How to convert the below filter-branch command ?

[DilLip-Chowdary-Codes]

Checked docs but was not able to find a way

git filter-branch --commit-filter 'if [ $(git show --no-patch --no-notes --pretty='%ct' $GIT_COMMIT) -ge 1682879400 ]; then git commit-tree "$@"; else skip_commit "$@"; fi' -- --all

[andry81]

Looks like the command tries to cut off all commits younger than a commit date-time retaining all files and changes in the first commit in each branch. You can just find all these commits manually and if there is not many of them, then just cut off all the parents in each commit through:

git replace --graft commit
git filter-repo --force
git for-each-ref --format='delete %(refname)' refs/replace | git update-ref --stdin

[DilLip-Chowdary-Codes]

Yeah, basically we trying to solve this problem is it possible with our filter-repo without any manual work?

At least by modifying the python code in git-filter-repo, is there any possibility?

[andry81]

This is what you actually need to do on your own. Go over all branches and find the first commit:

IFS=$'\r\n'
for show_ref_line in `git show-ref --heads`; do
  IFS=$' \t'; read prefix head_ref <<< $show_ref_line
  echo $head_ref
  git log --until=2022-10-28T -1 --format='%H' $head_ref
done

And run replace command on each commit.

But, there is a catch. If you have a merge commits between branches, then git log can return for different branches the same commit hash.
To workaround that you can change the time before the rewritten commit to select different branch commit to cut off before it.

[newren]

Untested, but if you add --full-tree to the list of options that filter-repo passes to fast-export (search for --show-original-ids and add --full-tree to that list), then I think you could translate your filter-branch command to the following:

git filter-repo --commit-callback '
    epoch = int(commit.committer_date.split()[0])
    if epoch < 1682879400:
        commit.skip()
'