers 
objected at the time that this might conflict with their own Git 
settings so I changed autogen.sh to not alter Git settings except on 
explicit request. Given the confusion on this topic on the meantime, 
though, I think the change was mistaken, and in that sense I believe 
that you and I are in agreement.

To help improve the situation I propose the attached patch, which 
reverts the default ./autogen.sh behavior to be './autogen.sh all', the 
way it used to be (briefly). Developers who want autogen.sh to leave 
their Git settings alone can continue to use './autogen.sh autoconf'.