Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Pass Manager Visualization Tests fail locally

See original GitHub issue

Information

Qiskit Terra version: Master
Python version: 3.7
Operating system: Linux
dot - graphviz version 2.40.1 (0)

What is the current behavior?

When running the terra unit tests locally from a fresh master with no local modification the image comparison in the pass manager visualization tests fail for me locally. Either we need to relax the comparison threshold or make the image drawing deterministic

test.python.visualization.test_pass_manager_drawer.TestPassManagerDrawer.test_pass_manager_drawer_basic
-------------------------------------------------------------------------------------------------------

Captured traceback:
~~~~~~~~~~~~~~~~~~~
    b'Traceback (most recent call last):'
    b'  File "/home/computertreker/git/qiskit/qiskit-core/test/python/visualization/test_pass_manager_drawer.py", line 52, in test_pass_manager_drawer_basic'
    b"    self.assertImagesAreEqual(filename, path_to_diagram_reference('pass_manager_l0_ref.png'))"
    b'  File "/home/computertreker/git/qiskit/qiskit-core/test/python/visualization/visualization.py", line 62, in assertImagesAreEqual'
    b'    .format(diff_tolerance * 100))'
    b'  File "/usr/lib64/python3.7/unittest/case.py", line 692, in assertTrue'
    b'    raise self.failureException(msg)'
    b'AssertionError: False is not true : The images are different by more than a 0.1%'
    b''

test.python.visualization.test_pass_manager_drawer.TestPassManagerDrawer.test_pass_manager_drawer_style
-------------------------------------------------------------------------------------------------------

Captured traceback:
~~~~~~~~~~~~~~~~~~~
    b'Traceback (most recent call last):'
    b'  File "/home/computertreker/git/qiskit/qiskit-core/test/python/visualization/test_pass_manager_drawer.py", line 69, in test_pass_manager_drawer_style'
    b"    path_to_diagram_reference('pass_manager_l1_style_ref.png'))"
    b'  File "/home/computertreker/git/qiskit/qiskit-core/test/python/visualization/visualization.py", line 62, in assertImagesAreEqual'
    b'    .format(diff_tolerance * 100))'
    b'  File "/usr/lib64/python3.7/unittest/case.py", line 692, in assertTrue'
    b'    raise self.failureException(msg)'
    b'AssertionError: False is not true : The images are different by more than a 0.1%'
    b''

Steps to reproduce the problem

Run unit tests with graphviz installed

What is the expected behavior?

The tests pass

Issue Analytics

State:
Created 4 years ago
Comments:6 (6 by maintainers)

Top GitHub Comments

1reaction

kdkcommented, May 31, 2019

What about dumping the pydot graph to DOT and asserting based on that? We can still verify that we are building the graph correctly, and leave pydot to draw the graph consistently. It should be more robust to changes in environment (and would be diffable!)

0reactions

1ucian0commented, May 30, 2019

Fonts might vary a lot in each platform. Ideally, we should try to find the same font for every platform, but it might not be possible. We should consider alternatives if that’s a dead-end.