Web lists-archives.com

Re: Pull without fetch

On Mon, Apr 8, 2019 at 7:51 PM Ævar Arnfjörð Bjarmason <avarab@xxxxxxxxx> wrote:
> On Mon, Apr 08 2019, Duy Nguyen wrote:
> > On Mon, Apr 8, 2019 at 8:34 AM Junio C Hamano <gitster@xxxxxxxxx> wrote:
> >>
> >> Damien Robert <damien.olivier.robert@xxxxxxxxx> writes:
> >>
> >> > is there a way to do a git pull without it running git fetch?
> >> > Looking at the source in builtin/pull.c does not seem to indicate so.
> >>
> >> The reason behind that is because it does not make any sense for
> >> "pull", which is meant as a quick short-cut to say "fetch && merge",
> >> not to run fetch, especially back then when 'git pull' was designed,
> >> the world was much simpler.  There was no "fetch && rebase", our
> >> branches did not know what their @{upstream}s were.  In that simpler
> >> world, what you are trying to do would have been:
> >>
> >>         git fetch
> >>         # did I get anything worth integrating?
> >>         git merge FETCH_HEAD
> >>
> >> That obviously would not work for those with "pull.rebase", and I do
> >> not think it makes much sense to teach "git rebase" the same trick
> >> to read FETCH_HEAD as "git merge" does in the above sequence.
> >>
> >> Others may have a better idea, but I do not immediately see any
> >> solution better than inventing a new option to "git pull".
> >>
> >> Another and better option that may be harder to arrange is to make
> >> sure that a no-op "git fetch" incurs very low cost.  If you did so,
> >
> > Not exactly related. But I often wish to see the list of branch
> > updates since the last fetch. There's no easy way (that I know) to do
> > this unless you copy the last fetch's output somewhere. If this "fetch
> > at low cost" could simply read FETCH_HEAD and summarizes it like a
> > normal fetch, that would be great. And it should also be very low cost
> > because we only replay the last part (making summary) of normal fetch.
> The ability to have this is something reftables will provide (from my
> memory of a comment by Stefan Beller), which Christian Couder is working
> on implementing these days.

I don't think we even need reftables to implement it. The list of new
SHA-1 is available in FETCH_HEAD (at least until the next fetch), and
from reflog we know the old SHA-1 (or a new branch/tag, I think we
know too).