Re: Bug#961224: python-cobra: test failures on s390x

Hi,

On Sat, May 23, 2020 at 02:56:22PM +0300, Adrian Bunk wrote:
> [ debian-s390 added ]
> 
> On Thu, May 21, 2020 at 05:03:19PM +0100, peter green wrote:
> > Source: python-cobra
> > Version: 0.18.0-1
> > Severity: serious
> > 
> > Three of the last four builds of python-cobra on s390x have failed with.
> > 
> > > =================================== FAILURES ===================================
> > > ___________ test_model_summary_to_frame_with_fva[optlang-glpk-0.95] ____________
> > > 
> > > model = <Model e_coli_core at 0x3ff6e63b3d0>, opt_solver = 'optlang-glpk'
> > > fraction = 0.95
> > > 
> > >      @pytest.mark.parametrize("fraction", [0.95])
> > >      def test_model_summary_to_frame_with_fva(model, opt_solver, fraction):
> > >          """Test model summary.to_frame() (using FVA)."""
> > >          if opt_solver == "optlang-gurobi":
> > >              pytest.xfail("FVA currently buggy")
> > >          # test non-fva version (these should be fixed for textbook model)
> > >          expected_in_fluxes = ['o2_e', 'glc__D_e', 'nh4_e', 'pi_e', np.nan,
> > >                                np.nan, np.nan, np.nan, np.nan, np.nan, np.nan,
> > >                                np.nan]
> > >          expected_out_fluxes = ['h2o_e', 'co2_e', 'h_e', 'for_e', 'ac_e', 'acald_e',
> > >                                 'pyr_e', 'etoh_e', 'lac__D_e', 'succ_e', 'akg_e',
> > >                                 'glu__L_e']
> > >          model.solver = opt_solver
> > >          solution = model.optimize()
> > >          out_df = model.summary(solution, fva=fraction).to_frame()
> > >          assert out_df[('IN_FLUXES', 'ID')].tolist() == expected_in_fluxes
> > > >       assert out_df[('OUT_FLUXES', 'ID')].tolist() == expected_out_fluxes
> > > E       AssertionError: assert ['h2o_e', 'co... 'pyr_e', ...] == ['h2o_e', 'co2...acald_e', ...]
> > > E         At index 5 diff: 'pyr_e' != 'acald_e'
> > > E         Use -v to get the full diff
> > 
> 
> Can an s390x porter please have a look?

The issue remains in next upstream version.  I will exclude s390x from
the list of architectures since it seems nobody manages to care for this
issue.

Kind regards

        Andreas.