I don't understand why using ->next is correct here, how does that make sense?

In D14762#396950, @Hans Goudey (HooglyBoogly) wrote:

I don't understand why using ->next is correct here, how does that make sense?

Before the commit, it worked like this:

id_to_inspection_string((ID *)*value.get<Object *>(), ID_OB);

After the commit, it works like this:

const void *buffer = value.get();
id_to_inspection_string((ID *)buffer, ID_OB);

Which is equivalent this:

id_to_inspection_string((ID *)value.get<Object *>(), ID_OB);

Notice the sneaky dereference in the original is lost.

This patch, works like this:

const void *buffer = value.get();
id_to_inspection_string((ID *)((ID *)buffer)->next, ID_OB);

Which is equivalent to the original.

Either that, or I don't know what I am talking about.

Hmm, I don't think ->next is the same as de-referencing the ID pointer. I don't think the next and previous pointers are expected to be valid for IDs outside of Main.

What about this? *static_cast<ID **>(buffer)

@Hans Goudey (HooglyBoogly)

In D14762#396967, @Hans Goudey (HooglyBoogly) wrote:

Hmm, I don't think ->next is the same as de-referencing the ID pointer. I don't think the next and previous pointers are expected to be valid for IDs outside of Main.

ID is defined like this.

/* There's a nasty circular dependency here.... 'void *' to the rescue! I
 * really wonder why this is needed. */
typedef struct ID {
  void *next, *prev;
  struct ID *newid;

  struct Library *lib;
⋮

*next is first in the struct, so the deref gets that pointer and then casts it to ID *.

What about this? *static_cast<ID **>(buffer)

Does not work.

Severity	Code	Description	Project	File	Line	Suppression State
Error	C2440	'static_cast': cannot convert from 'const void *' to 'ID **'	bf_editor_space_node	C:\Users\Ethan\Documents\blender-git\blender\source\blender\editors\space_node\node_draw.cc	790

@Hans Goudey (HooglyBoogly) This works: (ID *)*(static_cast<const Object *const *>(buffer))

Note: This patch does not fix the name update issue for collections. (When a collection is renamed, the old name is still displayed in the socket. Problem since 3.0.0)

I haven't posted a bug report on it yet because this bug covers it up, and I only found the bug while testing this patch.

Looks better now! I think it could be a little simpler though if we removed the second cast to ID * and used static_cast to make a const ID ** in the first place. That would review id_to_inspection_string to have a const ID * argument, which seems fine.