Unable to install sfepy 2017.1 in python 3.6 environment using anaconda3

[Emanuel Woiski]

Hi,

I've been unable to install sfepy 2017.1 using conda-forge in the python 3.6 environment.
 

The error was:

"UnsatisfiableError: The following specifications were found to be in conflict:
  - python 3.6*
  - sfepy 2017.1* -> python 2.7*
Use "conda info <package>" to see the dependencies for each package."

After using conda info sfepy this is what I've got:

"sfepy 2017.1 np111py35_0
------------------------
file name   : sfepy-2017.1-np111py35_0.tar.bz2
name        : sfepy
version     : 2017.1
build string: np111py35_0
build number: 0
channel     : conda-forge
size        : 7.3 MB
arch        : x86_64
has_prefix  : True
license     : BSD-3-Clause
md5         : b38a92b0ecdf0ead6b390cd984e00de1
noarch      : None
platform    : linux
requires    : ()
subdir      : linux-64
url         : https://conda.anaconda.org/conda-forge/linux-64/sfepy-2017.1-np111py35_0.tar.bz2
dependencies:
    matplotlib
    mpi4py
    numpy 1.11*
    pyparsing
    pytables
    python 3.5*
    scipy
    sympy"

What am I missing in order to install sfepy 2007.1 in python 3.6?

Thanks,

woiski

[Robert Cimrman]

Hi,

we have been skipping python 3.6 in the conda-forge recipe, as there were
probably some compatibility problems (I do not recall exactly). SfePy now works
with 3.6, so I am trying to update the recipe.

r.

[Emanuel Woiski]

Awesome, thanks.

Eventually I will try it again.

woiski

--
You received this message because you are subscribed to the Google Groups "sfepy-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sfepy-devel+unsubscribe@googlegroups.com.
To post to this group, send email to sfepy...@googlegroups.com.
Visit this group at https://groups.google.com/group/sfepy-devel.

[Robert Cimrman]

Hi Emanuel,

the python 3.6 installation should work now out of the box, except the two
tests that depend on igakit.

r.

>> email to sfepy-devel...@googlegroups.com.

[Emanuel Woiski]

Thanks!
There was no problem with the installation using 'conda install sfepy'.

However 'python run_tests.py' does not work. So I was trying to test using 'nosetests sfepy' and the following message was produced:

EEEE....EEEEEEEEEEEEEEEEEEEEEEEE..Intel MKL FATAL ERROR: Cannot load libmkl_avx2.so or libmkl_def.so.
Intel MKL FATAL ERROR: Cannot load libmkl_avx2.so or libmkl_def.so.
Intel MKL FATAL ERROR: Cannot load libmkl_avx2.so or libmkl_def.so.
Intel MKL FATAL ERROR: Cannot load libmkl_avx2.so or libmkl_def.so.
Intel MKL FATAL ERROR: Cannot load libmkl_avx2.so or libmkl_def.so.

According to conda, mkl is already installed.

Any further help or suggestion?

woiski

2017-03-28 4:17 GMT-03:00 Robert Cimrman <cimr...@ntc.zcu.cz>:

Hi Emanuel,

the python 3.6 installation should work now out of the box, except the two tests that depend on igakit.

r.

On 03/27/2017 05:47 PM, Emanuel Woiski wrote:

Awesome, thanks.
Eventually I will try it again.
woiski

2017-03-27 9:38 GMT-03:00 Robert Cimrman <cimr...@ntc.zcu.cz>:

Hi,

we have been skipping python 3.6 in the conda-forge recipe, as there were
probably some compatibility problems (I do not recall exactly). SfePy now
works with 3.6, so I am trying to update the recipe.

r.

​[snip]​

[Robert Cimrman]

On 29.3.2017 21:44, Emanuel Woiski wrote:
> Thanks!
> There was no problem with the installation using 'conda install sfepy'.
> However 'python run_tests.py' does not work. So I was trying to test using

'python run_tests.py' is meant to be used from the directory of sfepy sources
(e.g. if you clone the code from github). To test an installed sfepy, try using
'sfepy-run run_tests'. The instructions on the web seem to be wrong.

r.

[Robert Cimrman]

Concerning the MKL errors - do numpy and scipy work?

r.

[Emanuel Woiski]

yes, they work.

It seems that the installed sfepy is unable to find the mkl libraries.

woiski

--
You received this message because you are subscribed to the Google Groups "sfepy-devel" group.

To unsubscribe from this group and stop receiving emails from it, send an email to sfepy-devel+unsubscribe@googlegroups.com.

[Robert Cimrman]

Well, sfepy does not use MKL by itself anywhere, only through other packages,
like numpy.

"sfepy-run run_tests" also fails? If yes, could you send us the output of
"sfepy-run run_tests --raise"?

Thanks,
r.

>> email to sfepy-devel...@googlegroups.com.

[Emanuel Woiski]

Sure. Please find attached.

Thanks for looking into it.

woiski

email to sfepy-devel+unsubscribe@googlegroups.com.

To post to this group, send email to sfepy...@googlegroups.com.
Visit this group at https://groups.google.com/group/sfepy-devel.

--
You received this message because you are subscribed to the Google Groups "sfepy-devel" group.

To unsubscribe from this group and stop receiving emails from it, send an email to sfepy-devel+unsubscribe@googlegroups.com.

[Robert Cimrman]

On 30.3.2017 21:47, Emanuel Woiski wrote:
> Sure. Please find attached.
> Thanks for looking into it.
> woiski

So it fails when the code is about to solve a linear system. Do you have
scikit-umfpack installed? Does
https://docs.scipy.org/doc/scipy-0.14.0/reference/generated/scipy.sparse.linalg.spsolve.html
work?

r.

[Emanuel Woiski]

Good guess! After installing scikit-umfpack using conda(-forge?), I executed 'sfepy-run run_tests'.

The final message was:
98 test file(s) executed in 695.81 s, 2 failure(s) of 165 test(s).

The two failures (expected) are related to navier-stokes2d _iga.py and test_projection_iga_fem .

I think it is ok now!

I must thank you again for your support.

Best regards

woiski

[Robert Cimrman]

On 30.3.2017 22:38, Emanuel Woiski wrote:
> Good guess! After installing scikit-umfpack using conda(-forge?), I
> executed 'sfepy-run run_tests'.
> The final message was:
> 98 test file(s) executed in 695.81 s, 2 failure(s) of 165 test(s).
>
> The two failures (expected) are related to navier-stokes2d _iga.py and
> test_projection_iga_fem .
> I think it is ok now!
> I must thank you again for your support.
> Best regards
> woiski

OK, hth!

r.

> 2017-03-30 16:56 GMT-03:00 Robert Cimrman <cimr...@ntc.zcu.cz>:
>
>> On 30.3.2017 21:47, Emanuel Woiski wrote:
>>
>>> Sure. Please find attached.
>>> Thanks for looking into it.
>>> woiski
>>>
>>
>> So it fails when the code is about to solve a linear system. Do you have
>> scikit-umfpack installed? Does https://docs.scipy.org/doc/sci
>> py-0.14.0/reference/generated/scipy.sparse.linalg.spsolve.html work?
>>
>> r.
>>
>> 2017-03-30 8:33 GMT-03:00 Robert Cimrman <cimr...@ntc.zcu.cz>:
>>>
>>> Well, sfepy does not use MKL by itself anywhere, only through other
>>>> packages, like numpy.
>>>>
>>>> "sfepy-run run_tests" also fails? If yes, could you send us the output of
>>>> "sfepy-run run_tests --raise"?
>>>>
>>>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "sfepy-devel" group.
>> To unsubscribe from this group and stop receiving emails from it, send an

>> email to sfepy-devel...@googlegroups.com.