You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Even if this was moved to Actions the test environment is setup differently. i.e. these tests don't use tox, cibuildwheel or the pipelines script installs the dist and then runs pytest. So I don't think moving this to Actions would help at all.
This is not an absolute blocker since in principle we can do this manually but... yeah, it's a huge pain that way. So definitely should try to get it fixed
=================================== FAILURES ===================================
________________________ test_write_extra_SkyCoord_cols ________________________
def test_write_extra_SkyCoord_cols():
"""
Tests output for cases when table contains multiple ``SkyCoord``
columns.
"""
exp_output = '''\
================================================================================
Byte-by-byte Description of file: table.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 7 A7 --- name Description of name
9-24 A16 --- Unknown Description of Unknown
26-29 F4.1 h RAh Right Ascension (hour)
31-33 F3.1 min RAm Right Ascension (minute)
35-52 F18.15 s RAs Right Ascension (second)
54 A1 --- DE- Sign of Declination
55-58 F5.1 deg DEd Declination (degree)
60-63 F4.1 arcmin DEm Declination (arcmin)
65-80 F16.13 arcsec DEs Declination (arcsec)
--------------------------------------------------------------------------------
Notes:
--------------------------------------------------------------------------------
HD81809 330.564 -61.6596 22.0 2.0 15.450000000007265 -61.0 39.0 34.5999960000006
''' # noqa: W291
import pytest
t = Table()
t['name'] = ['HD81809']
coord = SkyCoord(330.564375, -61.65961111, unit=u.deg)
t['coord1'] = coord
t['coord2'] = coord
out = StringIO()
with pytest.raises(UserWarning):
t.write(out, format='ascii.cds')
lines = out.getvalue().splitlines()
i_bbb = lines.index('=' * 80)
lines = lines[i_bbb:] # Select Byte-By-Byte section and later lines.
# Check the written table.
> assert lines == exp_output.splitlines()
E AssertionError: assert ['===========...me ', ...] == ['===========...me ', ...]
E At index 5 diff: ' 1- 7 A7 --- name Description of name ' != ' 1- 7 A7 --- name Description of name '
E Left contains 6 more items, first extra item: '104-107 F4.1 arcmin DEm Declination (arcmin) '
E Use -v to get the full diff
/opt/hostedtoolcache/Python/3.9.7/x64/lib/python3.9/site-packages/astropy/io/ascii/tests/test_cds.py:489: AssertionError
@Cadair et al - is this possible to reproduce (or try reproducing) outside of the dist build system? Personally I have no idea even where to start if it is passing our CI.
is this possible to reproduce (or try reproducing) outside of the dist build system
You could try and run the test command in a clean venv with a fresh tarball installed (with the test extra). So something like:
<clean venv>
pip install build
cd <astropy_checkout_dir>
python -m build -s .
pip install "dist/astropy-<...>.tar.gz[test]"
cd <anywhere_else>
pytest -p no:warnings --astropy-header -k "not test_data_out_of_range and not test_datetime_difference_agrees_with_timedelta and not test_wcsapi_extension and not test_timedelta_datetime_roundtrip" --pyargs astropy
Hi,
It looks the behaviour of the function is a bit different in this particular case. Need to look up the full diff. However, the extra items may have risen if somehow the extra SkyCoord column is not bypassed. The expected behaviour is to bypass extra SkyCoord columns. Dunno why it shouldn't work in Azure, if CI gives no errors.
If it helps you debug it, this is the generated output vs the expected output prettified better than pytest does it:
The last line is a floating point comparison and should be treated as such; fails for me on macOS as well.
EDIT: It does fail earlier on the multiple RA/Dec columns as above.
It looks the behaviour of the function is a bit different in this particular case. Need to look up the full diff. However, the extra items may have risen if somehow the extra SkyCoord column is not bypassed. The expected behaviour is to bypass extra SkyCoord columns. Dunno why it shouldn't work in Azure, if CI gives no errors.
On the affected systems like @Cadair's or mine, or apparently azure, indeed both SkyCoord columns are created in full
Activity
Cadair commentedon Sep 15, 2021
Also looks like #12129 introduced a second fail: https://dev.azure.com/astropy-project/astropy/_build/results?buildId=1915&view=logs&j=f7656e4a-aab1-5409-7a28-08f0d640eba6&t=d1ac6189-42a2-5a09-839e-9c6a924261dc&l=591
pllim commentedon Sep 15, 2021
Why is the behavior different from CI?
pllim commentedon Sep 15, 2021
Is it time to move this to Actions so everything is more consistent? I don't even know where to begin with this one.
Cadair commentedon Sep 15, 2021
Even if this was moved to Actions the test environment is setup differently. i.e. these tests don't use tox, cibuildwheel or the pipelines script installs the dist and then runs pytest. So I don't think moving this to Actions would help at all.
pllim commentedon Sep 19, 2021
I am guessing this is a blocker for 5.0 release? Who is going to fix this? Please assign that person to this issue. Thanks!
eteq commentedon Sep 20, 2021
This is not an absolute blocker since in principle we can do this manually but... yeah, it's a huge pain that way. So definitely should try to get it fixed
pllim commentedon Oct 21, 2021
@Cadair , the posted logs expired. Do you have updated link(s)? Thanks!
Cadair commentedon Oct 26, 2021
With #12287 merged we are down to one test failure (from #11897): https://dev.azure.com/astropy-project/astropy/_build/results?buildId=2106&view=logs&j=f7656e4a-aab1-5409-7a28-08f0d640eba6&t=d1ac6189-42a2-5a09-839e-9c6a924261dc&l=554
taldcroft commentedon Oct 26, 2021
@Suyog7130 @hamogu @aaryapatil - any thoughts on this?
@Cadair et al - is this possible to reproduce (or try reproducing) outside of the dist build system? Personally I have no idea even where to start if it is passing our CI.
Cadair commentedon Oct 26, 2021
You could try and run the test command in a clean venv with a fresh tarball installed (with the test extra). So something like:
that should get you close?
Suyog7130 commentedon Oct 26, 2021
Hi,
It looks the behaviour of the function is a bit different in this particular case. Need to look up the full diff. However, the extra items may have risen if somehow the extra SkyCoord column is not bypassed. The expected behaviour is to bypass extra SkyCoord columns. Dunno why it shouldn't work in Azure, if CI gives no errors.
Cadair commentedon Oct 26, 2021
Using the steps above I am able to reproduce this error locally. I have no idea why the normal CI using tox isn't hitting it.
Cadair commentedon Oct 26, 2021
If it helps you debug it, this is the generated output vs the expected output prettified better than pytest does it:
dhomeier commentedon Oct 26, 2021
The last line is a floating point comparison and should be treated as such; fails for me on macOS as well.
EDIT: It does fail earlier on the multiple
RA/Deccolumns as above.dhomeier commentedon Oct 26, 2021
On the affected systems like @Cadair's or mine, or apparently azure, indeed both
SkyCoordcolumns are created in fulland repeated after the floating point values in the
Unknowncolumn.hamogu commentedon Oct 26, 2021
It is certainly strange, but unfortunately, I won't have time today to look into it. I can try tomorrow.
pllim commentedon Oct 26, 2021
Well, Stuart is going to fix the broken builds by ignoring this failing test for now. Please discuss it further at #12299 going forward.