From: "H.Yamamoto" <ocean@...2.ccsnet.ne.jp>
Date: 2006-01-11T20:01:42+09:00
Subject: Re: core dump with ruby 1.9.0 (2006-01-10) and bdb-0.5.8

Hi.

ts <decoux@moulon.inra.fr> wrote:
(2006/01/11 18:53)

>>>>>> "T" == Tanaka Akira <akr@m17n.org> writes:
>
>T> GC.stress = true
>
> This is this that it don't like. I'll see what I can do.
>
>Guy Decoux

This method was introduced into HEAD (ruby1.9) yesterday or today
for the sake of GC debuging. If this is turned on, every call of
xmalloc (or xcalloc or xrealloc) causes GC.

void *ruby_xmalloc(size_t size)
{
    if (GC.stress) {
        rb_gc(); /* every time */
        return malloc(size);
    }
    else {
        void *p = malloc(size);
        if (p) return p;
        rb_gc(); /* only when memory is exhausted */
        return malloc(size);
    }
}

Frequent GC smokes out GC (memory) bug.
(I used this when debuging syck memory bug)