Wouldn't it make sense if "git pull" would by default also pull the
branch with the same name from the remote, in case no upstream is
configured?
If I can push to a remote with a simple "git push", then I'd also expect
to be able to pull from that same remote with a simple "git pull".
Does anything speak against this?
Example:
git clone $url
git checkout -b fix-1
# do commits
git push # push to origin/fix-1 (works)
git push origin # push to origin/fix-1 (works)
# other people push to origin/fix-1
git pull # pull from origin/fix-1 (fails)
git pull origin # pull from origin/fix-1 (fails)
Hi Mathias,
On 31/05/21 16.18, Mathias Kunter wrote:
Wouldn't it make sense if "git pull" would by default also pull the
branch with the same name from the remote, in case no upstream is
configured?
If I can push to a remote with a simple "git push", then I'd also expect
to be able to pull from that same remote with a simple "git pull".
Does anything speak against this?
Example:
git clone $url
git checkout -b fix-1
# do commits
git push # push to origin/fix-1 (works)
git push origin # push to origin/fix-1 (works)
# other people push to origin/fix-1
git pull # pull from origin/fix-1 (fails)
git pull origin # pull from origin/fix-1 (fails)
IME, I did git fetch first before I did git pull, unless I have repos
that I didn't intentionally want to contribute to (just collecting
them). When I choose to work, I always create a branch, then submit
PR/patches from that against mainline.
Since you do centralized workflow like above, I advise you to integrate
from remote with git fetch + git merge.
And you asked whether plain git pull can work. It is yes, provided that
you don't do any local work on remote-tracking branches (such as
mainline or hotfixes).
--
An old man doll... just what I always wanted! - Clara
Hi Bagas,
thanks for your reply.
Since you do centralized workflow like above, I advise you to integrate
from remote with git fetch + git merge.
I'm aware of this. However, if I know that I didn't commit any changes
to the local branch since I last pushed it (as in the given example),
then it should be possible to use a simple "git pull" IMHO.
It seems unnecessarily complex to me that users would have to use "git
pull origin fix-1" in this situation, when a simple "git pull" could
also work.
Am 31.05.21 um 13:27 schrieb Bagas Sanjaya:
Hi Mathias,
On 31/05/21 16.18, Mathias Kunter wrote:
quoted
Wouldn't it make sense if "git pull" would by default also pull the
branch with the same name from the remote, in case no upstream is
configured?
If I can push to a remote with a simple "git push", then I'd also
expect to be able to pull from that same remote with a simple "git pull".
Does anything speak against this?
Example:
git clone $url
git checkout -b fix-1
# do commits
git push # push to origin/fix-1 (works)
git push origin # push to origin/fix-1 (works)
# other people push to origin/fix-1
git pull # pull from origin/fix-1 (fails)
git pull origin # pull from origin/fix-1 (fails)
IME, I did git fetch first before I did git pull, unless I have repos
that I didn't intentionally want to contribute to (just collecting
them). When I choose to work, I always create a branch, then submit
PR/patches from that against mainline.
Since you do centralized workflow like above, I advise you to integrate
from remote with git fetch + git merge.
And you asked whether plain git pull can work. It is yes, provided that
you don't do any local work on remote-tracking branches (such as
mainline or hotfixes).
Hi Mathias,
I hadn't seen this mail.
Mathias Kunter wrote:
Wouldn't it make sense if "git pull" would by default also pull the
branch with the same name from the remote, in case no upstream is
configured?
I agree.
In fact, back in 2014 I wrote a builtin called `git update` to fix all
the problems with `git pull`. That command worked exactly as you
suggest.
I recently sent a new proposal for `git update` [1] asking for feedback on
how different arguments should be handled. Nobody has suggested
anything.
For the next iteration I will implement what you suggested, in my
opinion it makes sense.
Cheers.
[1] https://lore.kernel.org/git/20210705123209.1808663-1-felipe.contreras@gmail.com/
--
Felipe Contreras