[Seaside-dev] Re: [Metacello] A newly discovered Metacello bug (yet to be characterized)

Thu Jun 2 00:49:56 UTC 2011

Ocompletion-ul.102 was loaded because ConfigurationOfPharo 1.2.2 loads ConfigurationOfOmniBrowser 1.2.5 which loads ConfigurationOfOCompletion 1.3 which is not correct for Pharo 1.2.2 ... It looks like ConfigurationOfPharo 1.2.2 was introduced for Pharo 1.2.2 and previous versions of Pharo-1.2 were using ConfigurationOfPharo 1.2-beta2 which calls for ConfigurationOfOmniBrowser 1.2.4, which is presumably the version that should be used in Pharo 1.2.

ConfigurationOfOmniBrowser 1.2.5 looks like a direct copy of ConfigurationOfOmniBrowser 1.3 and 1.3 _is_ targetted at Pharo1.3 ...

With the way things are configured, any attempt to load a configuration that either directly or indirectly references ConfigurationOfOmniBrowser will likely fail...

....Now on to characterizing the Metacello bug:)

Dale

----- Original Message -----
> The Seaside3.0.5 load failure in Pharo1.2.2-12353 is due to the fact
> that Ocompletion-ul.102 is loaded in the image which is a package
> that is part of ConfigurationOfOCompletion 1.4 (presumably targetted
> at Pharo1.3 not Pharo1.2.2) ... In the earlier version of Pharo 1.2,
> Ocompletion-ul.92 is loaded and it is part of
> ConfigurationOfOCompletion 1.2.2 ...
> 
> Not sure why Ocompletion-ul.102 was loaded as ConfigurationOfPharo
> points to the correct version of ConfigurationOfOCompletion, so I'll
> have to do more detective work...
> 
> Dale
> 
> ----- Original Message -----
> > Okay the odd behavior between Pharo1.2-12339, Pharo1.2.1-12345 and
> > Pharo1.2.2-12353 can be attributed to a known Metacello issue:
> > 
> >   http://code.google.com/p/metacello/issues/detail?id=112
> > 
> > So we are only facing two bugs:
> > 
> >   - the issue reported by Lukas, where version OmniBrowser 1.2.3 is
> >   loaded on
> >     top of version 1.2.4 (when nothing should have been loaded).
> >     Which appears
> >     to be fixed by Metacello 1.0-beta.29 (still need a test case to
> >     reproduce
> >     the bug)
> >   - the Seaside3.0.5 load failure Pharo1.2.2-12353 that doesn't
> >   show
> >   up in earlier
> >     versions. This error has to do with
> >     OcompletionPharoCompatibility
> >     and is likely
> >     due to a configuration error somewhere.
> > 
> > I'll keep on digging...
> > 
> > Dale
> > 
> > 
> > ----- Original Message -----
> > > 
> > > In the mail to Lukas below, I was able to identify a previously
> > > unreported bug in Metacello1.0-beta.28.3.3 when loading
> > > Seaside3.0.5
> > > into Pharo1.2-12339.  In the process of characterizing the bug, I
> > > discovered that the Seaside load was correct when using Metacello
> > > 1.0-beta.29 (an unreleased version).
> > > 
> > > However, when I tested Seaside3.0.5 and Metacello1.0-beta.29
> > > using
> > > Pharo1.2.2-12353, I discovered more odd behavior and Seaside3.0.5
> > > hit errors during the load. There is something fishy going on and
> > > at
> > > this point in time the root cause is not obvious.
> > > 
> > > The Metacello unit tests are passing in all three versions:
> > > Pharo1.2-12339, Pharo1.2.1-12345 and Pharo1.2.2-12353.
> > > 
> > > I am now in the process of attempting to characterize the bugs.
> > > 
> > > I will keep you posted on my progress.
> > > 
> > > Dale
> > > 
> > > ----- Original Message -----
> > > > Lukas,
> > > > 
> > > > I have looked specifically at Seaside3.0.5 and
> > > > Pharo1.2-12339/PharoCore1.2-12336
> > > >  and have found two things:
> > > > 
> > > >   1. #stable version for ConfigurationOfOmniBrowser for 1.2 is
> > > >   1.2.3
> > > >   and it should be 1.2.4.
> > > >   2. if OB 1.2.4 is already loaded then the version is
> > > >   downgraded
> > > >   to
> > > >   1.2.3
> > > > 
> > > > Item #1 is a configuration issue and item #2 is a bug ...
> > > > 
> > > > I had not heard that "many projects could no longer be loaded
> > > > into
> > > > Pharo1.2", so I was (and still am) blissfully unaware of the
> > > > problems that folks are having.
> > > > 
> > > > Seaside3.0.5 loads into Pharo1.2/PharoCore1.2 and passes all of
> > > > the
> > > > unit tests ... so there are no "load issues." The Seaside
> > > > control
> > > > panel works fine and I have not noticed any other
> > > > OmniBrowser-related issues ... but then I don't live in the
> > > > images
> > > > day in and day out so there could be subtle issues.
> > > > 
> > > > I am currently trying to characterized the "downgrade version"
> > > > bug
> > > > and will push out a fix as soon as I have one.
> > > > 
> > > > Dale
> > > >  
> > > > 
> > > > ----- Original Message -----
> > > > > > Here's the current state of 3.0.5: I saved the images from
> > > > > > Jenkins
> > > > > > and
> > > > > > uploaded them to the web site. The change log has also been
> > > > > > updated
> > > > > > [1]. Everything but the Metacello configuration seems to
> > > > > > work
> > > > > > fine.
> > > > > > If
> > > > > > I load the Metacello configuration into a Pharo 1.2 it
> > > > > > tries
> > > > > > to
> > > > > > load a
> > > > > > version of OB that does not work in Pharo 1.2.
> > > > > 
> > > > > As far as I know most configurations have been updated to
> > > > > Pharo
> > > > > 1.3
> > > > > already and many projects (I learned that with Glamour,
> > > > > Mondrian,
> > > > > Moose, Spy, ...) cannot be loaded into latest stable (Pharo
> > > > > 1.2)
> > > > > anymore. What surprised me more is that Metacello does not
> > > > > allow
> > > > > to
> > > > > load old versions (at least this is true for the
> > > > > configurations
> > > > > I
> > > > > had
> > > > > problems with). So I guess we have the same problem here.
> > > > > 
> > > > > I suggest that you load OB with this script:
> > > > > https://github.com/renggli/builder/blob/master/scripts/omnibrowser.st
> > > > > 
> > > > > Lukas
> > > > > 
> > > > > --
> > > > > Lukas Renggli
> > > > > www.lukas-renggli.ch
> > > > > _______________________________________________
> > > > > seaside-dev mailing list
> > > > > seaside-dev at lists.squeakfoundation.org
> > > > > http://lists.squeakfoundation.org/mailman/listinfo/seaside-dev
> > > > > 
> > > > _______________________________________________
> > > > seaside-dev mailing list
> > > > seaside-dev at lists.squeakfoundation.org
> > > > http://lists.squeakfoundation.org/mailman/listinfo/seaside-dev
> > > > 
> > > 
> > 
> 