#11564 closed defect (fixed)
new py-scipy fails due to fortran dependencies
Reported by: | erin.sheldon@… | Owned by: | erickt@… |
---|---|---|---|
Priority: | High | Milestone: | |
Component: | ports | Version: | |
Keywords: | Cc: | ||
Port: |
Description
Starting with clean macports install
sudo port install py-scipy (which is version 0.5.2)
fails with the following error trying to run f95. The fortran dependency is incorrect; scipy should depend on gfortran the same way numpy does.
f95:f77: Lib/fftpack/dfftpack/dcosqb.f sh: line 1: f95: command not found sh: line 1: f95: command not found error: Command "f95 -fixed -O4 -target=native -c -c Lib/fftpack/dfftpack/dcosqb.f -o build/temp.darwin-8.9.1-i386-2.4/Lib/fftpack/dfftpack/dcosqb.o" failed with exit status 127
Change History (7)
comment:1 Changed 18 years ago by pipping@…
Milestone: | → Available Ports |
---|
comment:2 Changed 18 years ago by pipping@…
Milestone: | Available Ports → Port Bugs |
---|
comment:3 Changed 18 years ago by jochen@…
comment:4 Changed 18 years ago by david.m.cooke@…
This should be fixed in NumPy 1.0.2.
Also, NumPy itself doesn't require Fortran at all (on purpose -- it's targeted to a wider audience). The Fortran support included in NumPy is for SciPy and other packages like it that do require Fortran.
-- your friendly upstream developer
comment:5 Changed 17 years ago by erickt@…
Owner: | changed from macports-dev@… to erickt@… |
---|---|
Status: | new → assigned |
numpy is now updated to 1.0.3 and uses g95 instead of gcc41/gcc42. Let me know if this fixes your problems.
comment:6 Changed 17 years ago by erickt@…
Resolution: | → fixed |
---|---|
Status: | assigned → closed |
Solution to a similar problem I hd with installing py-scipy:
Current scipy (0.5.2) depends on py-numpy
That's good!
py-numpy depends on gcc42 on Intel, which is also good.
However, there is a mismatch that manifests itself when using numpy, to e.g. install py-scipy:
In numpy/distutils/fcompiler/gnu.py a string match is done against the output of
gfortran --version
:but
gfortran-dp-4.2 --version
gives "GNU Fortran (GCC) 4.2.0 20070307 (prerelease)" (thus: no "95"!)Using the following patch to numpy fixes the problem for me and I can successfully compile and use py-scipy on Intel then. However, the real fix might be to correct the gfortran version output.