Question about Universes

> On Wed, Feb 18, 2009 at 5:11 PM, Alexandre Bergel
> <[hidden email]> wrote:
>> Omnibrowser gets installed, where it should not. Where is the
>> definition of this Universe stored?
>
> Without thinking and checking: ShoutOmnibrowser depends on
> Omnibrowser, that's why OB is installed.
>
> --
> Damien Cassou
> http://damiencassou.seasidehosting.st
>
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>

>
>>> Omnibrowser gets installed, where it should not. Where is the
>>> definition of this Universe stored?
>>
>> Without thinking and checking: ShoutOmnibrowser depends on
>> Omnibrowser, that's why OB is installed.
>
> What I don't understand: Why does Universe load the OmniBrowser  
> dependency when there
> is already a newer version of OmniBrowser installed in this image? I  
> think it
> shouldn't do that. Does Universe not check whether the version  
> number installed is
> higher than the latest one defined in Universe? Wouldn't it make  
> sense to do this?
>
> David
>
>
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>

>> You are wrong about Sake I think.
>
> I know Sake. I've written a generator for builder.seaside.st. The
> result (or a modified version of it) is included with the Sake
> distribution as Seaside29Builder.
>
>> Sake can declare a universe of
>> known-to-work-together packages. Just take a class, declare the sake
>> tasks you know work together in the image you target and things will
>> be as in universe.
>
> Sure, you can use Sake to build something very similar to a Universe.
> However tasks are not declarative, but instead use a script to perform
> some actions. In most cases they call Installer to find and load an
> appropriate version (what is already scary in itself).
>
> I understand that Package Universe is too restrictive for some cases,
> however I wouldn't dare to replace it with Sake. Sake depends on a
> stack of hacks, it even uses its own compiler to allow uppercase
> method names. In the small codebase Code Critics finds 14 serious bugs
> like non existing inst-variable references and message sents that are
> not implemented anywhere. Of course there is not a single test. The
> way Sake calculates dependencies is totally strange, I am not sure if
> it is even correct.
>
> Lukas
>
> --
> Lukas Renggli
> http://www.lukas-renggli.ch
>
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>

>> You are wrong about Sake I think.
>
> I know Sake. I've written a generator for builder.seaside.st. The
> result (or a modified version of it) is included with the Sake
> distribution as Seaside29Builder.
>
>> Sake can declare a universe of
>> known-to-work-together packages. Just take a class, declare the sake
>> tasks you know work together in the image you target and things will
>> be as in universe.
>
> Sure, you can use Sake to build something very similar to a Universe.
> However tasks are not declarative, but instead use a script to perform
> some actions. In most cases they call Installer to find and load an
> appropriate version (what is already scary in itself).
>
> I understand that Package Universe is too restrictive for some cases,
> however I wouldn't dare to replace it with Sake. Sake depends on a
> stack of hacks, it even uses its own compiler to allow uppercase
> method names. In the small codebase Code Critics finds 14 serious bugs
> like non existing inst-variable references and message sents that are
> not implemented anywhere. Of course there is not a single test. The
> way Sake calculates dependencies is totally strange, I am not sure if
> it is even correct.
>
> Lukas
>
> --
> Lukas Renggli
> http://www.lukas-renggli.ch
>
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>

> Thanks Lukas for sharing your experience with us.
> I have a naive question. Personally, I like the way packages get
> installed using class side methods on ScriptLoader. I have similar
> scripts for Mondrian and other projects.
> Is there something not convenient in that?
>
> Cheers,
> Alexandre
>
>
> On 19 Feb 2009, at 14:38, Lukas Renggli wrote:
>
>>> You are wrong about Sake I think.
>>
>> I know Sake. I've written a generator for builder.seaside.st. The
>> result (or a modified version of it) is included with the Sake
>> distribution as Seaside29Builder.
>>
>>> Sake can declare a universe of
>>> known-to-work-together packages. Just take a class, declare the sake
>>> tasks you know work together in the image you target and things will
>>> be as in universe.
>>
>> Sure, you can use Sake to build something very similar to a Universe.
>> However tasks are not declarative, but instead use a script to perform
>> some actions. In most cases they call Installer to find and load an
>> appropriate version (what is already scary in itself).
>>
>> I understand that Package Universe is too restrictive for some cases,
>> however I wouldn't dare to replace it with Sake. Sake depends on a
>> stack of hacks, it even uses its own compiler to allow uppercase
>> method names. In the small codebase Code Critics finds 14 serious bugs
>> like non existing inst-variable references and message sents that are
>> not implemented anywhere. Of course there is not a single test. The
>> way Sake calculates dependencies is totally strange, I am not sure if
>> it is even correct.
>>
>> Lukas
>>
>> --
>> Lukas Renggli
>> http://www.lukas-renggli.ch
>>
>> _______________________________________________
>> Pharo-project mailing list
>> [hidden email]
>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>>
>
> --
> _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:
> Alexandre Bergel  http://www.bergel.eu
> ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;.
>
>
>
>
>
>
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>

>> You are wrong about Sake I think.
>>    
>
> I know Sake. I've written a generator for builder.seaside.st. The
> result (or a modified version of it) is included with the Sake
> distribution as Seaside29Builder.
>
>  
>> Sake can declare a universe of
>> known-to-work-together packages. Just take a class, declare the sake
>> tasks you know work together in the image you target and things will
>> be as in universe.
>>    
>
> Sure, you can use Sake to build something very similar to a Universe.
> However tasks are not declarative, but instead use a script to perform
> some actions. In most cases they call Installer to find and load an
> appropriate version (what is already scary in itself).
>  

> > I don't think it is fair to use code critics as an argument.
>
> I think it's perfectly fair, be it code critic or lint, if it finds  
> bad code then it finds bad code.  It's not like you don't know about  
> code critics and Lint is already in Squeak, so it's your choice if  
> you choose not to avail yourself of tools that point out badly  
> written code.  He writes about the tools he's working on at his blog http://www.lukas-renggli.ch/blog 
>  which is easily subscribed to.  There's no reason anyone can say  
> they're unaware of what he's doing unless they're choosing to be  
> unaware.
>
> Ramon Leon
> http://onsmalltalk.com
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

>
>> In the small codebase Code Critics finds 14 serious bugs
>> like non existing inst-variable references and message sents that are
>>
> I don't think it is fair to use code critics as an argument. To me it
> appears that you have developed all of these tools for your own use. I
> have never even seen you announce their existence to squeak-dev.  
> Where
> is the documentation?
>> not implemented anywhere. Of course there is not a single test.
> What are you talking about: Sake-Tests.
>> The
>> way Sake calculates dependencies is totally strange, I am not sure if
>> it is even correct.
>>
> Again, what do you mean? There are two algorithms in use. The first is
> lifted directly from rake. The second is stolen directly from  
> universes.
>
> Keith
>
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>

> Lukas Renggli wrote:
>>> You are wrong about Sake I think.
>>>
>>
>> I know Sake. I've written a generator for builder.seaside.st. The
>> result (or a modified version of it) is included with the Sake
>> distribution as Seaside29Builder.
>>
>>
>>> Sake can declare a universe of
>>> known-to-work-together packages. Just take a class, declare the sake
>>> tasks you know work together in the image you target and things will
>>> be as in universe.
>>>
>>
>> Sure, you can use Sake to build something very similar to a Universe.
>> However tasks are not declarative, but instead use a script to  
>> perform
>> some actions. In most cases they call Installer to find and load an
>> appropriate version (what is already scary in itself).
>>
> Why?
>
> They are "mostly" declarative, following feedback from Andreas.  If  
> the
> default script is used it simply analyses the url data given, which is
> the case for 99% of packages.
>> I understand that Package Universe is too restrictive for some cases,
>> however I wouldn't dare to replace it with Sake. Sake depends on a
>> stack of hacks, it even uses its own compiler to allow uppercase
>> method names.
> No it doesnt.
>
> It doesn't depend upon that compiler hack. That hack provides the
> ability to put textual data in as an appendix to a method. It is an
> entirely independent facility that could be perhaps better included  
> as a
> trait for those that want it.
>
> I use it when I want to use methods to be a mini database, in the  
> Mantis
> package, and for Bob to be able to manages build scripts in the image
> without having to enclose in quotes and escape quotes.
>> In the small codebase Code Critics finds 14 serious bugs
>> like non existing inst-variable references and message sents that are
>>
> I don't think it is fair to use code critics as an argument. To me it
> appears that you have developed all of these tools for your own use. I
> have never even seen you announce their existence to squeak-dev.  
> Where
> is the documentation?
>> not implemented anywhere. Of course there is not a single test.
> What are you talking about: Sake-Tests.
>> The
>> way Sake calculates dependencies is totally strange, I am not sure if
>> it is even correct.
>>
> Again, what do you mean? There are two algorithms in use. The first is
> lifted directly from rake. The second is stolen directly from  
> universes.
>
> Keith
>
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>