Unable to compile with GLM on Ubuntu, using Clang++ and C++11 due to errors

[AndyEsser]

I'm working on an OpenGL graphics display project, and using GLM as the mathematics library to save me a world of pain. I've used GLM previously and thing it's fantastic. However I'm not able to compile my project currently after trying to include GLM.

I am having a number of errors being thrown about reinterpret cast, or old style casts. I am compiling with Warnings treated as Errors due to the nature of the project as well as a number of other strict compilation flags. Below is the compilation output, and the command line invocation of clang++. I have also included a minimal project zipped up that demonstrates this issue.

Ubuntu 17.04
Clang++ 4.0.0-1ubuntu1

clang++ src/*.cpp -I include/ -o output -g -Weverything -Wpedantic -Werror -Wno-padded -Wno-c++98-compat -Wno-documentation -std=c++11

In file included from src/main.cpp:1: In file included from include/main.hpp:3: In file included from include/glm/glm.hpp:71: In file included from include/glm/detail/../mat2x2.hpp:6: In file included from include/glm/detail/type_mat2x2.hpp:182: In file included from include/glm/detail/type_mat2x2.inl:4: In file included from include/glm/detail/func_matrix.hpp:149: In file included from include/glm/detail/func_matrix.inl:4: In file included from include/glm/detail/../geometric.hpp:6: In file included from include/glm/detail/func_geometric.hpp:113: In file included from include/glm/detail/func_geometric.inl:5: In file included from include/glm/detail/func_common.hpp:426: include/glm/detail/func_common.inl:699:10: error: reinterpret_cast from 'float' to 'int &' has undefined behavior [-Werror,-Wundefined-reinterpret-cast] return reinterpret_cast<int&>(const_cast<float&>(v)); ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ include/glm/detail/func_common.inl:710:10: error: reinterpret_cast from 'float' to 'unsigned int &' has undefined behavior [-Werror,-Wundefined-reinterpret-cast] return reinterpret_cast<uint&>(const_cast<float&>(v)); ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ include/glm/detail/func_common.inl:721:10: error: reinterpret_cast from 'int' to 'float &' has undefined behavior [-Werror,-Wundefined-reinterpret-cast] return reinterpret_cast<float&>(const_cast<int&>(v)); ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ include/glm/detail/func_common.inl:732:10: error: reinterpret_cast from 'uint' (aka 'unsigned int') to 'float &' has undefined behavior [-Werror,-Wundefined-reinterpret-cast] return reinterpret_cast<float&>(const_cast<uint&>(v)); ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ In file included from src/main.cpp:1: In file included from include/main.hpp:3: In file included from include/glm/glm.hpp:71: In file included from include/glm/detail/../mat2x2.hpp:6: In file included from include/glm/detail/type_mat2x2.hpp:182: In file included from include/glm/detail/type_mat2x2.inl:4: In file included from include/glm/detail/func_matrix.hpp:149: In file included from include/glm/detail/func_matrix.inl:4: In file included from include/glm/detail/../geometric.hpp:6: In file included from include/glm/detail/func_geometric.hpp:113: In file included from include/glm/detail/func_geometric.inl:5: In file included from include/glm/detail/func_common.hpp:426: In file included from include/glm/detail/func_common.inl:848: In file included from include/glm/detail/func_common_simd.inl:6: include/glm/detail/../simd/common.h:115:57: error: implicit conversion changes signedness: 'unsigned int' to 'int' [-Werror,-Wsign-conversion] glm_vec4 const sgn0 = _mm_castsi128_ps(_mm_set1_epi32(0x80000000)); ~~~~~~~~~~~~~~ ^~~~~~~~~~ include/glm/detail/../simd/common.h:147:56: error: implicit conversion changes signedness: 'unsigned int' to 'int' [-Werror,-Wsign-conversion] glm_vec4 const sgn0 = _mm_castsi128_ps(_mm_set1_epi32(0x80000000)); ~~~~~~~~~~~~~~ ^~~~~~~~~~ include/glm/detail/../simd/common.h:223:38: error: implicit conversion changes signedness: 'unsigned int' to 'int' [-Werror,-Wsign-conversion] glm_ivec4 const t3 = _mm_set1_epi32(0xFF000000); // exponent mask ~~~~~~~~~~~~~~ ^~~~~~~~~~ include/glm/detail/../simd/common.h:237:61: error: implicit conversion changes signedness: 'unsigned int' to 'int' [-Werror,-Wsign-conversion] return _mm_castsi128_ps(_mm_cmpeq_epi32(t2, _mm_set1_epi32(0xFF000000))); // exponent is all 1s, fraction is 0 ~~~~~~~~~~~~~~ ^~~~~~~~~~ In file included from src/main.cpp:1: In file included from include/main.hpp:3: In file included from include/glm/glm.hpp:71: In file included from include/glm/detail/../mat2x2.hpp:6: In file included from include/glm/detail/type_mat2x2.hpp:182: In file included from include/glm/detail/type_mat2x2.inl:4: In file included from include/glm/detail/func_matrix.hpp:149: In file included from include/glm/detail/func_matrix.inl:4: In file included from include/glm/detail/../geometric.hpp:6: In file included from include/glm/detail/func_geometric.hpp:113: In file included from include/glm/detail/func_geometric.inl:5: In file included from include/glm/detail/func_common.hpp:426: In file included from include/glm/detail/func_common.inl:848: include/glm/detail/func_common_simd.inl:194:40: error: use of old-style cast [-Werror,-Wold-style-cast] __m128i const Load = _mm_set_epi32(-(int)a.w, -(int)a.z, -(int)a.y, -(int)a.x); ^ ~~~ include/glm/detail/func_common_simd.inl:194:51: error: use of old-style cast [-Werror,-Wold-style-cast] __m128i const Load = _mm_set_epi32(-(int)a.w, -(int)a.z, -(int)a.y, -(int)a.x); ^ ~~~ include/glm/detail/func_common_simd.inl:194:62: error: use of old-style cast [-Werror,-Wold-style-cast] __m128i const Load = _mm_set_epi32(-(int)a.w, -(int)a.z, -(int)a.y, -(int)a.x); ^ ~~~ include/glm/detail/func_common_simd.inl:194:73: error: use of old-style cast [-Werror,-Wold-style-cast] __m128i const Load = _mm_set_epi32(-(int)a.w, -(int)a.z, -(int)a.y, -(int)a.x); ^ ~~~ In file included from src/main.cpp:1: In file included from include/main.hpp:3: In file included from include/glm/glm.hpp:71: In file included from include/glm/detail/../mat2x2.hpp:6: In file included from include/glm/detail/type_mat2x2.hpp:182: In file included from include/glm/detail/type_mat2x2.inl:4: In file included from include/glm/detail/func_matrix.hpp:149: In file included from include/glm/detail/func_matrix.inl:399: include/glm/detail/func_matrix_simd.inl:22:6: error: use of old-style cast [-Werror,-Wold-style-cast] *(glm_vec4 const (*)[4])&x[0].data, ^ ~~~~~~~~~~ include/glm/detail/func_matrix_simd.inl:23:6: error: use of old-style cast [-Werror,-Wold-style-cast] *(glm_vec4 const (*)[4])&y[0].data, ^ ~~~~~~~~~~ include/glm/detail/func_matrix_simd.inl:24:6: error: use of old-style cast [-Werror,-Wold-style-cast] *(glm_vec4(*)[4])&result[0].data); ^ ~~~~~~~~~~~~~~~ include/glm/detail/func_matrix_simd.inl:36:6: error: use of old-style cast [-Werror,-Wold-style-cast] *(glm_vec4 const (*)[4])&m[0].data, ^ ~~~~~~~~~~ include/glm/detail/func_matrix_simd.inl:37:6: error: use of old-style cast [-Werror,-Wold-style-cast] *(glm_vec4(*)[4])&result[0].data); ^ ~~~~~~~~~~~~~~~ include/glm/detail/func_matrix_simd.inl:67:42: error: dereference of type '__m128 (*)[4]' that was reinterpret_cast from type 'typename detail::storage<float, sizeof(float) * 4, detail::is_aligned<(precision)5U>::value>::type *' (aka '__m128 *') has undefined behavior [-Werror,-Wundefined-reinterpret-cast] glm_mat4_outerProduct(c.data, r.data, *reinterpret_cast<__m128(*)[4]>(&m[0].data)); ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ include/glm/detail/func_matrix_simd.inl:75:42: error: dereference of type '__m128 (*)[4]' that was reinterpret_cast from type 'typename detail::storage<float, sizeof(float) * 4, detail::is_aligned<(precision)4U>::value>::type *' (aka '__m128 *') has undefined behavior [-Werror,-Wundefined-reinterpret-cast] glm_mat4_outerProduct(c.data, r.data, *reinterpret_cast<__m128(*)[4]>(&m[0].data)); ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ fatal error: too many errors emitted, stopping now [-ferror-limit=] 20 errors generated. makefile:11: recipe for target 'build' failed make: *** [build] Error 1

glm-issue.tar.gz

[Groovounet]

Hi,
Could you #define GLM_FORCE_MESSAGES before including GLM, build, and report here what GLM says?
Thanks!

[AndyEsser]

Hi,

Apologies for the slow response.

I have #define'd GLM_FORCE_MESSAGES before the include and rebuilt (I've done this on the ZIP file project I included to avoid any extra issues being dragged in). Below is the output:

GLM: Linux platform detected
GLM: version 0.9.8.4
GLM: Clang compiler detected
GLM: 64 bits model
GLM: SSE2 instruction set
GLM: C++11
GLM: Swizzling operators disabled, #define GLM_SWIZZLE to enable swizzle operators
GLM: Depth clip space: negative one to one
GLM: Coordinate system: right handed
GLM: .length() returns glm::length_t, a typedef of int following the GLSL specification
GLM: Core library included

Edited for readability

[Groovounet]

Hi,

I think the issue is that these arguments -g -Weverything -Wpedantic -Werror -Wno-padded -Wno-c++98-compat -Wno-documentation -std=c++11 are just way too strict for GLM.

Thanks for reporting,
Christophe

[AndyEsser]

Many thanks for even bothering to take a look at it, it's much appreciated :).

Unfortunately for this project I'll have to ditch GLM then and roll my own. I shall continue using GLM for my other less strict projects, many thanks

[Groovounet]

A lot of warnings has been resolved in master branch. Still quite a few warnings explicitly ignored in the CMakeLists.

Thanks,
Christophe