Thanks for your pull request, @n8sh!

Bugzilla references

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 + phobos#7923"

I suggest using UDA as the API for type author to override the default policy:

import core.memory : GC;

@(GC.ShouldScan.no)
struct MyType
{
    // ...
}

@n8sh Please add the word "Fix" at the beginning of your commit message so that the bot understands that you are actually fixing the issue.

BTW, we should check if hasIndirections(T) can be optimized via https://dlang.org/spec/traits.html#getPointerBitmap, as IIRC @rainers added this trait, because the equivalent recursive template implementation was way too slow to compile. Also we may consider integrating the handling of GC.ShouldScan UDA inside getPointerBitmap.

ping @n8sh

@RazvanN7 recently I was traveling. Updated commit message as requested.

Would be nice with a short description what it does and why it's needed. Like pseudo

Why? Improve performance... etc
How? Annotate types with uda to enable... etc

I suggest using UDA as the API for type author to override the default policy:

import core.memory : GC;

@(GC.ShouldScan.no)
struct MyType
{
    // ...
}

@PetarKirov Is there a way to do this conditionally from inside the definition of the struct?

Would be nice with a short description what it does and why it's needed.

@Imperatorn did you see the comment above shouldGCScan?

Code that manually allocates memory without using the GC is responsible for calling GC.addRange if there is the possibility the memory might contain a pointer through which GC-allocated memory is reachable. Conservatively determining this via std.traits.hasIndirections can lead to false positives when nesting multiple types that manage their own memory (for instance a refcounted array of refcounted items). This template addresses that problem by acting like std.traits.hasIndirections but adding an escape hatch: any struct or union that defines enum bool __GC_NO_SCAN = true is regarded as having no pointers that can be used to reach GC-allocated memory. Names beginning with two underscores are reserved for use by the language so there is not a danger of conflicting with an enum that is identically-named by coincidence.

Bear in mind that this template is not public so the explanation is written for Phobos developers.

Is there a way to do this conditionally from inside the definition of the struct?

If MyType is a template you can do

template Mytype(T)
{
    alias whatever = logic goes here;
    @whatever
    struct MyType { ... }
}

not sure about for a non-template

@n8sh is it possible that this could fix this issue: dlang-community/libdparse#387 (comment) ?

@n8sh how should we proceed here?