Status: New Owner: ---- CC: stephane...@gmail.com, anquetil...@gmail.com, usman.bh...@gmail.com Labels: Type-Defect Priority-Critical Component-VerveineJ
New issue 728 by tu...@tudorgirba.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
For some reason, some methods are placed in the wrong classes.
I did not manage to identify a standalone case, but it looks like if we have a case like below:
public class A { ... public void m() { ... x() } }
public class B { public void x() { ... } }
public class C { ... }
we sometimes get x as belonging to class C instead of B. The interesting part is that the invocation will go from A::m() to C::x().
This is a critical issue that stops us from using VerveineJ because the information is too unreliable.
Updates: Status: WontFix
Comment #1 on issue 728 by anquetil...@gmail.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
nothing we can do here until we have a concrete case to investigate. Close it until it can be reproduced.
Comment #2 on issue 728 by alexandr...@gmail.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
In two weeks I will have 13 students using Moose. It would be good to have this issue sorted out.
I would really love to solve this problem. Like any other one for which you can provide me something to work on.
I don't need to explain to you how difficult it is to solve a bug that occur from time to time whitout any concrete example that produce the error ...
nicolas
----- Mail original -----
De: moose-technology@googlecode.com À: moose-dev@iam.unibe.ch Envoyé: Mercredi 19 Octobre 2011 14:09:31 Objet: [Moose-dev] Re: Issue 728 in moose-technology: VerveineJ places methods in the wrong classes Comment #2 on issue 728 by alexandr...@gmail.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
In two weeks I will have 13 students using Moose. It would be good to have this issue sorted out.
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Comment #3 on issue 728 by tu...@tudorgirba.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
I will certainly not close it :). I reproduced it several times on proprietary case studies. I will try to create a simpler case, but maybe in the meantime others have experienced something similar.
I did not close it because I think it is irrelevant, but because without more information it is close to useless. As I said I would be more than happy to solve it.
As for the example, I don't even ask for a simple one (although the simpler is definitely the better), I just need one example.
Anyway, you know how it is ...
nicolas
----- Mail original -----
De: moose-technology@googlecode.com À: moose-dev@iam.unibe.ch Envoyé: Mercredi 19 Octobre 2011 14:25:40 Objet: [Moose-dev] Re: Issue 728 in moose-technology: VerveineJ places methods in the wrong classes Comment #3 on issue 728 by tu...@tudorgirba.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
I will certainly not close it :). I reproduced it several times on proprietary case studies. I will try to create a simpler case, but maybe in the meantime others have experienced something similar.
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
can't you rename all things C1, C2, C3, ... m1, m2, m3, ... v1, v2, v3, ...
and remove everything that appears to be unrelated (arithmetic expressions, I/O, conditions and loops)
obfuscate it ...
I understand that this can represent some amount of work. But I need something to work on.
nicolas
----- Mail original -----
De: moose-technology@googlecode.com À: moose-dev@iam.unibe.ch Envoyé: Mercredi 19 Octobre 2011 14:25:40 Objet: [Moose-dev] Re: Issue 728 in moose-technology: VerveineJ places methods in the wrong classes Comment #3 on issue 728 by tu...@tudorgirba.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
I will certainly not close it :). I reproduced it several times on proprietary case studies. I will try to create a simpler case, but maybe in the meantime others have experienced something similar.
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
I am looking into it right now. Actually, I think the issue is related to a larger context that I do not see clearly at this moment.
Doru
On Wed, Oct 19, 2011 at 2:41 PM, Nicolas Anquetil <nicolas.anquetil@inria.fr
wrote:
can't you rename all things C1, C2, C3, ... m1, m2, m3, ... v1, v2, v3, ...
and remove everything that appears to be unrelated (arithmetic expressions, I/O, conditions and loops)
obfuscate it ...
I understand that this can represent some amount of work. But I need something to work on.
nicolas
----- Mail original -----
De: moose-technology@googlecode.com À: moose-dev@iam.unibe.ch Envoyé: Mercredi 19 Octobre 2011 14:25:40 Objet: [Moose-dev] Re: Issue 728 in moose-technology: VerveineJ places
methods in the wrong classes
Comment #3 on issue 728 by tu...@tudorgirba.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
I will certainly not close it :). I reproduced it several times on proprietary case studies. I will try to create a simpler case, but maybe in the meantime others have experienced something similar.
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
I am happy to announce that thanks to the very valuable help of Doru, the bug has been clearly identified and subsequently removed. nicolas ----- Mail original -----
De: "Tudor Girba" tudor@tudorgirba.com À: "Moose-related development" moose-dev@iam.unibe.ch Envoyé: Mercredi 19 Octobre 2011 14:53:41 Objet: [Moose-dev] Re: Issue 728 in moose-technology: VerveineJ places methods in the wrong classes I am looking into it right now. Actually, I think the issue is related to a larger context that I do not see clearly at this moment. Doru On Wed, Oct 19, 2011 at 2:41 PM, Nicolas Anquetil < nicolas.anquetil@inria.fr > wrote:
can't you rename all things C1, C2, C3, ... m1, m2, m3, ... v1, v2, v3, ... and remove everything that appears to be unrelated (arithmetic expressions, I/O, conditions and loops) obfuscate it ... I understand that this can represent some amount of work. But I need something to work on. nicolas ----- Mail original -----
De: moose-technology@googlecode.com À: moose-dev@iam.unibe.ch Envoyé: Mercredi 19 Octobre 2011 14:25:40 Objet: [Moose-dev] Re: Issue 728 in moose-technology: VerveineJ places methods in the wrong classes Comment #3 on issue 728 by tu...@tudorgirba.com : VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
I will certainly not close it :). I reproduced it several times on proprietary case studies. I will try to create a simpler case, but maybe in the meantime others have experienced something similar.
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
--
www.tudorgirba.com "Every thing has its own flow" _______________________________________________ Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
So what was the nature of the problem?
On Fri, Oct 21, 2011 at 11:39 PM, Nicolas Anquetil < nicolas.anquetil@inria.fr> wrote:
I am happy to announce that thanks to the very valuable help of Doru, the bug has been clearly identified and subsequently removed.
nicolas
*De: *"Tudor Girba" tudor@tudorgirba.com *À: *"Moose-related development" moose-dev@iam.unibe.ch *Envoyé: *Mercredi 19 Octobre 2011 14:53:41
*Objet: *[Moose-dev] Re: Issue 728 in moose-technology: VerveineJ places methods in the wrong classes
I am looking into it right now. Actually, I think the issue is related to a larger context that I do not see clearly at this moment.
Doru
On Wed, Oct 19, 2011 at 2:41 PM, Nicolas Anquetil < nicolas.anquetil@inria.fr> wrote:
can't you rename all things C1, C2, C3, ... m1, m2, m3, ... v1, v2, v3, ...
and remove everything that appears to be unrelated (arithmetic expressions, I/O, conditions and loops)
obfuscate it ...
I understand that this can represent some amount of work. But I need something to work on.
nicolas
----- Mail original -----
De: moose-technology@googlecode.com À: moose-dev@iam.unibe.ch Envoyé: Mercredi 19 Octobre 2011 14:25:40 Objet: [Moose-dev] Re: Issue 728 in moose-technology: VerveineJ places
methods in the wrong classes
Comment #3 on issue 728 by tu...@tudorgirba.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
I will certainly not close it :). I reproduced it several times on proprietary case studies. I will try to create a simpler case, but maybe in the meantime others have experienced something similar.
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
--
www.tudorgirba.com
"Every thing has its own flow"
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Yeah! U rock guys!
Le 21 oct. 2011 à 18:39, Nicolas Anquetil nicolas.anquetil@inria.fr a écrit :
I am happy to announce that thanks to the very valuable help of Doru, the bug has been clearly identified and subsequently removed.
nicolas
De: "Tudor Girba" tudor@tudorgirba.com À: "Moose-related development" moose-dev@iam.unibe.ch Envoyé: Mercredi 19 Octobre 2011 14:53:41 Objet: [Moose-dev] Re: Issue 728 in moose-technology: VerveineJ places methods in the wrong classes
I am looking into it right now. Actually, I think the issue is related to a larger context that I do not see clearly at this moment.
Doru
On Wed, Oct 19, 2011 at 2:41 PM, Nicolas Anquetil nicolas.anquetil@inria.fr wrote: can't you rename all things C1, C2, C3, ... m1, m2, m3, ... v1, v2, v3, ...
and remove everything that appears to be unrelated (arithmetic expressions, I/O, conditions and loops)
obfuscate it ...
I understand that this can represent some amount of work. But I need something to work on.
nicolas
----- Mail original -----
De: moose-technology@googlecode.com À: moose-dev@iam.unibe.ch Envoyé: Mercredi 19 Octobre 2011 14:25:40 Objet: [Moose-dev] Re: Issue 728 in moose-technology: VerveineJ places methods in the wrong classes Comment #3 on issue 728 by tu...@tudorgirba.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
I will certainly not close it :). I reproduced it several times on proprietary case studies. I will try to create a simpler case, but maybe in the meantime others have experienced something similar.
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
--
www.tudorgirba.com
"Every thing has its own flow"
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Updates: Status: Fixed Cc: -stephane...@gmail.com -anquetil...@gmail.com -usman.bh...@gmail.com
Comment #4 on issue 728 by anquetil...@gmail.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
(No comment was entered for this change.)
Updates: Status: Started Owner: anquetil...@gmail.com
Comment #5 on issue 728 by tudor.gi...@gmail.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
Great job.
However, we still have another case of misplaced methods that is still present.
It is different than the one I sent you in that it seems to only involve super calls (while the other one was related to calls to a different class). Essentially, something like:
public class A extends SomeClass { ... public void x() { super.m() } }
leads to m() being placed in an unrelated class B.
I suspect it is the same issue as the one you fixed only present in a different place. I am not sure if my description is clear enough.
I will try again to reproduce on a smaller case study.
Updates: Status: Fixed
Comment #6 on issue 728 by anquetil...@gmail.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
I believe this second case is solved too.
Updates: Status: Started
Comment #7 on issue 728 by tudor.gi...@gmail.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
Hmm, I just tried. First, I fixed the shell script (verveinej.sh) to collect the jars automatically from the lib folder. Then I removed the obsolete jars.
But, still, if I run the following experiment, I get 575 potentially problematic methods:
- download ArgoUML 0.32 from: http://argouml-downloads.tigris.org/argouml-0.32/ - parse, load in Moose, and then execute: (MooseModel root allModels first allMethods select: [:each | each sourceAnchor isNil and: [ each parentType sourceAnchor notNil and: [ '<Initializer>' ~= each name and: [ each name ~= each parentType name ]]]]) openInMoose
Am I doing something wrong?
what I did:
- download ArgoUML-0.32-src.tar.gz (sources of ArgoUML) from the web site you gave - download ArgoUML-0.32-libs.tar.gz (libraries needed by ArgoUML) from the web site you gave - update verveinej to get your new shell script - rm -f output.mse - create MSE file:"verveinej.sh -autocp ArgoUML-0.32-libs ArgoUML-0.32-src" - load MSE in Moose panel - ran your script in workspace
When calling verveineJ, "-autocp ArgoUML-0.32-libs" is important. Without it, JDT is not able to resolve a lot of types names, and as a consequence, a lot of source methods that use these types (e.g. as parameter type). And without JDT's name resolution, verveineJ is not able to link a method call to a method declaration.
With the autocp, I found 13 methods with problems. Without it, I found 473 (or something like that) methods with problems. Not your 575 ones, but closer to it.
No idea how we can get so different results with the same data and program !!!
nicolas
----- Mail original -----
De: moose-technology@googlecode.com À: moose-dev@iam.unibe.ch Envoyé: Mardi 1 Novembre 2011 07:40:20 Objet: [Moose-dev] Re: Issue 728 in moose-technology: VerveineJ places methods in the wrong classes Updates: Status: Started
Comment #7 on issue 728 by tudor.gi...@gmail.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
Hmm, I just tried. First, I fixed the shell script (verveinej.sh) to collect the jars automatically from the lib folder. Then I removed the obsolete jars.
But, still, if I run the following experiment, I get 575 potentially problematic methods:
- download ArgoUML 0.32 from:
http://argouml-downloads.tigris.org/argouml-0.32/
- parse, load in Moose, and then execute:
(MooseModel root allModels first allMethods select: [:each | each sourceAnchor isNil and: [ each parentType sourceAnchor notNil and: [ '<Initializer>' ~= each name and: [ each name ~= each parentType name ]]]]) openInMoose
Am I doing something wrong?
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Thanks. Indeed, it is strange. Btw, I am on a Mac OS X Lion.
I will try again and get back to you later.
But, coming back to the issue, the goal of the script is to provide a set of possible methods that are misplaced (that is methods that should appear in another class than where they are actually defined). The issue is that if we have the source for the class, it means we should have parsed it and have the proper information about its methods. So, how could then methods appear inside without source code?
Cheers, Doru
On 1 Nov 2011, at 10:35, Nicolas Anquetil wrote:
what I did:
- download ArgoUML-0.32-src.tar.gz (sources of ArgoUML) from the web site you gave
- download ArgoUML-0.32-libs.tar.gz (libraries needed by ArgoUML) from the web site you gave
- update verveinej to get your new shell script
- rm -f output.mse
- create MSE file:"verveinej.sh -autocp ArgoUML-0.32-libs ArgoUML-0.32-src"
- load MSE in Moose panel
- ran your script in workspace
When calling verveineJ, "-autocp ArgoUML-0.32-libs" is important. Without it, JDT is not able to resolve a lot of types names, and as a consequence, a lot of source methods that use these types (e.g. as parameter type). And without JDT's name resolution, verveineJ is not able to link a method call to a method declaration.
With the autocp, I found 13 methods with problems. Without it, I found 473 (or something like that) methods with problems. Not your 575 ones, but closer to it.
No idea how we can get so different results with the same data and program !!!
nicolas
----- Mail original -----
De: moose-technology@googlecode.com À: moose-dev@iam.unibe.ch Envoyé: Mardi 1 Novembre 2011 07:40:20 Objet: [Moose-dev] Re: Issue 728 in moose-technology: VerveineJ places methods in the wrong classes Updates: Status: Started
Comment #7 on issue 728 by tudor.gi...@gmail.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
Hmm, I just tried. First, I fixed the shell script (verveinej.sh) to collect the jars automatically from the lib folder. Then I removed the obsolete jars.
But, still, if I run the following experiment, I get 575 potentially problematic methods:
- download ArgoUML 0.32 from:
http://argouml-downloads.tigris.org/argouml-0.32/
- parse, load in Moose, and then execute:
(MooseModel root allModels first allMethods select: [:each | each sourceAnchor isNil and: [ each parentType sourceAnchor notNil and: [ '<Initializer>' ~= each name and: [ each name ~= each parentType name ]]]]) openInMoose
Am I doing something wrong?
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
-- www.tudorgirba.com
"We cannot reach the flow of things unless we let go."
Yes it does seem strange, and it took me a while to accept it.
The point is that, in Java, you can have different methods with the same name and different parameters. So, without a clear handle on the type of all parameters, you cannot decide for sure that a method call is the same as a method declaration. Another point is that JDT is made to work in an IDE, with incomplete code.
That's why, even with the source declaration of the method, JDT does not link it to its calls unless it could resolve all the parameters' types.
That's why it is also important to have all the needed libraries when parsing with verveineJ
nicolas
----- Mail original -----
De: "Tudor Girba" tudor@tudorgirba.com À: "Moose-related development" moose-dev@iam.unibe.ch Cc: codesite-noreply@google.com Envoyé: Mardi 1 Novembre 2011 11:07:26 Objet: [Moose-dev] Re: Issue 728 in moose-technology: VerveineJ places methods in the wrong classes Thanks. Indeed, it is strange. Btw, I am on a Mac OS X Lion.
I will try again and get back to you later.
But, coming back to the issue, the goal of the script is to provide a set of possible methods that are misplaced (that is methods that should appear in another class than where they are actually defined). The issue is that if we have the source for the class, it means we should have parsed it and have the proper information about its methods. So, how could then methods appear inside without source code?
Cheers, Doru
On 1 Nov 2011, at 10:35, Nicolas Anquetil wrote:
what I did:
- download ArgoUML-0.32-src.tar.gz (sources of ArgoUML) from the web
site you gave
- download ArgoUML-0.32-libs.tar.gz (libraries needed by ArgoUML)
from the web site you gave
- update verveinej to get your new shell script
- rm -f output.mse
- create MSE file:"verveinej.sh -autocp ArgoUML-0.32-libs
ArgoUML-0.32-src"
- load MSE in Moose panel
- ran your script in workspace
When calling verveineJ, "-autocp ArgoUML-0.32-libs" is important. Without it, JDT is not able to resolve a lot of types names, and as a consequence, a lot of source methods that use these types (e.g. as parameter type). And without JDT's name resolution, verveineJ is not able to link a method call to a method declaration.
With the autocp, I found 13 methods with problems. Without it, I found 473 (or something like that) methods with problems. Not your 575 ones, but closer to it.
No idea how we can get so different results with the same data and program !!!
nicolas
----- Mail original -----
De: moose-technology@googlecode.com À: moose-dev@iam.unibe.ch Envoyé: Mardi 1 Novembre 2011 07:40:20 Objet: [Moose-dev] Re: Issue 728 in moose-technology: VerveineJ places methods in the wrong classes Updates: Status: Started
Comment #7 on issue 728 by tudor.gi...@gmail.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
Hmm, I just tried. First, I fixed the shell script (verveinej.sh) to collect the jars automatically from the lib folder. Then I removed the obsolete jars.
But, still, if I run the following experiment, I get 575 potentially problematic methods:
- download ArgoUML 0.32 from:
http://argouml-downloads.tigris.org/argouml-0.32/
- parse, load in Moose, and then execute:
(MooseModel root allModels first allMethods select: [:each | each sourceAnchor isNil and: [ each parentType sourceAnchor notNil and: [ '<Initializer>' ~= each name and: [ each name ~= each parentType name ]]]]) openInMoose
Am I doing something wrong?
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
-- www.tudorgirba.com
"We cannot reach the flow of things unless we let go."
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Updates: Status: Fixed Labels: Milestone-5.0
Comment #8 on issue 728 by tu...@tudorgirba.com: VerveineJ places methods in the wrong classes http://code.google.com/p/moose-technology/issues/detail?id=728
(No comment was entered for this change.)
-
Alexandre Bergel -
moose-technology＠googlecode.com -
Nicolas Anquetil -
Tudor Girba -
Usman Bhatti