Numpy: Failed to build v1.17.0 on CentOS7 with default gcc (4.8)
Cannot build numpy v1.17.0 on CentOS 7.
Previous version works fine.
I'm seeing a lot of C99/C11 features in error message, and avx512f.
Not sure if the required --std option is set but I don't think gcc 4.8.5 (the system default compiler on CentOS 7) can handle all of that.
Reproducing code example:
Simply build the wheel on CentOS 7 with python3.
Error message:
Some error log attached.
The entire story is too long to be pasted here.
#6 356.1 compile options: '-Inumpy/core/src/common -Inumpy/core/src -Inumpy/core -Inumpy/core/src/npymath -Inumpy/core/src/multiarray -Inumpy/core/src/umath -Inumpy/core/src/npysort -I/usr/include/python3.6m -c'
#6 356.1 gcc: _configtest.c
#6 356.1 _configtest.c: In function 'main':
#6 356.1 _configtest.c:5:3: warning: implicit declaration of function '__builtin_mul_overflow' [-Wimplicit-function-declaration]
#6 356.1 __builtin_mul_overflow(5, 5, (int*)5);
#6 356.1 ^
#6 356.1 gcc -pthread _configtest.o -o _configtest
#6 356.1 _configtest.o: In function `main':
#6 356.1 /tmp/pip-req-build-poiltggx/_configtest.c:5: undefined reference to `__builtin_mul_overflow'
#6 356.1 collect2: error: ld returned 1 exit status
#6 356.1 failure.
#6 356.1 removing: _configtest.c _configtest.o _configtest.o.d
#6 356.1 C compiler: gcc -pthread -Wno-unused-result -Wsign-compare -DDYNAMIC_ANNOTATIONS_ENABLED=1 -DNDEBUG -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -m64 -mtune=generic -D_GNU_SOURCE -fPIC -fwrapv -fPIC
#6 356.1
#6 356.1 compile options: '-Inumpy/core/src/common -Inumpy/core/src -Inumpy/core -Inumpy/core/src/npymath -Inumpy/core/src/multiarray -Inumpy/core/src/umath -Inumpy/core/src/npysort -I/usr/include/python3.6m -c'
#6 356.1 gcc: _configtest.c
#6 356.1 _configtest.c: In function 'main':
#6 356.1 _configtest.c:7:16: warning: unused variable 'r' [-Wunused-variable]
#6 356.1 volatile int r = __builtin_cpu_supports("sse");
#6 356.1 ^
#6 356.1 gcc -pthread _configtest.o -o _configtest
#6 356.1 success!
#6 356.1 removing: _configtest.c _configtest.o _configtest.o.d _configtest
#6 356.1 C compiler: gcc -pthread -Wno-unused-result -Wsign-compare -DDYNAMIC_ANNOTATIONS_ENABLED=1 -DNDEBUG -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -m64 -mtune=generic -D_GNU_SOURCE -fPIC -fwrapv -fPIC
#6 356.1
#6 356.1 compile options: '-Inumpy/core/src/common -Inumpy/core/src -Inumpy/core -Inumpy/core/src/npymath -Inumpy/core/src/multiarray -Inumpy/core/src/umath -Inumpy/core/src/npysort -I/usr/include/python3.6m -c'
#6 356.1 gcc: _configtest.c
#6 356.1 _configtest.c: In function 'main':
#6 356.1 _configtest.c:7:3: error: Parameter to builtin not valid: avx512f
#6 356.1 volatile int r = __builtin_cpu_supports("avx512f");
#6 356.1 ^
#6 356.1 _configtest.c:7:16: warning: unused variable 'r' [-Wunused-variable]
#6 356.1 volatile int r = __builtin_cpu_supports("avx512f");
#6 356.1 ^
#6 356.1 failure.
#6 356.1 removing: _configtest.c _configtest.o
#6 356.1 C compiler: gcc -pthread -Wno-unused-result -Wsign-compare -DDYNAMIC_ANNOTATIONS_ENABLED=1 -DNDEBUG -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -m64 -mtune=generic -D_GNU_SOURCE -fPIC -fwrapv -fPIC
#6 356.1
#6 356.1 compile options: '-Inumpy/core/src/common -Inumpy/core/src -Inumpy/core -Inumpy/core/src/npymath -Inumpy/core/src/multiarray -Inumpy/core/src/umath -Inumpy/core/src/npysort -I/usr/include/python3.6m -c'
#6 356.1 gcc: _configtest.c
#6 356.1 gcc -pthread _configtest.o -o _configtest
#6 356.1 success!
#6 356.1 removing: _configtest.c _configtest.o _configtest.o.d _configtest
#6 356.1 C compiler: gcc -pthread -Wno-unused-result -Wsign-compare -DDYNAMIC_ANNOTATIONS_ENABLED=1 -DNDEBUG -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -m64 -mtune=generic -D_GNU_SOURCE -fPIC -fwrapv -fPIC
#6 356.1
#6 356.1 compile options: '-Inumpy/core/src/common -Inumpy/core/src -Inumpy/core -Inumpy/core/src/npymath -Inumpy/core/src/multiarray -Inumpy/core/src/umath -Inumpy/core/src/npysort -I/usr/include/python3.6m -c'
#6 356.1 gcc: _configtest.c
#6 356.1 gcc -pthread _configtest.o -o _configtest
#6 356.1 success!
#6 356.1 removing: _configtest.c _configtest.o _configtest.o.d _configtest
#6 358.3 C compiler: gcc -pthread -Wno-unused-result -Wsign-compare -DDYNAMIC_ANNOTATIONS_ENABLED=1 -DNDEBUG -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -m64 -mtune=generic -D_GNU_SOURCE -fPIC -fwrapv -fPIC
#6 358.3
#6 358.3 compile options: '-Ibuild/src.linux-x86_64-3.6/numpy/core/src/common -Inumpy/core/include -Ibuild/src.linux-x86_64-3.6/numpy/core/include/numpy -Inumpy/core/src/common -Inumpy/core/src -Inumpy/core -Inumpy/core/src/npymath -Inumpy/core/src/multiarray -Inumpy/core/src/umath -Inumpy/core/src/npysort -I/usr/include/python3.6m -Ibuild/src.linux-x86_64-3.6/numpy/core/src/common -Ibuild/src.linux-x86_64-3.6/numpy/core/src/npymath -Ibuild/src.linux-x86_64-3.6/numpy/core/src/common -Ibuild/src.linux-x86_64-3.6/numpy/core/src/npymath -c'
#6 358.3 gcc: build/src.linux-x86_64-3.6/numpy/core/src/npysort/radixsort.c
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'radixsort_bool':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:112:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src:112:5: note: use option -std=c99 or -std=gnu99 to compile your code
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'aradixsort_bool':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:202:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'radixsort_byte':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:112:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'aradixsort_byte':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:202:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'radixsort_ubyte':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:112:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'aradixsort_ubyte':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:202:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'radixsort_short':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:112:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'aradixsort_short':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:202:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'radixsort_ushort':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:112:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'aradixsort_ushort':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:202:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'radixsort_int':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:112:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'aradixsort_int':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:202:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'radixsort_uint':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:112:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'aradixsort_uint':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:202:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'radixsort_long':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:112:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'aradixsort_long':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:202:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'radixsort_ulong':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:112:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'aradixsort_ulong':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:202:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'radixsort_longlong':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:112:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'aradixsort_longlong':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:202:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'radixsort_ulonglong':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:112:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 numpy/core/src/npysort/radixsort.c.src: In function 'aradixsort_ulonglong':
#6 358.3 numpy/core/src/npysort/radixsort.c.src:202:5: error: 'for' loop initial declarations are only allowed in C99 mode
#6 358.3 for (npy_intp i = 1; i < num; i++) {
#6 358.3 ^
#6 358.3 error: Command "gcc -pthread -Wno-unused-result -Wsign-compare -DDYNAMIC_ANNOTATIONS_ENABLED=1 -DNDEBUG -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -m64 -mtune=generic -D_GNU_SOURCE -fPIC -fwrapv -fPIC -Ibuild/src.linux-x86_64-3.6/numpy/core/src/common -Inumpy/core/include -Ibuild/src.linux-x86_64-3.6/numpy/core/include/numpy -Inumpy/core/src/common -Inumpy/core/src -Inumpy/core -Inumpy/core/src/npymath -Inumpy/core/src/multiarray -Inumpy/core/src/umath -Inumpy/core/src/npysort -I/usr/include/python3.6m -Ibuild/src.linux-x86_64-3.6/numpy/core/src/common -Ibuild/src.linux-x86_64-3.6/numpy/core/src/npymath -Ibuild/src.linux-x86_64-3.6/numpy/core/src/common -Ibuild/src.linux-x86_64-3.6/numpy/core/src/npymath -c build/src.linux-x86_64-3.6/numpy/core/src/npysort/radixsort.c -o build/temp.linux-x86_64-3.6/build/src.linux-x86_64-3.6/numpy/core/src/npysort/radixsort.o -MMD -MF build/temp.linux-x86_64-3.6/build/src.linux-x86_64-3.6/numpy/core/src/npysort/radixsort.o.d" failed with exit status 1
#6 358.3 ----------------------------------------
#6 358.4 ERROR: Command errored out with exit status 1: /usr/bin/python3 -u -c 'import sys, setuptools, tokenize; sys.argv[0] = '"'"'/tmp/pip-req-build-poiltggx/setup.py'"'"'; __file__='"'"'/tmp/pip-req-build-poiltggx/setup.py'"'"';f=getattr(tokenize, '"'"'open'"'"', open)(__file__);code=f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, __file__, '"'"'exec'"'"'))' install --record /tmp/pip-record-50nzprh1/install-record.txt --single-version-externally-managed --compile Check the logs for full command output.
Numpy/Python version information:
CentOS 7
Default python 3 (3.6)
Default gcc (4.8)
xkszltl
All 23 comments
Full log attached: numpy.log
xkszltl
on 29 Jul 2019
Numpy 1.17 needs at least --std=c99.
I think the decision was to leave passing that flag up to the distributor rather than doing it automatically, but I forget the rationale. @rgommers, were you part of that discussion?
eric-wieser
on 29 Jul 2019
That sounds terrible if it's not set by default.
However, it somehow works with Ubuntu 18.04.
Besides, what's the correct way to pass in compiler flags?
I tried export CC="gcc --std=c11" and export CFLAGS="--std=c11" but both were ignored.
xkszltl
on 29 Jul 2019
Looks like on Ubuntu 18.04, gcc 7.4.0 defaults to gnu11 while on CentOS 7, gcc 4.8.5 defaults to gnu90.
I run . scl_source enable devtoolset-8 to switch to gcc 8, which uses gnu17 by default.
But seems all related env vars were ignored by numpy.
xkszltl
on 29 Jul 2019
Never mind, numpy can pick up the env correctly.
There's a sudo before pip install which drops everything.
But the issue remains: Do we still have the default support for CentOS 7 and its gcc 4.8.5?
xkszltl
on 29 Jul 2019
I don't know if it's related but I got a similar problem trying to install on Amazon Linux. The following work around helped me:
update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc48-c99 100 --slave /usr/bin/c89 c89 /usr/bin/gcc48-c89 --slave /usr/bin/c99 c99 /usr/bin/gcc48-c99 --slave /usr/bin/cc cc /usr/bin/gcc48 --slave /usr/bin/gcov gcov /usr/bin/gcov48
update-alternatives --config gcc
paulofbmarcon
on 31 Jul 2019
@xkszltl You should pass --std=c99 in the CFLAGS environmental variable. Following solved build issues for me.
CFLAGS="--std=c99" python3.6 -mpip wheel "numpy>=1.17,<1.18"
sashkab
on 31 Jul 2019
@sashkab I got it resolved for me by simply using gcc-8. But should --std=c99 be added automatically during build if there's no user setting detected?
xkszltl
on 1 Aug 2019
numpy 1.17 is also failing to build on pypy, using a ubuntu 14.04 image: https://travis-ci.org/antocuni/pypy-wheels/jobs/566563056
it seems to be that numpy should try harder to compile even on older systems, unless you have a good reason not to.
antocuni
on 1 Aug 2019
I think passing in the std flags by default in a compiler and platform-agnostic manner can be tricky & as with many distutils things, hard to test completely.
tylerjereddy
on 5 Aug 2019
Numpy 1.17 needs at least
--std=c99.I think the decision was to leave passing that flag up to the distributor rather than doing it automatically, but I forget the rationale. @rgommers, were you part of that discussion?
That rings a bell. I attempted to fix this before, and concluded there's no good place to insert the fix: https://github.com/numpy/numpy/pull/12620#issuecomment-451348138. This attempt has some comment as to what the problem is: https://github.com/rgommers/numpy/commit/8aaf4d8ff4828cd7302788af05f65f667ff1ba45
Taking a step back, the numpy.distutils design problem we run into here is that all flags are (a) taken over from Python distutils, then hardcoded per compiler in numpy/distutils/<compilername>.py. And at the time we do that hardcoding, we don't know what comes from distutils. And inserting flags dynamically later on (what I tried) is equally tricky.
It's possible I missed a sane way to fix this. Suggestions/PRs welcome.
IIRC the minimum gcc version that works for building numpy at all is 4.8. For gcc <5.1, users need to add --std=c99.
This is a duplicate of gh-12766.
rgommers
on 5 Aug 2019
This is biting enough people that I'll mark it for 1.17.1 - should be possible to insert a hacky fix somehow.
rgommers
on 6 Aug 2019
Most linux distributions should have a c99 compiler script that sets the flag
charris@fc [~]$ c99 --version
gcc (GCC) 9.1.1 20190503 (Red Hat 9.1.1-1)
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
@xkszltl What does that do for you? There ought to be a way we can use that on linux platforms.
charris
on 6 Aug 2019
@charris
Good to know! It's part of gcc 4.8.5 rpm in CentOS so I think it should work.
Also available for SCL (I checked devtoolset-6/7/8).
And...Wow you guys already have gcc-9 ?!
Waiting for SCL on CentOS...
xkszltl
on 6 Aug 2019
BTW it'll be better if we can find a way to inject $CFLAGS automatically instead of searching for c99 executable as $CC, because $CC may be further customized to gcc-xxx, clang, ccache gcc, distcc...
xkszltl
on 6 Aug 2019
Pushed this off to 1.17.2. Tempted to remove the milestone as there is a workaround and the question seems to be "is there a better way."
charris
on 20 Aug 2019
Similar issue here on CentOS 7, python 3.6.3+ (configured with --with-pydebug), and gcc 4.8.5. This worked for me when activated into my venv
CFLAGS=-std=c99 pip install numpy==1.17.0
oolongtea
on 22 Aug 2019
This issue also affects RHEL and RHEL container images, pip installing numpy fails on non-intel architectures.
torsava
on 29 Aug 2019
Fixed in gh-14771, which added both the flag and a CI run with gcc4.8
mattip
on 11 Nov 2019
The issue is reproducible on nvidia/cuda-ppc64le:10.0-cudnn7-devel-centos7
sh1ng
on 18 Mar 2020
@sh1ng with what version of NumPy? Are you adding the -std=c99 flag?
mattip
on 18 Mar 2020
Without adding the flag
| Numpy version | Status |
| ------------- | -----:|
|1.17.0|Fail|
|1.17.1|Fail|
|1.17.2|Fail|
|1.17.3|Fail|
|1.17.4|Fail|
|1.17.5|Fail|
|1.18.0|Success|
|1.18.1|Fail|
|1.18.2|Fail|
Adding it do solves the issue, but makes the installation process platform dependent.
sh1ng
on 18 Mar 2020
@mattip the issue occurs when miniconda is used, not a centos python package. Most likely conda's issue.
sh1ng
on 18 Mar 2020
Related issues
marcocaccin
·
4Comments
kevinzhai80
·
4Comments
perezpaya
·
4Comments
ghost
·
4Comments
Foadsf
·
3Comments
Most helpful comment
@xkszltl You should pass
--std=c99in theCFLAGSenvironmental variable. Following solved build issues for me.