Fix druntime -preview=nosharedaccess errors

[atilaneves]

This makes druntime build and run its unittest target with -preview=nosharedaccess enabled.

[dlang-bot]

Thanks for your pull request and interest in making D better, @atilaneves! We are looking forward to reviewing it, and you should be hearing from a maintainer soon.
Please verify that your PR follows this checklist:

• My PR is fully covered with tests (you can see the coverage diff by visiting the details link of the codecov check)
• My PR is as minimal as possible (smaller, focused PRs are easier to review than big ones)
• I have provided a detailed rationale explaining my changes
• New or modified functions have Ddoc comments (with Params: and Returns:)

Please see CONTRIBUTING.md for more information.

---

If you have addressed all reviews or aren't sure how to proceed, don't hesitate to ping us with a simple comment.

Bugzilla references

Your PR doesn't reference any Bugzilla issue.

If your PR contains non-trivial changes, please reference a Bugzilla issue or create a manual changelog.

Testing this PR locally

If you don't have a local development environment setup, you can use Digger to test this PR:

dub run digger -- build "master + dmd#23056"

[thewilsonator]

Do ping when this is no longer draft

[TurkeyMan]

Thanks for moving on this! I added some comments.

[atilaneves]

Do ping when this is no longer draft

ping

[TurkeyMan]

I don't think this is ready; the shared(bool[10]) nonsense exposed an implementation bug.

---

Repeated for visibility:

So...

import core.atomic : cas;

alias A = shared(bool)[10];
alias B = shared(bool[10]);

static assert(is(A == B));
static assert(A.stringof == "shared(bool[10])");
static assert(B.stringof == "shared(bool[10])");

That's surprising on impact... but I can see how it manifests. Is it correct though? Let's think this through.

It starts shared(int)* vs shared(int*); most certainly NOT the same thing.
And then shared(int)[] vs shared(int[]); should be the same as the pointer case.

These are obvious, because the unshared pointer/slice are a variable in its own right.
But... we have the static array case (which is inhibiting you from indexing it?), where that metadata is implicit, and denying you access to its address... which can't be right.

shared(int) i;
shared(int)* pi = &i; // obviously, we can handle the address of a shared thing!

// so, this should be legal:
shared(int[10]) s;
shared(int[10])* y = &s; // <- obviously acceptable
shared(int)[] x = s[];   // <- if it's not, then that's a bug; valid to take the address of a shared thing, 10 is just an int literal

shared(int)* p = x.ptr + 5; // <- clearly legal, x.ptr/length are not shared
shared(int)* q = &x[5];     // <- so, same operation? or does the compiler 'imagine' a value resolution before the `&`?

// and then these cases:
shared(int)* r = &s[5];  // obviously legal, and valid code can arrive here as shown above
shared(int)* s = &y[5];  // ditto

So, I think I'm convinced that is(shared(bool)[10] == shared(bool[10])), because the distinction between terms is purely metadata; but the bug is that you can't index a shared static-array.
You wouldn't be able to index a shared(T[]) (the pointer itself is shared), but you CAN index a shared(T[N]); the pointer is implicit information, so can't carry any notion of shared-ness.

Needs a language fix, then remove this mess :)

[TurkeyMan]

Also, the problem with class is substantial. Outside the fact that I think D got class totally wrong and it never never should have been a pointer implicitly, that's a serious problem here.

How do we distinguish a shared(class*) from a shared(class)*? (repeat for const and every other attribute!)
We can't, and we have to assume the conservative former expression, even though the latter expression is what the user wants 99 times out of 100.

With this in mind; shared(class) is a mostly useless concept, and the fact those sync primitives are classes (!!) is not workable. They need to be re-implemented as struct's. Probably new names as struct need to be created, and the class names deprecated. Functions that handle those sync primitives need overloads for the new type :/

We also DESPERATELY need struct inheritence! We've needed it forever, we need it more than ever.
Any argument that goes "but D avoided object slicing"; a problem you read about in the first 5-minutes of education, and no C++ programmer has EVER had a problem with in their career in reality... but our reality is that the cure is so much worse than the disease.

[rikkimax]

A lot of these problems originate in a single decision, trying to force shared an already questionable in what it "proves" into representing atomics.

There is a reason why I have long held and suggested that we needed a dedicated atomic storage class.

The atomic nature of a cell does not reflect the atomic nature of any contained cells that are stored via a pointer.

Just keep in mind those sync classes, which support synchronized won't be supported as structs.
I tried to get approval to get synchronized converted to be operator overload based, as part of properly making the monitor a fully user defined field. But that was not approved.