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

Re: [Pharo-project] How I build squeak-vm with FT2Plugin on Linux

by Mariano Martinez Peck

2010/3/22 laurent laffont <laurent.laffont(a)gmail.com> > > > 2010/3/22 Mariano Martinez Peck <marianopeck(a)gmail.com> > > >> >> On Mon, Mar 22, 2010 at 9:22 PM, Bryce Kampjes <bryce(a)kampjes.demon.co.uk >> > wrote: >> >>> On Thu, 2010-03-11 at 10:28 +0100, Mariano Martinez Peck wrote: >>> > Both really interesting posts. Now...Some people reported that the >>> > Exupery Linux VM has "nicer" fonts than the standard VM, even if both >>> > use true type. >>> > Then Bryce Kampjes said in the mailing list: >>> > >>> > "If it's font related then the Exupery VMs also have a patched bitlbt >>> > for >>> > subpixel antialiasing. The true type code uses that to make fonts look >>> > a >>> > bit nicer." >>> > >>> > Do you know how that can be included ? >>> >>> Apply the .cs in the following email to your copy of VMMaker and also >>> build the truetype plugin: >>> >>> >>> http://n4.nabble.com/Fwd-Pharo-project-New-Pharo-based-on-core-10309-with-a… >>> >>> The truetype work was done by Andy Tween. Googling for him, truetype, >>> plugin, and VM should find a bit more information. >>> >>> >> Thanks Bryce!!! would be cool to have the "official" Linux VM for Pharo >> 1.0 release with this feature. >> > > > I've built one here > http://lolgzs.free.fr/pharo/Squeak-3.11.3.2135-src-pached.tar.gz with > FT2Plugin and the patched bitlbt. > > Excellent!!! do you know if this vm has the "Gnuification" ??? I saw an email from Adrian in VM maling list saying he has problem with that. > If someone can test it on another machine ... > > I will try later (lazy to open virtual box hahah) Note that I have several tests failing with this VM on RC3 image (but it > doesn't crash, so that's far better than my previous custom VM :) > > Laurent Laffont > > >> >> Cheers >> >> Mariano >> >> >> Bryce >>> >>> >>> _______________________________________________ >>> 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 >

March 22, 2010

Re: [Pharo-project] bug in the code of CompiledMethod >> =

by Eliot Miranda

On Mon, Mar 22, 2010 at 2:54 PM, Nicolas Cellier < nicolas.cellier.aka.nice(a)gmail.com> wrote: > You might also want to check last code published in trunk. > I saw that. You're eliminating closeTo: for float comparison which seems fine to me :) Thanks Nicolas! However, it is indicative that a recompile all is necessary after redefining the number scanning facilities. I guess that this is part of the standard release build process. If not, we should add it. > > Nicolas > > 2010/3/22 Eliot Miranda <eliot.miranda(a)gmail.com>: > > > > > > On Mon, Mar 22, 2010 at 2:25 PM, StÃ©phane Ducasse > > <stephane.ducasse(a)inria.fr> wrote: > >> > >> eliot > >> > >> Your changes fixed some of the problems. Now I tried to fix the test > >> testAnalogousCodeTo in MethodPropertiesTest and > >> the expression returns false. > >> > >> (#zork->'hello') analogousCodeTo: (#zork->'hello') -> false > >> > >> (#zork->'hello') = (#zork->'hello') -> true > > > > For me > > #(#zork -> 'hello') size 3 > > In Pharo is it true that > > #(#zork -> 'hello') size 1 > > ??!?! > > In that case I would implement analogousCodeTo: in Association, e.g. > > analogousCodeTo: anObject > > ^anObject class == self class > > and: [key = anObject key > > and: [value = anObject value]] > > > > But if Association isn't a literal I would simply check > > MethodPropertiesTests. It could be obsolete. > >> > >> Stef > >> > >> On Mar 22, 2010, at 6:20 PM, Eliot Miranda wrote: > >> > >> > Oops. Find attached. > >> > > >> > 2010/3/22 Cyrille Delaunay <cy.delaunay(a)gmail.com> > >> > try in a Pharo-1.0-10515-rc3 image : > >> > > >> > |set| > >> > set := Set new. > >> > Collection withAllSubclasses do: [:aClass | > >> > set addAll: aClass methods > >> > ]. > >> > > >> > It will raise an exception telling: "MessageNotUnderstood: > >> > ByteSymbol>>analogousCodeTo:". > >> > Indeed, in the code of CompiledMethod >> = , the message > >> > 'analogousCodeTo:' is send to a > >> > symbol. > >> > > >> > > >> > I opened an Issue: > >> > http://code.google.com/p/pharo/issues/detail?id=2185 > >> > > >> > > >> > > >> > _______________________________________________ > >> > Pharo-project mailing list > >> > Pharo-project(a)lists.gforge.inria.fr > >> > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > >> > > >> > > >> > > <Object-analogousCodeTo.st>_______________________________________________ > >> > 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 > > > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >

March 22, 2010

Re: [Pharo-project] How I build squeak-vm with FT2Plugin on Linux

by laurent laffont

2010/3/22 Mariano Martinez Peck <marianopeck(a)gmail.com> > > > On Mon, Mar 22, 2010 at 9:22 PM, Bryce Kampjes <bryce(a)kampjes.demon.co.uk>wrote: > >> On Thu, 2010-03-11 at 10:28 +0100, Mariano Martinez Peck wrote: >> > Both really interesting posts. Now...Some people reported that the >> > Exupery Linux VM has "nicer" fonts than the standard VM, even if both >> > use true type. >> > Then Bryce Kampjes said in the mailing list: >> > >> > "If it's font related then the Exupery VMs also have a patched bitlbt >> > for >> > subpixel antialiasing. The true type code uses that to make fonts look >> > a >> > bit nicer." >> > >> > Do you know how that can be included ? >> >> Apply the .cs in the following email to your copy of VMMaker and also >> build the truetype plugin: >> >> >> http://n4.nabble.com/Fwd-Pharo-project-New-Pharo-based-on-core-10309-with-a… >> >> The truetype work was done by Andy Tween. Googling for him, truetype, >> plugin, and VM should find a bit more information. >> >> > Thanks Bryce!!! would be cool to have the "official" Linux VM for Pharo 1.0 > release with this feature. > I've built one here http://lolgzs.free.fr/pharo/Squeak-3.11.3.2135-src-pached.tar.gz with FT2Plugin and the patched bitlbt. If someone can test it on another machine ... Note that I have several tests failing with this VM on RC3 image (but it doesn't crash, so that's far better than my previous custom VM :) Laurent Laffont > > Cheers > > Mariano > > > Bryce >> >> >> _______________________________________________ >> 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 >

March 22, 2010

Re: [Pharo-project] bug in the code of CompiledMethod >> =

by Nicolas Cellier

You might also want to check last code published in trunk. Nicolas 2010/3/22 Eliot Miranda <eliot.miranda(a)gmail.com>: > > > On Mon, Mar 22, 2010 at 2:25 PM, StÃ©phane Ducasse > <stephane.ducasse(a)inria.fr> wrote: >> >> eliot >> >> Your changes fixed some of the problems. Now I tried to fix the test >> testAnalogousCodeTo in MethodPropertiesTest and >> the expression returns false. >> >> (#zork->'hello') analogousCodeTo: (#zork->'hello') Â -> false >> >> (#zork->'hello') = (#zork->'hello') -> true > > For me > Â Â Â #(#zork -> 'hello') size 3 > In Pharo is it true that > #(#zork -> 'hello') size 1 > ??!?! > In that case I would implement analogousCodeTo: in Association, e.g. > analogousCodeTo: anObject > Â Â Â ^anObject class == self class > Â Â Â and: [key = anObject key > Â Â Â and: [value = anObject value]] > > But if Association isn't a literal I would simply check > MethodPropertiesTests. Â It could be obsolete. >> >> Stef >> >> On Mar 22, 2010, at 6:20 PM, Eliot Miranda wrote: >> >> > Oops. Â Find attached. >> > >> > 2010/3/22 Cyrille Delaunay <cy.delaunay(a)gmail.com> >> > try in a Pharo-1.0-10515-rc3 image : >> > >> > |set| >> > set := Set new. >> > Collection withAllSubclasses do: [:aClass | >> > Â Â Â set addAll: aClass methods >> > Â Â Â ]. >> > >> > It will raise an exception telling: "MessageNotUnderstood: >> > ByteSymbol>>analogousCodeTo:". >> > Indeed, in the code of CompiledMethod >> = , the message >> > 'analogousCodeTo:' is send to a >> > symbol. >> > >> > >> > I opened an Issue: >> > http://code.google.com/p/pharo/issues/detail?id=2185 >> > >> > >> > >> > _______________________________________________ >> > Pharo-project mailing list >> > Pharo-project(a)lists.gforge.inria.fr >> > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> > >> > >> > <Object-analogousCodeTo.st>_______________________________________________ >> > 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 >

March 22, 2010

Re: [Pharo-project] bug in the code of CompiledMethod >> =

by Eliot Miranda

On Mon, Mar 22, 2010 at 2:25 PM, StÃ©phane Ducasse <stephane.ducasse(a)inria.fr > wrote: > eliot > > Your changes fixed some of the problems. Now I tried to fix the test > testAnalogousCodeTo in MethodPropertiesTest and > the expression returns false. > > (#zork->'hello') analogousCodeTo: (#zork->'hello') -> false > > (#zork->'hello') = (#zork->'hello') -> true > For me #(#zork -> 'hello') size 3 In Pharo is it true that #(#zork -> 'hello') size 1 ??!?! In that case I would implement analogousCodeTo: in Association, e.g. analogousCodeTo: anObject ^anObject class == self class and: [key = anObject key and: [value = anObject value]] But if Association isn't a literal I would simply check MethodPropertiesTests. It could be obsolete. > Stef > > On Mar 22, 2010, at 6:20 PM, Eliot Miranda wrote: > > > Oops. Find attached. > > > > 2010/3/22 Cyrille Delaunay <cy.delaunay(a)gmail.com> > > try in a Pharo-1.0-10515-rc3 image : > > > > |set| > > set := Set new. > > Collection withAllSubclasses do: [:aClass | > > set addAll: aClass methods > > ]. > > > > It will raise an exception telling: "MessageNotUnderstood: > ByteSymbol>>analogousCodeTo:". > > Indeed, in the code of CompiledMethod >> = , the message > 'analogousCodeTo:' is send to a > > symbol. > > > > > > I opened an Issue: > > http://code.google.com/p/pharo/issues/detail?id=2185 > > > > > > > > _______________________________________________ > > Pharo-project mailing list > > Pharo-project(a)lists.gforge.inria.fr > > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > > > > > <Object-analogousCodeTo.st>_______________________________________________ > > 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 >

March 22, 2010

Re: [Pharo-project] bug in the code of CompiledMethod >> =

by StÃ©phane Ducasse

not sure that it does not introduce side effect but Association>>isLiteral ^ self key isLiteral and: [self value isLiteral] Fix the problem. stef On Mar 22, 2010, at 10:25 PM, StÃ©phane Ducasse wrote: > eliot > > Your changes fixed some of the problems. Now I tried to fix the test testAnalogousCodeTo in MethodPropertiesTest and > the expression returns false. > > (#zork->'hello') analogousCodeTo: (#zork->'hello') -> false > > (#zork->'hello') = (#zork->'hello') -> true > > Stef > > On Mar 22, 2010, at 6:20 PM, Eliot Miranda wrote: > >> Oops. Find attached. >> >> 2010/3/22 Cyrille Delaunay <cy.delaunay(a)gmail.com> >> try in a Pharo-1.0-10515-rc3 image : >> >> |set| >> set := Set new. >> Collection withAllSubclasses do: [:aClass | >> set addAll: aClass methods >> ]. >> >> It will raise an exception telling: "MessageNotUnderstood: ByteSymbol>>analogousCodeTo:". >> Indeed, in the code of CompiledMethod >> = , the message 'analogousCodeTo:' is send to a >> symbol. >> >> >> I opened an Issue: >> http://code.google.com/p/pharo/issues/detail?id=2185 >> >> >> >> _______________________________________________ >> Pharo-project mailing list >> Pharo-project(a)lists.gforge.inria.fr >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> >> <Object-analogousCodeTo.st>_______________________________________________ >> 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

March 22, 2010

Re: [Pharo-project] bug in the code of CompiledMethod >> =

by StÃ©phane Ducasse

eliot Your changes fixed some of the problems. Now I tried to fix the test testAnalogousCodeTo in MethodPropertiesTest and the expression returns false. (#zork->'hello') analogousCodeTo: (#zork->'hello') -> false (#zork->'hello') = (#zork->'hello') -> true Stef On Mar 22, 2010, at 6:20 PM, Eliot Miranda wrote: > Oops. Find attached. > > 2010/3/22 Cyrille Delaunay <cy.delaunay(a)gmail.com> > try in a Pharo-1.0-10515-rc3 image : > > |set| > set := Set new. > Collection withAllSubclasses do: [:aClass | > set addAll: aClass methods > ]. > > It will raise an exception telling: "MessageNotUnderstood: ByteSymbol>>analogousCodeTo:". > Indeed, in the code of CompiledMethod >> = , the message 'analogousCodeTo:' is send to a > symbol. > > > I opened an Issue: > http://code.google.com/p/pharo/issues/detail?id=2185 > > > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > > <Object-analogousCodeTo.st>_______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

March 22, 2010

Re: [Pharo-project] Thanks martin :)

by Marcus Denker

On Mar 22, 2010, at 9:33 PM, stephane ducasse wrote: > hi martin > > marcus and me integrated your fixes. Marcus should send a nice update mail. done. Marcus -- Marcus Denker -- http://www.marcusdenker.de INRIA Lille -- Nord Europe. Team RMoD.

March 22, 2010

[Pharo-project] [update 1.1] #11284

by Marcus Denker

#11284 ====== Hashed collection performance improvement, phase 4 of 4. Changes identityHash to cover most of the positive SmallInteger range, and uses good prime table sizes. This makes large hashed collections much faster, and does not penalize the small ones. Targeted at Pharo Core 1.1-11280 Load the slice for each phase in order. Allow maybe half an hour to load all four phases; the intermediate phases load very slowly due to existing hashed collections running in "slow but safe" mode while their hash function changes under them. In phase 4: * All of the scanFor: methods modified in phase 1 are loaded into their final form (sometimes a revert to original) which does not scan the entire table before giving up. * The rehashing class and temporary methods are removed, having done their job at the end of phase 3. -- Marcus Denker -- http://www.marcusdenker.de INRIA Lille -- Nord Europe. Team RMoD.

March 22, 2010

[Pharo-project] [update 1.1] #11283

by Marcus Denker

#11283 ====== Hashed collection performance improvement, phase 3 of 4. Changes identityHash to cover most of the positive SmallInteger range, and uses good prime table sizes. This makes large hashed collections much faster, and does not penalize the small ones. Targeted at Pharo Core 1.1-11280 Load the slice for each phase in order. Allow maybe half an hour to load all four phases; the intermediate phases load very slowly due to existing hashed collections running in "slow but safe" mode while their hash function changes under them. In phase 3: * Identity hash values are changed. * Set>>growSize, rendered obsolete by phase 2, is removed. * A "Rehasher" class is added * Rehasher class>>initialize rehashes all hashed collections. Phase 3 takes several minutes to load, due to load and rehashing taking place in "slow but safe" mode. -- Marcus Denker -- http://www.marcusdenker.de INRIA Lille -- Nord Europe. Team RMoD.

March 22, 2010