sci-libs/scipy-0.15.1 fails emerge with the following output - Processing scipy/special/_ellip_harm_2.pyx /usr/bin/python3.3 setup.py build config_fc --noopt --noarch Traceback (most recent call last): File "setup.py", line 249, in <module> setup_package() File "setup.py", line 237, in setup_package from numpy.distutils.core import setup File "/usr/lib64/python3.3/site-packages/numpy/__init__.py", line 170, in <module> from . import add_newdocs File "/usr/lib64/python3.3/site-packages/numpy/add_newdocs.py", line 13, in <module> from numpy.lib import add_newdoc File "/usr/lib64/python3.3/site-packages/numpy/lib/__init__.py", line 17, in <module> from . import scimath as emath ImportError: cannot import name scimath Posting build.log Reproducible: Always
Created attachment 407044 [details] build.log
Created attachment 407046 [details] emerge-info
which numpy are you using?
(In reply to Justin Lecher from comment #3) > which numpy are you using? Numpy is currently :dev-python/numpy-1.9.2::gentoo
I cannot reproduce this at all. Could try to rebuild numpy and scipy?
(In reply to Justin Lecher from comment #5) > I cannot reproduce this at all. Could try to rebuild numpy and scipy? My scipy was still at sci-math/scipy-0.14.0-r1 when I tried this update, the sci-libs/* has been pruned so I'm stuck at this point for now. dev-python/numpy-1.9.2 builds fine.
(In reply to Jason Mours from comment #6) > (In reply to Justin Lecher from comment #5) > > I cannot reproduce this at all. Could try to rebuild numpy and scipy? > > My scipy was still at sci-math/scipy-0.14.0-r1 when I tried this update, the > sci-libs/* has been pruned so I'm stuck at this point for now. > > dev-python/numpy-1.9.2 builds fine. I came across the same problem while installing sagemath, rebuild numpy works! thanks very much!
still hitting this. I'll rebuild numpy as seems to be the suggestion
so this fails reliably in catalyst when both scipy and numpy are not installed and trying to be installed together. the deps of scipy appear correct, as numpy is in DEPEND and should be available when portage starts trying to install scipy. Is it possible that maybe some environment setup isn't complete and /etc/profile needs to be re-read or something?
:-( nevermind, that didn't work either... mine is just plain broken
(In reply to Rick Farina (Zero_Chaos) from comment #10) > :-( nevermind, that didn't work either... mine is just plain broken You are gentoo, it is broken... I'm going to start over again tho. Some other chip to -Os. I filed this from a 3.18 gentoo. If you're still seeing this from a 4.x gentoo. It is scipy, not just you. Like I was thinking. Python 3.3 is a '3' it's weak but essential. They never finished 3.4 so your '3' doesn't '4' because python changed hands. It is hard to make a 3 -4, as it were. The there is Python 3.5 which doesn't coil right, I have a bug that is still open that proves that. IMOH It's essential too, It's our '5' since we threw away 2.5.... stupid. But it's unfinished. If your still seeing this in the maintained 3.6 or 3.7 then is circular throughout all of python. Something missing, not docs. Undefined. I may have more if I get to it in a month or two. Maybe not just scipy, dunno.
I can confirm the problem with the current stable packages (numpy-1.14.5, scipy-0.19.1). Rebuilding numpy does not resolve the issue. However, it occurs with python 2.7 and for a different file for me: Processing scipy/linalg/_solve_toeplitz.pyx python2.7 setup.py build config_fc --noopt --noarch Traceback (most recent call last): File "setup.py", line 416, in <module> setup_package() File "setup.py", line 396, in setup_package from numpy.distutils.core import setup File "/usr/lib64/python2.7/site-packages/numpy/__init__.py", line 142, in <module> from . import add_newdocs File "/usr/lib64/python2.7/site-packages/numpy/add_newdocs.py", line 13, in <module> from numpy.lib import add_newdoc File "/usr/lib64/python2.7/site-packages/numpy/lib/__init__.py", line 18, in <module> from . import scimath as emath ImportError: cannot import name scimath
Version 1.x fail with Bug 681786 -> I am stuck here
(In reply to Till Schäfer from comment #13) > Version 1.x fail with Bug 681786 -> I am stuck here This isn't 'docs' this is <module>. They tore apart portage. My original bug still had a gentoo with portage 2.2. Which means to python. LOTS of dependencies. I did and still believe in yesterdays code. Down with the ++ up with the c89. Brains don't bit shift. We are all going to have code generators anyway. Get it? Generators? Generational Code? I digress. This install was over a year old and it was TRASHEed to ++ SOME x2011-2099 seriezzXXYYZZ...standard...ANSSI somehting...ERRoRRR...DAAARRRWIIINNN I'm was bent. I did give sweet traces though. The code to find for <module> is in there.....somewhere.. <sparse> <-something other than But it might have been amputated when they transitioned portage to python3. Which means the WHOLE TREE NEEDS REBUILT!! On one box. Probably in 'C' with ! all of python !, just to be a jagoff. But this is history right? I'm not mad at the distro. I just got a pair of e5-2680 to -0s. So I still believe in GNU, I bought a tshirt. But I had compiled IceT over 600 times. Glibc over 1000. on an 8 core... and we just had to move to linux4. The 3.33333333 was NOT done yet. The 4 did fall apart 'kit-kat' did it not? Alot of 'kids' are lost. Projects languish. linux5 is years always. And Till is stuck? It will be awhile OR we will lose something inherited generational from some other compiler or library. Astrophysics is my guess. This is regression Astor, it's always painful in generational compile. Especially when your Gentoo is ML, i.e. portage. My bug, my pollution.
I think it is related to Bug 675542 but for some reason it is not showing the full error message with the broken fortran link errors. Therefore rebuild your blas and lapack. Assuming you are using reference like my, these are the commands: emerge -1 blas-reference lapack-reference emerge -1 numpy scipy It fixed for me :)
I answered this in another similar bug. There are two issues that can hit people here: 1) gfortran upgrade 2) numpy as a runtime dependency on dev-python/future (not futures) which is missing from the ebuild. In turn, since scipy (and any other ebuild depending on numpy) needs a working numpy, dev-python/future should be a build time dependency of scipy.
The import failure from this bug title is due to the missing dev-python/future dependency.
(In reply to Derek Dolney from comment #15) > I think it is related to Bug 675542 but for some reason it is not showing > the full error message with the broken fortran link errors. Therefore > rebuild your blas and lapack. Assuming you are using reference like my, > these are the commands: > > emerge -1 blas-reference lapack-reference > emerge -1 numpy scipy > > It fixed for me :) This is it! Thank you!
(In reply to Derek Dolney from comment #15) > I think it is related to Bug 675542 but for some reason it is not showing > the full error message with the broken fortran link errors. Therefore > rebuild your blas and lapack. Assuming you are using reference like my, > these are the commands: > > emerge -1 blas-reference lapack-reference > emerge -1 numpy scipy > > It fixed for me :) Thanks Derek.
(In reply to Jason Mours from comment #18) > (In reply to Derek Dolney from comment #15) > > I think it is related to Bug 675542 but for some reason it is not showing > > the full error message with the broken fortran link errors. Therefore > > rebuild your blas and lapack. Assuming you are using reference like my, > > these are the commands: > > > > emerge -1 blas-reference lapack-reference > > emerge -1 numpy scipy > > > > It fixed for me :) > > This is it! Thank you! This isn't a future dependency. emerge -1 blas-reference lapack-reference emerge -1 numpy scipy This isn't a fix the import. Things are missing from LA. Like what right? Linear has yet to be proven. So says Cambridge, as I heard, in passing... Trying to pack all of FORTRAN into it...when it starts to parallel. Stuff just isn't in there. This is exotic, not to say rare. It's not scipy....not numpy. It's scimath. Who has scimath? Do you hate me? Do you remember this code from 80 something? It might be 90's. GCC what? SCO? I'm sure it's codebase. Was it a pull?
(In reply to Jason Mours from comment #20) > (In reply to Jason Mours from comment #18) > > (In reply to Derek Dolney from comment #15) > > > I think it is related to Bug 675542 but for some reason it is not showing > > > the full error message with the broken fortran link errors. Therefore > > > rebuild your blas and lapack. Assuming you are using reference like my, > > > these are the commands: > > > > > > emerge -1 blas-reference lapack-reference > > > emerge -1 numpy scipy > > > > > > It fixed for me :) > > > > This is it! Thank you! > > This isn't a future dependency. > > emerge -1 blas-reference lapack-reference > emerge -1 numpy scipy > > This isn't a fix the import. Things are missing from LA. Like what right? > Linear has yet to be proven. So says Cambridge, as I heard, in passing... > Trying to pack all of FORTRAN into it...when it starts to parallel. Stuff > just isn't in there. This is exotic, not to say rare. It's not scipy....not > numpy. It's scimath. Who has scimath? Do you hate me? Do you remember this > code from 80 something? It might be 90's. GCC what? SCO? I'm sure it's > codebase. Was it a pull? These are phone books. Shelves of them. Rows of shelves. '++' ugh unicode...it's ansi. I'm sure it's stateside. I cannot procure function with what I have. SCO? TI? SGI?
Closing bugs for old versions no longer in ::gentoo. Please reopen if it still happens.
