Re: bug?: ORIG_HEAD incorrect after reset during git-rebase -i

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

 



Hi Philippe & Erik

On 05/01/2023 00:11, Philippe Blain wrote:
Hi Phillip and Erik,

Le 2021-12-16 à 11:44, Erik Cervin Edin a écrit :
Hi Phillip,

Yes, I know.
It's just that I was under the impression ORIG_HEAD was to be reverted
to .git/rebase-merge/orig-head at the finish of the rebase.
Personally, it's the behavior I would expect.

Thanks for the tips.

I just hit the same bug (I think it qualifies as one). In fact git-rebase(1) explicitely mentions
that ORIG_HEAD is set to the branch tip before the rebase starts:

Strictly speaking that is what we do so we're documentation the implemented behavior. What's not clear from the documentation is that if the user run 'git reset' while rebasing then ORIG_HEAD will be overwritten. We could update ORIG_HEAD at the end of the rebase as you suggested but I wouldn't be surprised if some else complains that ORIG_HEAD no longer points to the commit that the reset while running rebase. I also wonder if users would expect 'git rebase --continue' to update ORIG_HEAD to point to the pre-rebase HEAD so it is consistent each time rebase stops. Basically I think the situation is confusing and I don't have a clear idea as to how to make it better. If someone submits a patch to try and clean things up I'll happily look at it but unless I'm hit by a bright idea as to how to fix it I probably wont work on it myself.

Best Wishes

Phillip

$ git grep -C2 ORIG_HEAD Documentation/git-rebase.txt
Documentation/git-rebase.txt-36-The current branch is reset to `<upstream>` or `<newbase>` if the
Documentation/git-rebase.txt-37-`--onto` option was supplied.  This has the exact same effect as
Documentation/git-rebase.txt:38:50:`git reset --hard <upstream>` (or `<newbase>`). `ORIG_HEAD` is set
Documentation/git-rebase.txt-39-to point at the tip of the branch before the reset.
Documentation/git-rebase.txt-40-

Here is my runnable reproducer. It is slightly more complicated than Erik's, since
I split the second commit in two, but this is not necessary to trigger the bug; just
running 'git reset HEAD^' as Erik wrote is enough.

```bash
#!/bin/bash
rm -rf repro
git init repro
(
cd repro
# Create 3 commits
cat << EOF >test
hello
every
one
EOF
git add test
git commit -m initial
cat << EOF >test
hello

add new lines

every

and also here

one
EOF
git commit -am second
cat << EOF >test
hello

add new lines

every

and also here

one

still more changes
EOF
git commit -am third
# Rebase to split the second commit
GIT_SEQUENCE_EDITOR="sed -ie '1 s/^p /e /'" git rebase -i HEAD~2
git reset HEAD^
cat << EOF >test
hello

add new lines

every
one
EOF
git ci -am "second 1/2"
cat <<EOF >test
hello

add new lines

every

and also here

one
EOF
git ci -am "second 2/2"
# Finish rebase and demonstrate bug
git rebase --continue
echo ---
echo "@{1} is :"
git log -1 @{1}
echo "ORIG_HEAD is :"
git log -1 ORIG_HEAD

)
```

Cheers,

Philippe.



[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