Issue 5407 in pharo: about garbage collection

Status: New
Owner: ----
Labels: Milestone-1.3 Type-Bug

New issue 5407 by [hidden email]: about garbage collection
http://code.google.com/p/pharo/issues/detail?id=5407

Hi,

I thought I would just mention this in case other users were struggling  
with similar problems. I'm using 1.3 on Windows, with the latest Cog VM  
(13307). However, I've had a similar problem on Pharo since version 1.0.

The problem is that the GC doesn't deallocate memory effectively at times.  
This is particularly a problem with the Windows version, as this is limited  
to a max of 512mB of RAM. In particular, the problem seems to crop up with  
WideString. It has been very difficult to resolve, as my code does not in  
any way "hang onto" these objects. SpaceTally effectively reports that I  
have several thousand of these objects hanging around, but even working out  
a strategy to manually free them would be difficult. At times I've been  
forced to virtually start over with an old image with change updates, as i  
needed to resolve the situation quickly.

Recently this happened again. I ended up with over 3000 WideString objects.  
I ran garbage collection, including a version that cleans up after WA and  
so forth. No joy. I was about to rollback and update, but I played around  
with the objects a little, calling allInstances and  a couple of other  
things on them. I checked the memory again, and suddenly the WideString  
objects were deallocated, reduced from 3000 to less than 400, which is  
about right.

Anyway, I just thought I would post this note for others to ponder, and for  
people in a similar situation looking for a quick fix path to explore.

I'm sure you can tell from the above that I really  like Pharo (though I  
miss some things from Dolphin) but these memory problems have been a bit of  
an anxiety for me.

regards,
Scott Lewis


_______________________________________________
Pharo-bugtracker mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-bugtracker

Comment #1 on issue 5407 by marianopeck: about garbage collection
http://code.google.com/p/pharo/issues/detail?id=5407

Hi Scott. I have never seen this problem before. I have large images of  
60MB and WideString instanceCount -> 132. So...I maybe you have something  
in particular that holds into WideString instances?  Can you give us more  
information?  Do you have an image with all those WideStrings that we can  
take a look?  it would be interesting to search the pointers to those  
WideStrings to see who is holding into them. Did you check that?

Thanks for the report


_______________________________________________
Pharo-bugtracker mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-bugtracker

Comment #2 on issue 5407 by [hidden email]: about garbage collection
http://code.google.com/p/pharo/issues/detail?id=5407

Hello,

What do we do with this report? The problem is that I do not see how we can  
take action, as I can't reproduce
the problem.
*Reporting* the problem was the good thing! Because now at least two active  
people will remember this report as soon as the topic GC and WideStrings  
will come up again.
In addition, the tracker database is searchable, so people will find it  
when later they encounter the same problem and can then comment to add more  
to it.

But this report is not *actionable*, that is, I can not do anything to  
close it. This means, it would just stay here
for the next X years.
Is that what we want? Or is it enough to have it in the database, but not  
taking active space in the TODO list?

-> let it open?
-> close it?


_______________________________________________
Pharo-bugtracker mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-bugtracker

Comment #3 on issue 5407 by [hidden email]: about garbage collection
http://code.google.com/p/pharo/issues/detail?id=5407

I would close it for now, it is currently not reproduceable.


_______________________________________________
Pharo-bugtracker mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-bugtracker

Updates:
        Status: Closed

Comment #4 on issue 5407 by [hidden email]: about garbage collection
http://code.google.com/p/pharo/issues/detail?id=5407

(No comment was entered for this change.)


_______________________________________________
Pharo-bugtracker mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-bugtracker