I am generating code with Roslyn and loading it into a collectible AssemblyLoadContext. Then I try to generate a Castle dynamic proxy for a class from the generated assembly in the collectible load context, but get this error:
A non-collectible assembly may not reference a collectible assembly
I guess the issue is that the load context can't be unloaded if another non-collectible assembly is referencing it.
I have tried to load Castle.Core.dll into the collectible load context and then invoke proxy generation via reflection and hoped this would solve the problem. But I still get the same error. I have double-checked that all castle-types comes from the plugin-assembly-path. I assumed that all assemblies loaded into the collectible load context would become collectible themselves?
Any hope I could get this to work somehow?
Related
I have a micronaut API like this:
#Get("/")
List<Club> listClubs()
#Get("/{id}")
Club show(Long id)
In my unit test, when I invoke the show method, the listClubs() method is actually getting invoked, instead.
Why is this happening?
Details:
Thinking that my URL mappings must be wrong, I started debugging into Netty to try to understand how the framework constructs URLs.
In HttpClientIntroductionAdvice, the context shows the API method like this:
Club show(Long param0)
The interceptor is setting param0 in the parameter map, which doesn't match the actual parameter name of my method. When the URI template is expanded, this causes the ID to get dropped (thus the URI becomes / instead of /1).
I am trying to follow this example:
https://github.com/alvarosanchez/micronaut-workshop/tree/master/ex02/solution/clubs
There is one important different in my project, which is that the endpoint is set at "/club" instead of at "/":
#Controller("/club")
#Client("/club")
I am using a diff tool to compare my project to the sample, but I am struggling to find any other difference (besides package name changes).
Why is this happening? What should I be looking for?
Thanks
Update:
Tested the target endpoint with the browser - looks fine.
Gradle clean does not resolve the issue.
I switched from debugging the Application class with IntelliJ to using "gradlew run" and in the process, I made a change to build.gradle (adding JVM properties pass-through from the gradle CLI). I also played with enabling/disabling the annotation processor in the IDE.
(note: In the previous project, I enabled annotation processing as soon as I imported into the IDE. On this project, I didn't enable it until I started having issues.)
I think the build.gradle alteration caused the problem to go away. Since the issue shows up unreliably, it's hard to tell for certain if this is the change that caused it to be fixed.
I have a WebApp on Azure that uses a dll. This library needs Interop libraries x86 and x64.
Sometimes, at the restart of the App (I suppose), the App fails due to an exception:
System.EntryPointNotFoundException: Unable to find an entry point named 'sqlite3_config' in DLL 'SQLite.Interop.dll'. at System.Data.SQLite.UnsafeNativeMethods.sqlite3_config_none(SQLiteConfigOpsEnum op) at System.Data.SQLite.SQLite3.StaticIsInitialized() at System.Data.SQLite.SQLiteLog.Initialize() at System.Data.SQLite.SQLiteConnection..ctor(String connectionString, Boolean parseViaFramework) at T_Dox.WebService.SQLiteDb.CreateConnection() at WebService.CeDb.Connect()
The SQLite used is the SQLCipher's one.
What am I missing here? I don't understand why the app stops working suddenly even if I don't make any changes.
The App is a Web Service (.asmx file) that uses a data access layer to perform some business logic.
It was under a web site project, then we moved it into another project, a webapi\mvc project.
The routing bypasses this extension, so it works as before, a simple web service call.
The called web method initializes a business class loaded from another .net library (a VB.Net library).
Inside, this class uses a wrapper to a sqlConnection, in this case the SQLiteConnection.
In its constructor it starts an SQLiteConnection, and normally it works.
Then it performs some CRUD operations ...
So I can represent the operation this way:
[WebService(Namespace = "...")]
[WebServiceBinding(ConformsTo = WsiProfiles.BasicProfile1_1)]
public class SampleService : System.Web.Services.WebService
{
[WebMethod]
public ServerInfo Test()
{
var sampleBusinessClass = new SampleBusinessCLass();
sampleBusinessClass.DoSomething();
using(var connection = new SQLiteConnection()) //the constructor is the parameterless one
{
//...
}
}
}
And the stack will be this (this is not the real one):
System.EntryPointNotFoundException: Unable to find an entry point named 'sqlite3_config' in DLL 'SQLite.Interop.dll'.
at System.Data.SQLite.UnsafeNativeMethods.sqlite3_config_none(SQLiteConfigOpsEnum op)
at System.Data.SQLite.SQLite3.StaticIsInitialized()
at System.Data.SQLite.SQLiteLog.Initialize()
at System.Data.SQLite.SQLiteConnection..ctor(String connectionString, Boolean parseViaFramework)
at xxx.WebService.SampleService.Test()
It always works, but sometimes it starts to launch this error until the stop and start of the web application on iss (in our case: Azure).
Inspecting the System.Data.SQlite.dll I can clearly see the entry point and actually it always passes this internal code (no conditions that can bypass this part) and it generally works.
The System.Data.SQlite.dll (1.0.96.0 version) is provided by SqlCypher product. I think it is the original System.Data.SQLite one because at first sight I can see the same assembly manifest and content.
The interop System.Data.SQLite uses is probably modified by SqlCypher team to give their features.
To avoid possible issues we put the interop in the path /bin/x64, then we compile our web app ONLY in x64 and it runs on a x64 environment.
I have an Azure Mobile Service project which has a dependency to another (persistence) project which is referencing FluentNHibernate. Locally everything is running correctly in Release and Debug mode. Publishing the project seems to be successful (blue smiley). The problems starts when I make a request, where FluentNHibernate is used. I get the following error message:
Unable to find assembly 'FluentNHibernate, Version=2.0.3.0, Culture=neutral, PublicKeyToken=null'.
I already tried a lot of things:
Reinstalling packages
A new plain vanilla Mobile Service
Adding dependentAssembly in Web.config in the main project and App.config in the persistence project.
A little confusing for me is following fact: When I change the version of the FluentNHibernate package, I can see in the publish preview window that this dll will not be updated.
I am really not sure if this problem is depended to this specific package (FluentNHibernate). For example, what means: PublicKeyToken=null?
What else can I try to make the service running in the cloud?
The code below worked for my solution. It wires up a handler to the AppDomain's AssemblyResolve event, which is raised if an assembly cannot be found. In this case, I tell it to check the currently loaded assemblies and return one if there is a match, which there should be for FluentNHibernate. Try sticking this as the first line in WebApiConfig.Register
public static void Register()
{
AppDomain.CurrentDomain.AssemblyResolve += (sender, args) =>
{
return AppDomain.CurrentDomain.GetAssemblies()
.Where(a => a.FullName == args.Name).FirstOrDefault();
};
// the rest of WebApiConfig.Register...
}
I'm having serious issues trying to send share custom objects between portlets in liferay. I have a Hook Plugin, with a servlet filter, which loads an object of Type MyCustomClass and inserts it into the request object as a parameter.
When i try to read this object in a portlet's render() i get a ClassCastException, though i am casting the object to the same class.
I understand that liferay plugins have different contexts, and i already tried to change the classloader before loading the object in the bean and portlet like this:
ClassLoader portalcl = PortalClassLoaderUtil.getClassLoader();
ClassLoader currentcl = Thread.currentThread().getContextClassLoader();
Thread.currentThread().setContextClassLoader(portalcl);
//do my stuff
Thread.currentThread().setContextClassLoader(currentcl);
however, it did not solved the problem, and the only way i found to solve the problem is to serialize the object into a json string, and deserialize it whenever i need it.
Isn't this kinda lame ? Does anyone know a better solution ?
Regards, DS
It sounds like the main problem you're seeing is that two different class loaders are loading the class which techncally makes them different classes (which it seems like you've already determined).
I haven't used LifeRay much but this has been a problem I've seen on other platforms as well. We were using WebSphere and solved this problem by putting the common MyCustomClass into a shared library that was on the server classpath. This way the server will load the class and make it available to all applications on the server through the server's single classloader. If you let each application load the class then you'll keep seeing this exception.
Hopefully someone can help me with this.
It is my understanding that using a ClassLoader is the most reliable way to load in content.
public class Pipeline{
public static URL getResource(String filename) {
return ClassLoader.getSystemResource(filename);
}
public static InputStream getResourceAsStream(String filename) {
return ClassLoader.getSystemResourceAsStream(filename);
}
}
If you had a file at "[jar bundle]/resources/abc.png" ..You would load it by:
URL url = Pipeline.getResource("resources/abc.png");
Loading is simple.
Saving is what's getting me.
I have a program that collects data while running, saves that data on exit, and then loads the data back in next time and keeps adding to it.
Easiest solution I think would be to save back into the jar bundle so that ClassLoader can get at them. Is this even possible? Or recommended?
I don't mind having my resources outside of the jar, just as long as I don't have to resort to 'File' to get at them and save to them. (Unless it can be done cleanly)
folder/application.jar
folder/resources/abc.png
If you could ../ back one from where the ClassLoader is looking it would be easy to cleanly get data from the directory that actually contains the jar file
Pipeline.getResource("../resources/abc.png");
Any ideas?
This isn't really what class loaders are meant for. Loading resources from the class loader is meant so that you can bundle up your application as one package and components can read each other without worrying about how the system you're deploying to is setup.
If the file in the JAR is meant to be changed by the app, then it isn't part of the app and thus probably shouldn't be in the JAR.
I don't have a lot of context on your app, but hopefully my suggestion will be valid for your situation.
I recommend setting a requirement in your app that it has a work area to which it is allowed to read and write and accept a configuration setting that specifies where this directory is. Typical ways to do this in Java are with environment variables, system properties or JNDI settings (for container deployments).
Examples:
Tomcat's startup scripts figure out where it is installed and sets a system property called catalina.home and allows you to over-ride it with an environment variable called CATALINA_HOME.
JBoss looks for JBOSS_HOME
Java application servers typically look for JAVA_HOME to find the JDK.