Re: test failure with current tip

[Jaroslav Hajek]

On Wed, Jan 13, 2010 at 1:00 PM, John W. Eaton <address@hidden> wrote:
> With the current sources, I'm seeing the following failure:
>
>  test_bc_overloads.m .................................... PASS    0/1    FAIL 
> 1
>
> The info from the fntests.log file is
>
>  >>>>> processing test_bc_overloads
>    ***** test
>   f = fieldnames (ex);
>   n = numel (f);
>   s = c1 = c2 = cell (n);
>   for i = 1:n
>     for j = 1:n
>       c1{i,j} = ex.(f{i});
>       c2{i,j} = ex.(f{j});
>       s{i,j} = tbcover (ex.(f{i}), ex.(f{j}));
>     endfor
>   endfor
>   assert (cellfun (@tbcover, c1, c2, "uniformoutput", false), s);
>  !!!!! test failed
>  assert (cellfun (@tbcover, c1, c2, "uniformoutput", false),s) expected
>  {
>    [1,1] = double
>    [2,1] = single
>    ...
>  }
>  but got
>  {
>    [1,1] = none
>    [2,1] = none
>    ...
>    all the rest are the same
>    ...
>  }
>  shared variables {
>    ex =
>    {
>      double =  1
>      single =  1
>      logical =  1
>      char = char
>      int8 = 1
>      int16 = 1
>      int32 = 1
>      int64 = 1
>      uint8 = 1
>      uint16 = 1
>      uint32 = 1
>      uint64 = 1
>      cell = {}(0x0)
>      struct =
>      {
>        1x1 struct array containing the fields:
>
>      }
>
>      function_handle =
>
>  numel
>
>    }
>
>  }
>
> Is anyone else seeing this problem, or is it just me?
>
> I'm building in a separate directory from the sources and the
> @TYPE/tbcover.m files in the build-tree/test directory appear to have
> been created correctly.
>
> The failure only happens when I run "make check".  If I cd to the
> build-tree/test directory, then do
>
>  ../run-octave
>  octave> test test_bc_overloads
>
> the test passes.  So maybe it is a path problem.  Hmm, I tried
> changing the run-octave script to add the builddir/test directory to
> the path, verified that it is adding the directory correctly when run
> with "make check", but then it is still failing.
>
> OK, one more thing.  If I remove script_tree from the list of fundirs
> in test/fntests.m, then make check passes the test_bc_overloads test.
> But I have no clue why.
>
> Anyone?
>
> jwe
>

I think one problem is that build_bc_overload_tests also reads the
file test/bc_overloads_expected that contains the overload results as
computed by Matlab (using the script build_bc_overloads_expected.m,
which is included just for completeness). So, if it is run from a
different directory, it will fail to generate files correctly, which
also explains why you see just one test while there should be more
than 400. I suppose a solution is to pass the source directory to the
script. I'm not sure this explains the failure you see, however.

-- 
RNDr. Jaroslav Hajek, PhD
computing expert & GNU Octave developer
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz