V5.8.0rc

[rjleveque]

I think the remaining open PRs can wait for now and we should release v5.8.0 with several important bug fixes.

[rjleveque]

@mandli and @ketch: This seems to be failing in pyclaw with many more errors than what you saw in clawpack/pyclaw#658.

Can you take a look and see if there's a real problem or if I can ignore this?

[mandli]

It looks like there are slight differences in some of the test results. I have not see this before so it bears looking into.

[rjleveque]

I just tried running the tests on my laptop, cloning a new copy via:

git clone git://github.com/clawpack/clawpack.git clawpack_v5.8.0rc
cd clawpack_v5.8.0rc
git checkout -b rjleveque-v5.8.0rc master
git pull https://github.com/rjleveque/clawpack.git v5.8.0rc
git submodule init
git submodule update

It gives the same errors as on travis, e.g.

cd pyclaw/examples/acoustics_1d_homogeneous/
nosetests -sv .


======================================================================
ERROR: examples.acoustics_1d_homogeneous.acoustics_1d({'solver_type': 'classic', 'disable_output': True})
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/rjl/miniconda/envs/geo5/lib/python3.6/site-packages/nose/case.py", line 198, in runTest
    self.test(*self.arg)
  File "/Users/rjl/Downloads/clawpack_v5.8.0rc/pyclaw/src/pyclaw/util.py", line 174, in <lambda>
    test = lambda: test_app(application, verifier, test_kwargs)
  File "/Users/rjl/Downloads/clawpack_v5.8.0rc/pyclaw/src/pyclaw/util.py", line 262, in test_app
    raise VerifyError(err)
clawpack.pyclaw.util.VerifyError: /Users/rjl/Downloads/clawpack_v5.8.0rc/pyclaw/examples/acoustics_1d_homogeneous/acoustics_1d.py
********************************************************************************
verification function
        def acoustics_verify(claw):
            from clawpack.pyclaw.util import check_diff
            import numpy as np

            # tests are done across the entire domain of q normally
            q0 = claw.frames[0].state.get_q_global()
            qfinal = claw.frames[claw.num_output_times].state.get_q_global()

            # and q_global is only returned on process 0
            if q0 is not None and qfinal is not None:
                q0 = q0.reshape([-1])
                qfinal = qfinal.reshape([-1])
                dx = claw.solution.domain.grid.delta[0]
                test = dx*np.sum(np.abs(qfinal-q0))
                return check_diff(expected, test, abstol=1e-4)
            else:
                return

args                 : {'use_petsc': False, 'kernel_language': 'Fortran', 'solver_type': 'classic', 'disable_output': True}
norm of expected data: 0.001049
norm of test data    : 0.0017203753108528746
test error           : 0.0006713753108528746
abstol  : 0.0001
********************************************************************************


======================================================================
ERROR: examples.acoustics_1d_homogeneous.acoustics_1d({'ptwise': True, 'solver_type': 'classic', 'disable_output': True})
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/Users/rjl/miniconda/envs/geo5/lib/python3.6/site-packages/nose/case.py", line 198, in runTest
    self.test(*self.arg)
  File "/Users/rjl/Downloads/clawpack_v5.8.0rc/pyclaw/src/pyclaw/util.py", line 174, in <lambda>
    test = lambda: test_app(application, verifier, test_kwargs)
  File "/Users/rjl/Downloads/clawpack_v5.8.0rc/pyclaw/src/pyclaw/util.py", line 262, in test_app
    raise VerifyError(err)
clawpack.pyclaw.util.VerifyError: /Users/rjl/Downloads/clawpack_v5.8.0rc/pyclaw/examples/acoustics_1d_homogeneous/acoustics_1d.py
********************************************************************************
verification function
        def acoustics_verify(claw):
            from clawpack.pyclaw.util import check_diff
            import numpy as np

            # tests are done across the entire domain of q normally
            q0 = claw.frames[0].state.get_q_global()
            qfinal = claw.frames[claw.num_output_times].state.get_q_global()

            # and q_global is only returned on process 0
            if q0 is not None and qfinal is not None:
                q0 = q0.reshape([-1])
                qfinal = qfinal.reshape([-1])
                dx = claw.solution.domain.grid.delta[0]
                test = dx*np.sum(np.abs(qfinal-q0))
                return check_diff(expected, test, abstol=1e-4)
            else:
                return

args                 : {'use_petsc': False, 'kernel_language': 'Fortran', 'ptwise': True, 'solver_type': 'classic', 'disable_output': True}
norm of expected data: 0.001049
norm of test data    : 0.0017203753108528746
test error           : 0.0006713753108528746
abstol  : 0.0001
********************************************************************************


----------------------------------------------------------------------
Ran 9 tests in 0.696s

FAILED (errors=2)

[ketch]

I tried the same (fresh clone and running the same tests, actually by just copy/pasting the commands posted by @rjleveque . The tests pass for me. Could this be due to a compiler change?

[ketch]

Oops, I forgot to actually install the freshly-cloned version, so I was running an old version. Now I get the same errors. I'll see if I can track down the change that caused this.

[ketch]

Never mind my last message (now deleted). This was a very serious bug that I introduced. It should be fixed by clawpack/pyclaw#660.

[rjleveque]

Ah, good to know the regression tests are still useful in finding bugs and not just a nuisance!