From: Eric Wong <normalperson@...>
Date: 2014-02-24T01:19:50+00:00
Subject: [ruby-core:61036] Re: is RB_GC_GUARD needed in rb_io_syswrite?

Eric Wong <normalperson@yhbt.net> wrote:
> Nobuyoshi Nakada <nobu@ruby-lang.org> wrote:
> > And we, ko1 an I, have talked whether we can guarantee that arguments
> > of cfunc should be marked, today.
> 
> Did you come up with any conclusion?

Ping?

> I've also been thinking about exporting dispose_string() (from parse.y)
> to wider use, maybe even public C API.  Using that would give both GC
> safety and reduce memory use.

I cannot think of anything compatible other than using manual
management/guard right now :<

I suppose using RB_GC_GUARD in more places makes switching to a precise
GC easier.  In the face of stronger compiler/linker optimizations, I
wonder if precise and conservative GCs end up being equal in
difficulty-of-use.