pyFFTW import segfaults on armhf

[itdaniher]

Processor : ARMv7 Processor rev 5 (v7l)
Features : swp half thumb fastmult vfp edsp thumbee neon vfpv3 tls vfpv4 idiva idivt
gcc version 4.9.2 (Debian 4.9.2-10)
Python 3.6.0 (default, Dec 31 2016, 21:20:16)
Linux debian 3.4.113-sun8i #50 SMP PREEMPT Mon Nov 14 08:41:55 CET 2016 armv7l GNU/Linux

Using built-in specs.
COLLECT_GCC=gcc-4.9
COLLECT_LTO_WRAPPER=/usr/lib/gcc/arm-linux-gnueabihf/4.9/lto-wrapper
Target: arm-linux-gnueabihf
Configured with: ../src/configure -v --with-pkgversion='Debian 4.9.2-10' --with-bugurl=file:///usr/share/doc/gcc-4.9/README.Bugs --enable-languages=c,c++,java,go,d,fortran,objc,obj-c++ --prefix=/usr --program-suffix=-4.9 --enable-shared --enable-linker-build-id --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --with-gxx-include-dir=/usr/include/c++/4.9 --libdir=/usr/lib --enable-nls --with-sysroot=/ --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-gnu-unique-object --disable-libitm --disable-libquadmath --enable-plugin --with-system-zlib --disable-browser-plugin --enable-java-awt=gtk --enable-gtk-cairo --with-java-home=/usr/lib/jvm/java-1.5.0-gcj-4.9-armhf/jre --enable-java-home --with-jvm-root-dir=/usr/lib/jvm/java-1.5.0-gcj-4.9-armhf --with-jvm-jar-dir=/usr/lib/jvm-exports/java-1.5.0-gcj-4.9-armhf --with-arch-directory=arm --with-ecj-jar=/usr/share/java/eclipse-ecj.jar --enable-objc-gc --enable-multiarch --disable-sjlj-exceptions --with-arch=armv7-a --with-fpu=vfpv3-d16 --with-float=hard --with-mode=thumb --enable-checking=release --build=arm-linux-gnueabihf --host=arm-linux-gnueabihf --target=arm-linux-gnueabihf
Thread model: posix

libfftw built from http://fftw.org/fftw-3.3.6-pl1.tar.gz the requisite three times, one of which with --enable-neon for simd-accelerated singlefp work.

./configure --enable-shared --enable-threads --enable-long-double
./configure --enable-shared --enable-threads --enable-neon --enable-single
./configure --enable-shared --enable-threads

---

I think this is an issue with libfftw and (presumably) my GCC version. I'm continuing to explore this - opening issue here as there's a nontrivial amount of discussion about pyfftw on armhf, #128, #127, #70

[itdaniher]

Odds are high that this is an issue merely with my toolchain... close the issue if this discussion is unwelcome.

FFTW's test fails:

./tests/bench --report-verbose -s irc32
[2] 20400 segmentation fault ./tests/bench --report-verbose -s irc32

[itdaniher]

Issue's with FFTW!

FFTW/fftw3#30 (comment)

[hgomersall]

Oh sorry, I totally missed this issue! Interesting to know about the problem, so thanks.

[itdaniher]

@hgomersall NP, hope it helps. PyFFTW is great, FFTW just sometimes feels like it's rooted firmly in the past. IE: A release claiming to officially support ARM+NEON that segfaults when compiled on multiple versions of widely-deployed GCC, unless you patch out a feature? sigh

Other issue I encountered here was just the under-documented "you must compile fftw by hand three times" which I've seen mentioned in several different places here on GH Issues.

[hgomersall]

@itdaniher Yes, I'm acutely aware of that issue. I've been very busy for quite a while. I would genuinely value assistance with any of the issues, but I've no doubt I will eventually reach them!