Issue 5988 - Template accepts instantiating an already-instantiated template type

[9rnsr]

http://d.puremagic.com/issues/show_bug.cgi?id=5988

[ghost]

LGTM.

But: Since recently we can declare templates inside of functions. Test-cases would improve in readability if we tried to avoid encoding the bug name in the template names. Your test-case could become:

void test5988()
{
    template Templ(alias T)
    {
        alias T!int Templ;
    }

    class A(T) { Templ!A foo; }
    //Templ!A fooA;    // Commented version

    void testA()
    {
        A!int a; // Was error, now works
    }

    class B(T) { Templ!B foo; }
    Templ!B fooB;      // Uncomment version

    void testB()
    {
        B!int a; // Works
    }

    testA();
    testB();
}

Maybe this is a small issue because these test-cases are mostly used to avoid regressions, but it helps with reviewing. Currently the test-case is unreadable.

[9rnsr]

@AndrejMitrovic Such rewriting would not be correct always. Some test case will need to declare symbols in DeclDefs scope (e.g. forward reference bugs). I'll do it only when I'm sure that such rewriting is definitely unrelated to the testing bug.

[9rnsr]

So I'll care test code style, but don't care test code structure, even if readability is not good. If the test code would reproduce bug, that is enough good test case.

[ghost]

Ok.

[CyberShadow]

This pull request introduced a regression:
https://issues.dlang.org/show_bug.cgi?id=15069