Preventing Xamarin(/Mono?) from optimizing on iOS - ios

When I use reflection in ViewDidLoad like this:
foreach (var m in this.GetType().GetMethods()) {
Console.WriteLine (m.Name);
}
without any other code, methods like
Add
set_View
are not there; when I add;
public void PreventOptimizing() {
var x = this.View;
this.View = x;
this.Add (null);
}
to the class and without calling that method, they are there. So I assume the AOT compilation optimizes these methods away as they are not called. I don't know which methods it adds away so I would like the compiler, for my experiment, not to optimize anything away. How can that be done? Or is there another trick preventing automated removal of method?
Edit: so full code, if this isn't clear enough;
result output will not contain 'Add' and 'set_View';
public class TestController : UIViewController
{
public override void ViewDidLoad ()
{
base.ViewDidLoad ();
foreach (var m in this.GetType().GetMethods()) {
Console.WriteLine (m.Name);
}
}
}
Output does contain Add and set_View;
public class TestController : UIViewController
{
public void PreventOptimizing() {
var x = this.View;
this.View = x;
this.Add (null);
}
public override void ViewDidLoad ()
{
base.ViewDidLoad ();
foreach (var m in this.GetType().GetMethods()) {
Console.WriteLine (m.Name);
}
}
}

You either set the linker option to "Link SDK assemblies only" or use the PreserveAttribute on the class.
Details are in Xamarin website:
http://developer.xamarin.com/Guides/ios/Advanced_Topics/Linker/
Linker Behavior
The linking process can be customized in different ways. The primary
mechanism for controlling the linker in Xamarin Studio is the Linker
Behavior drop-down in the iOS Build Xamarin Studio's Project Options
dialog box. In Visual Studio, this is located in project Properties,
under iOS Build.
Preserving Code
When you use the linker it can sometimes remove code
that you might have called dynamically either using
System.Reflection.MemberInfo.Invoke, or by exporting your methods to
Objective-C using the [Export] attribute and then invoking the
selector manually.
In those cases, you can instruct the linker to consider either entire
classes to be used or individual members to be preserved by applying
the [Xamarin.iOS.Foundation.Preserve] attribute either at the
class-level or the member-level. Every member that is not statically
linked by the application is subject to be removed. This attribute is
hence used to mark members that are not statically referenced, but
that are still needed by your application.
For instance, if you instantiate types dynamically, you may want to
preserve the default constructor of your types. If you use XML
serialization, you may want to preserve the properties of your types.
You can apply this attribute on every member of a type, or on the type
itself. If you want to preserve the whole type, you can use the syntax
[Preserve (AllMembers = true)] on the type.
Sometimes you want to preserve certain members, but only if the
containing type was preserved. In those cases, use [Preserve
(Conditional=true)]

Related

Should subclasses inherit private mixin variables in Dart?

Should I get the following error:
class.dart:11:11: Error: The getter '_privateID' isn't defined for the class 'Y'.
- 'Y' is from 'class.dart'.
Try correcting the name to the name of an existing getter, or defining a getter or field named '_privateID'.
From the following code?
mixin.dart:
class Mixin {
static int _nextID = 0;
int publicID = _nextID++; // I only need one of these lines
int _privateID = _nextID++; // but this variable is inaccessible
}
class.dart:
import 'mixin.dart';
class X with Mixin {
void run() {
print(publicID); // no error here
}
}
class Y with Mixin {
void run() {
print(_privateID); // Error: _privateID not defined
}
}
void main() {
Y().run();
}
Or is this a bug? If it's not a bug, I'd like to understand why this behavior is reasonable.
When I instead define the mixin in the same file as the above classes, I get no error.
(Dart SDK 2.4.1.)
It is not a bug.
The private field is inherited, but you cannot access it because its name is private to a different library.
Dart's notion of "privacy" is library private names.
The name _privateID in the mixin.dart library introduces a library private name. This name is special in that it can only be written inside the same library.
If someone writes _privateID in a different library, it is a different name, one unique to that library instead.
It is as if private names includes the library URI of the library it is written in, so what you really declare is a name _privateID#mixin.dart.
When you try to read that field in class.dart, you write ._privateID, but because it is in a different library, what you really write is ._privateID#class.dart, a completely different name, and the classs does not have any declarations with that name.
So, if one class needs to access a private member of another class (or mixin, or anything), then the two needs to be declared in the same library, because otherwise they cannot even write the name of that variable.
That is why the code works if you write the mixin in the same library.
If you want to move the mixin to a separate file, but not necessarily a separate library, you can use a part file.

No default instance or named instance 'Default' for requested plugin type

I'm trying to avoid referencing the concrete type library in my main project, but I'm getting this error:
No default instance or named instance 'Default' for requested plugin type StackExchangeChatInterfaces.IClient
1.) Container.GetInstance(StackExchangeChatInterfaces.IClient ,{username=; password=; defaultRoomUrl=; System.Action`2[System.Object,System.Object]=System.Action`2[System.Object,System.Object]})
I've setup my container to scan for assemblies, like so:
var container = new Container(x =>
{
x.Scan(scan =>
{
scan.AssembliesFromApplicationBaseDirectory();
scan.ExcludeNamespace("StructureMap");
scan.WithDefaultConventions();
scan.AddAllTypesOf<IMessageHandlers>();
});
//x.For<IClient>().Use<Client>(); //GetInstance will work if this line is not commented out.
});
When I try to get an instance, I get the error, my code for getting an instance is here:
chatInterface = container
.With("username").EqualTo(username)
.With("password").EqualTo(password)
.With("defaultRoomUrl").EqualTo(roomUrl)
.With<Action<object, object>>(delegate(object sender, object messageWrapper)
{
string message = ((dynamic)messageWrapper).Message;
Console.WriteLine("");
Console.WriteLine(message);
foreach (var item in messageHandlers)
{
item.MessageHandler.Invoke(message, chatInterface);
}
}).GetInstance<IClient>();
If I explicitly map the concrete class to the interface, everything works hunky dory, but that means I need to reference the project that Client is in, which I don't want to do.
This is really interesting. Looks like default conventions are not able to register types with such constructor (tried on both versions 2.6.3 and 3+). I was only registered when only parameterless constructor was specified. Looking at sources of both versions it is really suspicious as it should be registered. Deeper dive into the code would be needed...
Anyway try using custom registration convention:
public class ClientConvention : IRegistrationConvention
{
public void Process(Type type, Registry registry)
{
if (type.IsClass && !type.IsAbstract && !type.IsGenericType &&
type.GetInterfaces().Contains(typeof(IClient)))
{
registry.For(typeof(IClient)).Use(type);
}
}
}
Configure it like this:
var container = new Container(
c => c.Scan(
s =>
{
s.ExcludeNamespace("StructureMap");
s.WithDefaultConventions();
s.Convention<ClientConvention>();
s.AddAllTypesOf<IMessageHandlers>();
}));
and this should work just fine.
The default type scanning will not pick up concrete types whose constructor functions contain primitive arguments like strings, numbers, or dates. The thinking is that you'd effectively have to explicitly configure those inline dependencies anyway.
"but that means I need to reference the project that Client is in, which I don't want to do."
Does that actually matter? I think you're making things harder than they have to be by trying to eliminate the assembly reference.

Role-based navigation display in MVC4 Bootstrap Sample

How are you supposed to conditionally display menu items based on roles in the Bootstrap Sample project? I was thinking of doing the following
Implement INavigatonRouteFilter - really just implementing the shouldRemove(Route navigationRoutes) method - by getting the default controller/action for the route and seeing if the user is authorized
Call NavigationRoutes.Filters.Add(myAuthorizationFilter) after configuring the NavigationRoutes in App_Start
There are two problems I see with this approach:
I don't actually know how to do the first step unless I add in a bunch of conditional statements to check for Controller's name explicitly
This seems like it could make NavigationRoutes.Filters very hard to deal with once there are a lot of filters or a desire for more modularity later on
I don't know that I've explained the problem clearly enough, but basically I want to use what is provided in the Bootstrap sample to implement authorization-based navigation menu display if at all possible. Using INavigationRouteFilter just seemed like the most natural way to do so.
For those looking for an answer or at least a quick fix.
Here's what I've come up with after 5 minutes and I most certainly haven't though about any side effects this may have.
routes.MapNavigationRoute<HomeController>("Index", c => c.Index())
.FilterRoute(() => !WebSecurity.IsAuthenticated);
You can either do all your filtering in your call to FilterRoute() or you can add more extension methods to save you some characters.
I'm thinking of .RequireRole("Adiministrators"); that calls WebSecurity.RequireRoles() in turn (or HttpContext.Current.User.IsInRole()) etc.
public static NavigationRouteBuilder FilterRoute(this NavigationRouteBuilder builder, Func<bool> func)
{
var currentRoute = builder._parent;
NavigationRoutes.Filters.Add(new BootstrapAuthorizationFilter(builder, x =>
{
if (x == currentRoute)
return func();
else
return false;
}));
return builder;
}
and BootstrapAuthorizationFilter is just a class implementing INavigationRouteFilter that calls func() in its ShouldRemove() method
public class BootstrapAuthorizationFilter : INavigationRouteFilter
{
private NavigationRouteBuilder builder;
private Func<NamedRoute, bool> func;
public BootstrapAuthorizationFilter(NavigationRouteBuilder builder, Func<NamedRoute, bool> func)
{
this.builder = builder;
this.func = func;
}
public bool ShouldRemove(Route navigationRoutes)
{
if (navigationRoutes is NamedRoute)
return func(navigationRoutes as NamedRoute);
return false;
}
}
Clearly nothing fancy and I'm not sure if I'd use it in production.
But I think is simple enough and works (for the cases I tested).
Having said that, I hope the new routing functionality is going to be released soon :)

Unexpected Validate behavior with Moq

Moq has been driving me a bit crazy on my latest project. I recently upgraded to version 4.0.10827, and I'm noticing what seems to me to be a new behavior.
Basically, when I call my mocked function (MakeCall, in this example) in the code I am testing, I am passing in an object (TestClass). The code I am testing makes changes to the TestClass object before and after the call to MakeCall. Once the code has completed, I then call Moq's Verify function. My expectation is that Moq will have recorded the complete object that I passed into MakeCall, perhaps via a mechanism like deep cloning. This way, I will be able to verify that MakeCall was called with the exact object I am expecting it to be called with. Unfortunately, this is not what I'm seeing.
I attempt to illustrate this in the code below (hopefully, clarifying it a bit in the process).
I first create a new TestClass object. Its Var property is set to "one".
I then create the mocked object, mockedObject, which is my test subject.
I then call the MakeCall method of mockedObject (by the way, the Machine.Specifications framework used in the example allows the code in the When_Testing class to be read from top to bottom).
I then test the mocked object to ensure that it was indeed called with a TestClass with a Var value of "one". This succeeds, as I expected it to.
I then make a change to the original TestClass object by re-assigning the Var property to "two".
I then proceed to attempt to verify if Moq still thinks that MakeCall was called with a TestClass with a value of "one". This fails, although I am expecting it to be true.
Finally, I test to see if Moq thinks MakeCall was in fact called by a TestClass object with a value of "two". This succeeds, although I would initially have expected it to fail.
It seems pretty clear to me that Moq is only holding onto a reference to the original TestClass object, allowing me to change its value with impunity, adversely affecting the results of my testing.
A few notes on the test code. IMyMockedInterface is the interface I am mocking. TestClass is the class I am passing into the MakeCall method and therefore using to demonstrate the issue I am having. Finally, When_Testing is the actual test class that contains the test code. It is using the Machine.Specifications framework, which is why there are a few odd items ('Because of', 'It should...'). These are simply delegates that are called by the framework to execute the tests. They should be easily removed and the contained code placed into a standard function if that is desired. I left it in this format because it allows all Validate calls to complete (as compared to the 'Arrange, Act Assert' paradigm). Just to clarify, the below code is not the actual code I am having problems with. It is simply intended to illustrate the problem, as I have seen this same behavior in multiple places.
using Machine.Specifications;
// Moq has a conflict with MSpec as they both have an 'It' object.
using moq = Moq;
public interface IMyMockedInterface
{
int MakeCall(TestClass obj);
}
public class TestClass
{
public string Var { get; set; }
// Must override Equals so Moq treats two objects with the
// same value as equal (instead of comparing references).
public override bool Equals(object obj)
{
if ((obj != null) && (obj.GetType() != this.GetType()))
return false;
TestClass t = obj as TestClass;
if (t.Var != this.Var)
return false;
return true;
}
public override int GetHashCode()
{
int hash = 41;
int factor = 23;
hash = (hash ^ factor) * Var.GetHashCode();
return hash;
}
public override string ToString()
{
return MvcTemplateApp.Utilities.ClassEnhancementUtilities.ObjectToString(this);
}
}
[Subject(typeof(object))]
public class When_Testing
{
// TestClass is set up to contain a value of 'one'
protected static TestClass t = new TestClass() { Var = "one" };
protected static moq.Mock<IMyMockedInterface> mockedObject = new moq.Mock<IMyMockedInterface>();
Because of = () =>
{
mockedObject.Object.MakeCall(t);
};
// Test One
// Expected: Moq should verify that MakeCall was called with a TestClass with a value of 'one'.
// Actual: Moq does verify that MakeCall was called with a TestClass with a value of 'one'.
// Result: This is correct.
It should_verify_that_make_call_was_called_with_a_value_of_one = () =>
mockedObject.Verify(o => o.MakeCall(new TestClass() { Var = "one" }), moq.Times.Once());
// Update the original object to contain a new value.
It should_update_the_test_class_value_to_two = () =>
t.Var = "two";
// Test Two
// Expected: Moq should verify that MakeCall was called with a TestClass with a value of 'one'.
// Actual: The Verify call fails, claiming that MakeCall was never called with a TestClass instance with a value of 'one'.
// Result: This is incorrect.
It should_verify_that_make_call_was_called_with_a_class_containing_a_value_of_one = () =>
mockedObject.Verify(o => o.MakeCall(new TestClass() { Var = "one" }), moq.Times.Once());
// Test Three
// Expected: Moq should fail to verify that MakeCall was called with a TestClass with a value of 'two'.
// Actual: Moq actually does verify that MakeCall was called with a TestClass with a value of 'two'.
// Result: This is incorrect.
It should_fail_to_verify_that_make_call_was_called_with_a_class_containing_a_value_of_two = () =>
mockedObject.Verify(o => o.MakeCall(new TestClass() { Var = "two" }), moq.Times.Once());
}
I have a few questions regarding this:
Is this expected behavior?
Is this new behavior?
Is there a workaround that I am unaware of?
Am I using Verify incorrectly?
Is there a better way of using Moq to avoid this situation?
I thank you humbly for any assistance you can provide.
Edit:
Here is one of the actual tests and SUT code that I experienced this problem with. Hopefully it will act as clarification.
// This is the MVC Controller Action that I am testing. Note that it
// makes changes to the 'searchProjects' object before and after
// calling 'repository.SearchProjects'.
[HttpGet]
public ActionResult List(int? page, [Bind(Include = "Page, SearchType, SearchText, BeginDate, EndDate")]
SearchProjects searchProjects)
{
int itemCount;
searchProjects.ItemsPerPage = profile.ItemsPerPage;
searchProjects.Projects = repository.SearchProjects(searchProjects,
profile.UserKey, out itemCount);
searchProjects.TotalItems = itemCount;
return View(searchProjects);
}
// This is my test class for the controller's List action. The controller
// is instantiated in an Establish delegate in the 'with_project_controller'
// class, along with the SearchProjectsRequest, SearchProjectsRepositoryGet,
// and SearchProjectsResultGet objects which are defined below.
[Subject(typeof(ProjectController))]
public class When_the_project_list_method_is_called_via_a_get_request
: with_project_controller
{
protected static int itemCount;
protected static ViewResult result;
Because of = () =>
result = controller.List(s.Page, s.SearchProjectsRequest) as ViewResult;
// This test fails, as it is expecting the 'SearchProjects' object
// to contain:
// Page, SearchType, SearchText, BeginDate, EndDate and ItemsPerPage
It should_call_the_search_projects_repository_method = () =>
s.Repository.Verify(r => r.SearchProjects(s.SearchProjectsRepositoryGet,
s.UserKey, out itemCount), moq.Times.Once());
// This test succeeds, as it is expecting the 'SearchProjects' object
// to contain:
// Page, SearchType, SearchText, BeginDate, EndDate, ItemsPerPage,
// Projects and TotalItems
It should_call_the_search_projects_repository_method = () =>
s.Repository.Verify(r => r.SearchProjects(s.SearchProjectsResultGet,
s.UserKey, out itemCount), moq.Times.Once());
It should_return_the_correct_view_name = () =>
result.ViewName.ShouldBeEmpty();
It should_return_the_correct_view_model = () =>
result.Model.ShouldEqual(s.SearchProjectsResultGet);
}
/////////////////////////////////////////////////////
// Here are the values of the three test objects
/////////////////////////////////////////////////////
// This is the object that is returned by the client.
SearchProjects SearchProjectsRequest = new SearchProjects()
{
SearchType = SearchTypes.ProjectName,
SearchText = GetProjectRequest().Name,
Page = Page
};
// This is the object I am expecting the repository method to be called with.
SearchProjects SearchProjectsRepositoryGet = new SearchProjects()
{
SearchType = SearchTypes.ProjectName,
SearchText = GetProjectRequest().Name,
Page = Page,
ItemsPerPage = ItemsPerPage
};
// This is the complete object I expect to be returned to the view.
SearchProjects SearchProjectsResultGet = new SearchProjects()
{
SearchType = SearchTypes.ProjectName,
SearchText = GetProjectRequest().Name,
Page = Page,
ItemsPerPage = ItemsPerPage,
Projects = new List<Project>() { GetProjectRequest() },
TotalItems = TotalItems
};
Ultimately, your question is whether a mocking framework should take snapshots of the parameters you use when interacting with the mocks so that it can accurately record the state the system was in at the point of interaction rather than the state the parameters might be in at the point of verification.
I would say this is a reasonable expectation from a logical point of view. You are performing action X with value Y. If you ask the mock "Did I perform action X with value Y", you expect it to say "Yes" regardless of the current state of the system.
To summarize the problem you are running into:
You first invoke a method on a mock object with a reference type parameter.
Moq saves information about the invocation along with the reference type parameter passed in.
You then ask Moq if the method was called one time with an object equal to the reference you passed in.
Moq checks its history for a call to that method with a parameter that matches the supplied parameter and answers yes.
You then modify the object that you passed as the parameter to the method call on the mock.
The memory space of the reference Moq is holding in its history changes to the new value.
You then ask Moq if the method was called one time with an object that isn't equal to the reference its holding.
Mock checks its history for a call to that method with a parameter that matches the supplied parameter and reports no.
To attempt to answer your specific questions:
Is this expected behavior?
I would say no.
Is this new behavior?
I don't know, but it's doubtful the project would have at one time had behavior that facilitated this and was later modified to only allow the simple scenario of only verifying a single usage per mock.
Is there a workaround that I am unaware of?
I'll answer this two ways.
From a technical standpoint, a workaround would be to use a Test Spy rather than a Mock. By using a Test Spy, you can record the values passed and use your own strategy for remembering the state, such as doing a deep clone, serializing the object, or just storing the specific values you care about to be compared against later.
From a testing standpoint, I would recommend that you follow the principle "Use The Front Door First". I believe there is a time for state-based testing as well as interaction-based testing, but you should try to avoid coupling yourself to the implementation details unless the interaction is an important part of the scenario. In some cases, the scenario you are interested in will be primarily about interaction ("Transfer funds between accounts"), but in other cases all you really care about is getting the correct result ("Withdraw $10"). In the case of the specification for your controller, this seems to fall into the query category, not the command category. You don't really care how it gets the results you want as long as they are correct. Therefore, I would recommend using state-based testing in this case. If another specification concerns issuing a command against the system, there may still end up being a front door solution which you should consider using first, but it may be necessary or important to do interaction based testing. Just my thoughts though.
Am I using Verify incorrectly?
You are using the Verify() method correctly, it just doesn't support the scenario you are using it for.
Is there a better way of using Moq to avoid this situation?
I don't think Moq is currently implemented to handle this scenario.
Hope this helps,
Derek Greer
http://derekgreer.lostechies.com
http://aspiringcraftsman.com
#derekgreer
First, you can avoid the conflict between Moq and MSpec by declaring
using Machine.Specifications;
using Moq;
using It = Machine.Specifications.It;
Then you'll only need to prefix with Moq. when you want to use Moq's It, for example Moq.It.IsAny<>().
Onto your question.
Note: This is not the original answer but an edited one after the OP added some real example code to the question
I've been trying out your sample code and I think it's got more to do with MSpec than Moq. Apparently (and I didn't know this either), when you modify the state of your SUT (System Under Test) inside an It delegate the changes gets remembered. What is happening now is:
Because delegate is run
It delegates are run, one after the other. If one changes the state, the following It will never see the set up in the Because. Hence your failed test.
I've tried marking your spec with the SetupForEachSpecificationAttribute:
[Subject(typeof(object)), SetupForEachSpecification]
public class When_Testing
{
// Something, Something, something...
}
The attribute does as its name says: It will run your Establish and Because before every It. Adding the attribute made the spec behave as expected: 3 Successes, one fail (the verification that with Var = "two").
Would the SetupForEachSpecificationAttribute solve your problem or is resetting after every It not acceptable for your tests?
FYI: I'm using Moq v4.0.10827.0 and MSpec v0.4.9.0
Free tip #2: If you're testing ASP.NET MVC apps with Mspec you might want to take a look at James Broome's MSpec extensions for MVC

Error "More than one matching bindings are available" when using Ninject.Web.Mvc 2.0 and ASP.NET MVC 1.0

Recently I've switched to Ninject 2.0 release and started getting the following error:
Error occured: Error activating SomeController
More than one matching bindings are available.
Activation path:
1) Request for SomeController
Suggestions:
1) Ensure that you have defined a binding for SomeController only once.
However, I'm unable to find certain reproduction path. Sometimes it occurs, sometimes it does not.
I'm using NinjectHttpApplication for automatic controllers injection. Controllers are defined in separate assembly:
public class App : NinjectHttpApplication
{
protected override IKernel CreateKernel()
{
INinjectModule[] modules = new INinjectModule[] {
new MiscModule(),
new ProvidersModule(),
new RepositoryModule(),
new ServiceModule()
};
return new StandardKernel(modules);
}
protected override void OnApplicationStarted()
{
RegisterRoutes(RouteTable.Routes);
RegisterAllControllersIn("Sample.Mvc");
base.OnApplicationStarted();
}
/* ............. */
}
Maybe someone is familiar with this error.
Any advice?
I finally figured this issue out recently. Apparently, the NinjectHttpApplication.RegisterAllControllersIn() function doesn't do all of the proper bindings needed. It binds your concrete controller implementations to IController requests. For example, if you have a controller class called SampleMvcController, which inherits from System.Web.Mvc.Controller. It would do the following named binding during application start:
kernel.Bind<IController>().To(SampleMvcController).InTransientScope().Named("SampleMvc");
But when debugging the NinjectControllerFactory, I find that request are being made for the Ninject Kernel to return an object for the class "SampleMvcController", not for a concrete implementation of IController, using the named binding of "SampleMvc".
Because of this, when the first web request that involves the SampleMvcController is made, it creates a binding of SampleMvcController to itself. This is not thread safe though. So if you have several web requests being made at once, the bindings can potentially happen more than once, and now you are left with this error for having multiple bindings for the SampleMvcController.
You can verify this by quickly refreshing an MVC URL, right after causing your web application to restart.
The fix:
The simplest way to fix this issue is to create a new NinjectModule for your controller bindings, and to load this module during application start. Within this module, you self bind each of your defined controllers, like so:
class ControllerModule : StandardModule {
public override Load() {
Bind<SampleMvcController>().ToSelf();
Bind<AnotherMvcController>().ToSelf();
}
}
But if you don't mind changing the Ninject source code, you can modify the RegisterAllControllersIn() function to self bind each controller it comes across.
I have been dealing with this problem for months. I tried so many options but was unable to come to a solution. I knew that it was a threading problem because it would only occur when there was a heavy load on my site. Just recently a bug was reported and fixed in the ninject source code that solves this problem.
Here is a reference to the issue. It was fixed in build 2.1.0.70 of the Ninject source. The key change was in KernelBase.cs by removing the line
context.Plan = planner.GetPlan(service);
and replacing it with
lock (planner)
{
context.Plan = planner.GetPlan(service);
}
To use this new build with MVC you will need to get the latest build of Ninject then get the latest build of ninject.web.mvc. Build ninject.web.mvc with the new Ninject build.
I have been using this new build for about a week with a heavy load and no problems. That is the longest it has gone without a problem so I would consider this to be a solution.
Are you sure you really are creating a single completely new Kernel from scratch in your OnApplicationStarted every time it's invoked ? If you're not and you're actually creating it once but potentially running the registration bit twice. Remember that you're not guaranteed to only ever have one App class instantiated ever within a given AppDomain.
My answer was a bit more obvious.
I had declared the binding for one of my controllers more than once during refactor of my code.
I added this to my global.ascx.cs file:
public void RegisterAllControllersInFix(Assembly assembly)
{
RegisterAllControllersInFix(assembly, GetControllerName);
}
public void RegisterAllControllersInFix(Assembly assembly, Func<Type, string> namingConvention)
{
foreach (Type type in assembly.GetExportedTypes().Where(IsController))
Kernel.Bind(type).ToSelf();
}
private static bool IsController(Type type)
{
return typeof(IController).IsAssignableFrom(type) && type.IsPublic && !type.IsAbstract && !type.IsInterface;
}
private static string GetControllerName(Type type)
{
string name = type.Name.ToLowerInvariant();
if (name.EndsWith("controller"))
name = name.Substring(0, name.IndexOf("controller"));
return name;
}
Then called it from my OnApplicationStarted() method as follows:
RegisterAllControllersIn(Assembly.GetExecutingAssembly());
RegisterAllControllersInFix(Assembly.GetExecutingAssembly());
Difficult to know whether this fixed it though because it's so intermittent.

Resources