In #131, first, and then in #156, a change was proposed to support the same functionality as CircleCI, where a terminal is open concurrent to the running job's steps. And as I mentioned in the latter PR, I have found that feature quite useful (implementing it "manually", by having two shell: bash steps).
Here is a very small PR to propose a much simpler UI but essentially offer the same feature: by setting the input detached: true, the action-tmate step will start a new, detached tmate session, print the information how to connect to that, and then continue with the workflow job's next steps, keeping the tmate session running. In the post-job phase, action-tmate will then wait for the session to end (or, if nobody connected to the session within 10 minutes of that post-job Action, terminate anyway).
Sadly, this PR results in warnings about the usage of the save-state command that would only be solved by upgrading to @actions/core 1.10.0. However, in https://github.com/mxschmitt/action-tmate/pull/152 we tried that and the same blockers are still there. So I'll punt on that, for now.
In #131, first, and then in #156, a change was proposed to support the same functionality as CircleCI, where a terminal is open concurrent to the running job's steps. And as I mentioned in the latter PR, I have found that feature quite useful (implementing it "manually", by having two
shell: bash
steps).Here is a very small PR to propose a much simpler UI but essentially offer the same feature: by setting the input
detached: true
, theaction-tmate
step will start a new, detachedtmate
session, print the information how to connect to that, and then continue with the workflow job's next steps, keeping thetmate
session running. In the post-job phase,action-tmate
will then wait for the session to end (or, if nobody connected to the session within 10 minutes of that post-job Action, terminate anyway).Sadly, this PR results in warnings about the usage of the
save-state
command that would only be solved by upgrading to@actions/core
1.10.0. However, in https://github.com/mxschmitt/action-tmate/pull/152 we tried that and the same blockers are still there. So I'll punt on that, for now.