Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

libgfortran=3.0 should not be install with numpy <= 1.9 #2177

Closed
jreback opened this issue Mar 5, 2016 · 13 comments

Comments

Projects
None yet
7 participants
@jreback
Copy link

commented Mar 5, 2016

libgfortran=1.0 works just fine, but since libgfortran=3.0 just came out, its not compatible with older numpies (using 1.8.1 here), see here.

So this totally breaks on ubuntu.

jreback@ubuntu:~/pandas3.4$ conda create -n test python=3.4 numpy=1.8.1 libgfortran --dry
Using Anaconda Cloud api site https://api.anaconda.org
Fetching package metadata: ......
Solving package specifications: ............
Package plan for installation in environment /home/jreback/miniconda/envs/test:

The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    numpy-1.8.1                |           py34_0         7.2 MB

The following NEW packages will be INSTALLED:

    libgfortran: 3.0-0        
    numpy:       1.8.1-py34_0 
    openssl:     1.0.2g-0     
    pip:         8.0.3-py34_0 
    python:      3.4.4-0      
    readline:    6.2-2        
    setuptools:  20.1.1-py34_0
    sqlite:      3.9.2-0      
    tk:          8.5.18-0     
    wheel:       0.29.0-py34_0
    xz:          5.0.5-1      
    zlib:        1.2.8-0      

Dry run: exiting
jreback@ubuntu:~/pandas3.4$ 

cc @ilanschnell @msarahan @mcg1969

@zym1010

This comment has been minimized.

Copy link

commented Mar 5, 2016

same problem. tried numpy 1.9.3 without MKL and failed just upon importing.

(test3)yimengzh_everyday@leelabgpu:~$ miniconda2/bin/conda create --name test4 python nomkl numpy=1.9
Fetching package metadata: ....
Solving package specifications: ...............
Package plan for installation in environment /home/yimengzh_everyday/miniconda2/envs/test4:

The following NEW packages will be INSTALLED:

    libgfortran: 3.0-0
    nomkl:       1.0-0
    numpy:       1.9.3-py27_nomkl_1 [nomkl]
    openblas:    0.2.14-4
    openssl:     1.0.2g-0
    pip:         8.0.3-py27_0
    python:      2.7.11-0
    readline:    6.2-2
    setuptools:  20.1.1-py27_0
    sqlite:      3.9.2-0
    tk:          8.5.18-0
    wheel:       0.29.0-py27_0
    zlib:        1.2.8-0

Proceed ([y]/n)? y

Linking packages ...
[      COMPLETE      ]|######################################################################################################################| 100%
#
# To activate this environment, use:
# $ source activate test4
#
# To deactivate this environment, use:
# $ source deactivate
#
(test3)yimengzh_everyday@leelabgpu:~$ source activate test4
discarding /home/yimengzh_everyday/miniconda2/envs/test3/bin from PATH
prepending /home/yimengzh_everyday/miniconda2/envs/test4/bin to PATH
(test4)yimengzh_everyday@leelabgpu:~$ python -c "import numpy; numpy.test()"
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/home/yimengzh_everyday/miniconda2/envs/test4/lib/python2.7/site-packages/numpy/__init__.py", line 170, in <module>
    from . import add_newdocs
  File "/home/yimengzh_everyday/miniconda2/envs/test4/lib/python2.7/site-packages/numpy/add_newdocs.py", line 13, in <module>
    from numpy.lib import add_newdoc
  File "/home/yimengzh_everyday/miniconda2/envs/test4/lib/python2.7/site-packages/numpy/lib/__init__.py", line 18, in <module>
    from .polynomial import *
  File "/home/yimengzh_everyday/miniconda2/envs/test4/lib/python2.7/site-packages/numpy/lib/polynomial.py", line 19, in <module>
    from numpy.linalg import eigvals, lstsq, inv
  File "/home/yimengzh_everyday/miniconda2/envs/test4/lib/python2.7/site-packages/numpy/linalg/__init__.py", line 51, in <module>
    from .linalg import *
  File "/home/yimengzh_everyday/miniconda2/envs/test4/lib/python2.7/site-packages/numpy/linalg/linalg.py", line 29, in <module>
    from numpy.linalg import lapack_lite, _umath_linalg
ImportError: libgfortran.so.1: cannot open shared object file: No such file or directory
@zym1010

This comment has been minimized.

Copy link

commented Mar 5, 2016

currently, I guess a simple solution is to specify libgfortran=1 in conda install.

seems to me recently these kinds of problems appear pretty often... can Continuum Analytics be more strict on QA? I'd like to pay some fee for quality bundles, instead of compiling everything myself...

@jreback jreback changed the title libgfortran=3.0 should not be install with numpy < 1.9 libgfortran=3.0 should not be install with numpy <= 1.9 Mar 5, 2016

@ilanschnell

This comment has been minimized.

Copy link
Member

commented Mar 5, 2016

numpy 1.8.1 did not explicitly list libgfortran as a dependency, but 1.8.2 does:

$ conda create -n test python=3.4 numpy=1.8.2

works fine for me.

@jreback

This comment has been minimized.

Copy link
Author

commented Mar 5, 2016

thanks @zym1010 updated to be <= 1.9 (we had a CI run that specifically tested with 1.8.1)

@jreback

This comment has been minimized.

Copy link
Author

commented Mar 5, 2016

ahh, ok @ilanschnell thanks.

@ilanschnell

This comment has been minimized.

Copy link
Member

commented Mar 5, 2016

Better, just say:

$ conda create -n test python=3.4 numpy=1.8

@ilanschnell ilanschnell closed this Mar 5, 2016

@jreback

This comment has been minimized.

Copy link
Author

commented Mar 5, 2016

yeah, I had this hard-coded in the CI, so just left it. Looks ok now that I fixed the libgfortran=1.0

thanks @ilanschnell

@zym1010

This comment has been minimized.

Copy link

commented Mar 5, 2016

@ilanschnell seems that it's still broken for 1.9?

@mcg1969

This comment has been minimized.

Copy link
Contributor

commented Mar 5, 2016

Unless there is a reason to believe this is the solver's fault---and not faulty metadata---this isn't something I can fix. Leave it to the metadata manipulators ;-)

@insertinterestingnamehere

This comment has been minimized.

Copy link

commented Mar 5, 2016

It looks like we're missing a build of scipy 0.16.1 with the latest bugfix release of numpy 1.9. That's causing an older build of scipy to get pulled in by the solver. See scipy/scipy#5917 (comment).

The 1.9 failure here is specific to the nomkl package. Getting a build of that one with the newer libgfortran should fix it.

@ilanschnell

This comment has been minimized.

Copy link
Member

commented Mar 5, 2016

I've updated numpy 1.9 as well as scipy 0.17.0 np19 to link against libgfortran 3.0.

@tritemio

This comment has been minimized.

Copy link

commented Mar 15, 2016

This (or a similar) issue is causing RTD to fail in importing scipy when using conda envs: rtfd/readthedocs.org#2074

The error is triggered when importing scipy.stats:

ImportError: libgfortran.so.1: cannot open shared object file: No such file or directory

See the RTD issue for full logs.

@dprada

This comment has been minimized.

Copy link

commented Mar 22, 2016

@tritemio In my case this was solved updating conda to 4.0. After that numpy and scypi were also updated to:

    numexpr:      2.4.4-np110py27_p0  [mkl] --> 2.5-np110py27_0   
    numpy:        1.10.2-py27_p0      [mkl] --> 1.10.4-py27_1     
    pytables:     3.2.2-np110py27_0   --> 3.2.2-np110py27_1 
    scikit-learn: 0.17-np110py27_p1   [mkl] --> 0.17.1-np110py27_0
    scipy:        0.16.1-np110py27_p0 [mkl] --> 0.17.0-np110py27_2

pdebuyl added a commit to pdebuyl/scipy-lecture-notes that referenced this issue Jun 17, 2016

add conda install of libgfortran=1
The dependence of NumPy on libgfortran has changed in conda, see
conda/conda#2177

pdebuyl added a commit to pdebuyl/scipy-lecture-notes that referenced this issue Jun 17, 2016

add conda install of libgfortran=1
The dependence of NumPy on libgfortran has changed in conda, see
conda/conda#2177

pdebuyl added a commit to pdebuyl/scipy-lecture-notes that referenced this issue Jun 17, 2016

add conda install of libgfortran=1
The dependence of NumPy on libgfortran has changed in conda, see
conda/conda#2177
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.