This package provides one command, watch_gha_runs. It displays the status
of the latest GitHub Action runs on your current branch. If any of the runs
are in progress, it will refresh the display repeatedly with the latest status.
If you like, the name can be pronounced, "Watching? Ha!"
I suggest installing with pipx:
$ pipx install watchghaNow you have a command watch_gha_runs available. It will check all GitHub
remotes for the current directory's repo, and find action runs for the current
branch.
For complex defaulting, you can use a git alias. For example, this provides similar defaults, but can be adapted:
[alias]
runs = "!f() { \
watch_gha_runs $@ \
\"$(git remote get-url origin)\" \
\"$(git rev-parse --abbrev-ref HEAD)\"; \
}; f"Now git runs will show a live display of the current runs on your branch.
You can authenticate against GitHub if needed using either an entry in your
.netrc file, or by setting the GITHUB_TOKEN environment variable.
No authentication is needed for public repos. For private repos, OAuth or
classic tokens need the repo scope, and fine-grained tokens need the
"Actions (read)" repository permission.
If you use GitHub Enterprise, you can set the environment variables
GITHUB_SERVER_URL (default: https://github.com) and
GITHUB_API_URL (default: https://api.github.com)
to match your instance.
$ watch_gha_runs --help
Usage: watch_gha_runs [OPTIONS] [REPO] [BRANCH]
Watch GitHub Action runs.
Repeatedly gets the latest status and redraws the screen, until all
of the jobs are complete.
REPO is a local directory or GitHub URL, defaulting to ".".
BRANCH is defaulted from the git repo.
Options:
--sha TEXT The commit SHA to use. Must be a full SHA.
--poll INTEGER How many seconds between refreshes. [default: 15]
--wait-for-start Wait for jobs to start.
--only TEXT Words to limit the workflows shown. Only workflows
with these comma separated case insensitive
substrings in their names will be shown.
--help Show this message and exit.
The output shows runs and jobs. The current step of each job is shown, with a row of bullets indicating the number of steps, and which is current:
fix: recent pypy3.9 now omits lines after jumps nedbat/fix-pypy-nightly 53923268e8f9 @08:32am
⏲ queued Tests view 4397455341
3.7 on ubuntu ↻ •••••••• Run tox for 3.7
3.8 on ubuntu ↻ •••••••• Run tox for 3.8
3.9 on ubuntu ↻ •••••••• Run tox for 3.9
3.10 on ubuntu ↻ •••••••• Run tox for 3.10
3.11 on ubuntu ↻ •••••••• Run tox for 3.11
pypy-3.7 on ubuntu ↻ •••••••• Run tox for pypy-3.7
pypy-3.9 on ubuntu ↻ •••••••• Run tox for pypy-3.9
3.7 on macos ↻ •••••••• Run tox for 3.7
3.8 on macos ↻ •••••••• Run tox for 3.8
3.9 on macos ⏲ queued
3.10 on macos ↻ •••••••• Run tox for 3.10
3.11 on macos ↻ •••••••• Run tox for 3.11
pypy-3.7 on macos ⏲ queued
pypy-3.9 on macos ⏲ queued
3.7 on windows ⏲ queued
3.8 on windows ⏲ queued
3.9 on windows ⏲ queued
3.10 on windows ↻ ••••••• Check out the repo
3.11 on windows ⏲ queued
pypy-3.7 on windows ⏲ queued
↻ in_progress Quality view 4397455342
Check types ✓ success
Build docs ↻ ••••••• Tox doc
Pylint etc ↻ ••••••• Tox lint
↻ in_progress Python Nightly Tests view 4397455346
Python 3.10-dev ↻ •••◦•••• Run tox
Python 3.11-dev ↻ •••◦•••• Run tox
Python 3.12-dev ↻ •••◦•••• Run tox
Python pypy-3.7-nightly ↻ ••◦•••••• Run tox
Python pypy-3.8-nightly ↻ ••◦•••••• Run tox
Python pypy-3.9-nightly ↻ ••◦•••••• Run tox
Jobs and runs are collapsed once all of their children are successful:
fix: recent pypy3.9 now omits lines after jumps nedbat/fix-pypy-nightly 53923268e8f9 @08:32am
✓ success Tests view 4397455341
↻ in_progress Quality view 4397455342
Check types ✓ success
Build docs ↻ ••••••• Tox doc
Pylint etc ✓ success
✗ failure Python Nightly Tests view 4397455346
Python 3.10-dev ✓ success
Python 3.11-dev ✓ success
Python 3.12-dev ✓ success
Python pypy-3.7-nightly ✓ success
Python pypy-3.8-nightly ✓ success
Python pypy-3.9-nightly ✗ failure Run tox
Once all the runs are completed, the command ends, displaying the final status:
fix: recent pypy3.9 now omits lines after jumps nedbat/fix-pypy-nightly [push] 53923268e8f9 @08:32am
✓ success Tests view 4397455341
✓ success Quality view 4397455342
✗ failure Python Nightly Tests view 4397455346
Python 3.10-dev ✓ success
Python 3.11-dev ✓ success
Python 3.12-dev ✓ success
Python pypy-3.7-nightly ✓ success
Python pypy-3.8-nightly ✓ success
Python pypy-3.9-nightly ✗ failure Run tox
- GITHUB_SERVER_URL's like "[email protected]" are now correctly parsed, closing issue 22.
- Added a stop sign emoji for jobs in the Waiting state.
- Most fatal errors now result in a status code of 1. It was mistakenly 2.
- Workflows with many jobs could be truncated. There is still a limit of 100 jobs, but that is better than the earlier limit of 30.
- GitHub Enterprise is supported via
GITHUB_SERVER_URLandGITHUB_API_URLenvironment variables. Thanks, Colin Marquardt. - Fix: in unusual cases, GitHub can return strange statuses for job steps. Those are now displayed as question marks.
- Fix: steps can be in a "pending" state, and are now displayed with a dot instead of "pending".
- Fix: don't fail if a .netrc file can't be found. Fixes issue 18.
- Fix: in the odd case of duplicate remotes, don't list workflow runs twice. Fixes issue 19.
- Now all GitHub remotes are checked for jobs. Previously, only one was checked, so you wouldn't see jobs running on an upstream fork.
- Added option
--onlyto limit which workflows are displayed as requested in issue 17. - The output is now redrawn immediately when the terminal window is resized (on Mac or Linux). Thanks, Bill Mill.
- Implicit .netrc authentication stopped working, but has been fixed. Thanks, Rob Weir.
- The default polling interval is now 15 seconds.
- Now the GitHub repo location and branch name are defaulted from the current git repo. The repo location can be a local directory or GitHub URL. Closes issue 7.
- A new option,
--wait-for-startwill make watch_gha_runs wait until jobs are in progress. This fixes a problem with using watch_gha_runs programmatically: it can check the run status before any new runs have started, and simply report the done state of the last bunch of runs, then quit. - Fix: if a .yml workflow file couldn't be parsed, its "run" would persist in the list of runs for longer than it should. Now those unparsable runs aren't displayed at all.
- Fix: skipped runs are considered finished, and don't need their jobs shown.
- Error reporting is improved, removing unneeded noisy tracebacks in some cases, and providing more information for GitHub API errors. Closes issue 8.
- More operations are retried on failure, fixing issue 10.
- Interrupting with ctrl-C will set the exit status to 2.
- The
--polloption sets the number of seconds to wait between refreshes. - Requests to GitHub are now made asynchronously, speeding execution.
- Redirections from GitHub (for example, if a repo is renamed or moved) are followed transparently.
- The exit code is now 1 if any runs failed, 0 if all were successful.
- Long lines are no longer wrapped too short.
- Runs can be selected by a commit SHA by using
--shaon the command line. - Retry if GitHub returns "502 - Bad Gateway".
- Uses a
GITHUB_TOKENenvironment variable for authentication if it is defined.
- Support more forms of repo URLs:
[email protected]:, without.git, etc. - Better error messages if the repo URL can't be parsed.
First version
The code is a bit messy and undocumented, and there are no tests. If you want to change the code, open an issue and let's talk about it.
Contributors:
- Ned Batchelder
- Bill Mill
- Hugo van Kemenade
- Rob Weir
This started as a formatter for the output of gh run list from the gh
run command. Then I tried gh run watch, but wasn't happy with its
choices. So I wrote my own.