[Pharo-project] 10164 Image freezing
Hi all I loaded OB shout ecompletion and we were apply coding with damien. And the image froze (deeply) when we redefined a class. So the thread changes may be a problem. Stef
Apparently this is really when we define a class On Nov 21, 2008, at 3:30 PM, Stéphane Ducasse wrote:
Hi all
I loaded OB shout ecompletion
and we were apply coding with damien. And the image froze (deeply) when we redefined a class.
So the thread changes may be a problem.
Stef
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
On Nov 21, 2008, at 3:30 PM, Stéphane Ducasse wrote:
Hi all
I loaded OB shout ecompletion
and we were apply coding with damien. And the image froze (deeply) when we redefined a class.
So the thread changes may be a problem.
Stef
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
My intuition is that the debugger is asking a method coming from a trait its methodClass and then loops. See the attached stack We can nearly reproduce the bug when we define a class using traits and excluding a method. Stef On Nov 21, 2008, at 3:48 PM, Stéphane Ducasse wrote:
Apparently this is really when we define a class
On Nov 21, 2008, at 3:30 PM, Stéphane Ducasse wrote:
Hi all
I loaded OB shout ecompletion
and we were apply coding with damien. And the image froze (deeply) when we redefined a class.
So the thread changes may be a problem.
Stef
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
On 21.11.2008, at 15:30, Stéphane Ducasse wrote:
Hi all
I loaded OB shout ecompletion
and we were apply coding with damien. And the image froze (deeply) when we redefined a class.
So the thread changes may be a problem.
The only change done was to add one variable to Process (at the end)... I don't think this is the problem. What else could it be? The only other large change was a) ifNotNil with argument, but that was already earlier b) OldSocket removal. anything else? Marcus -- Marcus Denker -- denker@iam.unibe.ch http://www.iam.unibe.ch/~denker
If you can reproduce the problem, can you go back to an earlier image version to see if it exists there too? Could you provide an explanation of how to reproduce the problem? Adrian On Nov 21, 2008, at 16:04 , Marcus Denker wrote:
On 21.11.2008, at 15:30, Stéphane Ducasse wrote:
Hi all
I loaded OB shout ecompletion
and we were apply coding with damien. And the image froze (deeply) when we redefined a class.
So the thread changes may be a problem.
The only change done was to add one variable to Process (at the end)... I don't think this is the problem. What else could it be? The only other large change was
a) ifNotNil with argument, but that was already earlier b) OldSocket removal.
anything else?
Marcus
-- Marcus Denker -- denker@iam.unibe.ch http://www.iam.unibe.ch/~denker
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
I cannot predictably reproduce the bug. It happens really when editing class having traits with OB. We tried with TCloneTest -{#foo} and it did not break (we though that if we put a wrong selector this could be the problem but this was not). I breaks when you put a real selector corresponding to a compiled method My hypothesis was that may be the compiled get a methodClass but this is strange that it does not understand this message I have the impression that we get a nil somehwere ;-/ Stef On Nov 21, 2008, at 4:16 PM, Adrian Lienhard wrote:
If you can reproduce the problem, can you go back to an earlier image version to see if it exists there too? Could you provide an explanation of how to reproduce the problem?
Adrian
On Nov 21, 2008, at 16:04 , Marcus Denker wrote:
On 21.11.2008, at 15:30, Stéphane Ducasse wrote:
Hi all
I loaded OB shout ecompletion
and we were apply coding with damien. And the image froze (deeply) when we redefined a class.
So the thread changes may be a problem.
The only change done was to add one variable to Process (at the end)... I don't think this is the problem. What else could it be? The only other large change was
a) ifNotNil with argument, but that was already earlier b) OldSocket removal.
anything else?
Marcus
-- Marcus Denker -- denker@iam.unibe.ch http://www.iam.unibe.ch/~denker
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
I cannot predictably reproduce the bug. It happens really when editing class having traits with OB.
We tried with TCloneTest -{#foo}
and it did not break (we though that if we put a wrong selector this could be the problem but this was not). I breaks when you put a real selector corresponding to a compiled method
What I can tell from your debugger picture is that it searches for a selector in SymbolTest that does not exist. Probably what happens is that you remove a trait method in the class definition, then the browser updates, but still refers to that removed trait method which is actually not present anymore. So the browser model is not in sync with the system model in this case. But the strange thing is that the browser always asks the class organizer for the methods, so maybe it's the organizer that does not reflect the change to the class (yet). So I don't know whether the browser is to blame or not. What I did now is to check first whether the class indeed defines a particular method before accessing it. This should avoid the problem, but maybe the real problem lies somewhere else. David
My hypothesis was that may be the compiled get a methodClass but this is strange that it does not understand this message I have the impression that we get a nil somehwere ;-/ Stef
On Nov 21, 2008, at 4:16 PM, Adrian Lienhard wrote:
If you can reproduce the problem, can you go back to an earlier image version to see if it exists there too? Could you provide an explanation of how to reproduce the problem?
Adrian
On Nov 21, 2008, at 16:04 , Marcus Denker wrote:
On 21.11.2008, at 15:30, Stéphane Ducasse wrote:
Hi all
I loaded OB shout ecompletion
and we were apply coding with damien. And the image froze (deeply) when we redefined a class.
So the thread changes may be a problem.
The only change done was to add one variable to Process (at the end)... I don't think this is the problem. What else could it be? The only other large change was
a) ifNotNil with argument, but that was already earlier b) OldSocket removal.
anything else?
Marcus
-- Marcus Denker -- denker@iam.unibe.ch http://www.iam.unibe.ch/~denker
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Ok thanks I will continue to improve the tests with traits so we will see if the bug show up again. In that case I will save the image if I can with the debugger. stef On Nov 24, 2008, at 3:48 PM, David Röthlisberger wrote:
I cannot predictably reproduce the bug. It happens really when editing class having traits with OB. We tried with TCloneTest -{#foo} and it did not break (we though that if we put a wrong selector this could be the problem but this was not). I breaks when you put a real selector corresponding to a compiled method
What I can tell from your debugger picture is that it searches for a selector in SymbolTest that does not exist. Probably what happens is that you remove a trait method in the class definition, then the browser updates, but still refers to that removed trait method which is actually not present anymore. So the browser model is not in sync with the system model in this case. But the strange thing is that the browser always asks the class organizer for the methods, so maybe it's the organizer that does not reflect the change to the class (yet). So I don't know whether the browser is to blame or not.
What I did now is to check first whether the class indeed defines a particular method before accessing it. This should avoid the problem, but maybe the real problem lies somewhere else.
David
My hypothesis was that may be the compiled get a methodClass but this is strange that it does not understand this message I have the impression that we get a nil somehwere ;-/ Stef On Nov 21, 2008, at 4:16 PM, Adrian Lienhard wrote:
If you can reproduce the problem, can you go back to an earlier image version to see if it exists there too? Could you provide an explanation of how to reproduce the problem?
Adrian
On Nov 21, 2008, at 16:04 , Marcus Denker wrote:
On 21.11.2008, at 15:30, Stéphane Ducasse wrote:
Hi all
I loaded OB shout ecompletion
and we were apply coding with damien. And the image froze (deeply) when we redefined a class.
So the thread changes may be a problem.
The only change done was to add one variable to Process (at the end)... I don't think this is the problem. What else could it be? The only other large change was
a) ifNotNil with argument, but that was already earlier b) OldSocket removal.
anything else?
Marcus
-- Marcus Denker -- denker@iam.unibe.ch http://www.iam.unibe.ch/~denker
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
_______________________________________________ Pharo-project mailing list Pharo-project@lists.gforge.inria.fr http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
participants (4)
-
Adrian Lienhard -
David Röthlisberger -
Marcus Denker -
Stéphane Ducasse