Skip to content

f2py build hangs after running gcc w/1.15.0 #11649

@briantoby

Description

@briantoby

Sorry for a rather incomplete bug report, but isolating this to a code fragment may not be possible. I am happy to provide more help in diagnosis, though.

I have an automated Jenkins build procedure that uses f2py to generate .so files from a series of Fortran files. With anaconda packages

numpy-1.14.5               |   py27h1b885b7_4          35 KB
numpy-base-1.14.5          |   py27hdbf6ddf_4         4.1 MB

running on VM with Red Hat Enterprise Linux Server release 7.5 (Maipo), my build runs fine. If I update to

numpy                     1.15.0           py27h1b885b7_0  
numpy-base                1.15.0           py27h3dfced4_0  

The build hangs in this command:

jenkins  23248 23087  0 12:10 ?        00:00:02 /var/lib/jenkins/jobs/Py2_L64_Build-copy/workspace/conda27/bin/python /var/lib/jenkins/jobs/Py2_L64_Build-copy/workspace/conda27/bin/f2py -c fellipse.for -m fellipse libspsubs.a libtexturesubs.a libDIFFaXsubs.a libpowsubs.a --fcompiler=gnu95 --f77exec=gfortran --f77flags=-fno-range-check -m64

with this following output:

building 'fellipse' extension
compiling C sources
C compiler: gcc -pthread -B /var/lib/jenkins/jobs/Py2_L64_Build-copy/workspace/conda27/compiler_compat -Wl,--sysroot=/ -fno-strict-aliasing -g -O2 -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -fPIC

creating /tmp/tmpnkv0Hd/tmp
creating /tmp/tmpnkv0Hd/tmp/tmpnkv0Hd
creating /tmp/tmpnkv0Hd/tmp/tmpnkv0Hd/src.linux-x86_64-2.7
compile options: '-I/tmp/tmpnkv0Hd/src.linux-x86_64-2.7 -I/var/lib/jenkins/jobs/Py2_L64_Build-copy/workspace/conda27/lib/python2.7/site-packages/numpy/core/include -I/var/lib/jenkins/jobs/Py2_L64_Build-copy/workspace/conda27/include/python2.7 -c'
gcc: /tmp/tmpnkv0Hd/src.linux-x86_64-2.7/fellipsemodule.c
gcc: /tmp/tmpnkv0Hd/src.linux-x86_64-2.7/fortranobject.c

It apparently completes the gcc step and then hangs. It is not in an infinite loop, since the CPU time is not increasing. This is problem reproducible and restoring the previous numpy version addresses it.

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions