Regression on terraform validate since version 0.15.2

[BzSpi]

The terraform validate command behavior seems to have changed since version 0.15.2 and do not fail anymore for unknown attributes in a dynamic block that depends from a variable.

Terraform Version

0.15.2 and above

Terraform Configuration Files

provider "azurerm" {
  features {}
}

variable "test" {
  type = bool
}

resource "azurerm_app_service_plan" "example" {
  name                = "test-plan"
  location            = "West Europe"
  resource_group_name = "test-rg"

  dynamic "sku" {
    for_each = var.test ? ["_"] : []

    content {
      tier = "Standard"
      size = "S1"

      unknown_attribute = "foobar"
    }
  }
}
...

Expected Behavior

Test with Terraform 0.15.1

$ terraform version
Terraform v0.15.1
on linux_amd64
+ provider registry.terraform.io/hashicorp/azurerm v2.66.0

Your version of Terraform is out of date! The latest version
is 1.0.1. You can update by downloading from https://www.terraform.io/downloads.html

$ terraform validate
╷
│ Error: Unsupported argument
│ 
│   on test2.tf line 21, in resource "azurerm_app_service_plan" "example":
│   21:       unknown_attribute = "foobar"
│ 
│ An argument named "unknown_attribute" is not expected here.
╵

Actual Behavior

Test with Terraform 0.15.2

$ terraform version
Terraform v0.15.2
on linux_amd64
+ provider registry.terraform.io/hashicorp/azurerm v2.66.0

Your version of Terraform is out of date! The latest version
is 1.0.1. You can update by downloading from https://www.terraform.io/downloads.html

$ terraform validate
Success! The configuration is valid.

Same with Terraform 1.0.1

$ terraform version
Terraform v1.0.1
on linux_amd64
+ provider registry.terraform.io/hashicorp/azurerm v2.66.0

$ terraform validate
Success! The configuration is valid.

Steps to Reproduce

1. `terraform init`
2. `terraform validate`

[pdecat]

Probably related to:

core: Update HCL to allow better planning of dynamic blocks (#28424)

https://github.com/hashicorp/terraform/releases/tag/v0.15.2

[jbardin]

Hi @BzSpi, thanks for filing the issue. The change @pdecat pointed to is the change which would have caused this. While the use of dynamic will effectively create a block which only expects for_each, iterator and content, I suspect there is still a way to validate the structure earlier.

[github-actions]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.
If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.