What is your MakeGenericSelf() extension method? I did not write it, I'm
quite positive about it (the linked code only contains various Reference*
extension methods, that I did write).
Anyway, I see only one problem on this side (although you haven't shown the
code which emits the IL that actually uses the Utils class you inject here):
utils = new TypeDefinition(declaringType.FullName, "<>_Utils",
TypeAttributes.NestedAssembly | TypeAttributes.Sealed | TypeAttributes.Class
| TypeAttributes.Abstract, declaringType.Module.GetType<Object>())
{
DeclaringType = resolvedType // here! DeclaringType should be
set to the original, generic TypeReference, NOT the resolved TypeDefinition
};
2011/5/25 Stanislav <[email protected]>
> I have a problem making generics. I cannot insert nongeneric class
> into generic class as nested. For ex:
>
> Orriginally I have class GenericClass<T>
>
> public class Something
> {
> public class GenericClass<T>
> {
> public SomeProperty<T> { get; set; }
> }
> }
> but when I try to insert new nested class into GenericClass<T>, I
> whant to get result:
>
> public class Something
> {
> public class GenericClass<T>
> {
> private static class Utils
> {
> private static Utils() { /* ... */ }
> public static string field;
> }
> public SomeProperty<T> {
> get { return field; }
> set { /* ... */ }
> }
> }
> }
>
> But in SomeProperty getter I have wrong accessing to GenericClass
> field:
>
> Something.GenericClass<>.Utils.field;
>
> generic class GenericClass<> is not instanciated by type.
>
> How do I add nested type (here I use Gabor Kozar's method
> MakeGenericSelf http://pastebin.com/BTX3AvpV)?
>
> var genericOwnerType = declaringType.HasGenericParameters ?
> declaringType.MakeGenericSelf() : null;
> var resolvedType = genericOwnerType == null ?
> declaringType.Resolve() : genericOwnerType.Resolve();
>
> utils = new TypeDefinition(declaringType.FullName, "<>_Utils",
> TypeAttributes.NestedAssembly |
> TypeAttributes.Sealed | TypeAttributes.Class |
> TypeAttributes.Abstract,
> declaringType.Module.GetType<Object>())
> {
> DeclaringType = resolvedType
> };
>
> resolvedType.NestedTypes.Add(utils);
> var utilsCctor = new MethodDefinition(".cctor",
> MethodAttributes.SpecialName |
> MethodAttributes.RTSpecialName |
> MethodAttributes.HideBySig |
> MethodAttributes.Static,
> declaringType.Module.GetVoidType());
>
> utilsCctor.Body.Instructions.Add(Instruction.Create(OpCodes.Ret));
> utils.Methods.Add(utilsCctor);
> m_utilsClasses.Add(declaringType, utils);
>
> Whats wrong?
>
> On 25 май, 07:23, Stanislav <[email protected]> wrote:
> > No, no, all is ok )))) Thank you very much for great help!
> >
> > On 24 май, 14:12, Gábor Kozár <[email protected]> wrote:
> >
> >
> >
> >
> >
> >
> >
> > > I agree, it's not very intuitive, but that's how it works. You may
> write
> > > extension methods, or use ones that I've shown in the discussion to
> ease the
> > > problem.
> > > If you can implement it better, please go ahead.
> >
> > > 2011/5/24 Stanislav <[email protected]>
> >
> > > > > when you Resolve() a TypeReference or MethodReference (or any
> > > > > other reference), you loose all generic arguments. You have to then
> > > > restore
> > > > > those generic arguments by essentially cloning the MethodReference
> (in
> > > > your
> > > > > case), except setting its DeclaringType to the original
> TypeReference
> > > > (which
> > > > > still contains the generic arguments).
> >
> > > > It looks like bug, not as feature )
> >
> > > > --
> > > > --
> > > > mono-cecil
>
> --
> --
> mono-cecil
>
--
--
mono-cecil
