[Glass] Another metacello loading puzzle: xml

[Dale Henrichs via Glass]

Iwan,

Ah, I found it. In the method 
MetacelloPackagesSpec>>sortPackageSpecs:packageSpec:groupLoops:for: 
there is a `false ifTrue:` ... changing the `false` to `true` will dump 
additional logging to the Transcript ... If the additional debugging 
info doesn't help you figure out your circularity then send me a copy of 
the Transcript and I'll take a crack at decoding it ..

Dale


On 01/31/2018 10:31 AM, Dale Henrichs wrote:
> Iwan,
>
> The first thing I would ask you to do is to validate your baseline 
> with the following:
>
>   MetacelloToolBox validateBaseline: BaselineOf???
>
> given that you have a circularity in your dependencies, the validate 
> may not complete either, but if you do get some issues reported fixing 
> those issues might clear up the circularity.
>
> It's been awhile since I've had to debug a circularity and I don't 
> remember the drill of the top of my head ... so I will need to spend a 
> little time refreshing my memory as to the best way to solve the 
> problem ... I don't think that I have an explicit validation for 
> circular dependencies and this may be an opportunity to add one ...
>
> Dale
>
>
> On 01/31/2018 07:05 AM, Iwan Vosloo via Glass wrote:
>> Hi Dale,
>>
>> On 26/01/2018 15:58, Dale Henrichs via Glass wrote:
>>>> I am ignoring pharo for the moment so that I can first sort out 
>>>> issues in our own code (this also upgrades seaside and magritte and 
>>>> the XML projects for us)..... and will then bug the Pharo-users 
>>>> list when I am ready to tackle that issue.
>>> It is likely to be a Metacello/Monticello issue in pharo as well --- 
>>> the actual error and stack would be critical in determining this --- 
>>> and if it is I am probably the best person to handle it for Pharo as 
>>> well :)
>>
>> I am now turning to Pharo... and I managed to get Pharo 3.0 and 6.1 
>> to load what I can load in Gemstone3.4 with/without a network 
>> connection, BUT... I now have issues loading our own project. Our own 
>> project is loaded via a BaselineOf which only contains internal 
>> packages on the local filetree repo.
>>
>> It breaks with a 'Apparent loop in before/after dependency 
>> definitions' while trying to do 
>> MetacelloMCVersionSpec(MetacelloVersionSpec)>>packageSpecsInLoadOrder
>>
>> I have put the (very short transcript) and stack from Pharo6.1 on 
>> pastebin [1]. The MetacelloPackagesSpec asString is also useful - but 
>> it contains information I'd rather keep confidential, so I will email 
>> that off-list.  (I have also pushed the dependencies as per the 
>> output of that MetacelloPackagesSpec asString through a topological 
>> sort and found no circularity.)
>>
>> Can you shed any light?
>>
>> [1] https://pastebin.com/hUsQ4eSK
>>
>>
>> Thanks
>> Iwan
>>
>>
>