Re: merge branch with updated submodule leaves repo in modified state

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, Dec 10, 2019 at 11:58 AM Daniel Bosk <dbosk@xxxxxx> wrote:
>
> Hi all!
>
> `git merge`: if a submodule is updated in the merged branch, the new
> version is not checked out after the merge. Is this correct behaviour?
>
> `git switch` correctly updates the submodules when switching between the
> branches. I would expect `git merge` to leave the things in the most
> up-to-date state.
>
> I tested this with the following settings.
>
>   $ git version
>   git version 2.24.0
>
>   $ git config -l
>   push.default=matching
>   diff.submodule=log
>   diff.colorwords=true
>   diff.algorithm=minimal
>   diff.exit-code=true
>   commit.gpgsign=true
>   submodule.fetchjobs=0
>   submodule.recurse=true
>   log.follow=true
>
> MWE:
>
> ```
> mkdir repoA repoB
>
> # init repos with READMEs
> cd repoA
> git init
> echo "A" > README
> git add README
> git commit -m "Adds repoA README" README
>
> cd ../repoB
> git init
> echo "B" > README
> git add README
> git commit -m "Adds repoB README" README
>
> # construct nested structure: repoA/repoB
> cd ../repoA
> git submodule add ../repoB/.git
> git commit -m "Adds repoB submodule"
>
> # update repoB in branch
> git switch -c update-repoB-submodule
> cd repoB
> echo "B updated" > README
> git commit -m "Updates repoB README" README
> cd ..
> git commit -m "Updates repoB submodule" repoB
> ```
>
> Now, doing `git switch master` will have the correct version of repoB,
> where `cat repoB/README` outputs "B". A `git switch update-repoB-submodule`
> from there will yield `cat repoB/README` output "B updated".
>
> However, switching to master and then merging will leave repoB at the
> old commit:
> ```
> git switch master
> git merge update-repoB-submodule
> ```
> A `git status` says there are new commits in repoB, it's actually a
> rewind which excludes the new commit (changing `repoB/README`), so `cat
> repoB/README` outputs "B" instead of "B updated". It feels unintended to
> get a repo which is in a modified state after a clean merge. Bug?

Don't know where it falls on the spectrum from bug to unimplemented
feature, but it's certainly a known issue.  The last time I remember
discussing this was here:

https://lore.kernel.org/git/CABPp-BHDrw_dAESic3xK7kC3jMgKeNQuPQF69OpbVYhRkbhJsw@xxxxxxxxxxxxxx/


Hope that helps,
Elijah



[Index of Archives]     [Linux Kernel Development]     [Gcc Help]     [IETF Annouce]     [DCCP]     [Netdev]     [Networking]     [Security]     [V4L]     [Bugtraq]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux SCSI]     [Fedora Users]

  Powered by Linux