gopull

Recursively find git repositories under a directory and git pull them all, sequentially or in parallel.

Discovery walks the tree and stops descending once a repository is found, so repos nested inside other repos (vendored checkouts, submodule working trees) are ignored. Pulls run git pull --ff-only with terminal prompts disabled, so a repo with diverged branches or broken credentials shows up as a failure in the summary instead of merging or hanging.

Install

go install github.com/NickName-AM/gopull@latest

Or from a local checkout:

go build -o gopull .

Requires the git binary on PATH - pulls use your normal git config, SSH keys, and credential helpers.

Usage

gopull [flags] [root]        root defaults to "."

  -p, --parallel   pull repositories concurrently
  -j, --jobs N     worker count in parallel mode (default: number of CPUs)
  -l, --list       only list discovered repositories, don't pull
  -d, --depth N    max directory depth to search (0 = unlimited)
  -r, --remote R      remote to pull from (default: each repo's upstream)
  -b, --branch B      branch to pull (default: each repo's current branch)
  -f, --force-branch  with -b, merge the branch even into repos checked out
                      on a different branch

By default each repo pulls its own current branch from its configured upstream. -r/-b name an explicit source (git pull origin main style); -b alone implies -r origin.

With -b main, a repo that is not on main is never merged with origin/main. Instead it falls back to a plain pull of its own configured upstream, or is reported as skipped if it has none (detached HEADs are skipped too). Pass -f/--force-branch to bypass this guard and merge the named branch into whatever branch each repo has checked out.

Examples:

gopull ~/Development/projects          # sequential
gopull -p -j 8 ~/Development/projects  # parallel with 8 workers
gopull --list ~/Development/projects   # just show what would be pulled

Interrupting

Press Ctrl-C to stop a run. No further repositories are dispatched, the pulls already running are stopped, and you still get the summary for everything that completed - repos whose pull was cut short are reported as canceled, and the ones never started are counted as not pulled. A second Ctrl-C quits on the spot, in case a pull refuses to wind down.

An interrupted pull is safe: git is asked to terminate rather than killed outright, so it removes its own lock files on the way out. Each pull runs in its own process group and the whole group is signaled, so the fetch and merge that git pull spawns - the ones actually holding index.lock - are stopped too, rather than being left behind to finish on their own. A pull that ignores the interrupt and completes anyway is still reported as updated.

Exit code is 0 when every repo pulled cleanly (or was already up to date), 1 if any repo failed or was cut short without an interrupt, and 130 if the run was interrupted.

Output

Pulling 3 repositories with 3 workers...
· uptodate (up to date)
✓ work/behind (updated, 0.1s)
✗ work/deep/diverged (failed)

1 updated, 1 up to date, 1 failed

✗ work/deep/diverged:
    fatal: Not possible to fast-forward, aborting.

Interrupted with Ctrl-C partway through:

Pulling 8 repositories with 3 workers...
· fast1 (up to date)
· fast2 (up to date)

interrupt: stopping (Ctrl-C again to force quit)
· slow1 (canceled)

0 updated, 2 up to date, 0 skipped, 0 failed, 1 canceled, 5 not pulled
interrupted