Re: Latent bugs?

Re: Latent bugs?

[Lynn Winebarger]

Redirected to guile-devel from bug-guile because this list gets read 
more often.

On Saturday 28 December 2002 13:35, Marko Rauhamaa wrote:
> Lynn Winebarger <owinebar@free-expression.org>:
> 
> > On Saturday 28 December 2002 03:49, Marko Rauhamaa wrote:
> > > 
> > >   return scm_list_2(scm_cons(scm_long2num(old_timer.it_interval.tv_sec),
> > >                              scm_long2num(old_timer.it_interval.tv_usec)),
> > >                     scm_cons(scm_long2num(old_timer.it_value.tv_sec),
> > >                              scm_long2num(old_timer.it_value.tv_usec)));
> > > 
> > > If I understand it correctly, this statement contains several
> > > allocations, each of which can potentially trigger garbage collection
> > > and delete any of the intermediate results.
> >
> >      The code above, though it is not explicit, will store the
> > intermediate return values (references) on the stack, where the
> > garbage collector will see them (and thus not free the newly created
> > object). It's just how the traditional model of a C machine works.
> 
> I took a glance at gc_os_dep.c and I was surprised. It appears the GC
> actually sweeps through all data segments, the stack segment and CPU
> registers and heuristically locates pointers into the heap. I've used
> that technique myself before, but I wouldn't have thought that was an
> effective way of implementing a Lisp machine.
> 
> The Guile manual doesn't state that basically all writable memory of a
> process participates in GC:
> 
>   http://www.gnu.org/software/guile/docs/guile-ref/Memory-Management.html#Memory%20Management

      Guile's GC got an overhaul earlier this year.  I don't know how
well the documentation reflects that.
       Anybody?

Lynn


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://mail.gnu.org/mailman/listinfo/guile-devel

Re: Latent bugs?

[Neil Jerram]

>>>>> "Lynn" == Lynn Winebarger <owinebar@free-expression.org> writes:

    Lynn> On Saturday 28 December 2002 13:35, Marko Rauhamaa wrote:
    >> 
    >> I took a glance at gc_os_dep.c and I was surprised. It appears the GC
    >> actually sweeps through all data segments, the stack segment and CPU
    >> registers and heuristically locates pointers into the heap. I've used
    >> that technique myself before, but I wouldn't have thought that was an
    >> effective way of implementing a Lisp machine.
    >> 
    >> The Guile manual doesn't state that basically all writable memory of a
    >> process participates in GC:
    >> 
    >> http://www.gnu.org/software/guile/docs/guile-ref/Memory-Management.html#Memory%20Management

    Lynn>       Guile's GC got an overhaul earlier this year.  I don't know how
    Lynn> well the documentation reflects that.
    Lynn>        Anybody?

Barring a more major overhaul than I'm aware of, it is not the case
that all writable memory participates in GC.  Just the stack, heap
segments (of SCM cells) allocated by Guile, and any specially
registered roots.

gc_os_dep.c (which comes from the Boehm GC) is only there because it
includes code for finding the base of the stack on lots of OS's.
Other things in gc_os_dep.c don't necessarily describe how Guile's GC
works.

        Neil



_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://mail.gnu.org/mailman/listinfo/guile-devel

Re: Latent bugs?

[Marko Rauhamaa]

Marius Vollmer <mvo@zagadka.ping.de>:

> Only the stacks (of all threads) and the registers are searched in
> this manner.

I think mentioning this behavior in the Guile documentation would be,
shall we say, worthwhile.


Marko

-- 
Marko Rauhamaa      mailto:marko@pacujo.net     http://pacujo.net/marko/


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://mail.gnu.org/mailman/listinfo/guile-devel

Re: Latent bugs?

[Han-Wen Nienhuys]

neil@ossau.uklinux.net writes:
> Barring a more major overhaul than I'm aware of, it is not the case
> that all writable memory participates in GC.  Just the stack, heap
> segments (of SCM cells) allocated by Guile, and any specially
> registered roots.
> 
> gc_os_dep.c (which comes from the Boehm GC) is only there because it
> includes code for finding the base of the stack on lots of OS's.

This is exactly right.  Perhaps gc_os_dep.c should be pruned to remove
dead code.

-- 

Han-Wen Nienhuys   |   hanwen@cs.uu.nl   |   http://www.cs.uu.nl/~hanwen 


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://mail.gnu.org/mailman/listinfo/guile-devel

Re: Latent bugs?

[Marius Vollmer]

Han-Wen Nienhuys <hanwen@cs.uu.nl> writes:

> This is exactly right.  Perhaps gc_os_dep.c should be pruned to remove
> dead code.

When I first installed it, I did remove some code, but maybe not
enough, or maybe too much.  It would be nice to be able to easily pick
up improvements in the 'upstream' Boehm-GC code.

-- 
GPG: D5D4E405 - 2F9B BCCC 8527 692A 04E3  331E FAF8 226A D5D4 E405


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://mail.gnu.org/mailman/listinfo/guile-devel