Why are changes to a Lambda Function code not considered an update?

[dashmug]

I encountered the following where code for a Lambda Function was updated. cdk diff shows the diff correctly but the summary created by this action is incorrect, "0 to add, 0 to update, 0 to destroy".

I checked out the code to see if I can fix it and submit a PR but then I found out that it is intentional.

In src/diff.ts#L117,

const resourceType = change.oldValue?.Type ?? change.newValue?.Type;
switch (resourceType) {
  case 'AWS::CDK::Metadata':
    return;
  case 'AWS::Lambda::Function':
    const keys = Object.keys(change.propertyUpdates);
    if (
      (keys.length <= 2 && keys.includes('Code')) ||
      keys.includes('Metadata')
    ) {
      return;
    }
}

Any reason why we drift from cdk diff's result? If it's 0 to add, 0 to update, 0 to destroy, I expect it to be a no-op CDK deploy.

[laazyj]

Cross-referencing #162: this is the same root cause seen from the other side. The Lambda Code/Metadata exemption in src/diff.ts only mutates the counts (evaluateDiff), while the comment body is rendered from the raw TemplateDiff via formatDifferences — hence the diff shows but the summary says 0 to update.

The flip side (in #162) is that resources not covered by that exemption, e.g. Custom::CDKBucketDeployment SourceObjectKeys, are classified MAY_REPLACE and trip the destructive gate on routine content changes. A filter applied to the TemplateDiff itself (rather than just the counts) would make counts, body, and destructive classification all consistent.