Re: [PATCH 3/5] gnu: Enable tests in Python 3.

[Ludovic Courtès]

Cyril Roelandt <address@hidden> skribis:

> * gnu/packages/python.scm: enable tests for Python 3
> * gnu/packages/python-fix-tests.patch: New file.
> * gnu/packages/gnu-sysem.am (dist_patch_DATA): add it.

Typo here.

> diff --git a/gnu/packages/patches/python-fix-tests.patch 
> b/gnu/packages/patches/python-fix-tests.patch
> new file mode 100644
> index 0000000..23ef17f
> --- /dev/null
> +++ b/gnu/packages/patches/python-fix-tests.patch
> @@ -0,0 +1,62 @@
> +--- Lib/test/test_shutil.py  2014-03-01 04:56:37.768311000 +0100
> ++++ Lib/test/test_shutil.py  2014-03-01 03:02:36.088311000 +0100
> +@@ -1053,7 +1053,6 @@
> +         self.assertRaises(ValueError, make_archive, base_name, 'xxx')
> +

Could you add a couple of lines of summary and (possibly) and a link to
the upstream report/discussion (or a statement on the upstream status)?

> +     @requires_zlib
> +-    @unittest.skipIf(True, "getgrgid(0)[0] raises a KeyError on Guix")

It seems that this part of the patch is reversed, no?

> +--- Lib/test/test_posixpath.py       2014-03-01 05:46:56.984311000 +0100
> ++++ Lib/test/test_posixpath.py       2014-03-01 06:20:50.704311000 +0100
> +@@ -319,7 +319,11 @@
> +                 del env['HOME']
> +                 home = pwd.getpwuid(os.getuid()).pw_dir
> +                 # $HOME can end with a trailing /, so strip it (see #17809)
> +-                self.assertEqual(posixpath.expanduser("~"), 
> home.rstrip("/"))
> ++            # The Guix builders have '/' as a home directory, so
> ++            # home.rstrip("/") will be an empty string and the test will
> ++            # fail. Let's just disable it since it does not really make
> ++            # sense with such a bizarre setup.
> ++                # self.assertEqual(posixpath.expanduser("~"), 
> home.rstrip("/"))

I see that the recipe’s ‘pre-check’ phase does:

  (setenv "HOME" (getcwd))

so $HOME should actually be /tmp/nix-build-xxx, not just /, no?

> +--- Lib/test/test_socket.py  2014-03-02 22:14:12.264311000 +0100
> ++++ Lib/test/test_socket.py  2014-03-03 01:12:21.360311000 +0100
> +@@ -819,6 +819,8 @@
> +             self.assertRaises(OverflowError, socket.htonl, k)
> +             self.assertRaises(OverflowError, socket.htons, k)
> +
> ++    @unittest.skipUnless(os.path.exists("/etc/services"),
> ++                         "getservbyname uses /etc/services, which is not in 
> the chroot")

(I think that should go upstream.)

> ++    @unittest.skipUnless(os.path.exists("/etc/gai.conf"),
> ++                         "getaddrinfo uses /etc/gai.conf, which is not in 
> the chroot")

I don’t think gai.conf is the problem.  Instead, the problem is that one
cannot rely on name lookups in the build environment at all, and even in
general.

So this patch may be OK for ourselves (except the comment), but upstream
Python should really just gracefully skip the test if name lookup
fails.

> +    (arguments
> +     (let ((args `(#:modules ((guix build gnu-build-system)
> +                              (guix build utils)
> +                              (srfi srfi-1)
> +                              (srfi srfi-26))
> +                   ,@(package-arguments python-2))))
> +       (substitute-keyword-arguments args
> +         ((#:tests? _)
> +          `(list #t)))))

#t instead of `(list #t) should do the job.  :-)

Thanks for going through all this!

Ludo’.