When using Pulumi Deployments with github integration, when I create a PR to a branch that is associated with a given stack deployment, then it is expected that a "pulumi update" only runs onces that PR is merged.
Current behavior
By running re-run check in Github for a failed PR-pulumi-preview, Pulumi Deployments runs it as if a "Github Push" occurred and thus runs a pulumi update.
Steps to reproduce
Set up a github deployment configuration for a stack. I had a dev stack connected to the main branch of it's repo.
I tested this by pushing a change directly to main to confirm Pulumi Deployments ran an update on the dev stack.
Create a new branch.
Make a change on that branch that will fail a preview. I added a non-existent environment to my Pulumi.dev.yaml.
Create a PR to merge that branch to the main branch.
This will correctly trigger a Deployment to run a pulumi preview
The preview will fail.
In Github, go to the PR and click on "Details" link for the failed check.
Click on the "re-run" link under "Pulumi"
Go to Pulumi Cloud and see that a Deployment is triggered but it sees the trigger as a Github Push and thus wants to run a pulumi update instead of a pulumi preview
Context (Environment)
Although, the assumption is that all re-runs of Deployments would be via github updates and not by clicking "re-run" in Github itself, it is not a guarantee and we should react to this case correctly.
Expected behavior
When using Pulumi Deployments with github integration, when I create a PR to a branch that is associated with a given stack deployment, then it is expected that a "pulumi update" only runs onces that PR is merged.
Current behavior
By running
re-run
check in Github for a failed PR-pulumi-preview, Pulumi Deployments runs it as if a "Github Push" occurred and thus runs apulumi update
.Steps to reproduce
pulumi preview
pulumi update
instead of apulumi preview
Context (Environment)
Although, the assumption is that all re-runs of Deployments would be via github updates and not by clicking "re-run" in Github itself, it is not a guarantee and we should react to this case correctly.
Affected feature