I am wondering if there is already a way to run a test suite inside a workflow without pushing a commit.
Here is what I got so far… is all open source https://github.com/chevah/brink/pull/120 , but is based on the legacy buildbot code designed to run on AIX 5.3, HPUX and Solaris 10 on Sparc.
I am using buildbot try (https://docs.buildbot.net/current/manual/cmdline.html#try) command to help with TDD for multi-OS code.
Most of the time my usage is as follows.
I develop on my local OS (ex Linux). I make sure all the tests pass.
I then push a commit.
I find out that some tests are failing on Windows or MacOS.
I then start using
buildbot try to just re-run the selected tests for a patch without a commit
The tests are run on private hosted VM which are not reset at the end of the test run.
So is like a remote command execution.
I did a proof of concept tool for GitHub actions using the workflow_dispatch event which has a
diff as input.
I create a local diff for uncommitted changed, encode it in base64 and then send it to the workflow.
Then the workflow has the required rules for decoding and applying the diff.
For now is a big hack, as it looks like the current API is not designed for such an usage.
When I trigger the workflow_dispatch via the REST API I don’t get the run ID as part of the response.
What do you think?