Update symengine, symengine_py to 0.8.1; drop dependency on nose

[mkoeppe]

Update symengine, symengine_py to 0.8.1; drop dependency on nose

Upstream report for Sage objects: symengine/symengine.py#373

Upstream: Fixed upstream, but not in a stable release.

CC: @isuruf @sagetrac-tmonteil

Component: packages: optional

Author: Matthias Koeppe, Isuru Fernando

Branch/Commit: 6dd3e5b

Reviewer: Thierry Monteil, Matthias Koeppe

Issue created by migration from https://trac.sagemath.org/ticket/32603

[mkoeppe]

Branch: u/mkoeppe/update_symengine__symengine_py__drop_dependency_on_nose

[mkoeppe]

Author: Matthias Koeppe

[mkoeppe]

Commit: f929448

[mkoeppe]

comment:2

The pytest does not go through for me, but it's possible that my installation is messed up

[symengine_py-0.8.1] Running the test suite for symengine_py-0.8.1...
[symengine_py-0.8.1] ============================= test session starts ==============================
[symengine_py-0.8.1] platform darwin -- Python 3.9.7, pytest-6.2.5, py-1.10.0, pluggy-1.0.0 -- /Users/mkoeppe/s/sage/sage-rebasing/local/var/lib/sage/venv-3.9/bin/python3
[symengine_py-0.8.1] cachedir: .pytest_cache
[symengine_py-0.8.1] rootdir: /Users/mkoeppe/s/sage/sage-rebasing/local/var/lib/sage/venv-3.9/var/tmp/sage/build/symengine_py-0.8.1/src
[symengine_py-0.8.1] collecting ... collected 0 items / 23 errors
[symengine_py-0.8.1] 
[symengine_py-0.8.1] ==================================== ERRORS ====================================
[symengine_py-0.8.1] ________________ ERROR collecting symengine/tests/test_arit.py _________________
[symengine_py-0.8.1] ImportError while importing test module '/Users/mkoeppe/s/sage/sage-rebasing/local/var/lib/sage/venv-3.9/var/tmp/sage/build/symengine_py-0.8.1/src/symengine/tests/test_arit.py'.
[symengine_py-0.8.1] Hint: make sure your test modules/packages have valid Python names.
[symengine_py-0.8.1] Traceback:
[symengine_py-0.8.1] /usr/local/Cellar/python@3.9/3.9.7/Frameworks/Python.framework/Versions/3.9/lib/python3.9/importlib/__init__.py:127: in import_module
[symengine_py-0.8.1]     return _bootstrap._gcd_import(name[level:], package, level)
[symengine_py-0.8.1] symengine/__init__.py:12: in <module>
[symengine_py-0.8.1]     import symengine.lib.symengine_wrapper as wrapper
[symengine_py-0.8.1] E   ModuleNotFoundError: No module named 'symengine.lib.symengine_wrapper'
[symengine_py-0.8.1] _________________ ERROR collecting symengine/tests/test_cse.py _________________

---

New commits:

b3fcbd8	build/pkgs/symengine[_py]: Update to 0.8.1
c4faa5f	build/pkgs/symengine/patches/1749.patch: Remove
f929448	build/pkgs/symengine_py: Switch from nose to pytest

[sagetrac-tmonteil]

comment:4

I can reproduce the error.

[isuruf]

New commits:

fc0e37c

fix symengine_py tests

[isuruf]

Changed branch from u/mkoeppe/update_symengine__symengine_py__drop_dependency_on_nose to u/isuruf/update_symengine__symengine_py__drop_dependency_on_nose

[isuruf]

Changed author from Matthias Koeppe to Matthias Koeppe, Isuru Fernando

[isuruf]

Changed commit from f929448 to fc0e37c

[sagetrac-tmonteil]

Reviewer: Thierry Monteil

[sagetrac-tmonteil]

comment:6

It does not work for me, see the attached log.

[sagetrac-tmonteil]

Attachment: symengine_py-0.8.1.log

[sagetrac-git]

Changed commit from fc0e37c to 1b2bdbb

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

1b2bdbb

Fix getting string representation of sage objects

[sagetrac-tmonteil]

comment:9

Still some problem, see the attached log.

[sagetrac-tmonteil]

Upstream: Reported upstream. No feedback yet.

[sagetrac-tmonteil]

comment:10

Attachment: symengine_py-0.8.1.2.log

[sagetrac-git]

Changed commit from 1b2bdbb to 6dd3e5b

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

6dd3e5b

update version for a rebuild

[isuruf]

comment:12

I think the issue was that symengine_py was not rebuilt. I've bumped the version to trigger a rebuild

[sagetrac-tmonteil]

comment:14

The issue persists. The reason is that the doctest look at files in $SAGE_ROOT/local/lib/python3.9/site-packages/symengine/tests/ but when Sage tests a spkg, it only installs the spkg AFTER the test succeeds. So, if you install the spkg first and then install/test the spkg, currently what is tested is the previous installation.

So, changing package version may not fix the issue. In spkg-check.in, we should ask the testing framework to test files in $SAGE_ROOT/local/var/tmp/sage/build/symengine_py-0.8.1/src/symengine/tests not in $SAGE_ROOT/local/lib/python3.9/site-packages/symengine/tests/

By the way, looking at symengine.py-0.8.1/bin/test_travis.sh (in the tarball), it seems that tests from pytest and tests from test_python.py are not the same (both are run one after the other), so we should perhaps test both.

[isuruf]

comment:15

By the way, looking at symengine.py-0.8.1/bin/test_travis.sh (in the tarball), it seems that tests from pytest and tests from test_python.py are not the same (both are run one after the other), so we should perhaps test both.

Not really. They check that the tests run fine on inplace builds and on installed symengine.
Since sage is not doing inplace builds, it doesn't matter.

Thanks for the info. I'll try to fix this.

[isuruf]

comment:16

The reason is that the doctest look at files in $SAGE_ROOT/local/lib/python3.9/site-packages/symengine/tests/ but when Sage tests a spkg, it only installs the spkg AFTER the test succeeds. So, if you install the spkg first and then install/test the spkg, currently what is tested is the previous installation.

At this point where is symengine_py installed to?

[mkoeppe]

comment:17

Replying to @sagetrac-tmonteil:

The reason is that the doctest look at files in $SAGE_ROOT/local/lib/python3.9/site-packages/symengine/tests/ but when Sage tests a spkg, it only installs the spkg AFTER the test succeeds. So, if you install the spkg first and then install/test the spkg, currently what is tested is the previous installation.

That's not true. Take a look at build/bin/sage-spkg: spkg-check is run after the staged installation has been copied to the final installation directory.

[mkoeppe]

comment:18

(Also note that since #32361, Python packages are not actually staged using SAGE_DESTDIR any more. We just build and then install a wheel. This happens before running spkg-check.)

[isuruf]

comment:19

I'm confused as to how the issue mentioned by @sagetrac-tmonteil happens then.

[isuruf]

Changed upstream from Reported upstream. No feedback yet. to Fixed upstream, but not in a stable release.

[mkoeppe]

comment:21

In #32595 I am currently running portability testing on GH Actions, including this ticket. We'll see if these tests can reproduce the issue.

[mkoeppe]

comment:22

This has run successfully on https://github.com/mkoeppe/sage/actions/runs/1298853742; so I am setting this to positive review

[mkoeppe]

Changed reviewer from Thierry Monteil to Thierry Monteil, Matthias Koeppe

[vbraun]

Changed branch from u/isuruf/update_symengine__symengine_py__drop_dependency_on_nose to 6dd3e5b