[TIP] tox offline usage

holger krekel holger at merlinux.eu
Sun Jun 26 06:25:42 PDT 2011


Hi Alexander, Aaron,

On Mon, Jun 20, 2011 at 15:35 -0600, Aaron Meurer wrote:
> On Mon, Jun 20, 2011 at 11:28 AM, Alexander Steinert <sputnik at termo.net> wrote:
> > Dear list,
> >
> > I would like to use tox in an environment that has no internet access.
> > In my trivial example, I specify a dependency in the testenv section:
> >
> >    [tox]
> >    envlist = py26
> >    [testenv]
> >    deps = pytest==2.0.3
> >    commands = true
> >
> > I want tox to install the pytest-2.0.3.zip distribution that exists in
> > my file system, say ./lib/pytest-2.0.3.zip.
> >
> > With pip, I can achieve that by using --no-index and --find-links, but I
> > found no way to tell tox to call pip that way.
> >
> > Using the indexserver setting would require me to have a PyPI server
> > running in my network. I'd like to avoid that.
> >
> > Apart from patching tox' code, is there any other way to achieve my
> > goal?

you could try to copy pytest-2.0.3 and others dependencies into
$HOME/.tox/distshare - which is looked up by tox.

See the docs for more info:
http://readthedocs.org/docs/tox/en/latest/example/general.html?highlight=distshare#access-package-artifacts-between-multiple-tox-runs

> > While we're at it: When I run tox (configured as above) for the first
> > time, it fails (as expected) with "could not install deps
> > [pytest==2.0.3]". When run again, it is happily "reusing existing
> > matching virtualenv py26" and running /bin/true but silently ignoring
> > the missing dependency. Should I file a bug report?
> 
> I've also noticed this.  Is this an intended behavior?  It's let me to
> be confused a couple of times regarding dependencies that don't
> install properly (see my previous threads on this mailing list).

Would be good if one of you provides a minimal example and
post an issue on the tracker.

> By the way, if you're wondering, I would prefer that it didn't do
> this, because if it can't install all the dependencies correctly, then
> I consider that particular test environment to be broken, and the way
> it is now, it falsely makes it appear that they are not.

right.

holger



More information about the testing-in-python mailing list