August 2010 - Pharo-dev - lists.pharo.org

Re: [Pharo-project] Why has MethodFinder a fixed list of results?

by StÃ©phane Ducasse

Ok if I understand you mean you interpret everything on the copy of the inputs and you run until you either get an ok primitives or you stop on not ok one. Could be a nice way to stress runSimulated: :) Stef >> > Ah, and a negative list was I think not used as it is too dangerous wrt. to completenes. If the positive list is incomplete, you miss > a hit. If the negative list is incomplete, you woud crash in some nasty way. > > Just look at the lists and how many methods you find that have been removed years ago... > > It's not just methods like #quitPrimitive. In general, with the image based nature you could end up modifying state of objects > in the image by accident, and realize it weeks later. > e.g. this means one would very carefully protect the reflective model, to not accidentally modify the methodDictionary, for example. > > So if someone did try my simulation approach above the only lists needed would be a positive list of those primitives that were safe to apply to any object and a positive list of those primitives it would be safe to apply to the objects allocated during the simulation. > > best, > Eliot > > > Marcus > > > -- > Marcus Denker -- http://www.marcusdenker.de > INRIA Lille -- Nord Europe. Team RMoD. > > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

Aug. 27, 2010

Re: [Pharo-project] Why has MethodFinder a fixed list of results?

by Bart Gauquie

Eliot, I think the simulation path is good suggestion. Did a quick test for instance on: ContextPart class>>trace: aBlock on: aStream and modified ^ thisContext sender runSimulated: aBlock contextAtEachStep: [:current | (current receiver == Smalltalk) "added" ifTrue: [self error: 'you cant do this ...']. "added" Sensor anyButtonPressed ifTrue: [^ nil]. and then evaluate: ContextPart trace: ['test' perform: #firstLetter]. => works fine, just prints the trace, ContextPart trace: [Smalltalk printString]. => throws the error. and with a nested piece of code that is also to be illegal : String>>illegalFirstLetter Smalltalk printString. ^self copyFrom: 1 to: 1 ContextPart trace: ['test' perform: #illegalFirstLetter]. => also throws the error :-). Meaning if we keep a negative list of combinations of receiver / or messages, (instead of just the (current receiver == Smalltalk) "added" ifTrue: [self error: 'you cant do this ...']. "added") it is possible to throw errors on nested code that is not be executed ... . Something to try out ... Kind Regards, Bart 2010/8/27 Eliot Miranda <eliot.miranda(a)gmail.com> > > > 2010/8/27 Marcus Denker <marcus.denker(a)inria.fr> > >> >> Smalltalk.true. >> >> takes a while, but it does not try to call #quitPrimitive. Which it would >> if Methodfinder would not be based on a positive list. >> >> >> Ah, and a negative list was I think not used as it is too dangerous wrt. >> to completenes. If the positive list is incomplete, you miss >> a hit. If the negative list is incomplete, you woud crash in some nasty >> way. >> >> Just look at the lists and how many methods you find that have been >> removed years ago... >> >> It's not just methods like #quitPrimitive. In general, with the image >> based nature you could end up modifying state of objects >> in the image by accident, and realize it weeks later. >> e.g. this means one would very carefully protect the reflective model, to >> not accidentally modify the methodDictionary, for example. >> > > So if someone did try my simulation approach above the only lists needed > would be a positive list of those primitives that were safe to apply to any > object and a positive list of those primitives it would be safe to apply to > the objects allocated during the simulation. > > best, > Eliot > > >> Marcus >> >> >> -- >> Marcus Denker -- http://www.marcusdenker.de >> INRIA Lille -- Nord Europe. Team RMoD. >> >> >> _______________________________________________ >> Pharo-project mailing list >> Pharo-project(a)lists.gforge.inria.fr >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> > > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > -- imagination is more important than knowledge - Albert Einstein Logic will get you from A to B. Imagination will take you everywhere - Albert Einstein Learn from yesterday, live for today, hope for tomorrow. The important thing is not to stop questioning. - Albert Einstein The true sign of intelligence is not knowledge but imagination. - Albert Einstein However beautiful the strategy, you should occasionally look at the results. - Sir Winston Churchill It's not enough that we do our best; sometimes we have to do what's required. - Sir Winston Churchill

Aug. 27, 2010

Re: [Pharo-project] could we fix MorphicModel

by Gary Chambers

Indeed, one of the more beautiful bits of Etoys... Regards, Gary Sent from my iPad On 27 Aug 2010, at 10:29, stephane ducasse <stephane.ducasse(a)free.fr> wrote: > gary? > > Do you have surgeon knife? > Because I do not see why slider need to depend on that wonderfully designed code. > > use: cachedSelector orMakeModelSelectorFor: selectorBody in: selectorBlock > | selector | > model ifNil: [^ nil]. > cachedSelector ifNil: > ["Make up selector from slotname if any" > selector := (slotName ifNil: [selectorBody] > ifNotNil: [slotName , selectorBody]) asSymbol. > (model class canUnderstand: selector) ifFalse: > [(self confirm: 'Shall I compile a null response for' > , Character cr asString > , model class name , '>>' , selector) > ifFalse: [self halt]. > model class compile: (String streamContents: > [:s | selector keywords doWithIndex: > [:k :i | s nextPutAll: k , ' arg' , i printString]. > s cr; nextPutAll: '"Automatically generated null response."'. > s cr; nextPutAll: '"Add code below for appropriate behavior..."'.]) > classified: 'input events' > notifying: nil]] > ifNotNil: > [selector := cachedSelector]. > ^ selectorBlock value: selector > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

Aug. 27, 2010

Re: [Pharo-project] Experimental Cocoa OS-X based Squeak Cog JIT VM 5.8b3.

by Sven Van Caekenberghe

It works now: I tried closing all open windows in the image and now it runs headless. I thought that wasn't necessary with other VMs but maybe I always did so without thinking about it. Thanks for answering, Eliot. Sven On 27 Aug 2010, at 20:46, Eliot Miranda wrote: > On Fri, Aug 27, 2010 at 1:59 AM, Sven Van Caekenberghe <sven(a)beta9.be> wrote: > > On 26 Aug 2010, at 21:44, Sven Van Caekenberghe wrote: > > > I hope the linux VM is equally fast. > > Is the COG VM capable of running headless on Linux ? > > We routinely run it headless on linux using -vm-display-null. > > sven@bladerunner:~/Smalltalk$ ./coglinux/bin/squeak -version > 3.9-7 #2 Wed Aug 18 19:45:40 GMT 2010 gcc 4.1.2 > Croquet Closure Cog VM [CoInterpreter VMMaker-oscog.26] > Linux cern 2.6.18-194.8.1.el5PAE #1 SMP Fri Jul 2 10:18:13 CEST 2010 i686 i686 i386 GNU/Linux > plugin path: /home/sven/Smalltalk/coglinux/lib/squeak/3.9-7/ [default: /home/sven/Smalltalk/coglinux/lib/squeak/3.9-7/] > > sven@bladerunner:~/Smalltalk$ ./coglinux/bin/squeak -vm-display-null -vm-sound-null seaside3.lr.cog.image > > This doesn't seem to work, no errors to be seen anywhere. Can I enable some debugging output ? > > Try -sendtrace, e.g. > ./coglinux/bin/squeak -vm-display-null -vm-sound-null -sendtrace seaside3.lr.cog.image > > This same image runs perfectly well with a display, as in > > sven@bladerunner:~/Smalltalk$ ./coglinux/bin/squeak seaside3.lr.cog.image > > What does your image do? Does it depend on any additional plugins or shared objects? Are these also present in your Cog installation? > > HTH > Eliot > > Thx, > > Sven

Aug. 27, 2010

Re: [Pharo-project] Why has MethodFinder a fixed list of results?

by Eliot Miranda

2010/8/27 Marcus Denker <marcus.denker(a)inria.fr> > > Smalltalk.true. > > takes a while, but it does not try to call #quitPrimitive. Which it would > if Methodfinder would not be based on a positive list. > > > Ah, and a negative list was I think not used as it is too dangerous wrt. to > completenes. If the positive list is incomplete, you miss > a hit. If the negative list is incomplete, you woud crash in some nasty > way. > > Just look at the lists and how many methods you find that have been removed > years ago... > > It's not just methods like #quitPrimitive. In general, with the image based > nature you could end up modifying state of objects > in the image by accident, and realize it weeks later. > e.g. this means one would very carefully protect the reflective model, to > not accidentally modify the methodDictionary, for example. > So if someone did try my simulation approach above the only lists needed would be a positive list of those primitives that were safe to apply to any object and a positive list of those primitives it would be safe to apply to the objects allocated during the simulation. best, Eliot > Marcus > > > -- > Marcus Denker -- http://www.marcusdenker.de > INRIA Lille -- Nord Europe. Team RMoD. > > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >

Aug. 27, 2010

Re: [Pharo-project] Why has MethodFinder a fixed list of results?

by Eliot Miranda

2010/8/27 Marcus Denker <marcus.denker(a)inria.fr> > > On Aug 27, 2010, at 8:59 PM, Bart Gauquie wrote: > > Dear list, > > I was showing off Pharo to a colleague today and was surprised that. > > My quick introduction of adding new methods on existing system classes > > String>>firstLetter > ^self copyFrom: 1 to: 1 > > - extension on String worked immediately. But that: > > in the MethodFinder evaluating: 'tried'.'t' > did not return my method I just added .... :-( > > Turns out that MethodFinder has a class side Set of Approved and > AddAndRemove methods. > What is the rationale of this? Should we not allow any method, but exclude > some tricky ones (doesNotUnderstand:, ... )? > > > The rational is that the MethodFinder would, if it would not have a > positive list, quite likely call methods that do bad things > to the state of the system. Remember that is just calls all methods to > check if some leads to the return value that you gave! > > e.g. > > Smalltalk.true. > > takes a while, but it does not try to call #quitPrimitive. Which it would > if Methodfinder would not be based on a positive list. > > It would of course be nice to be able to not need a positive list. > The question of how to control side-effects (and how to realize a secure > yet reflective language in general) is > of course a very interesting one... > I'd love for someone to try and implement MethodFinder using simulation. If one used something based on ContextPart runSimulated: to execute the tests the simulation could maintain the set of objects that get instantiated as execution proceeds and only allow writes to these objects. Any attempt to write an object outside the set would cause an error and abandon exploring that alternative. This *might* be fast enough for use, but the experiment would show whether this was the case or not. best Eliot > Marcus > > -- > Marcus Denker -- http://www.marcusdenker.de > INRIA Lille -- Nord Europe. Team RMoD. > > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >

Aug. 27, 2010

Re: [Pharo-project] [update 1.2] #12116

by Sven Van Caekenberghe

On 27 Aug 2010, at 21:19, StÃ©phane Ducasse wrote: > Strange > where is the other tests packages? > in the inbox? No, I mean the latest version in Andreas' WebClient repo.

Aug. 27, 2010

Re: [Pharo-project] Why has MethodFinder a fixed list of results?

by Marcus Denker

> > Smalltalk.true. > > takes a while, but it does not try to call #quitPrimitive. Which it would if Methodfinder would not be based on a positive list. Ah, and a negative list was I think not used as it is too dangerous wrt. to completenes. If the positive list is incomplete, you miss a hit. If the negative list is incomplete, you woud crash in some nasty way. Just look at the lists and how many methods you find that have been removed years ago... It's not just methods like #quitPrimitive. In general, with the image based nature you could end up modifying state of objects in the image by accident, and realize it weeks later. e.g. this means one would very carefully protect the reflective model, to not accidentally modify the methodDictionary, for example. Marcus -- Marcus Denker -- http://www.marcusdenker.de INRIA Lille -- Nord Europe. Team RMoD.

Aug. 27, 2010

Re: [Pharo-project] Why has MethodFinder a fixed list of results?

by StÃ©phane Ducasse

Now what we should do is probably to get a more dynamic way to tag methods because not methodFinder has a huge baglog of selectors and it is probably not in sync with the system We were thinking about using pragma to tag methods that should not be executed but we need a proof of concept. Stef On Aug 27, 2010, at 8:59 PM, Bart Gauquie wrote: > Dear list, > > I was showing off Pharo to a colleague today and was surprised that. > > My quick introduction of adding new methods on existing system classes > > String>>firstLetter > ^self copyFrom: 1 to: 1 > > - extension on String worked immediately. But that: > > in the MethodFinder evaluating: 'tried'.'t' > did not return my method I just added .... :-( > > Turns out that MethodFinder has a class side Set of Approved and AddAndRemove methods. > What is the rationale of this? Should we not allow any method, but exclude some tricky ones (doesNotUnderstand:, ... )? > > Kind Regards, > > Bart > > > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

Aug. 27, 2010

Re: [Pharo-project] Why is Stack extending from LinkedList?

by StÃ©phane Ducasse

On Aug 27, 2010, at 6:32 PM, Levente Uzonyi wrote: > On Thu, 26 Aug 2010, Bart Gauquie wrote: > >> Hi list, >> >> is there a specific reason Stack is extending from LinkedList, instead of >> using a LinkedList internally? >> To me, it seems that the protocol that LinkedList, SequenceableCollection, >> ... provides is not the protocol you expect from a Stack? >> >> Any thoughts on this? > > Originally it used encapsulation, when where? > but for some reason it was changed to use inheritance, probably for a "cleaner" system. IMHO the changes should be reverted. I was not aware that Stack was a subclass of linkedList, it looks again subclassing versus subtyping. And subtyping is better. Stef > > > Levente > >> >> Kind regards, >> >> Bart >> >> -- >> imagination is more important than knowledge - Albert Einstein >> Logic will get you from A to B. Imagination will take you everywhere - >> Albert Einstein >> Learn from yesterday, live for today, hope for tomorrow. The important thing >> is not to stop questioning. - Albert Einstein >> The true sign of intelligence is not knowledge but imagination. - Albert >> Einstein >> However beautiful the strategy, you should occasionally look at the results. >> - Sir Winston Churchill >> It's not enough that we do our best; sometimes we have to do what's >> required. - Sir Winston Churchill >> > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

Aug. 27, 2010