Vulkan tests fail on 32bit x86 systems

[tatokis]

The Vulkan tests fail on 32bit x86 systems with errors such as:

[root@arch32-builder3 glad-2.0.6]# gcc -Wall -Wextra -Werror -Wsign-conversion -Wcast-qual -Wstrict-prototypes -ansi -pedantic -save-temps /build/glad/src/glad-2.0.6/test/c/compile/vulkan/alias/001/test.c -o /build/glad/src/glad-2.0.6/build/test -I/build/glad/src/glad-2.0.6/build/include /build/glad/src/glad-2.0.6/build/src/vulkan.c -ldl
In file included from /build/glad/src/glad-2.0.6/test/c/compile/vulkan/alias/001/test.c:9:
/build/glad/src/glad-2.0.6/build/include/glad/vulkan.h:4453:1: error: integer constant is too large for 'long' type [-Werror=long-long]      
 4453 | static const VkAccessFlagBits2 VK_ACCESS_2_SHADER_SAMPLED_READ_BIT = 4294967296;

This is because the tests are compiled with -ansi, but on 32bit x86 the only way to get a 64 bit integer is by using long long, which requires C99.

Passing -std=c99 instead of -ansi resolves it. I am unsure as to how this should be properly resolved, as I assume you explicitly want to verify the code builds as C90 on other platforms.

[Dav1dde]

Hey, thanks for the report, seems like glad should generate the constant as:

 static const VkAccessFlagBits2 VK_ACCESS_2_SHADER_SAMPLED_READ_BIT = 4294967296ULL;

[tatokis]

Thanks for the quick response. I just tested it, and I'm not sure that works.

[root@arch32-builder3 /]# cat test.c 
#include <inttypes.h>
typedef uint64_t VkAccessFlagBits2;
static const VkAccessFlagBits2 VK_ACCESS_2_SHADER_SAMPLED_READ_BIT = 4294967296ULL;

int main() { return 0; }
[root@arch32-builder3 /]# gcc -ansi -pedantic test.c
test.c:3:70: warning: use of C99 long long integer constant [-Wlong-long]
    3 | static const VkAccessFlagBits2 VK_ACCESS_2_SHADER_SAMPLED_READ_BIT = 4294967296ULL;
      |                                                                      ^~~~~~~~~~~~~

[Dav1dde]

Yeah seems to be a C99 thing :(