Max,

I would put this text to a comment for clarity.

Feel free to ignore.

On 11/20/23 17:53, Igor Munkin wrote:

<snipped>
> The reason of two consequent <collectgarbage> calls is not for "mark and
> sweep" but rather for "full GC cycle + finalization of the resurrected
> objects". The first call implements full (even 1.5) GC cycle; as a
> result of this call some *objects* can be *released*, but the
> corresponding *memory* is preserved in another list (gc->mmudata, IIRC)
> to properly finalize the object (i.e. call __gc metamethod). Hence, the
> second call finally releases the memory for the objects resurrected
> within the sweep phase of the first call.
>
> Since Max allocates plain FFI objects with no <ffi.gc> finalizers,
> there is no need in the second <collectgarbage> call here.
>
<snipped>