http://mxr.mozilla.org/mozilla-central/source/build/virtualenv/populate_virtualenv.py#116 We set --system-site-packages which puts the system packages first in sys.path . This causes bugs like bug 907476 and otherwise makes behaviour less encapsulated and predictable. We should probably not set this. If we do set this, the reason should at least be documented IMHO.

This was added in bug 758739.

Yeah. I think the evident ways this can break things are argument enough to remove --with-system-ply.

I'd rather have a different fix for bug 758739 than removing --with-system-ply

Let's detect where ply.py is and add it to the virtualenv's sys.path somehow?

I think --system-ply being the thing that's forcing us to use system site-packages in our virtualenv is silly. We /really/ need to stop the virtualenv from using the system's site-packages. I don't think there's much room for debate. The way I see it, we have two options: 1) Kill --system-ply entirely. 2) Provide a backdoor for the configuration to inject its own paths into the virtualenv. e.g. --python-include-path=/path/to/some/other/path I would prefer we go with #1 so the virtualenv is reasonably guaranteed to be isolated from the host system. But, I'd settle for #2. AFAICT Mike Hommey is the only person who wants --system-ply. He possibly represents the larger Debian/packaging population for downstream distros. Mike: what are your actual requirements here?

I don't want ply to be shipped in the sdk shipped in the corresponding debian package. While I could just strip those file, i also want to ensure the sdk is not broken as a result.

(In reply to Mike Hommey [:glandium] from comment #6) > I don't want ply to be shipped in the sdk shipped in the corresponding > debian package. While I could just strip those file, i also want to ensure > the sdk is not broken as a result. I'm still trying to understand why we (upstream Mozilla) should care. This seems like a trivial downstream action (only wanted by Debian)? It's causing us hardship, so I'm inclined to punt the problem onto the people who have created it for themselves. I'm being overly generous in giving you the opportunity to convince me otherwise. I'm just not hearing anything...

This patch removes --system-ply from configure and removes system site packages from the virtualenv. Patch itself is trivial. Hard part is determining whether to unsupport --system-ply people.