Fix linker error in destroy under -betterC

[JinShil]

This is a followup to #2178

My fix was insufficient. Although it fixed the compiler error, there was still a linker error. Because we were copying two arrays, it was making a runtime call to _d_arraycopy. Since the runtime is not linked in -betterC, it resulted in an unresolved reference error at link time.

See the implementation of _d_arraycopy at

druntime/src/rt/arraycat.d

Line 22 in 4a1d176

void[] _d_arraycopy(size_t size, void[] from, void[] to)

. It basically just forwards to memcpy, so the array copy code was simply replaced with a direct call to memcpy.

The test case was also updated to compile and link.

Sorry for the mistake.

[dlang-bot]

Thanks for your pull request, @JinShil!

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 fetch digger
dub run digger -- build "master + druntime#2204"

[n8sh]

If dest[] = src[] doesn't work in betterC, that should be fixed rather than applying a bandaid here.

[JinShil]

It is possible to make dest[] = src[] work in -betterC, but that is an enormous undertaking. It requires rewriting the array lowerings in the compiler to templates (similar to what was done for __equals). I've been working on it, but is extremely difficult and filled with all kinds of problems.

For example, did you know you can set the length property of an array in @safe nothrow pure code, but it gets lowered to a runtime hook which is neither @safe, nothrow, nor pure? The compiler is lying to us. I attempted it already, but I discovered it required all the code in https://github.com/dlang/druntime/blob/098a08a9593a8e265a9b02da55bda53393f3d5c1/src/rt/lifetime.d to be rewritten. That is a long-term project and one, which if not done carefully, could cause quite a few problems. It's going to take years to get that done, but in the meantime users are expecting features like this to work.

[n8sh]

@JinShil Of course you're right.

I was also writing under the mistaken assumption that slice copying not working in betterC was a recent regression rather than a longstanding limitation, since the documentation claims that array slicing is supported in -betterC.

[wilzbach]

Though of course it would be great if betterC would support array copying out of the box, but well let's got step by step.