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

Re: [Pharo-project] image crash when saving a method

by Damien Pollet

On Wed, Sep 29, 2010 at 21:15, Henrik Sperre Johansen <henrik.s.johansen(a)veloxit.no> wrote: > Â On 29.09.2010 20:54, Damien Pollet wrote: >> >> I have an image that reliably crashes the VM whenever I save a method >> in a particular class. >> >> I tried opening a different browser, saving an empty method, restoring >> the method from changesâ¦ crash, and no PharoDebug.log >> It seems I can still modify other classes. >> >> This is the 1.1.1 image (#11414) with a few packages (PetitParser, >> Filesystem, Autotest) on John's 5.8b10 VM. >> >> What to do ? > > For the last time, Autotest uses Objects as Methods. > Cog VMs until recently crash when you use that. > 5.8b10 is one of those. > > So if the method you edit would trigger a test, the image will crash. OK, I closed the autotest window, and could save my method (which was indeed a unit test). -- Damien Pollet type less, do more [ | ] http://people.untyped.org/damien.pollet

Sept. 29, 2010

Re: [Pharo-project] How to subclass Process and have it used by spawned blocks?

by Stefan Marr

Stef, I had a look at the description of these fixes, but it does not look like there is something related in it. Eliot, the requested output is below. But as far as I can see, it is identical. Have verified it also in the Squeak 4.1 image, all three produce the same bytecode. Perhaps it not the compiler directly but some closure related code? Thanks Stefan Pharo 1.1: (Erlang class>>#spawn:) abstractSymbolic '<primitive: 19> pushLit: Process pushTemp: 0 closureNumCopied: 1 numArgs: 0 bytes 51 to 59 pushConstant: 1 send: #halt (0 args) pop pushTemp: 0 send: #value (0 args) pop pushLit: Processor send: #terminateActive (0 args) blockReturn send: #asContext (0 args) pushLit: Processor send: #activePriority (0 args) send: #forContext:priority: (2 args) popIntoTemp: 1 pushTemp: 1 send: #resume (0 args) pop pushTemp: 1 returnTop ' Pharo 1.1: (Erlang class>>#spawn:) symbolic. '<primitive: 19> 45 <41> pushLit: Process 46 <10> pushTemp: 0 47 <8F 10 00 09> closureNumCopied: 1 numArgs: 0 bytes 51 to 59 51 <76> pushConstant: 1 52 <D3> send: halt 53 <87> pop 54 <10> pushTemp: 0 55 <C9> send: value 56 <87> pop 57 <45> pushLit: Processor 58 <D4> send: terminateActive 59 <7D> blockReturn 60 <D2> send: asContext 61 <45> pushLit: Processor 62 <D6> send: activePriority 63 <F0> send: forContext:priority: 64 <69> popIntoTemp: 1 65 <11> pushTemp: 1 66 <D7> send: resume 67 <87> pop 68 <11> pushTemp: 1 69 <7C> returnTop ' Squeak 4.2: (Erlang class>>#spawn:) abstractSymbolic. '<primitive: 19> pushLit: Process pushTemp: 0 closureNumCopied: 1 numArgs: 0 bytes 51 to 59 pushConstant: 1 send: #halt (0 args) pop pushTemp: 0 send: #value (0 args) pop pushLit: Processor send: #terminateActive (0 args) blockReturn send: #asContext (0 args) pushLit: Processor send: #activePriority (0 args) send: #forContext:priority: (2 args) popIntoTemp: 1 pushTemp: 1 send: #resume (0 args) pop pushTemp: 1 returnTop ' Squeak 4.2: (Erlang class>>#spawn:) symbolic. '<primitive: 19> 45 <41> pushLit: Process 46 <10> pushTemp: 0 47 <8F 10 00 09> closureNumCopied: 1 numArgs: 0 bytes 51 to 59 51 <76> pushConstant: 1 52 <D3> send: halt 53 <87> pop 54 <10> pushTemp: 0 55 <C9> send: value 56 <87> pop 57 <45> pushLit: Processor 58 <D4> send: terminateActive 59 <7D> blockReturn 60 <D2> send: asContext 61 <45> pushLit: Processor 62 <D6> send: activePriority 63 <F0> send: forContext:priority: 64 <69> popIntoTemp: 1 65 <11> pushTemp: 1 66 <D7> send: resume 67 <87> pop 68 <11> pushTemp: 1 69 <7C> returnTop ' On 29 Sep 2010, at 20:29, StÃ©phane Ducasse wrote: > and http://code.google.com/p/pharo/issues/list?cursor=2818 > > Stef > > >>> spawn: aBlock >>> "Answer an ErlProcess running the code given in the parameter. The process isscheduled." >>> | proc | >>> <primitive: 19> "Simulation guard" >>> proc := Process >>> forContext: >>> [1 halt. aBlock value. >>> Processor terminateActive] asContext >>> priority: Processor activePriority. >>> proc resume. >>> ^ proc -- Stefan Marr Software Languages Lab Vrije Universiteit Brussel Pleinlaan 2 / B-1050 Brussels / Belgium http://soft.vub.ac.be/~smarr Phone: +32 2 629 2974 Fax: +32 2 629 3525

Sept. 29, 2010

Re: [Pharo-project] Alien Plugin for Linux

by Henrik Sperre Johansen

On 29.09.2010 20:11, Nicolas Cellier wrote: > 2010/9/29 Marcus Denker<marcus.denker(a)inria.fr>: >> On Sep 29, 2010, at 6:49 PM, Nicolas Cellier wrote: >> >>> 2010/9/29 Henrik Johansen<henrik.s.johansen(a)veloxit.no>: >>>> Uhm, this method (and many others in the extensions-category) have been present in their expected-by-Alien form in Pharo since at least 1.0 >>>> >>>> Probably a better idea to separate the Alien-"Core" into two packages, one which is actually needed, and one which you only need to load in really old Squeak/Cuis images in order to have it work. >>>> In which case you probably won't have method trailers either, and generate: #(0 0 0 0) is the proper extension. >>>> >>> CompiledMethodTrailers default or mysterious #(0 0 0 0) codes both are >>> mostly unnecessary. >>> If I want default behaviour, why should I bother to tell what this >>> default is exactly ? >>> Let just call #generate, and obfuscate those details. >>> Squeak Pharo Cuis will handle these better than us (didn't verify >>> Cuis, but quite sure Squeak/Pharo already implement generate). >>> >> I added #generate in 3.9 and refactored everything to use it. But the closure code un-did a lot of >> these as it was coming from an older version (3.7 or 3.8 was the basis of Croquet, I think). >> >> Marcus >> > Thank you Marcus, > A bit like http://en.wikipedia.org/wiki/Sisyphus, let just redo. > > Nicolas Yeah, my point was that Pharo 1.0, Pharo 1.1, Squeak 4.1, and Squeak 4.2 all contain proper versions of MethodNode>>generate. If it existed in 3.9 as well, it would probably only be needed in the scenario that you want to use Alien in 3.8 or earlier. Thus the override should be moved to an optional 3.8 package, or simply removed, as few VMs that will run 3.8 actually contains/support the Alien plugin. (it required specific changes to the core VM iirc) Simply updating the override of the method as included in Alien-Core, to reflect the method trailer changes done between Squeak 4.0 and 4.1, and Pharo 1.0 and 1.1, as Fernando did, would definetely be the wrong solution, at least in my eyes. Cheers, Henry

Sept. 29, 2010

Re: [Pharo-project] image crash when saving a method

by Henrik Sperre Johansen

On 29.09.2010 20:54, Damien Pollet wrote: > I have an image that reliably crashes the VM whenever I save a method > in a particular class. > > I tried opening a different browser, saving an empty method, restoring > the method from changesâ¦ crash, and no PharoDebug.log > It seems I can still modify other classes. > > This is the 1.1.1 image (#11414) with a few packages (PetitParser, > Filesystem, Autotest) on John's 5.8b10 VM. > > What to do ? For the last time, Autotest uses Objects as Methods. Cog VMs until recently crash when you use that. 5.8b10 is one of those. So if the method you edit would trigger a test, the image will crash.

Sept. 29, 2010

Re: [Pharo-project] image crash when saving a method

by Nicolas Cellier

I'd say peek latest binary COG compiled by Eliot and test again (can't remember exact address, you'll have to browse...). Then forward the problem to vm-dev with more details. Nicolas 2010/9/29 StÃ©phane Ducasse <stephane.ducasse(a)inria.fr>: > > can you file out this class and load it in a 12414 and does it crash? > >> I have an image that reliably crashes the VM whenever I save a method >> in a particular class. > > Strange. > >> I tried opening a different browser, saving an empty method, restoring >> the method from changesâ¦ crash, and no PharoDebug.log >> It seems I can still modify other classes. >> >> This is the 1.1.1 image (#11414) with a few packages (PetitParser, >> Filesystem, Autotest) on John's 5.8b10 VM. > > does it mean with cog? > >> >> What to do ? >> >> -- >> Damien Pollet >> type less, do more [ | ] http://people.untyped.org/damien.pollet >> >> _______________________________________________ >> 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 >

Sept. 29, 2010

Re: [Pharo-project] image crash when saving a method

by StÃ©phane Ducasse

can you file out this class and load it in a 12414 and does it crash? > I have an image that reliably crashes the VM whenever I save a method > in a particular class. Strange. > I tried opening a different browser, saving an empty method, restoring > the method from changesâ¦ crash, and no PharoDebug.log > It seems I can still modify other classes. > > This is the 1.1.1 image (#11414) with a few packages (PetitParser, > Filesystem, Autotest) on John's 5.8b10 VM. does it mean with cog? > > What to do ? > > -- > Damien Pollet > type less, do more [ | ] http://people.untyped.org/damien.pollet > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

Sept. 29, 2010

[Pharo-project] image crash when saving a method

by Damien Pollet

I have an image that reliably crashes the VM whenever I save a method in a particular class. I tried opening a different browser, saving an empty method, restoring the method from changesâ¦ crash, and no PharoDebug.log It seems I can still modify other classes. This is the 1.1.1 image (#11414) with a few packages (PetitParser, Filesystem, Autotest) on John's 5.8b10 VM. What to do ? -- Damien Pollet type less, do more [ | ] http://people.untyped.org/damien.pollet

Sept. 29, 2010

Re: [Pharo-project] How to subclass Process and have it used by spawned blocks?

by StÃ©phane Ducasse

and http://code.google.com/p/pharo/issues/list?cursor=2818 Stef On Sep 29, 2010, at 8:28 PM, StÃ©phane Ducasse wrote: > we still should integrate your changes but they require momentum and concentration which we do not have. > Stefan if you want to help you are welcome > > http://code.google.com/p/pharo/issues/detail?id=2897 > > Stef > > On Sep 29, 2010, at 7:49 PM, Eliot Miranda wrote: > >> Hi Stefan, >> >> this sounds a lot like an issue that bit someone in squeak 4.x recently and was due to a bytecode compiler bug that I fixed. I can't find or remember the email associated with this, so if anyone does remember please speak up. Stefan, you might try running this in an updated squeak 4.1 image and compare the bytecodes. If you get different bytecodes then we need to port the compiler fixes from 4.1 trunk to Pharo. >> >> HTH >> Eliot >> >> >> On Wed, Sep 29, 2010 at 10:22 AM, Stefan Marr <pharo(a)stefan-marr.de> wrote: >> Further testing reveals a strange behavior in the debugger. >> So, I am not sure where the problem is actually caused. >> >> I added a new local variable block to the method. >> block gets assigned aBlock (the parameter). >> >> If I executed Erlang spawn: [Transcript show: 'foo'], the '1 halt.' is executed as I mentioned before. >> However, the debugger does not allow me to access aBlock, there is a subscript out of bound exception. >> Interestingly, 'block' is accessible without error. >> >> >> If I now actually use 'block' instead of 'aBlock' in the executed block, then phenomena is reversed. >> Thus, the variable which is actually referenced in the code is not accessible anymore but gets somehow lost. They also get lost if I use both. >> >> >> I would still like to understand what is going on here. >> So any remarks are welcome. >> >> Thanks >> Stefan >> >> >> >> >> spawn: aBlock >> "Answer an ErlProcess running the code given in the parameter. The process isscheduled." >> | proc block | >> <primitive: 19> "Simulation guard" >> block := aBlock. >> proc := Process >> forContext: >> [1 halt. aBlock value. >> Processor terminateActive] asContext >> priority: Processor activePriority. >> proc resume. >> ^ proc >> >> >> >> On 29 Sep 2010, at 11:21, Stefan Marr wrote: >> >>> >>> On 29 Sep 2010, at 11:01, Igor Stasenko wrote: >>> >>>> On 29 September 2010 11:11, StÃ©phane Ducasse <stephane.ducasse(a)inria.fr> wrote: >>>>> no idea but I want to know that too so keep us posted. >>>>> >>>> Smalltalk specialObjectsArray at: 28. >>>> Thats it, Process is a special object, recognized by VM. >>>> No other Process (sub)classes can be accepted by it. >>> Could you elaborate on that? No other process can be accepted for what? >>> Being scheduled, or being executed? >>> >>> There is also a comment in the comment of Process which suggests that it should be possible: >>> "(If anyone ever makes a subclass of Process, be sure to use allSubInstances in anyProcessesAbove:.)" >>> But even after making the change, that does not seem to be the actual problem. >>> >>> >>> The process gets execute fine. With a '1 halt.' as the first statement of the aBlock block I even get a debugger. It only fails on accessing variables referenced by the closure with an index out of bound error or something similar. >>> >>>>> Erlang>>spawn: aBlock >>>>> "Answer an ErlProcess running the code given in the parameter. >>>>> The process is scheduled." >>>>> | proc | >>>>> proc := ErlProcess >>>>> forContext: >>>>> [1 halt. aBlock value. >>>>> Processor terminateActive] asContext >>>>> priority: Processor activePriority. >>>>> proc resume. >>>>> ^ proc >>> >>> >>> Thanks and best regards >>> Stefan >>> >>> >>> -- >>> Stefan Marr >>> Software Languages Lab >>> Vrije Universiteit Brussel >>> Pleinlaan 2 / B-1050 Brussels / Belgium >>> http://soft.vub.ac.be/~smarr >>> Phone: +32 2 629 2974 >>> Fax: +32 2 629 3525 >>> >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> Pharo-project(a)lists.gforge.inria.fr >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> >> -- >> Stefan Marr >> Software Languages Lab >> Vrije Universiteit Brussel >> Pleinlaan 2 / B-1050 Brussels / Belgium >> http://soft.vub.ac.be/~smarr >> Phone: +32 2 629 2974 >> Fax: +32 2 629 3525 >> >> >> _______________________________________________ >> 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 > > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

Sept. 29, 2010

Re: [Pharo-project] How to subclass Process and have it used by spawned blocks?

by StÃ©phane Ducasse

we still should integrate your changes but they require momentum and concentration which we do not have. Stefan if you want to help you are welcome http://code.google.com/p/pharo/issues/detail?id=2897 Stef On Sep 29, 2010, at 7:49 PM, Eliot Miranda wrote: > Hi Stefan, > > this sounds a lot like an issue that bit someone in squeak 4.x recently and was due to a bytecode compiler bug that I fixed. I can't find or remember the email associated with this, so if anyone does remember please speak up. Stefan, you might try running this in an updated squeak 4.1 image and compare the bytecodes. If you get different bytecodes then we need to port the compiler fixes from 4.1 trunk to Pharo. > > HTH > Eliot > > > On Wed, Sep 29, 2010 at 10:22 AM, Stefan Marr <pharo(a)stefan-marr.de> wrote: > Further testing reveals a strange behavior in the debugger. > So, I am not sure where the problem is actually caused. > > I added a new local variable block to the method. > block gets assigned aBlock (the parameter). > > If I executed Erlang spawn: [Transcript show: 'foo'], the '1 halt.' is executed as I mentioned before. > However, the debugger does not allow me to access aBlock, there is a subscript out of bound exception. > Interestingly, 'block' is accessible without error. > > > If I now actually use 'block' instead of 'aBlock' in the executed block, then phenomena is reversed. > Thus, the variable which is actually referenced in the code is not accessible anymore but gets somehow lost. They also get lost if I use both. > > > I would still like to understand what is going on here. > So any remarks are welcome. > > Thanks > Stefan > > > > > spawn: aBlock > "Answer an ErlProcess running the code given in the parameter. The process isscheduled." > | proc block | > <primitive: 19> "Simulation guard" > block := aBlock. > proc := Process > forContext: > [1 halt. aBlock value. > Processor terminateActive] asContext > priority: Processor activePriority. > proc resume. > ^ proc > > > > On 29 Sep 2010, at 11:21, Stefan Marr wrote: > > > > > On 29 Sep 2010, at 11:01, Igor Stasenko wrote: > > > >> On 29 September 2010 11:11, StÃ©phane Ducasse <stephane.ducasse(a)inria.fr> wrote: > >>> no idea but I want to know that too so keep us posted. > >>> > >> Smalltalk specialObjectsArray at: 28. > >> Thats it, Process is a special object, recognized by VM. > >> No other Process (sub)classes can be accepted by it. > > Could you elaborate on that? No other process can be accepted for what? > > Being scheduled, or being executed? > > > > There is also a comment in the comment of Process which suggests that it should be possible: > > "(If anyone ever makes a subclass of Process, be sure to use allSubInstances in anyProcessesAbove:.)" > > But even after making the change, that does not seem to be the actual problem. > > > > > > The process gets execute fine. With a '1 halt.' as the first statement of the aBlock block I even get a debugger. It only fails on accessing variables referenced by the closure with an index out of bound error or something similar. > > > >>> Erlang>>spawn: aBlock > >>> "Answer an ErlProcess running the code given in the parameter. > >>> The process is scheduled." > >>> | proc | > >>> proc := ErlProcess > >>> forContext: > >>> [1 halt. aBlock value. > >>> Processor terminateActive] asContext > >>> priority: Processor activePriority. > >>> proc resume. > >>> ^ proc > > > > > > Thanks and best regards > > Stefan > > > > > > -- > > Stefan Marr > > Software Languages Lab > > Vrije Universiteit Brussel > > Pleinlaan 2 / B-1050 Brussels / Belgium > > http://soft.vub.ac.be/~smarr > > Phone: +32 2 629 2974 > > Fax: +32 2 629 3525 > > > > > > _______________________________________________ > > Pharo-project mailing list > > Pharo-project(a)lists.gforge.inria.fr > > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > > -- > Stefan Marr > Software Languages Lab > Vrije Universiteit Brussel > Pleinlaan 2 / B-1050 Brussels / Belgium > http://soft.vub.ac.be/~smarr > Phone: +32 2 629 2974 > Fax: +32 2 629 3525 > > > _______________________________________________ > 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

Sept. 29, 2010

Re: [Pharo-project] Alien Plugin for Linux

by Nicolas Cellier

2010/9/29 Marcus Denker <marcus.denker(a)inria.fr>: > > On Sep 29, 2010, at 6:49 PM, Nicolas Cellier wrote: > >> 2010/9/29 Henrik Johansen <henrik.s.johansen(a)veloxit.no>: >>> Uhm, this method (and many others in the extensions-category) have been present in their expected-by-Alien form in Pharo since at least 1.0 >>> >>> Probably a better idea to separate the Alien-"Core" into two packages, one which is actually needed, and one which you only need to load in really old Squeak/Cuis images in order to have it work. >>> In which case you probably won't have method trailers either, and generate: #(0 0 0 0) is the proper extension. >>> >> >> CompiledMethodTrailers default or mysterious #(0 0 0 0) codes both are >> mostly unnecessary. >> If I want default behaviour, why should I bother to tell what this >> default is exactly ? >> Let just call #generate, and obfuscate those details. >> Squeak Pharo Cuis will handle these better than us (didn't verify >> Cuis, but quite sure Squeak/Pharo already implement generate). >> > > I added #generate in 3.9 and refactored everything to use it. But the closure code un-did a lot of > these as it was coming from an older version (3.7 or 3.8 was the basis of Croquet, I think). > > Â Â Â Â Marcus > Thank you Marcus, A bit like http://en.wikipedia.org/wiki/Sisyphus, let just redo. Nicolas > > -- > Marcus Denker Â -- 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 >

Sept. 29, 2010