Brett Gilio <brettg@posteo.net> writes:

> From 74c956284d5ff7b20e12ed4315404697320dc809 Mon Sep 17 00:00:00 2001
> From: Brett Gilio <brettg@posteo.net>
> Date: Fri, 24 May 2019 19:11:00 -0500
> Subject: [PATCH 4/4] gnu: python-hdf4: Remove check-phase argument.
>
> * gnu/packages/python-xyz.scm (python-hdf4): Update to .
> ---
>  gnu/packages/python-xyz.scm | 15 ---------------
>  1 file changed, 15 deletions(-)
>
> diff --git a/gnu/packages/python-xyz.scm b/gnu/packages/python-xyz.scm
> index 54cc0c2b27..43287d212e 100644
> --- a/gnu/packages/python-xyz.scm
> +++ b/gnu/packages/python-xyz.scm
> @@ -385,21 +385,6 @@ and verifies that it matches the intended target hostname.")
>      `(("hdf4" ,hdf4)
>        ("libjpeg" ,libjpeg)
>        ("zlib" ,zlib)))
> -   (arguments
> -    `(#:phases
> -      (modify-phases %standard-phases
> -        (replace 'check
> -          (lambda _
> -            ;; The 'runexamples' script sets PYTHONPATH to CWD, then goes
> -            ;; on to import numpy. Somehow this works on their CI system.
> -            ;; Let's just manage PYTHONPATH here instead.
> -            (substitute* "runexamples.sh"
> -              (("export PYTHONPATH=.*") ""))
> -            (setenv "PYTHONPATH"
> -                    (string-append (getcwd) ":"
> -                                   (getenv "PYTHONPATH")))
> -            (invoke "./runexamples.sh")
> -            (invoke "nosetests" "-v"))))))
>     (home-page "https://github.com/fhs/python-hdf4")
>     (synopsis "Python interface to the NCSA HDF4 library")
>     (description

While this makes the build succeed, the 'check' phase is now moot.  It
appears to simply build the library for a second time, and then prints
'Ran 0 tests in N seconds...'.

Can you look into what happened to the tests between version 0.9.0 and
0.9.2?