Doubt about GC

[Eustaquio Rangel de Oliveira Jr.]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello there!

I have this program:

- -------------------------------------------------------------------------
class Car
attr_reader :number
def initialize(n)
@number=n
puts "Creating car no. #{n}:#{object_id}"
end
def to_s
"Car no. #{@number}"
end
end

def makeCar(n)
c = Car.new(n)
end

def listCars
puts "\nListing cars:"
ObjectSpace.each_object(Car) {|o| puts "#{o}:#{o.object_id}"}
puts
end

c1 = makeCar(1)
makeCar(2)
makeCar(3)

listCars()
makeCar(4)
makeCar(5)

listCars();

puts "Firing gc."
GC.start
listCars();
- -------------------------------------------------------------------------

When I run, I get:

Creating car no. 1:-604870360
Creating car no. 2:-604870370
Creating car no. 3:-604870410

Listing cars:
Car no. 3:-604870410
Car no. 2:-604870370
Car no. 1:-604870360

Creating car no. 4:-604870500
Creating car no. 5:-604870490

Listing cars:
Car no. 4:-604870500
Car no. 5:-604870490
Car no. 3:-604870410
Car no. 2:-604870370
Car no. 1:-604870360

Firing gc.

Listing cars:
Car no. 4:-604870500
Car no. 1:-604870360

My doubt is why Car no. 4 is still there, after running gc. There is no
reference to it outside the function, right? So it should not have be
sweeped by GC?

Thanks!

- ----------------------------
Eustáquio "TaQ" Rangel
(e-mail address removed)
http://beam.to/taq
Usuário GNU/Linux no. 224050
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.7 (GNU/Linux)

iD8DBQFCiO4mb6UiZnhJiLsRAmJ7AJ9M8a1CsY/ksAW1fThdmIpeLbB7gwCcCz5w
+sh/g0F1Vork+DKH/xRj+9M=
=Tm+7
-----END PGP SIGNATURE-----

 

[Jason Foreman]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
=20
Hello there!
=20
I have this program:
=20
- -----------------------------------------------------------------------= --
class Car
attr_reader :number
def initialize(n)
@number=3Dn
puts "Creating car no. #{n}:#{object_id}"
end
def to_s
"Car no. #{@number}"
end
end
=20
def makeCar(n)
c =3D Car.new(n)
end
=20
def listCars
puts "\nListing cars:"
ObjectSpace.each_object(Car) {|o| puts "#{o}:#{o.object_id}"}
puts
end
=20
c1 =3D makeCar(1)
makeCar(2)
makeCar(3)
=20
listCars()
makeCar(4)
makeCar(5)
=20
listCars();
=20
puts "Firing gc."
GC.start
listCars();
- -----------------------------------------------------------------------= --
=20
When I run, I get:
=20
Creating car no. 1:-604870360
Creating car no. 2:-604870370
Creating car no. 3:-604870410
=20
Listing cars:
Car no. 3:-604870410
Car no. 2:-604870370
Car no. 1:-604870360
=20
Creating car no. 4:-604870500
Creating car no. 5:-604870490
=20
Listing cars:
Car no. 4:-604870500
Car no. 5:-604870490
Car no. 3:-604870410
Car no. 2:-604870370
Car no. 1:-604870360
=20
Firing gc.
=20
Listing cars:
Car no. 4:-604870500
Car no. 1:-604870360
=20
My doubt is why Car no. 4 is still there, after running gc. There is no
reference to it outside the function, right? So it should not have be
sweeped by GC?
=20
Thanks!
=20
- ----------------------------
Eust=E1quio "TaQ" Rangel
(e-mail address removed)
http://beam.to/taq
Usu=E1rio GNU/Linux no. 224050
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.7 (GNU/Linux)
=20
iD8DBQFCiO4mb6UiZnhJiLsRAmJ7AJ9M8a1CsY/ksAW1fThdmIpeLbB7gwCcCz5w
+sh/g0F1Vork+DKH/xRj+9M=3D
=3DTm+7
-----END PGP SIGNATURE-----
=20
=20

You script returns the expected result for me (only car 1). WinXP, Ruby 1.=
8.2.

I'm guessing garbage collection is happening asynchronously, and that
your listCars executed before car 4 was collected? Try sleep(2) in
between GC.start and listCars to see if that changes the result.

This is only a guess, and could be very wrong. I have not looked at
the Ruby source to verify how garbage collection works.


Jason

 

[Kent Sibilev]

When I run it, I get:

Creating car no. 1:20722128
Creating car no. 2:20722116
Creating car no. 3:20722068

Listing cars:
Car no. 3:20722068
Car no. 2:20722116
Car no. 1:20722128

Creating car no. 4:20721960
Creating car no. 5:20721972

Listing cars:
Car no. 4:20721960
Car no. 5:20721972
Car no. 3:20722068
Car no. 2:20722116
Car no. 1:20722128

Firing gc.

Listing cars:
Car no. 1:20722128

Kent.

 

[Eustaquio Rangel de Oliveira Jr.]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi.

| You script returns the expected result for me (only car 1). WinXP, Ruby
1.8.2.

Weird! Here I'm running Slackware 10.1-current, Ruby 1.8.2.

| I'm guessing garbage collection is happening asynchronously, and that
| your listCars executed before car 4 was collected? Try sleep(2) in
| between GC.start and listCars to see if that changes the result.

No change, even if I sleep more than 2 seconds, but if I make

GC.start
sleep(1)
GC.start

it works, only rest Car 1.

| This is only a guess, and could be very wrong. I have not looked at
| the Ruby source to verify how garbage collection works.

But it's a beggining and prove me that I'm not at least fully crazy ehehe.
There is something there.

Thanks!

- ----------------------------
Eustáquio "TaQ" Rangel
(e-mail address removed)
http://beam.to/taq
Usuário GNU/Linux no. 224050
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.7 (GNU/Linux)

iD8DBQFCiPM4b6UiZnhJiLsRAiQ4AJwPMxk7sMOzjdpZYb7GplFiPLTUiACfRkP6
I0xFnyqaZjyYoEUBm0NRxUw=
=jHMR
-----END PGP SIGNATURE-----

 

[Lothar Scholz]

Hello Eustaquio,


ERdOJ> My doubt is why Car no. 4 is still there, after running gc. There is no
ERdOJ> reference to it outside the function, right? So it should not have be
ERdOJ> sweeped by GC?

Unlikely but the stack marking is conservative. Maybe there is
something on the stack that looks like a pointer to Car no.4. In this
event Car no.4 will not be deleted.

 

[Eustaquio Rangel de Oliveira Jr.]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Lothar!

| Unlikely but the stack marking is conservative. Maybe there is
| something on the stack that looks like a pointer to Car no.4. In this
| event Car no.4 will not be deleted.

Kind of some "memory dirty" or something like that?

Thanks!

- ----------------------------
Eustáquio "TaQ" Rangel
(e-mail address removed)
http://beam.to/taq
Usuário GNU/Linux no. 224050
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.7 (GNU/Linux)

iD8DBQFCiQAbb6UiZnhJiLsRAkSTAJ9m21h3Dq4EKBhmxA9tQXUxDIqx/gCgtmRL
vbhMVGGS5wH6Kpuab78unx4=
=KwJz
-----END PGP SIGNATURE-----

 

[Caio Tiago Oliveira]

Eustaquio Rangel de Oliveira Jr., 16-05-2005 17:19:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Lothar!

| Unlikely but the stack marking is conservative. Maybe there is
| something on the stack that looks like a pointer to Car no.4. In this
| event Car no.4 will not be deleted.

Kind of some "memory dirty" or something like that?

I'm in such a different hardware and I got a result somehow different in
ruby1.8 and ruby1.9.

I've HT enabled and running the code in both versions left cars 1, 4 and
5 after GC.
If I sleep a lot of time (I've tried uo to 10 sec), in ruby1.8 it still
left the three cars and in ruby1.9 only cars 1 and 5.


$ ruby --version
ruby 1.8.2 (2005-04-11) [i386-linux]

$ruby Car.rb

Creating car no. 1:-604682514
Creating car no. 2:-604682524
Creating car no. 3:-604682564

Listing cars:
Car no. 3:-604682564
Car no. 2:-604682524
Car no. 1:-604682514

Creating car no. 4:-604682654
Creating car no. 5:-604682644

Listing cars:
Car no. 4:-604682654
Car no. 5:-604682644
Car no. 3:-604682564
Car no. 2:-604682524
Car no. 1:-604682514

Sleeping...
Firing gc.

Listing cars:
Car no. 5:-604682644
Car no. 1:-604682514

todos@tiago:~/caio/prog/ruby$ ruby1.8 Car.rb
Creating car no. 1:-604583790
Creating car no. 2:-604583800
Creating car no. 3:-604583840

Listing cars:
Car no. 3:-604583840
Car no. 2:-604583800
Car no. 1:-604583790

Creating car no. 4:-604583930
Creating car no. 5:-604583920

Listing cars:
Car no. 4:-604583930
Car no. 5:-604583920
Car no. 3:-604583840
Car no. 2:-604583800
Car no. 1:-604583790

Sleeping...
Firing gc.

Listing cars:
Car no. 4:-604583930
Car no. 5:-604583920
Car no. 1:-604583790



and
$ ruby1.9 --version
ruby 1.9.0 (2005-04-12) [i386-linux]

$ruby1.9 Car.rb
Creating car no. 1:-604682514
Creating car no. 2:-604682524
Creating car no. 3:-604682564

Listing cars:
Car no. 3:-604682564
Car no. 2:-604682524
Car no. 1:-604682514

Creating car no. 4:-604682654
Creating car no. 5:-604682644

Listing cars:
Car no. 4:-604682654
Car no. 5:-604682644
Car no. 3:-604682564
Car no. 2:-604682524
Car no. 1:-604682514

Sleeping...
Firing gc.

Listing cars:
Car no. 5:-604682644
Car no. 1:-604682514