Opened 10 years ago
Last modified 10 years ago
#46392 closed defect
gfortran-mp-4.[89] internal compiler error when building scipy-0.14.1 on a PPC — at Version 1
Reported by: | josephsacco | Owned by: | sean@… |
---|---|---|---|
Priority: | Normal | Milestone: | |
Component: | ports | Version: | 2.3.3 |
Keywords: | PPC, scipy, gfortran | Cc: | |
Port: | py-numpy |
Description (last modified by seanfarley (Sean Farley))
System: dual G4 powerMac running OS X 10.5.8
gfortran-mp-4.[89] fails to compile scipy/fftpack/src/fftpack/cfftb1.f when optimization is set to '-O3'. Reducing the optimization level to '-O2' allows the compilation to succeed. The attached file contains the relevant portion of the build log and a copy of cfftb1.f.
py(27|34)-scipy-0.14.0 built without incident with an earlier version of gcc-mp-4.8.
A quick workaround to get scipy to build is to modify the numpy file that sets the optimization level for the fortran compiler.
--- site-packages/numpy/distutils/fcompiler/gnu.py- 2015-01-04 12:14:16.000000000 -0500 +++ site-packages/numpy/distutils/fcompiler/gnu.py 2015-01-04 12:14:29.000000000 -0500 @@ -200,7 +200,7 @@ # with -O3 caused failures in lib.lapack heevr,syevr tests. opt = ['-O2'] else: - opt = ['-O3'] + opt = ['-O2'] opt.append('-funroll-loops') return opt
Change History (2)
Changed 10 years ago by josephsacco
Attachment: | fortranError.txt added |
---|
comment:1 Changed 10 years ago by seanfarley (Sean Farley)
Description: | modified (diff) |
---|---|
Owner: | changed from macports-tickets@… to sean@… |
Port: | py-numpy added; gcc48 gcc49 removed |
Status: | new → assigned |
This is a patch that needs to be applied to numpy, correct?
Note: See
TracTickets for help on using
tickets.
gfortran command + log file + fortran file