Re: Script to generate ChangeLogs automatically

[Carlos O'Donell]

On 12/1/18 6:42 PM, Richard Stallman wrote:
> [[[ To any NSA and FBI agents reading my email: please consider    ]]]
> [[[ whether defending the US Constitution against all enemies,     ]]]
> [[[ foreign or domestic, requires you to follow Snowden's example. ]]]
> 
>      > > We don't require git for anything in the GNU project, let alone being
>      > > able to use it to contribute a change.  You can get away happily
>      > > without learning any git commands and still be able to contribute to
>      > > the GNU project.  Just like being fluent in English is not an
>      > > requirement, it is unreasonable to expect that being fluent in git
>      > > magic is mandatory.
> 
>      > How do you checkout glibc if you're not using git?
> 
> Carlos, there is a gap in the argument which that response presents
> elliptically.  A lot of people know enough git to check out a
> program's master branch.  Even I can do that.  However, what is
> pertinent here is whether person knows advanced git commands _and
> how to check for when they give wrong information_.  Not the same
> thing at all.
> 
> Because you did not fully spell out the argument, people might easily
> not notice the gap in it.  Basically, your message invites people to
> overlook the gap and believe the argument due to that omission.
> 
> Please do not argue that way.   Show us the respect of spelling out
> your arguments so that people can tell whether they are valid.

My apologies, I was trying to limit my questions to ensure they got
answered in a sequential fashion. I had not considered the gap it
presented and how it might be interpreted.

Alfred wrote "We don't require git for anything in the GNU project", 
and the obvious proof by contradiction is that we do require it for
glibc development work on master. The answer "Tarball?" is not a
supportable one because you're quickly so far away from master that
you can't generate a diff that would apply.

The point that I'm trying to make is that you mast have a working
knowledge of git, rather than zero knowledge. This positions future
arguments about how much more git you need to know in a more favourable
position.

If I take a long view on this argument, and consider the pipeline of
future new GNU developers, I expect they will all have a working
knowledge of git, and have used common tooling to inspect the history
of a project (like gitk).

Curiously enough when I use gitk on glibc's sources I can, via
"show origin of this line" get down to the 'initial import' commit 
from Roland McGrath, and in that era history was messier and so
the ChangeLog's *were* useful as Jopseh Myers right points out. 
However, today that history is present and interactively 
available. This makes the argument for ChangeLog's weaker in 
my opinion.

I applaud Siddhesh's efforts and think we should continue to pursue
them in earnest, particularly since it seems they represent the
consensus between us.

-- 
Cheers,
Carlos.