Fixed StaticRoute working incorrectly with the same headerParam on mu…

[IrenejMarc]

…ltiple methods in interface

When the same interface had multiple functions all using the same headerParam, the comparator for AnySatisfy was reused from previously-compared functions, resulting in completely incorrect results

Some details in the code below.

Example:

import vibe.d;

interface API
{
    Collection!FooCollection foos();
}

class APIImpl : API
{
    private FooCollection _foos;

    this()
    {
        _foos = new FooCollectionImpl;
    }

    Collection!FooCollection foos()
    {
        return Collection!FooCollection(_foos);
    }
}

enum fooHeader = headerParam("foo", "FooHeader");
interface FooCollection
{
    struct CollectionIndices
    {
        long _id;
    }


// foo is the second param here
    @fooHeader
        string getSomething(long asd, string foo);

// BOTH the second param and the third one will now be headerParams("<fieldname>", "FooHeader")
    @fooHeader
        string get(long _id, int bar, string foo);
}

class FooCollectionImpl : FooCollection
{

    string getSomething(long asd, string foo)
    {
        return foo;
    }

    string get(long _id, int bar, string foo)
    {
        return foo;
    }
}

shared static this()
{
    auto settings = new HTTPServerSettings;
    settings.port = 8080;
    settings.bindAddresses = ["::1", "127.0.0.1"];

    auto router = new URLRouter;
    router.registerRestInterface(new APIImpl);

    listenHTTP(settings, router);
}

[s-ludwig]

I'm just wondering if using the manged name as an identifier is a good idea. For functions with D linkage (admittedly almost always the case), it should work, but for C++ linkage it might not for example. Safer might be to use fi.stringof.

[IrenejMarc]

Seeing as this is not a GenCmp change, but just a change in the StaticRoute, that applies to only REST interfaces, is it even possible to have C++ linkage for those methods?

[s-ludwig]

In theory it is, but admittedly highly unlikely. Ideally, GenCmp should be extended to support nested indices directly, I'll do that later when I get some time. But for now, the mangleof fix should do.