Re: Forgotten MFC

On Thu, 17 Apr 2025 23:49:36 -0600
Warner Losh <imp@bsdimp.com> wrote:

> On Thu, Apr 17, 2025, 11:14 PM Vadim Goncharov <vadimnuclight@gmail.com>
> wrote:
> 
> > On Thu, 17 Apr 2025 19:16:00 -0600
> > Warner Losh <imp@bsdimp.com> wrote:
> >
> >  
> > > We really need to go to a system where we can add Notes to commits after
> > > the fact (like we did with svn numbers in the conversion) so that we can
> > >  
> > tag  
> > > commits after the fact for merging (I forgot, or don't think it's
> > > important enough
> > > at the time, only to change my mind later) as well as say 'nope, this
> > > has a problem
> > > don't commit it' as well as gathering groups together to merge as a
> > > unit incrementally
> > > (again, for those that the committer might not have tagged it as
> > > x-mfc-with because
> > > the dependency was unknown at commit time).  The notes are optional, and
> > >  
> > one  
> > > can clone them or not, unless one wants the functionality, so it
> > > doesn't affect the
> > > instructions that we send out. And adding notes doesn't change the hash,
> > > though notes aren't versioned very well (but well enough to reconstruct
> > > accidental or malicious  
> >
> > If the migration was to Fossil instead of stupid Git, we'd already had such
> > notes for free (and amending without hash rewrites, it's just another
> > control
> > artifact in Fossil).
> >  
> 
> Notes are already a part of git. Fossil doesn't have the extra layer we
> need anyway.

Did you mean scripting layer? From object graph point of view, Fossil has much
more than dumb wooden git.

> But we were never going to pick fossil. It was too obsure. 

Sounds quite strange. NetBSD tried it and got many responses and fixes from
Fossil's developers. As for "obscure", it's Git which has more than hundred of
fat ill-defined commands, than clean Fossil's model.

> It doesn't have enough extra services built around it. 

What services do you mean here? Nothing had fundamentally changed for FreeBSD
here since CVS era.

GitHub? Laugh. I tried to do pull-requests there - took months for three-line
bug fix. If anybody stll tries to sell Git as "oh, we will get more
contributions!" - this is just marketing hype and in fact does not work.

> And arm chair quarterbacking isn't
> helpful. Wasting time on this doesn't move the ball forward.
> What i need is more people to show up and help me when i try the next push
> to relaunch our tooling push....

Real moving forward - that is, progress - would be moving *from* Git. It's not
why we chose Subversion in 2008 instead of Git just to loose features from it
later. And git's model is just inappropriate e.g. for ports (architectural
flaw with renames).

I understand this will unlikely to happen in next years due to lack of
resources, but to call building crutches and skids around Git a "moving
forward" is funny at best.

-- 
WBR, @nuclight