Issue for building for Raspberry PI4

[paullee2018]

hi,

I followed this tutorial to install the libepoxy.

https://www.raspberrypi.org/documentation/linux/software/libcamera/README.md

but I get the following error. It seems the header generated has khronos_uintptr_t declared twice. I deleted the last one and build/install works fine.

In file included from ../include/epoxy/gl.h:89, from ../src/dispatch_common.h:44, from ../src/dispatch_egl.c:28: include/epoxy/gl_generated.h:54:19: error: conflicting types for ‘khronos_uintptr_t’ typedef uintptr_t khronos_uintptr_t; ^~~~~~~~~~~~~~~~~ include/epoxy/gl_generated.h:42:32: note: previous declaration of ‘khronos_uintptr_t’ was here typedef unsigned long int khronos_uintptr_t; ^~~~~~~~~~~~~~~~~

Regards
Paul

[ebassi]

The generated files do not match these errors. The khronos_uintptr_t definition appears only once.

Have you tried building Epoxy multiple times? Care to try with:

git clean -xdf
meson setup _build .
meson compile -C _build
meson install -C _build

instead?

[paullee2018]

meson install -C _build

Still complains when I run meson compile -C _build after running the first 2 commands

[ebassi]

Care to attach your include/epoxy/gl_generated.h to this issue?

[YuriSedunov]

This issue reproduced for all 32-bit architectures.

20 #pragma once
21 #include <inttypes.h>
22 #include <stddef.h>
23
24 #include "epoxy/common.h"
25 #define _khrplatform_h 1
26 typedef int8_t khronos_int8_t;
27 typedef int16_t khronos_int16_t;
28 typedef int32_t khronos_int32_t;
29 typedef int64_t khronos_int64_t;
30 typedef uint8_t khronos_uint8_t;
31 typedef uint16_t khronos_uint16_t;
32 typedef uint32_t khronos_uint32_t;
33 typedef uint64_t khronos_uint64_t;
34 typedef float khronos_float_t;
35 #ifdef _WIN64
36 typedef signed long long int khronos_intptr_t;
37 typedef unsigned long long int khronos_uintptr_t;
38 typedef signed long long int khronos_ssize_t;
39 typedef unsigned long long int khronos_usize_t;
40 #else
41 typedef signed long int khronos_intptr_t;
42 typedef unsigned long int khronos_uintptr_t;
43 typedef signed long int khronos_ssize_t;
44 typedef unsigned long int khronos_usize_t;
45 #endif
46 typedef uint64_t khronos_utime_nanoseconds_t;
47 typedef int64_t khronos_stime_nanoseconds_t;
48 #define KHRONOS_MAX_ENUM 0x7FFFFFFF
49 typedef enum {
50 KHRONOS_FALSE = 0,
51 KHRONOS_TRUE = 1,
52 KHRONOS_BOOLEAN_ENUM_FORCE_SIZE = KHRONOS_MAX_ENUM
53 } khronos_boolean_enum_t;
54 typedef uintptr_t khronos_uintptr_t;
55 typedef unsigned int GLenum;
56 typedef unsigned char GLboolean;
57 typedef unsigned int GLbitfield;
58 typedef void GLvoid;
59 typedef khronos_int8_t GLbyte;
60 typedef khronos_uint8_t GLubyte;

[ebassi]

Ah, yes, found it.

There's a redeclaration of khronos_uintptr_t because I missed the typedef below. On IA and 64 bit, and with GCC or C11, the redeclaration is allowed because the type is the same, so I couldn't see the build breaking locally or on the CI.

It would be nice to have a 32bit build to check this stuff.

I'll open a MR with a possible fix.

[ebassi]

@YuriSedunov @paullee2018 could you try the build with #250 applied?

[YuriSedunov]

Yes, build passed for all architectures.
http://git.altlinux.org/tasks/271008/

[ebassi]

Thanks!