Helm4 upgrade with dry-run and server-side enabled doesn't catch errors

Hi.

Let's assume we have the following chart:

1. Chart.yaml

apiVersion: v2
name: helm4-test
version: 1.0.0

2. templates/deployment.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: bash
spec:
  replicas: 1
  selector:
    matchLabels:
      app.kubernetes.io/name: bash
      app.kubernetes.io/instance: bash
  template:
    metadata:
      labels:
        app.kubernetes.io/name: bash
        app.kubernetes.io/instance: bash
    spec:
      resources:
        requests:
          memory: "100Mi"
        limits:
          memory: "200Mi"
      containers:
        - image: bash
          imagePullPolicy: IfNotPresent
          name: bash
          command: ["bash", "-c", "echo 'Hello, World!'"]
      restartPolicy: Always

This is wrong spec - resources block is placed not under container.

Run helm dry-run with server-side:

helm upgrade test -n default --dry-run=server --server-side=true --install .

Result: No errors.

What is expected?

I expected that Helm4 should return error that resources is unknown field for spec.

Because if we run actual upgrade without dry-run, we will get the error:

╰─ helm upgrade test -n default --server-side=true --install .
Release "test" does not exist. Installing it now.
Error: failed to create typed patch object (default/bash; apps/v1, Kind=Deployment): .spec.template.spec.resources: field not declared in schema

Comparing to kubectl, if we use just deployment.yaml in apply command, it will catch the error:

╰─ kubectl apply --dry-run=server --server-side=true -f manifests.yaml
Error from server: failed to create typed patch object (default/bash; apps/v1, Kind=Deployment): .spec.template.spec.resources: field not declared in schema
╰─ kubectl apply --dry-run=server -f manifests.yaml 
Error from server (BadRequest): error when creating "manifests.yaml": Deployment in version "v1" cannot be handled as a Deployment: strict decoding error: unknown field "spec.template.spec.resources"

Should Helm4 with server-side work the same way?

Otherwise, dry-run doesn't catch such errors and it can be a surprise in actual install/upgrade.

Tested on 1.28 k8s server Helm 4.0.0