Re: [Patch] Move all stack info to gc.c

On Fri, 25 Jun 2004, Michal Rokos wrote:
> On Thursday 24 of June 2004 22:23, Mathieu Bouchard wrote:
> > On Wed, 23 Jun 2004, Yukihiro Matsumoto wrote:
> > > I don't think so.  Go ahead, or we will know if there's unseen
> > > reason.
> > If it's going to be global, it'd be better to have it start with the
> > "rb_" prefix...
> I assume you're talking about _stacksize...

Oh yes, sorry for my un-clarity...

> well it was non-static int before.... So I left it as it was, but just
> moved to gc.c. This was done intensionally, because I didn't know
> whether somebody does extern int _stacksize somewhere... I guess not.

I don't know about it... but I think there should be an API for
configuring the GC and stuff like stacksize, thread timeslice, and such.
It can be either global variables or functions, whichever is deemed
suitable. I asked for half of this a few years ago, and even wrote a
patch, but it didn't get into Ruby itself.

Now that I think of it, I could actually just bundle my own version of
Ruby with my software (as a source diff and a binary). I really didn't
want to get into this, but I guess that, if I have no other possibility,
that's what I'll do.

Another outstanding issue is that Ruby's support for embedding has a
problem with ruby_init being done on a non-empty stack. It took me months
to figure it out, and then I posted a fix a year ago... has anything been
done about it while I was not looking? At least I can fix it from inside
my own software using Init_stack and some black magic...

________________________________________________________________
Mathieu Bouchard                       http://artengine.ca/matju