Error updating universe list

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
19 messages Options
Reply | Threaded
Open this post in threaded view
|

Error updating universe list

Torsten Bergmann
Giuseppe wrote:
>I'm getting this error updating universe from network from different  
>images.

Looks like "Seaside-HTML5-beta version 2.8.411.5" (http://www.squeaksource.com/Seaside/HTML5-pmm.5.mcz) was added with a conflict - but the package defined in the conflict is empty:

<package>
<name>Seaside-HTML5-beta</name>
<version>2.8.411.5</version>
<category>Web Development beta</category>
<description>HTML5 support moved out of Seaside2.8</description>
<url>http://www.squeaksource.com/Seaside/HTML5-pmm.5.mcz</url>
<homepage>http://www.seaside.st/</homepage>
<maintainer>[hidden email]</maintainer>
<provides></provides>
<depends>
<packagename>Seaside-beta</packagename></depends>
<conflicts>
<packagename></packagename></conflicts></package>

--
Ist Ihr Browser Vista-kompatibel? Jetzt die neuesten
Browser-Versionen downloaden: http://www.gmx.net/de/go/browser

Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

Giuseppe
But I don't have it installed.

Fails on checking the list then?

El 25/07/2007, a las 22:44, Torsten Bergmann escribió:

> Giuseppe wrote:
>> I'm getting this error updating universe from network from different
>> images.
>
> Looks like "Seaside-HTML5-beta version 2.8.411.5" (http://
> www.squeaksource.com/Seaside/HTML5-pmm.5.mcz) was added with a  
> conflict - but the package defined in the conflict is empty:
>
> <package>
> <name>Seaside-HTML5-beta</name>
> <version>2.8.411.5</version>
> <category>Web Development beta</category>
> <description>HTML5 support moved out of Seaside2.8</description>
> <url>http://www.squeaksource.com/Seaside/HTML5-pmm.5.mcz</url>
> <homepage>http://www.seaside.st/</homepage>
> <maintainer>[hidden email]</maintainer>
> <provides></provides>
> <depends>
> <packagename>Seaside-beta</packagename></depends>
> <conflicts>
> <packagename></packagename></conflicts></package>
>
> --
> Ist Ihr Browser Vista-kompatibel? Jetzt die neuesten
> Browser-Versionen downloaden: http://www.gmx.net/de/go/browser
>


Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

keith1y
I deleted the offending package with a view to re-submitting it later.

I have been discussing with Lex the possibility of relaxing the
"editorial policy" on the development universe so that anyone who has
contributed a package can update other packages, so that the universe
can be socially policed.

Any thoughts or votes from users of the 'development' universe?

Keith

p.s. Installer now supports package universes including the
#answer:with: idiom for interaction free loading.


Installer universe
     answer: '*some dialog text* with: 'my answer';
     install: 'a package';
       install: 'b package(1.2.3)';
    install.

 

Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

NorbertHartl
On Wed, 2007-07-25 at 22:51 +0100, Keith Hodges wrote:
> I deleted the offending package with a view to re-submitting it later.
>
> I have been discussing with Lex the possibility of relaxing the
> "editorial policy" on the development universe so that anyone who has
> contributed a package can update other packages, so that the universe
> can be socially policed.
>
> Any thoughts or votes from users of the 'development' universe?

I like the universe stuff. Coming from a linux background using debian
for years it seems quite natural to have something like the universe.
What is unclear to me is what I can expect from a universe. It's noted
that things in the universe "should" work together. But how is it tested
that it is working? It is great to have dependencies which are being
resolved automatically. But then there are only a few combinations of
packages that work well together. I'm missing (of course) something
like the channels of stability which you have in debian.

I had a look at the universe editor. It is great the browser comes with
an editor and eases the way of updating the universe. The editor could
be a source of stability marks, too. Wouldn't it be feasible to vote for
stability or veto it? If I installed anything from the universe I
should be able to judge about stability. Most important in my opinion
is the ability to veto a combination (to mark it as broken).
On the other side it might be helpful to harvest additional data at the
moment of voting. One package that has dependencies builds a cluster
of packages. I think the "should work together" should be meant across
all clusters of packages which are in the universe. So, if I vote for
some package it is important to know what other packages are installed.
Imagine there is another relationship in the universe beside the
"depends on" relationship name it "known to work". That would build
something like a stable channel automatically. This would make it
possible that universe can propose versions to install based on my
selection of packages.

My two cents.

Norbert


Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

Damien Cassou-3
In reply to this post by keith1y
2007/7/25, Keith Hodges <[hidden email]>:
> I have been discussing with Lex the possibility of relaxing the
> "editorial policy" on the development universe so that anyone who has
> contributed a package can update other packages, so that the universe
> can be socially policed.
>
> Any thoughts or votes from users of the 'development' universe?

I agree with this. It's frustrating to me to always have to ask Lex
when I want to give a package to someone else as recently happened
with you and David.

Moreover, I tend to always prefer bazaar-like process where everyone
can do whatever he wants and the next user can undo the previous
commit (as in wikipedia, seaside...). I would like to try that kind of
process for the future release of Squeak but I may not have support
from the foundation and I'm not in any release team.

--
Damien Cassou

Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

Edgar J. De Cleene



El 7/26/07 7:20 AM, "Damien Cassou" <[hidden email]> escribió:

> Moreover, I tend to always prefer bazaar-like process where everyone
> can do whatever he wants and the next user can undo the previous
> commit (as in wikipedia, seaside...). I would like to try that kind of
> process for the future release of Squeak but I may not have support
> from the foundation and I'm not in any release team.
>
> --
> Damien Cassou


Anarchy... Not to far...


Edgar



Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

Ralph Johnson
In reply to this post by NorbertHartl
On 7/26/07, Norbert Hartl <[hidden email]> wrote:

 I had a look at the universe editor. It is great the browser comes with

> an editor and eases the way of updating the universe. The editor could
> be a source of stability marks, too. Wouldn't it be feasible to vote for
> stability or veto it? If I installed anything from the universe I
> should be able to judge about stability. Most important in my opinion
> is the ability to veto a combination (to mark it as broken).
> On the other side it might be helpful to harvest additional data at the
> moment of voting. One package that has dependencies builds a cluster
> of packages. I think the "should work together" should be meant across
> all clusters of packages which are in the universe. So, if I vote for
> some package it is important to know what other packages are installed.
> Imagine there is another relationship in the universe beside the
> "depends on" relationship name it "known to work". That would build
> something like a stable channel automatically. This would make it
> possible that universe can propose versions to install based on my
> selection of packages.

It is very important to make it easier for people to give feedback on
problems so that they can be fixed.  Lex, you ought to try things like
thia!

-Ralph

Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

keith1y
May I suggest a mailing list specifically for universe content
discussion to support the opening of the "editorial policy" so as to
adopt a "social model", rather than an "anarchic model". The server
could post to the mailinglist when it receives updates and users could
post with feedback.

The posts could be flagged with the universe to which it applies in
order to support appropriate filtering.

I have discovered that Universes do not appear to support Monticello
dependencies. This means that packages such as 'Magma Server' and
variants are not likely to work out of the box.

TimeZoneDatabase is not (as yet) compatible with 3.9 nor 3.10 since It
overrides DateAndTime fromSeconds: to return a DateAndTime instance.
Users of the subclass TimeStamp fromSeconds: are expecting a TimeStamp
instance.


Keith
>  
> It is very important to make it easier for people to give feedback on
> problems so that they can be fixed.  Lex, you ought to try things like
> thia!
>
> -Ralph
>
>


Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

NorbertHartl
On Thu, 2007-07-26 at 14:05 +0100, Keith Hodges wrote:

> May I suggest a mailing list specifically for universe content
> discussion to support the opening of the "editorial policy" so as to
> adopt a "social model", rather than an "anarchic model". The server
> could post to the mailinglist when it receives updates and users could
> post with feedback.
>
> The posts could be flagged with the universe to which it applies in
> order to support appropriate filtering.
>
> I have discovered that Universes do not appear to support Monticello
> dependencies. This means that packages such as 'Magma Server' and
> variants are not likely to work out of the box.
>
> TimeZoneDatabase is not (as yet) compatible with 3.9 nor 3.10 since It
> overrides DateAndTime fromSeconds: to return a DateAndTime instance.
> Users of the subclass TimeStamp fromSeconds: are expecting a TimeStamp
> instance.
>
>
+1

This is a solution which is easy to setup. Even if my proposal
(integrating feedback in the editor/browser) could be done it needs
a lot of work. A list would make it possible to develop any model
which will be sufficient.

And I wouldn't oppose "social" and "anarchy" that much. That is a
tension I often read between the lines (not yours!). The "social"
and "anarchy" are also only shades of grey. Having rules and
permissions doesn't create a "social" thing and less restrictions
do not create anarchy per default :)

What we need is to develop a policy "how" a universe is updated. I
wouldn't dare to update a package in the universe which is maintained
by another person. But if I would know that "it is working" means that
the maintainer tested A,B and C than I can do that on my own. This
plus the knowledge that my action is undoable would probably lead me
to updating the universe on my own. And this lessens the work of the
maintainer. And I think this i wanted, too. If this would work you
would get an "anarchy without harm" :)

Another 2 cents (still got 4 left :) )

Norbert


Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

NorbertHartl
In reply to this post by keith1y
On Thu, 2007-07-26 at 14:05 +0100, Keith Hodges wrote:

> May I suggest a mailing list specifically for universe content
> discussion to support the opening of the "editorial policy" so as to
> adopt a "social model", rather than an "anarchic model". The server
> could post to the mailinglist when it receives updates and users could
> post with feedback.
>
> The posts could be flagged with the universe to which it applies in
> order to support appropriate filtering.
>
> I have discovered that Universes do not appear to support Monticello
> dependencies. This means that packages such as 'Magma Server' and
> variants are not likely to work out of the box.
>
> TimeZoneDatabase is not (as yet) compatible with 3.9 nor 3.10 since It
> overrides DateAndTime fromSeconds: to return a DateAndTime instance.
> Users of the subclass TimeStamp fromSeconds: are expecting a TimeStamp
> instance.
>
>
Btw. how is the universe XML file being updated? Is the XML file updated
as a whole? Are there any concurrency problems?

thanks,

Norbert


Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

keith1y
I propose we could simply use the seemingly dormant Packages mailing list.
Since Universes are indeed collections of packages this would seem
appropriate.

The last meaningful post in this mailing list was in Dec 2005.

To subscribe: http://lists.squeakfoundation.org/mailman/listinfo/packages

Prefix messages with the [development] universe to which you are referring.

best regards

Keith

p.s. now we just need to ask Lex to open things up a bit.

Reply | Threaded
Open this post in threaded view
|

RE: Error updating universe list

J J-6
In reply to this post by Torsten Bergmann
> Date: Thu, 26 Jul 2007 14:05:13 +0100
> From: [hidden email]
> To: [hidden email]
> Subject: Re: Error updating universe list
>
> I have discovered that Universes do not appear to support Monticello
> dependencies. This means that packages such as 'Magma Server' and
> variants are not likely to work out of the box.

How would that work?  Monticello would need some way to flag branches as "stable", "dev" and so on so that the Universe would know which branch to use.


PC Magazine’s 2007 editors’ choice for best web mail—award-winning Windows Live Hotmail. Check it out!

Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

keith1y
J J wrote:

> > Date: Thu, 26 Jul 2007 14:05:13 +0100
> > From: [hidden email]
> > To: [hidden email]
> > Subject: Re: Error updating universe list
> >
> > I have discovered that Universes do not appear to support Monticello
> > dependencies. This means that packages such as 'Magma Server' and
> > variants are not likely to work out of the box.
>
> How would that work?  Monticello would need some way to flag branches
> as "stable", "dev" and so on so that the Universe would know which
> branch to use.
I am not saying that Monticello doesnt know how to provide resolution of
Universe dependencies.

I am saying that if you install a monticello package from a url, and
that package attempts to load other 'specific' packages due to its
internal dependencies, the url loading context may have no repositories
registered in which monticello can look to find those requirements.

Keith

Reply | Threaded
Open this post in threaded view
|

RE: Error updating universe list

J J-6
In reply to this post by Torsten Bergmann
Ah right.  My bad.  But I still like the idea of the revision system providing some kind of tagging so the Universe code can automatically look for updates in specific "guaranteed  to be stable" branches.

> Date: Thu, 26 Jul 2007 19:23:56 +0100
> From: [hidden email]
> To: [hidden email]
> Subject: Re: Error updating universe list
>
> I am not saying that Monticello doesnt know how to provide resolution of
> Universe dependencies.
>
> I am saying that if you install a monticello package from a url, and
> that package attempts to load other 'specific' packages due to its
> internal dependencies, the url loading context may have no repositories
> registered in which monticello can look to find those requirements.
>
> Keith
>


Don't get caught with egg on your face.    Play Chicktionary! 

Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

Damien Cassou-3
In reply to this post by keith1y
2007/7/26, Keith Hodges <[hidden email]>:

> J J wrote:
> > > I have discovered that Universes do not appear to support Monticello
> > > dependencies. This means that packages such as 'Magma Server' and
> > > variants are not likely to work out of the box.
> >
> > How would that work?  Monticello would need some way to flag branches
> > as "stable", "dev" and so on so that the Universe would know which
> > branch to use.
> I am not saying that Monticello doesnt know how to provide resolution of
> Universe dependencies.
>
> I am saying that if you install a monticello package from a url, and
> that package attempts to load other 'specific' packages due to its
> internal dependencies, the url loading context may have no repositories
> registered in which monticello can look to find those requirements.

You just have to use Universes all the way down. All packages must be
in the universe and dependencies must be explicit in the universe.

--
Damien Cassou

Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

David T. Lewis
In reply to this post by keith1y
On Thu, Jul 26, 2007 at 02:05:13PM +0100, Keith Hodges wrote:
>
> TimeZoneDatabase is not (as yet) compatible with 3.9 nor 3.10 since It
> overrides DateAndTime fromSeconds: to return a DateAndTime instance.
> Users of the subclass TimeStamp fromSeconds: are expecting a TimeStamp
> instance.

Lex reported this bug and I fixed it back in February. The up to date
version (1.2.4) is on SqueakMap (and SqueakSource), but the Universe has
not been updated.

I guess it should be my job to fix this, is that right? If so I'll
figure it out next week and let you know when it's done (I'm out of
town and short of time right now).

Dave


Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

Damien Cassou-3
2007/7/29, David T. Lewis <[hidden email]>:

> On Thu, Jul 26, 2007 at 02:05:13PM +0100, Keith Hodges wrote:
> >
> > TimeZoneDatabase is not (as yet) compatible with 3.9 nor 3.10 since It
> > overrides DateAndTime fromSeconds: to return a DateAndTime instance.
> > Users of the subclass TimeStamp fromSeconds: are expecting a TimeStamp
> > instance.
>
> Lex reported this bug and I fixed it back in February. The up to date
> version (1.2.4) is on SqueakMap (and SqueakSource), but the Universe has
> not been updated.

Ok, I've done it. It loads fine. However, there are 22 failures when I
run the tests.

--
Damien Cassou

Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

cdavidshaffer
Damien Cassou wrote:
>
> Ok, I've done it. It loads fine. However, there are 22 failures when I
> run the tests.
>
>  
If you switch to a VM which doesn't have the TimePlugin then this should
go down to 14 FWIW.  Here are the failures I get (ignore the 1 error, it
is in my own test):

http://www.shaffer-consulting.com/david/FailuresForDave.png

All of these failures seem to be "off by an hour" issues related to DST
transitions.

David


Reply | Threaded
Open this post in threaded view
|

Re: Error updating universe list

Chris Muller-3
In reply to this post by keith1y
> I have discovered that Universes do not appear to support Monticello
> dependencies. This means that packages such as 'Magma Server' and
> variants are not likely to work out of the box.

I'm late, but just to clarify..

Magma stopped using Monticello dependencies over a year ago and has
always been designed as a one-click install that works "out of the
box".