Re: LuaJIT on win32

["Adam D. Moss" <adam@...>]

Mike Pall wrote:
> Doing a complete GC is
> probably considered 'unfriendly' (and the OOM may hit you
> elsewhere).

The reason I suggested such a harsh solution is that it has
to be more friendly than panicking. :)

> I'm tempted to add lazy allocation/deallocation for the C stacks.
> The C stack can be deallocated once a thread returns with
> L->status != LUA_YIELD (either an error or 0 for completion).
> There's no need to wait until the GC deletes the thread object
> itself.

That sounds like a nifty approach from where I'm sitting.
It would rely on iterator callers bleeding the iterator dry
rather than abandoning it mid-lifetime to waste away, but
that's *probably* the common case.

--adam