On 9/11/15 6:09 AM, Usman Bhatti wrote:

Hi,

It seems that Metacello does not consider PackageName-xyz and PacakgeName-BranchName-xyz as the same package. I suspect it because when I look at the packages loaded by two different configurations with and without branch names as package name suffixes are not the same. Could someone knowing the internals of Metacello confirm it?

So, here is a log of Metacello of the loadDirective for ConfigOfGlamour with and without packages with branch names. I have truncated the output to highlight important parts of the log.

----------- Metacello log follows ----------------------

Without branching (Rubric and Glamour-Morphic-Renderer loaded only once):

linear load :

atomic load : 3.3.0 [ConfigurationOfGlamour]

linear load : 3.1.3.Moose51 [ConfigurationOfGlamourCore]

--------------------List of Packages----------------

atomic load : 1.2.14.Moose51 [ConfigurationOfRubric]

load : Rubric-AndreiChis.207

load : Glamour-Morphic-Renderer-AndreiChis.324

--------------------List of Packages----------------

linear load : 1.13 [ConfigurationOfRoassal2]

linear load : 3.1.4 [ConfigurationOfGlamourCore]

--------------------List of Packages----------------

With branching (old versions overwrites the fixes):

linear load :

atomic load : 3.3.0 [ConfigurationOfGlamour]

linear load : 3.1.3.Moose51 [ConfigurationOfGlamourCore]

--------------------List of Packages----------------

atomic load : 1.2.14.Moose51 [ConfigurationOfRubric]

load : Rubric-Moose51-usmanbhatti.207

load : Glamour-Morphic-Renderer-Moose51-usmanbhatti.324

--------------------List of Packages----------------

linear load : 1.13 [ConfigurationOfRoassal2]

linear load : 3.1.4 [ConfigurationOfGlamourCore]

atomic load : 1.2.15 [ConfigurationOfRubric]

load : Rubric-AlainPlantec.206

load : Glamour-Morphic-Renderer-AndreiChis.321

--------------------List of Packages----------------

---------------

What it actually changes is packages with branch name are overwritten with their predecessors and hence fixes are not correctly included in the resulting image. It seems that ensuring correct loading in the presence of branched packages requires no other configuration is loading previous versions of the branched packages.

While Thierry's comments (and my own) apply, I'd like to dive a little deeper into this for a minute ... I would like to compare the Metacello specs for Glamour-Morphic-Renderer in ConfigurationOfGlamourCore (that resolves to Glamour-Morphic-Renderer-Moose51-usmanbhatti.324) and the spec for Glamour-Morphic-Renderer in ConfigurationOfRubric (that resolves to Glamour-Morphic-Renderer-AndreiChis.321)...

I see that in Glamour-Morphic-Renderer-Moose51-usmanbhatti.324 the version number (324 ... which btw is another convention:) is greater than 321, so I do believe that if you had been using the branch naming conventions Glamour-Morphic-Renderer-AndreiChis.321 would not have been loaded .... and I suppose we'll find out shortly:)

I _am_ curious about the package meta information for Glamour-Morphic-Renderer-Moose51-usmanbhatti.324 ... when you crack open that package (inspect the ancestry of working copy) what package name is associated with the latest ancestor? I've seen cases where the package meta data has drifted significantly from the name of the file used to house the package and that has lead to interesting image state, to say the least.

Dale