an odd result with a python dependency

[galtgendo]

I'm not sure whether or not this would be a valid bug, but it's an odd result nevertheless.

For reasons, I wanted to check the state of my currently installed tree, not any particular repo.

So, I've changed self.repo = self.pm.repositories[repo_name] to self.repo = self.pm.installed in depgraph.py, then I've run qlist -Iv | PYTHONPATH=${PWD} ./depgraph.py.

Now, I haven't even really started on analyzing the result, but I've noticed something peculiar.

Among my packages there's net-wireless/crda.
The produced graph lists all of its deps except for dev-python/m2crypto.
m2crypto is only a BEPEND, but virtual/pkgconfig is listed...
Do you think that's a bug ?

[thesamesam]

Could you upload a bad graph and the full depgraph.py used to generate it (or a patch to apply)? Tree state would be useful as well (commit and so on).

Even better would be a set of steps to hit it from a clean stage3, but that might be pushing it. But yes, that sounds odd.

[galtgendo]

gpyutuils were at 67ea488 - the only change was that single line.
This is the graph: dep.graph.txt.
Given that it was generated from the current state, the other part is effectively unattainable - as you may see, there are quite a few packages that (quite likely) don't exist outside my private repo (for example the updated dev-ruby/ruby-gnome subpackges).

[galtgendo]

The command line may look a bit confusing - I've dropped master tarballs for gpyutils and gentoopm under one dir, then symlinked the script two levels up.

[mgorny]

Could you also attach /var/db/pkg/net-wireless/crda*? Or at least the *DEPEND files there.

[galtgendo]

...on unrelated note: do you also find 'limit by file extension' github misfeature for attachments 200% retarded ?
cdra-portage.tar.xz.zip

..though my minor mistake at creation: it's actually a tar.bz2 (I've typed 'j' instead of 'J')