Metacello update

I am in the process of cleaning this up. Right now I have released
ConfigurationOfSciSmalltalk with target version 1.02.
I reorganized some of the packages to try to separate dependencies which
had been preventing me from a clean load by packages. It's almost there,
right now DHB-Numerical is still including its extensions for Complex
and Quaternion, and raises a warning during load.

I tested this using Gofer in 2.0 and 1.4, and ammended the README file
to reflect the current situation. All but one test are green in both
clean loads (ImplicitSystemTest should fail for the time being.)

Dan Uber

On 06/05/2012 04:04 PM, Daniel Uber wrote:
> I am in the process of cleaning this up. Right now I have released
> ConfigurationOfSciSmalltalk with target version 1.02.
> I reorganized some of the packages to try to separate dependencies
> which had been preventing me from a clean load by packages. It's
> almost there, right now DHB-Numerical is still including its
> extensions for Complex and Quaternion, and raises a warning during load.
>
It looks like this is taken care of. I have tested this in several fresh
images. I am able to get a clean load package by package. I kept this as
version 1.02, and updated the readme file to reference #loadDevelopment

I removed Math-All package from Github, and don't recommend loading it
from squeaksource any longer.

Dan Uber

On Wed, Jun 6, 2012 at 4:04 AM, Daniel Uber <[hidden email]> wrote:
Sorry Dan, but I'm a bit lost with all these packages ...
I update from time to time from the github repo, but now i have
a lot duplicated packages. For example: Math-ComplexExtensions and
Math-Complex-Extensions

Why you separate the extensions in separate packages ?
I will try to reload everything in a fresh image.

Regards,
--
Serge Stinckwich
UMI UMMISCO 209 (IRD/UPMC), Hanoi, Vietnam
Every DSL ends up being Smalltalk
http://doesnotunderstand.org/

> Sorry Dan, but I'm a bit lost with all these packages ...
> I update from time to time from the github repo, but now i have
> a lot duplicated packages. For example: Math-ComplexExtensions and
> Math-Complex-Extensions
>
> Why you separate the extensions in separate packages ?
> I will try to reload everything in a fresh image.
>
> Regards,
Serge,

It may be that I just don't understand how to separate/package things
properly. The issue was that Math-Complex-Extensions was being included
into Math-Complex even though it had it's own extension and its own
package. This was causing Math-Complex to warn that it depended on
DhbVector since there were extensions to vector in the
Complex-Extensions. Changing the name to eliminate the hyphen corrected
the issue, now the extensions are only included in
Math-ComplexExtensions. Loading DHB first didn't help, since Vector had
extensions adaptToComplex:andSend: and adaptToQuaternion:andSend:, so it
warned that it depended on Complex and Quaternion packages. I did a
similar thing with the Tests (change the name of the package to prevent
inclusion/duplication by Monticello). I really prefer not to have
warnings during the load, since I'm not certain how the compiler deals
with references to classes that don't exist, and they're avoidable.
Werner hinted that the hyphen had special meaning to Monticello, and
that's why I removed the final hyphen. I think I'm just working around a
(quirk? feature?) in monticello.

Would setting a class method in Complex
  vectorClass
^DhbVector

solve this issue, or would the compiler still see a reference to a
non-existent class and warn?

Reloading into a fresh image does correct the small nuisance of having
duplicated packages. I've been reloading into a fresh image regularly to
ensure the load process is correct, and that nothing was lost in the
process. I also like to do this since I can have some leaks if I load
other packages (Moose includes a remove:at: method on OrderedCollection,
which if you load SciSmalltalk, then Moose, then commit the DHB package,
will be removed since the protocol changed when Moose loaded.) Lastly,
it's the best way to make sure the load is good on 1.4 and 2.0.

It may be that revisiting the packaging and naming is necessary again in
the future. In the meantime, I'm glad that this is working properly, and
the developerSupport methods in ConfigurationOfSciSmalltalk are easing
maintenance of changes a bit.