- Pharo-dev - lists.pharo.org

Re: [Pharo-dev] ConfigurationOfMorphicCore

by Thierry Goubier

2014-07-24 13:47 GMT+02:00 stepharo <stepharo(a)free.fr>: > > On 24/7/14 13:37, Pavel Krivanek wrote: > >> Hi, >> >> in latest update we integrated several patches needed for proper working >> of ConfiguratonOfMorphicCore. This simple configuration loads Morphic-Core >> package into the minimal image and initializes it so the result is a small >> standalone Morphic subset. >> >> But some issues that I would like to discuss with the community left: >> >> SystemWindows support >> ============================ >> SystemWindow is quite complex morph that requires many of other morph >> classes including text morphs, theming support etc. But in several >> situations the system windows play special role and the Morphic core >> classes need to be aware of them. So we have several options: >> >> - (current) do not have SystemWindow & co. in the image but let basic >> support of them. That will produce some unimplemented calls >> >> - create very simple SystemWindow superclass that will be part of the >> Morphic-Core but will not have look and possibilities of real SystemWindow >> > I would take this one. Having a minimal Morphic is to be able to do GUI stuff, and this requires some form of SystemWindow. Maybe a fullscreen SystemWindow / world abstraction then? If I only load Morphic-Core, then I can use it for full-screen demos. > >> - include SystemWindow and supporting classes >> > I had the same questions than you: > - I have the impression that systemWindow is really linked to widgets. > > > > Tools >> ============================ >> The current image generated by this configuraton will crash as soon as >> you try to press some keyboard button. The reason is in the method >> Morph>>#handleKeystroke. This method contains this code to support >> Keymapping >> >> Smalltalk tools shortcuts handleKeystroke: anEvent inMorph: self. >> > > I think that keymapping should not be a tool but the default way to change > and control shortcuts. > Do you have a take on this? Looks like an integration fault to me (hidden dependency on KeyMapping). handleKeystroke: was a Keymapping-Core-override in 2.0. In 3.0 and 4.0, it's a core method of Morph... If you go back to 2.0, no more problem :) Thierry

July 24, 2014

Re: [Pharo-dev] ConfigurationOfMorphicCore

by stepharo

On 24/7/14 13:37, Pavel Krivanek wrote: > Hi, > > in latest update we integrated several patches needed for proper > working of ConfiguratonOfMorphicCore. This simple configuration loads > Morphic-Core package into the minimal image and initializes it so the > result is a small standalone Morphic subset. > > But some issues that I would like to discuss with the community left: > > SystemWindows support > ============================ > SystemWindow is quite complex morph that requires many of other morph > classes including text morphs, theming support etc. But in several > situations the system windows play special role and the Morphic core > classes need to be aware of them. So we have several options: > > - (current) do not have SystemWindow & co. in the image but let basic > support of them. That will produce some unimplemented calls > > - create very simple SystemWindow superclass that will be part of the > Morphic-Core but will not have look and possibilities of real SystemWindow > > - include SystemWindow and supporting classes I had the same questions than you: - I have the impression that systemWindow is really linked to widgets. > Tools > ============================ > The current image generated by this configuraton will crash as soon as > you try to press some keyboard button. The reason is in the method > Morph>>#handleKeystroke. This method contains this code to support > Keymapping > > Smalltalk tools shortcuts handleKeystroke: anEvent inMorph: self. I think that keymapping should not be a tool but the default way to change and control shortcuts. Do you have a take on this? > The #tools message is part of Tool-Base package and need to be there > because it uses lazy evaluation that creates a ToolRegistry instance. > Because Tool-Base is not loaded in the minimal nor Morphic-Core image, > it raises DNU > > Solutions: > - introduce some safer form like Smalltalk withToolsDo: [:tools | > ...]. In this case it will not be able to look if the class variable > "Tools" is nil (because of lazy evaluation) but it will need to do > some ugly check of presence of ToolRegistry class or something > similar. In this case we will produce more unimplemented calls (on > several places in the image because withToolsDo: may be handy in other > cases too) > > - something different :) > > Cheers, > -- Pavel >

July 24, 2014

ConfigurationOfMorphicCore

by Pavel Krivanek

Hi, in latest update we integrated several patches needed for proper working of ConfiguratonOfMorphicCore. This simple configuration loads Morphic-Core package into the minimal image and initializes it so the result is a small standalone Morphic subset. But some issues that I would like to discuss with the community left: SystemWindows support ============================ SystemWindow is quite complex morph that requires many of other morph classes including text morphs, theming support etc. But in several situations the system windows play special role and the Morphic core classes need to be aware of them. So we have several options: - (current) do not have SystemWindow & co. in the image but let basic support of them. That will produce some unimplemented calls - create very simple SystemWindow superclass that will be part of the Morphic-Core but will not have look and possibilities of real SystemWindow - include SystemWindow and supporting classes Tools ============================ The current image generated by this configuraton will crash as soon as you try to press some keyboard button. The reason is in the method Morph>># handleKeystroke. This method contains this code to support Keymapping Smalltalk tools shortcuts handleKeystroke: anEvent inMorph: self. The #tools message is part of Tool-Base package and need to be there because it uses lazy evaluation that creates a ToolRegistry instance. Because Tool-Base is not loaded in the minimal nor Morphic-Core image, it raises DNU Solutions: - introduce some safer form like Smalltalk withToolsDo: [:tools | ...]. In this case it will not be able to look if the class variable "Tools" is nil (because of lazy evaluation) but it will need to do some ugly check of presence of ToolRegistry class or something similar. In this case we will produce more unimplemented calls (on several places in the image because withToolsDo: may be handy in other cases too) - something different :) Cheers, -- Pavel

July 24, 2014

Re: [Pharo-dev] File Outs: Can We Have Unix Line Endings?

by Damien Cassou

On Thu, Jul 24, 2014 at 9:32 AM, HernÃ¡n Morales Durand <hernan.morales(a)gmail.com> wrote: > Why you cannot detect OS and then format accordingly? OSPlatform current lineEnding -- Damien Cassou http://damiencassou.seasidehosting.st "Success is the ability to go from one failure to another without losing enthusiasm." Winston Churchill

July 24, 2014

Re: [Pharo-dev] Refactor Nautilus

by Damien Cassou

please On Thu, Jul 24, 2014 at 11:49 AM, Nicolai Hess <nicolaihess(a)web.de> wrote: > 2014-07-24 10:28 GMT+02:00 Baptiste Quide <baptiste.quide(a)inria.fr>: > >> Good luck for that. >> I will have a look because the tool which allows to watch package >> dependencies embeds the Nautilus Browser, and i'm one of the maintainer at >> this time. > > > Good to know! > >> >> >> Baptiste >> >> ________________________________ >> >> De: "Nicolai Hess" <nicolaihess(a)web.de> >> Ã: "Pharo Development List" <pharo-dev(a)lists.pharo.org> >> EnvoyÃ©: Jeudi 24 Juillet 2014 10:23:15 >> Objet: [Pharo-dev] Refactor Nautilus >> >> >> If no one raises objections, I would like to start on refactor / cleanup >> nautilus code. >> >> This includes: >> >> renaming / recategorizing >> solve code critics shown in Critics Browser >> split classes (AbstractNautilusUI has 427 methods 24 instvars ) >> stronger separation between the browser model, browser state >> and browser UI. >> >> review event handling and UI updating >> (sometimes there is a difference between what the UI shows >> as selected and the selection the model(s) holds >> >> maybe: >> create one package pane widget for package and groups >> (therefore move all that package vs groups handling from >> nautilus ui to that widget) >> merge Nautilus and PackageTreeNautilus >> >> (I don't consider all of Nautilus bad code, it has well >> designed parts, it just has grown meanwhile...) >> >> How to update the code? >> >> Working with slices makes it easier to find changes >> that introduces new bugs. >> Otherwise it takes time to do it in such small steps and I don't >> know if those changes are reviewed at all. >> >> So, I would try to group the changes. >> 1. code critics >> 2. renameing/ simple refactoring >> 3. .... >> >> >> nicolai >> >> >> >> > -- Damien Cassou http://damiencassou.seasidehosting.st "Success is the ability to go from one failure to another without losing enthusiasm." Winston Churchill

July 24, 2014

Re: [Pharo-dev] Komitter in 30852?

by Stephan Eggermont

Marcus wrote >https://pharo.fogbugz.com/f/cases/13539/receiver-of-NodeModel-is-Nil > >There is even a fix, but it needs to be reviewed and back ported. Thanks, I somehow missed that. I was a bit annoyed that I had just yesterday answered a question about closing lots of inspectors and then ran into the situation myself... Thierry wrote: >can you describe how you get it? I immediately had errors on 4.0 >when writing this fix (https://pharo.fogbugz.com/f/cases/13539/receiver-of-NodeModel-is-Nil) but it wasn't reproduced on 3.0. Open a fresh 30852 with 355 (Mac Mavericks 10.9.4). Open a Nautilus browser, create class Bla in category Bla. Open Komitter, 6 MNUs Close MNUs and Komitter add a class Blah in category Bla Open Komitter, 10 MNUs I tried it originally on an image that took over a minute to produce the MNUs... Stephan

July 24, 2014

Re: [Pharo-dev] Komitter in 30852?

by Thierry Goubier

Stephan, can you describe how you get it? I immediately had errors on 4.0 when writing this fix ( https://pharo.fogbugz.com/f/cases/13539/receiver-of-NodeModel-is-Nil) but it wasn't reproduced on 3.0. Thierry 2014-07-24 11:43 GMT+02:00 Stephan Eggermont <stephan(a)stack.nl>: > Anyone tried Komitter recently in Pharo 3? I get about 4 to 6 debugger > windows per change. > Don't try on an image with lots of open changes... > > Stephan > > > >

July 24, 2014

Re: [Pharo-dev] Komitter in 30852?

by Marcus Denker

On 24 Jul 2014, at 11:43, Stephan Eggermont <stephan(a)stack.nl> wrote: > Anyone tried Komitter recently in Pharo 3? I get about 4 to 6 debugger windows per change. > Don't try on an image with lots of open changesâ¦ > https://pharo.fogbugz.com/f/cases/13539/receiver-of-NodeModel-is-Nil There is even a fix, but it needs to be reviewed and back ported.

July 24, 2014

Re: [Pharo-dev] Refactor Nautilus

by Nicolai Hess

2014-07-24 10:28 GMT+02:00 Baptiste Quide <baptiste.quide(a)inria.fr>: > Good luck for that. > I will have a look because the tool which allows to watch package > dependencies embeds the Nautilus Browser, and i'm one of the maintainer at > this time. > Good to know! > > Baptiste > > ------------------------------ > > *De: *"Nicolai Hess" <nicolaihess(a)web.de> > *Ã: *"Pharo Development List" <pharo-dev(a)lists.pharo.org> > *EnvoyÃ©: *Jeudi 24 Juillet 2014 10:23:15 > *Objet: *[Pharo-dev] Refactor Nautilus > > > If no one raises objections, I would like to start on refactor / cleanup > nautilus code. > > This includes: > > renaming / recategorizing > solve code critics shown in Critics Browser > split classes (AbstractNautilusUI has 427 methods 24 instvars ) > stronger separation between the browser model, browser state > and browser UI. > > review event handling and UI updating > (sometimes there is a difference between what the UI shows > as selected and the selection the model(s) holds > > maybe: > create one package pane widget for package and groups > (therefore move all that package vs groups handling from > nautilus ui to that widget) > merge Nautilus and PackageTreeNautilus > > (I don't consider all of Nautilus bad code, it has well > designed parts, it just has grown meanwhile...) > > How to update the code? > > Working with slices makes it easier to find changes > that introduces new bugs. > Otherwise it takes time to do it in such small steps and I don't > know if those changes are reviewed at all. > > So, I would try to group the changes. > 1. code critics > 2. renameing/ simple refactoring > 3. .... > > > nicolai > > > > >

July 24, 2014

Re: [Pharo-dev] Komitter in 30852?

by stepharo

I just tried and I got some problems in 40 too. No time to look deeper into it. Now in 30 this is strange because I used it when working on moose. Stef On 24/7/14 11:43, Stephan Eggermont wrote: > Anyone tried Komitter recently in Pharo 3? I get about 4 to 6 debugger windows per change. > Don't try on an image with lots of open changes... > > Stephan > > > > >

July 24, 2014