DataContractSerializer caches collectible types in statics

[mburnell]

Description

The AssemblyLoadContext of any type that has undergone data contract serialization cannot be unloaded / collected. The specific offender is DataContractCriticalHelper, which has a few static caches and also retains a reference to the most-recently-serialized type.

Reproduction Steps

1. Create a collectible AssemblyLoadContext
2. Load a DataContractSerializable type into the collectible ALC
3. Instantiate the collectible type
4. Serialise the instance using something like:

var ser = new DataContractSerializer(type);
using (var ms = new System.IO.MemoryStream())
{
 ser.WriteObject(ms, document);
 var bytes = ms.ToArray();
 return System.Text.Encoding.UTF8.GetString(bytes, 0, bytes.Length);
}

5. Invoke Unload() on the collectible ALC
6. Trigger a garbage collection (collect / finalise / collect)
7. Review loaded assemblies using a debugger or something like AppDomain.CurrentDomain.GetAssemblies()
8. Observe that the collectible ALC / assembly has not unloaded.
9. Repeat steps but skip step 4 and observer that the ALC / assembly has unloaded properly

Expected behavior

Unloading and collecting results in proper collection of collectible types / assemblies / contexts.

Actual behavior

Collectible types / assemblies / contexts that have been data contract serialised cannot be collected.

Regression?

No response

Known Workarounds

The following code allows proper collection to occur:

private static void ClearDataContractCaches()
{
 var t = AppDomain.CurrentDomain.GetAssemblies()
 .SelectMany(x => x.GetTypes())
 .Single(x => x.Name == "DataContractCriticalHelper");

 var staticCache = t.GetFields(BindingFlags.NonPublic | BindingFlags.Static).Select(x => x.GetValue(t)).ToArray();
 
 foreach (var dictionary in staticCache.OfType<IDictionary>())
 dictionary.Clear();

 foreach (var array in staticCache.OfType<Array>())
 {
 for (var i = 0; i < array.Length; I++)
 {
 array.SetValue(null, i);
 }
 }
 
 var handleBreaker = new DataContractSerializer(typeof(object));
 var o = new object();
 using (var ms = new MemoryStream())
 {
 handleBreaker.WriteObject(ms, o);
 }
}

Configuration

.NET 6
macOS Ventura 13.0
ARM64

Other information

No response

Tagging subscribers to this area: @dotnet/area-system-collections
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Description

The AssemblyLoadContext of any type that has undergone data contract serialization cannot be unloaded / collected. The specific offender is DataContractCriticalHelper, which has a few static caches and also retains a reference to the most-recently-serialized type.

Reproduction Steps

1. Create a collectible AssemblyLoadContext
2. Load a DataContractSerializable type into the collectible ALC
3. Instantiate the collectible type
4. Serialise the instance using something like:

var ser = new DataContractSerializer(type);
using (var ms = new System.IO.MemoryStream())
{
 ser.WriteObject(ms, document);
 var bytes = ms.ToArray();
 return System.Text.Encoding.UTF8.GetString(bytes, 0, bytes.Length);
}

5. Invoke Unload() on the collectible ALC
6. Trigger a garbage collection (collect / finalise / collect)
7. Review loaded assemblies using a debugger or something like AppDomain.CurrentDomain.GetAssemblies()
8. Observe that the collectible ALC / assembly has not unloaded.
9. Repeat steps but skip step 4 and observer that the ALC / assembly has unloaded properly

Expected behavior

Unloading and collecting results in proper collection of collectible types / assemblies / contexts.

Actual behavior

Collectible types / assemblies / contexts that have been data contract serialised cannot be collected.

Regression?

No response

Known Workarounds

The following code allows proper collection to occur:

private static void ClearDataContractCaches()
{
 var t = AppDomain.CurrentDomain.GetAssemblies()
 .SelectMany(x => x.GetTypes())
 .Single(x => x.Name == "DataContractCriticalHelper");

 var staticCache = t.GetFields(BindingFlags.NonPublic | BindingFlags.Static).Select(x => x.GetValue(t)).ToArray();
 
 foreach (var dictionary in staticCache.OfType<IDictionary>())
 dictionary.Clear();

 foreach (var array in staticCache.OfType<Array>())
 {
 for (var i = 0; i < array.Length; I++)
 {
 array.SetValue(null, i);
 }
 }
 
 var handleBreaker = new DataContractSerializer(typeof(object));
 var o = new object();
 using (var ms = new MemoryStream())
 {
 handleBreaker.WriteObject(ms, o);
 }
}

Configuration

.NET 6
macOS Ventura 13.0
ARM64

Other information

No response

Author:	mburnell
Assignees:	-
Labels:	area-System.Collections
Milestone:	-