git for-each-repo

Multi-repo orchestration in one binary.

🎯 What & why

Runs a Git subcommand across many repositories listed in a multi-valued config key. It is multi-repo orchestration without a wrapper script - and it is what powers git maintenance schedules across all your clones.

🧠 Mental model

Read paths from git config --get-all <name>, chdir into each, run git <command>... in turn. That's it. No parallelism, no fancy filtering - just a deterministic loop the maintenance scheduler can rely on.

🛠️ Synopsis

git for-each-repo --config=<config> [--keep-going] [--] <command-args>

🎚️ Switches & options

Flag	What it does
`--config=<config-name>`	Required. Multi-valued config key whose values are repo paths to iterate (e.g. `maintenance.repo`).
`<command>...`	The Git command (and its args) to run in each repo - everything after `--` is forwarded verbatim.
`--keep-going`	Continue iterating even if one repo's command exits non-zero (otherwise it stops at the first failure).
`--config with a single-valued key`	⚠️ Will error or behave surprisingly - the config key must be multi-valued.

💡 Use cases

Centralized maintenance: pair maintenance.repo with git maintenance run so cron or systemd ticks one binary.
Bulk git fetch across all your work clones from a single command.
Cluster-wide git gc or git repack runs without a hand-rolled find | xargs.
Quick status sweeps: git for-each-repo --config=my.repos status -s to see who is dirty.

🧪 Examples

git config --global --add maintenance.repo /path/to/repo

Run scheduled maintenance across every registered repo

git for-each-repo --config=maintenance.repo maintenance run --schedule=daily

Fetch across a custom list of repos

git for-each-repo --config=my.repos -- fetch --all --prune

Don't bail on the first failure

git for-each-repo --keep-going --config=my.repos -- gc --auto

🎓 Recommendations

Use git maintenance register to populate maintenance.repo instead of editing config by hand - it also wires up the schedule.
Default to --keep-going for housekeeping commands; one corrupt repo shouldn't stop the rest from being maintained.
Put -- before the forwarded command's args to avoid having for-each-repo eat flags it does not recognize.

🪤 Common pitfalls

Iteration is sequential; this is not a parallel runner. Reach for xargs -P or a real orchestrator when you need concurrency.
Paths in the config are taken literally - no ~ expansion in older Git versions, and no glob support ever.
If the config key is empty or unset, the command runs zero times and exits 0 - silently doing nothing is easy to miss.

🔗 Related modules

📝 Quiz

Hit each option, then Check answers. Score is recorded; Next is always open.