__delete shouldn't segfault for null

[wilzbach]

See also: #2104 (comment)

[dlang-bot]

Thanks for your pull request, @wilzbach!

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.

[schveiguy]

I think there is a different problem here!

This does not segfault:

import core.memory;
void main()
{
   int *x = null;
   GC.free(x);
}

[schveiguy]

This seems wrong, shouldn't it be GC.free(x) ?

[wilzbach]

Yeah, looks like I was a bit too fast. Thanks!

[schveiguy]

OK, I see you fixed that problem. But now I'm slightly confused!

Because extracting out the problem:

void main()
{
    int *ptr = null;
    GC.free(&ptr);
}

This segfaults, but this:

void main()
{
   int *ptr = new int(1);
   GC.free(&ptr);
}

DOESN'T segfault. Neither is a valid heap pointer, so why does it segfault at all? The first thing GC.free should do is to find the pool the pointer is in, and it should fail in both cases.

In any case, a bug for another day. Your changes LGTM.

[schveiguy]

Hah, I think I know why it was segfaulting. Because of the new ProxyGC! Attempting to free a pointer before actually allocating any pointer will do this: https://github.com/dlang/druntime/blob/master/src/gc/impl/proto/gc.d#L140, which equates to a segfault when compiled in release mode.

[schveiguy]

This does fix a bug, though (__delete was not actually freeing the memory), so I'm pulling.