Avoid InvalidPhysicalResourceId for incorrectly configured src=DockerImageAsset

[athewsey]

Confession time: Today I was bad at reading the docs and paid the price...

I tried something along the lines of (in Python CDK):

from aws_cdk.aws_ecr_assets import DockerImageAsset
from cdk_ecr_deployment import ECRDeployment, DockerImageName

image = DockerImageAsset(
    self,
    "Image",
    directory="/home/CoolestPersonAlive/etc/container",
    ...,
)
deployment = ECRDeployment(
    self,
    "Deployment",
    src=image,
    dest=DockerImageName(destination_ecr_uri),
)

As diligent students of the readme will know already, I should've instead had src=DockerImageName(image.image_uri). The above breaks.

But, I'm raising this because it broke in a difficult-to-diagnose way: The synthesis completed and started CloudFormation deployment, to ultimately fail with a cryptic "Invalid PhysicalResourceId" message. It took some time and Google-fu to realise I needed to:

• Try cdk deploy again with --no-rollback to avoid obliterating the part-built resources
• Figure out which Lambda in my account was backing that particular custom resource based on the LogicalResourceId (this is a complex stack, so there was a bit of searching to do)
• Open the CloudWatch logs and find the2022/09/27 15:10:27 sending status failed: can't get SrcImage message hinting that something was up with the src configuration.

It feels like this was a pretty easy mistake to make, so 2 questions:

1. Could this construct maybe be extended to handle DockerImageAsset inputs directly & correctly?
2. Regardless, would it be possible to make the error handling a bit more informative here? Ideal if it could fail earlier in the process, but even a more understandable error later on would be useful!